**Section 3**

**Materials Technology** 

336 Infrared Spectroscopy – Materials Science, Engineering and Technology

Amerio, E.; Sangermano, M.; Colucci, G.; Malucelli, G.; Messori, M.; Taurino, R. & Fabbri, P.

Beyazit, S.; Aydogan, B.; Osken, I.; Ozturk, T. & Yagci, Y. (2011). Long wavelength

Decker, C. & Bendaika T. (1984). Photopolymérisation de macromères multifonctionnels - I. Étude Cinétique Original Research *Europ. Polym. J.*, 20, pp(753-758). Decker, C. & Moussa, K. (1988). A new method for monitoring ultra-fast

Fouassier, J.P. & Rabek J.F. (1993). *Radiation Curing in Polym. Sci. Tech.*, Vol. I–IV, Elsevier,

Lombardi, M.; Guerriero, A.; Kortaberria, G.; Mondragon, I.; Sangermano M. & Montanaro,

Lovelh, L.G.; Newman, S.M. & Bowman, C.N. (1999) The Effects of Light Intensity,

Sangermano, M.; Pegel, S.; Pötschke, P. & Voit B. (2008). Antistatic epoxy coatings with

Studer, K.; Decker*,* C.; Beck, E. & Schwalm, R. (2003) Overcoming oxygen inhibition in UV-

in the presence of onium salts *Polym. Chem.*, 2 pp(1185-1189).

BaTiO3-acrylic composites *Polym. Compos.*, 32, pp (1304-1312).

Dimethacrylate Dental Resins *J Dent Res*, 78, pp (1469-1476).

*Macromol. Mater. Eng.*, 293, pp(700-707).

189, pp(2381-2394).

London.

5 pp(396-400).

pp(92–100).

(2008) Organic-inorganic hybrid coatings containing polyhedral oligomeric silsesquioxane (POSS) blocks: UV curing and characterization of methacrylic films

photoinitiated free radical polymerization using conjugated thiophene derivatives

photopolymerizations by real-time infra-red (RTIR) spectroscopy *Makromol. Chem.*,

L. (2011) Effect of the ceramic filler features on the properties of photo-polymerized

Temperature, and Comonomer Composition on the Polymerization Behavior of

carbon nanotubes obtained by cationic photopolymerization *Macromol. Rap. Comm.*,

curing of acrylate coatings by carbon dioxide inerting, Part I *Prog. Org. Coat.*, 48

**4. References** 

**17** 

**Characterization of** 

Alata Hexig and Bayar Hexig\* *Department of Biomolecular Engineering,* 

*Midori-ku, Yokohama* 

*Japan* 

*Graduate School of Bioscience and Biotechnology, Tokyo Institute of Technology, Nagatsuta-cho,* 

**Compositional Gradient Structure of** 

**Polymeric Materials by FTIR Technology** 

Biomimetic and bioinspired optimal structures combining bioresorbable, bioactive and other advanced properties are expected for the next generation of biomaterials (Bruck et al., 2002, Hench et al., 2002, Akaike et al., 2005). Inspired by nature, to reveal the relationship between structure and functionality of biological materials has been emphasized in the biomaterials research field. In nature, gradient biological structures exist most commonly, such as the structure of bamboo (Amada et al., 1997), shells, teeth, bones, tendon and extracellular matrix (ECM) (Suresh et al., 2001). Man-made functionally gradient materials (FGMs) have been developed for combining irreconcilable properties within a single material and have been widely incorporated in metal/ceramic and organic/inorganic material fields for increasing the structural complexity and combining different functionality (Suganuma et al., 1983, Ishikawa et al., 2002, Czubarow et al., 1997). FGMs are spatial composites within which the composition or structure and thus the functions of the materials continuously or step-wisely vary along the specific coordinates. Examples of the structural change include

The concept of FGMs first originated in Japan in 1984-1985 in the spacecraft project (Kawasaki et al., 1987). FGMs are developed to give two conflicting properties such as good thermal conductivity and good thermal resistance into one material. In general, we see uniformed functions and properties within the materials we use. However, in FGMs we see different functions between one part and the other of them. In a metal-ceramic FGM, the metal-rich side is placed in the region where mechanical performance, such as toughness, needs to be stronger; and the ceramic-rich side, which has better thermal resistance, is exposed to high temperatures, or placed in the region where there is a potentially severe temperature variation (Dao et al., 1997). In FGMs, the properties change gradually with position due to the spatial gradients in composition, density, microstructural arrangement or atomic-order, which

those in crystal structure and orientation, porosity, particle size and so on.

**1. Introduction** 

 \*

Corresponding Author

## **Characterization of Compositional Gradient Structure of Polymeric Materials by FTIR Technology**

Alata Hexig and Bayar Hexig\*

*Department of Biomolecular Engineering, Graduate School of Bioscience and Biotechnology, Tokyo Institute of Technology, Nagatsuta-cho, Midori-ku, Yokohama Japan* 

## **1. Introduction**

Biomimetic and bioinspired optimal structures combining bioresorbable, bioactive and other advanced properties are expected for the next generation of biomaterials (Bruck et al., 2002, Hench et al., 2002, Akaike et al., 2005). Inspired by nature, to reveal the relationship between structure and functionality of biological materials has been emphasized in the biomaterials research field. In nature, gradient biological structures exist most commonly, such as the structure of bamboo (Amada et al., 1997), shells, teeth, bones, tendon and extracellular matrix (ECM) (Suresh et al., 2001). Man-made functionally gradient materials (FGMs) have been developed for combining irreconcilable properties within a single material and have been widely incorporated in metal/ceramic and organic/inorganic material fields for increasing the structural complexity and combining different functionality (Suganuma et al., 1983, Ishikawa et al., 2002, Czubarow et al., 1997). FGMs are spatial composites within which the composition or structure and thus the functions of the materials continuously or step-wisely vary along the specific coordinates. Examples of the structural change include those in crystal structure and orientation, porosity, particle size and so on.

The concept of FGMs first originated in Japan in 1984-1985 in the spacecraft project (Kawasaki et al., 1987). FGMs are developed to give two conflicting properties such as good thermal conductivity and good thermal resistance into one material. In general, we see uniformed functions and properties within the materials we use. However, in FGMs we see different functions between one part and the other of them. In a metal-ceramic FGM, the metal-rich side is placed in the region where mechanical performance, such as toughness, needs to be stronger; and the ceramic-rich side, which has better thermal resistance, is exposed to high temperatures, or placed in the region where there is a potentially severe temperature variation (Dao et al., 1997). In FGMs, the properties change gradually with position due to the spatial gradients in composition, density, microstructural arrangement or atomic-order, which

<sup>\*</sup> Corresponding Author

Characterization of Compositional Gradient

amorphous phase (Qiu et al., 2003).

**2.3 Results of ATR-FTIR measurements** 

observed at room temperature.

**spectroscopy** 

**2.1 Sample preparation and creation of gradient film** 

Structure of Polymeric Materials by FTIR Technology 341

with bacterial (PHB) (He et al., 2000). Poly(1,4-butylene succinate)(PBS) is a biodegradable aliphatic semi-crystalline polymer and flexible thermoplastic polyester (Uesaka et al., 2000). The miscibility and crystallization behavior in homogenous blends of PEO/PBS have been investigated by DSC and optical microscopy. PEO and PBS were found to be miscible in the

The films of PBS, PEO and the binary blend of PBS/PEO were prepared by solution casting method. Pure PBS and PEO films with thickness of 200 ±10 μm were laminated together under pressure of 5 Mpa to obtain the bilayered film, and then used for preparation of gradient film. The laminated PBS/PEO film was annealed at 140 oC for 24 hours in an oven in the vacuum condition. By this way, we could generate a continuous gradient structure in the thickness

ATR-FTIR spectra were recorded on a AIM-8800 FT-IR spectrometer ( Shimazu Co.Ltd. Japan) equipped with a Diamond EX'Press in order to analyze the chemical composition of the surfaces. The sum of 64 scans with a resolution of 4cm-1 was used to obtain the spectra.

Figure 1 shows the ATR-FTIR spectra of the film surface of pure PBS, PEO and 50/50 (wt%/wt%) PBS/PEO blend. The peak centered at wavenumber 1724 cm-1 corresponding to

Fig. 1. ATR-FTIR spectra of pure PBS, PEO, and 50/50(wt%/wt%) PBS/PEO binary blend

direction, which is difficult to achieve by a conventional stepped multi-layer process.

**2.2 Attenuated Total Reflectance Fourier Transform Infrared (ATR/FT-IR)** 

contribute to distribute thermal stresses, to reduce mechanical stress, and to improve interfacial bonding between dissimilar materials. This kind of materials also provides other functional properties depending on their constituents of composites, such as gradient optical polymers and multifocal lenses (Zuccarello et al., 2002, Kryszewski et al., 1998, Ma et al., 2002).

Recently, there have been many efforts to develop polymeric FGMs with unique properties and advanced functions that are inaccessible in conventional uniform systems (Agari et al., 1996, Kano et al., 1997, Xie et al., 1998,). In particular, compositional gradient biodegradable polymeric materials have many potential applications for biomedical devices and tissue engineering, and we consider they may lead to a wide range of new generation of biomedical materials. Polymeric FGMs have great potential to be used in various fields such as separation membrane, adhesive, and biomedical materials including artificial skin, artificial bone and teeth, drug delivery system, and so on. However, the difference in structure, functional group, miscibility, solvent, and thermal treatment, induced a large complexity for designing, preparing, confirming, and characterizing polymeric FGMs. Many preparation approaches have been developed to generate a polymeric functionally gradient structure during homogenization or segregation processes. Despite these efforts made recently to generate polymeric FGMs, characterization of their gradient structure, physicochemical properties and elucidation of formation mechanisms still remain to be explored. In this chapter, we will mainly discuss the confirming methods of compositional gradient structure of polymeric FGMs using FT-IR technology (such as ATR-FTIR, Mapping measurement of FTIR, and PAS-FTIR). FT-IR has been proven to be a powerful technique for characterizing the compositional differences of gradient film materials from surface to inside, and also on cross section. In our previous studies, four different compositional gradient polymeric materials have been prepared in miscible or immiscible blend systems, and in all these works, FT-IR has been applied as a main method for confirming the compositional gradient structure of the prepared materials (Hexig et al., 2005, 2010).

## **2. ATR-FTIR for confirming the compositional difference between two surfaces of polymeric FGM film**

The consentaneous confirming standard for polymeric materials with gradient structure has not been established yet. In general, ATR-FTIR can give the information about chemical composition in the range from the top surface into a few μm deep. We have been utilizing ATR-FTIR for confirming the compositional difference between two surfaces of polymeric FGMs.

We successfully prepared a novel polymeric material with compositional gradient in a binary miscible biodegradable polymer blend system, that is, poly(butylene succinate)(PBS)/poly(ethylene oxide)(PEO) blend through controlling interdiffusion process at a temperature above the melting points of both components. The semicrystalline polymer PEO has attracted much attention and has been extensively investigated both experimentally and theoretically during recent decades (Allen et al., 1999). PEO is biocompatible, biodegradable, and water-soluble polymer (Dormidontova et al., 2002). These specific features of PEO make it applicable for drug delivery purposes, and it also has potential for other biomedical applications. PEO is miscible with several amorphous polymers through hydrogen bonding interaction (Miyoshi et al., 1996), and is also miscible with crystalline polymers, such as, poly(ethylene succinate)(PES) (Chen et al., 2000), and poly(3-hydroxypropionate) (PHP) (Na et al., 2002), while it is partly miscible or immiscible

contribute to distribute thermal stresses, to reduce mechanical stress, and to improve interfacial bonding between dissimilar materials. This kind of materials also provides other functional properties depending on their constituents of composites, such as gradient optical polymers and multifocal lenses (Zuccarello et al., 2002, Kryszewski et al., 1998, Ma et al., 2002). Recently, there have been many efforts to develop polymeric FGMs with unique properties and advanced functions that are inaccessible in conventional uniform systems (Agari et al., 1996, Kano et al., 1997, Xie et al., 1998,). In particular, compositional gradient biodegradable polymeric materials have many potential applications for biomedical devices and tissue engineering, and we consider they may lead to a wide range of new generation of biomedical materials. Polymeric FGMs have great potential to be used in various fields such as separation membrane, adhesive, and biomedical materials including artificial skin, artificial bone and teeth, drug delivery system, and so on. However, the difference in structure, functional group, miscibility, solvent, and thermal treatment, induced a large complexity for designing, preparing, confirming, and characterizing polymeric FGMs. Many preparation approaches have been developed to generate a polymeric functionally gradient structure during homogenization or segregation processes. Despite these efforts made recently to generate polymeric FGMs, characterization of their gradient structure, physicochemical properties and elucidation of formation mechanisms still remain to be explored. In this chapter, we will mainly discuss the confirming methods of compositional gradient structure of polymeric FGMs using FT-IR technology (such as ATR-FTIR, Mapping measurement of FTIR, and PAS-FTIR). FT-IR has been proven to be a powerful technique for characterizing the compositional differences of gradient film materials from surface to inside, and also on cross section. In our previous studies, four different compositional gradient polymeric materials have been prepared in miscible or immiscible blend systems, and in all these works, FT-IR has been applied as a main method for confirming the

compositional gradient structure of the prepared materials (Hexig et al., 2005, 2010).

**2. ATR-FTIR for confirming the compositional difference between two** 

The consentaneous confirming standard for polymeric materials with gradient structure has not been established yet. In general, ATR-FTIR can give the information about chemical composition in the range from the top surface into a few μm deep. We have been utilizing ATR-FTIR for confirming the compositional difference between two surfaces of polymeric FGMs.

We successfully prepared a novel polymeric material with compositional gradient in a binary miscible biodegradable polymer blend system, that is, poly(butylene succinate)(PBS)/poly(ethylene oxide)(PEO) blend through controlling interdiffusion process at a temperature above the melting points of both components. The semicrystalline polymer PEO has attracted much attention and has been extensively investigated both experimentally and theoretically during recent decades (Allen et al., 1999). PEO is biocompatible, biodegradable, and water-soluble polymer (Dormidontova et al., 2002). These specific features of PEO make it applicable for drug delivery purposes, and it also has potential for other biomedical applications. PEO is miscible with several amorphous polymers through hydrogen bonding interaction (Miyoshi et al., 1996), and is also miscible with crystalline polymers, such as, poly(ethylene succinate)(PES) (Chen et al., 2000), and poly(3-hydroxypropionate) (PHP) (Na et al., 2002), while it is partly miscible or immiscible

**surfaces of polymeric FGM film** 

with bacterial (PHB) (He et al., 2000). Poly(1,4-butylene succinate)(PBS) is a biodegradable aliphatic semi-crystalline polymer and flexible thermoplastic polyester (Uesaka et al., 2000). The miscibility and crystallization behavior in homogenous blends of PEO/PBS have been investigated by DSC and optical microscopy. PEO and PBS were found to be miscible in the amorphous phase (Qiu et al., 2003).

## **2.1 Sample preparation and creation of gradient film**

The films of PBS, PEO and the binary blend of PBS/PEO were prepared by solution casting method. Pure PBS and PEO films with thickness of 200 ±10 μm were laminated together under pressure of 5 Mpa to obtain the bilayered film, and then used for preparation of gradient film. The laminated PBS/PEO film was annealed at 140 oC for 24 hours in an oven in the vacuum condition. By this way, we could generate a continuous gradient structure in the thickness direction, which is difficult to achieve by a conventional stepped multi-layer process.

#### **2.2 Attenuated Total Reflectance Fourier Transform Infrared (ATR/FT-IR) spectroscopy**

ATR-FTIR spectra were recorded on a AIM-8800 FT-IR spectrometer ( Shimazu Co.Ltd. Japan) equipped with a Diamond EX'Press in order to analyze the chemical composition of the surfaces. The sum of 64 scans with a resolution of 4cm-1 was used to obtain the spectra.

#### **2.3 Results of ATR-FTIR measurements**

Figure 1 shows the ATR-FTIR spectra of the film surface of pure PBS, PEO and 50/50 (wt%/wt%) PBS/PEO blend. The peak centered at wavenumber 1724 cm-1 corresponding to

Fig. 1. ATR-FTIR spectra of pure PBS, PEO, and 50/50(wt%/wt%) PBS/PEO binary blend observed at room temperature.

Characterization of Compositional Gradient

was confirmed by FT-IR mapping measurement.

**3.1 Sample preparation and creation of gradient film** 

laboratory press (Mini Test Press-10, Toyoseiki Co., Japan).

the laminated film was annealed at 140 º

adding TDP.

Structure of Polymeric Materials by FTIR Technology 343

a wide range of applications. In the case of immiscible polymer system, the contributions from favorable intermolecular interactions are not enough to overwhelm the unfavorable contribution arising from the solubility parameters (Coleman et al., 1991). The addition of low molecular-weight agent is a well-known way to modify polymeric material properties

Our interest is to develop a new strategy for creation of functionally gradient polymeric composite materials in immiscible polymer blend system utilizing the miscibilization effect of low molecular-weight agent. Two kinds of commercially available biodegradable polymers, namely, a random copolyester, poly(butylene adipate-co-44 mol% butylenes terephthalate)[P(BA-co-BT)] and water-soluble poly(ethylene-oxide)(PEO) are chosen for this purpose, because P(BA-co-BT) and PEO are immiscible over the whole range of blend composition, but two components become miscible in the presence of optimum amount of low molecular-weight component 4,4'-thiodiphenol (TDP) as reported in our previous study (Hexig et al., 2004). It has been found that a suitable content of TDP has the ability to improve the miscibility between P(BA-co-BT) and PEO through intermolecular hydrogen bonding interaction, as revealed by DSC and FT-IR measurements (Figure 3). In this work, we demonstrate the generation of compositional gradient phase structure in the immiscible P(BA-co-BT)/PEO system through improving the miscibility between two components by the formation of TDP-mediated hydrogen bonds. The formation of compositional gradient

The films of pure P(BA-co-BT), pure PEO, binary blends P(BA-co-BT)/TDP(92/8 wt%/wt%) and PEO/TDP (92/8 wt%/wt%), and the ternary blend of P(BA-co-BT)/PEO/TDP containing a constant composition(8wt%) of TDP with various P(BA-co-BT)/PEO ratios were prepared by conventional solution casting method. 5wt% Polymer solution in 1,4 dioxane was stirred for 6-8 h and cast on a Teflon dish. The solvent was allowed to evaporate slowly for 1 day at ambient temperature. The resulted films were then dried in a vacuum oven at 50 ºC for 2 days to remove residual solvent and subsequently compression molded between Teflon sheets for 3 min at 160 ºC under pressure of 5MPa by using

An *in situ* formation process that we used to create the gradient films is illustrated schematically in Figure 4. The low molecular weight TDP is expected to enhance the miscibility and induce the interdiffution process. Firstly, the films of P(BA-co-BT)/TDP(92/8 wt%/wt%) and PEO/TDP (92/8 wt%/wt%) with thickness of 200±10 μm were laminated together under pressure of 5 Mpa at room temperature to obtain a bilayered film, and then

structure. In order to investigate the effect of gravity on the interdiffusion process, two laminated films were annealed in the same time, the one consists of P(BA-co-BT)/TDP(92/8 wt%/wt%) and PEO/TDP(92/8wt%/wt%) at the top and the bottom side, respectively, and the other one has the arrangement reverse to the first one. During the annealing process, the gradient structure of the resulted films was analyzed for several times and the best gradient phase in the whole thickness range was obtained after 12 hours annealing for both the films. The same process was also performed for the laminated film of P(BA-co-BT)/PEO without

C in a vacuum oven to generate a gradient

and even improve the miscibility in case of immiscible polymer blends.

the carbonyl vibration region of PBS and that at 2887cm-1 corresponding to the C-H stretching region of PEO were detected for the pure PBS and PEO films, respectively, at room temperature. Both of these characteristic peaks also appeared in the 50/50(wt%/wt%)PBS/PEO blend film. There are also some bands in the C-H stretching region in the spectrum of pure PBS, but they are very weak and broad. Therefore, the two peaks at 1724 cm-1 and 2887 cm-1 can be used to analyze qualitatively the contents of the PBS and the PEO components at the arbitrary position in the compositional gradient film. The difference of the chemical composition between the two surfaces of the compositional gradient film was examined by means of ATR-FTIR, as shown in Figure 2. It shows that the PBS has notably diffused to the PEO side while a comparatively small fraction of PEO diffused to the PBS side, implying that the composition varies from one side to the other side of the film.

Fig. 2. ATR-FTIR spectra on both surfaces of the compositional gradient film observed at room temperature.

#### **3. FT-IR mapping measurements for characterizing the compositional change along the thickness direction of polymeric FGMs**

The fact that only a limited amount of miscible biodegradable polymer pairs are available constrains preparation of biodegradable polymeric gradient materials with high quality and

the carbonyl vibration region of PBS and that at 2887cm-1 corresponding to the C-H stretching region of PEO were detected for the pure PBS and PEO films, respectively, at room temperature. Both of these characteristic peaks also appeared in the 50/50(wt%/wt%)PBS/PEO blend film. There are also some bands in the C-H stretching region in the spectrum of pure PBS, but they are very weak and broad. Therefore, the two peaks at 1724 cm-1 and 2887 cm-1 can be used to analyze qualitatively the contents of the PBS and the PEO components at the arbitrary position in the compositional gradient film. The difference of the chemical composition between the two surfaces of the compositional gradient film was examined by means of ATR-FTIR, as shown in Figure 2. It shows that the PBS has notably diffused to the PEO side while a comparatively small fraction of PEO diffused to the PBS side, implying that the composition varies from one side to the other

Fig. 2. ATR-FTIR spectra on both surfaces of the compositional gradient film observed at

**along the thickness direction of polymeric FGMs** 

**3. FT-IR mapping measurements for characterizing the compositional change** 

The fact that only a limited amount of miscible biodegradable polymer pairs are available constrains preparation of biodegradable polymeric gradient materials with high quality and

side of the film.

room temperature.

a wide range of applications. In the case of immiscible polymer system, the contributions from favorable intermolecular interactions are not enough to overwhelm the unfavorable contribution arising from the solubility parameters (Coleman et al., 1991). The addition of low molecular-weight agent is a well-known way to modify polymeric material properties and even improve the miscibility in case of immiscible polymer blends.

Our interest is to develop a new strategy for creation of functionally gradient polymeric composite materials in immiscible polymer blend system utilizing the miscibilization effect of low molecular-weight agent. Two kinds of commercially available biodegradable polymers, namely, a random copolyester, poly(butylene adipate-co-44 mol% butylenes terephthalate)[P(BA-co-BT)] and water-soluble poly(ethylene-oxide)(PEO) are chosen for this purpose, because P(BA-co-BT) and PEO are immiscible over the whole range of blend composition, but two components become miscible in the presence of optimum amount of low molecular-weight component 4,4'-thiodiphenol (TDP) as reported in our previous study (Hexig et al., 2004). It has been found that a suitable content of TDP has the ability to improve the miscibility between P(BA-co-BT) and PEO through intermolecular hydrogen bonding interaction, as revealed by DSC and FT-IR measurements (Figure 3). In this work, we demonstrate the generation of compositional gradient phase structure in the immiscible P(BA-co-BT)/PEO system through improving the miscibility between two components by the formation of TDP-mediated hydrogen bonds. The formation of compositional gradient was confirmed by FT-IR mapping measurement.

## **3.1 Sample preparation and creation of gradient film**

The films of pure P(BA-co-BT), pure PEO, binary blends P(BA-co-BT)/TDP(92/8 wt%/wt%) and PEO/TDP (92/8 wt%/wt%), and the ternary blend of P(BA-co-BT)/PEO/TDP containing a constant composition(8wt%) of TDP with various P(BA-co-BT)/PEO ratios were prepared by conventional solution casting method. 5wt% Polymer solution in 1,4 dioxane was stirred for 6-8 h and cast on a Teflon dish. The solvent was allowed to evaporate slowly for 1 day at ambient temperature. The resulted films were then dried in a vacuum oven at 50 ºC for 2 days to remove residual solvent and subsequently compression molded between Teflon sheets for 3 min at 160 ºC under pressure of 5MPa by using laboratory press (Mini Test Press-10, Toyoseiki Co., Japan).

An *in situ* formation process that we used to create the gradient films is illustrated schematically in Figure 4. The low molecular weight TDP is expected to enhance the miscibility and induce the interdiffution process. Firstly, the films of P(BA-co-BT)/TDP(92/8 wt%/wt%) and PEO/TDP (92/8 wt%/wt%) with thickness of 200±10 μm were laminated together under pressure of 5 Mpa at room temperature to obtain a bilayered film, and then the laminated film was annealed at 140 º C in a vacuum oven to generate a gradient structure. In order to investigate the effect of gravity on the interdiffusion process, two laminated films were annealed in the same time, the one consists of P(BA-co-BT)/TDP(92/8 wt%/wt%) and PEO/TDP(92/8wt%/wt%) at the top and the bottom side, respectively, and the other one has the arrangement reverse to the first one. During the annealing process, the gradient structure of the resulted films was analyzed for several times and the best gradient phase in the whole thickness range was obtained after 12 hours annealing for both the films. The same process was also performed for the laminated film of P(BA-co-BT)/PEO without adding TDP.

Characterization of Compositional Gradient

gradient film.

during the scans.

**3.3 Results** 

Structure of Polymeric Materials by FTIR Technology 345

Fig. 4. Schematic illustration of the formation of P(BA-co-BT)/PEO/TDP compositional

The films of polymer blends used for the measurements of FT-IR were prepared by casting the polymer solution on the surface of a silicon wafer and dried under vacuum condition for 2 days. The film used in this study was thin enough to obey the Lambert-Beer law (<0.6 absorbance units). FT-IR spectra were recorded on a Perkin-Elmer Spectrum 2000 spectrometer using a minimum of 64 co-added scans at a resolution of 4cm-1. Nitrogen was used to purge CO2 and gaseous water in the detector and sample compartments prior to and

The sample for FT-IR mapping measurement of the film cross-section was prepared by slicing the film with microtome. FT-IR mapping measurements of the compositional gradient films were carried out by using a AIM-8800 FT-IR microscope (Shimazu Co.Ltd. Japan) equipped with a Diamond EX'Press. A square aperture of 30 х 30 μm was set on one side of the cross section and the aperture is automatically movable under a setting program. The 30 x 30 μm square aperture was displaced 50 times by a step of 10 μm. The sum of 64 scans with a resolution 4 cm-1 was averaged at each step. The spectra were then

Figure 5 shows the FT-IR spectra of pure TDP, pure PEO, pure P(BA-co-BT), and their blends of P(BA-co-BT)/TDP92/8(wt%/wt%), PEO/ TDP92/8(wt%/wt%), and P(BA-co-BT)/PEO/TDP46/46/8(wt%/wt%wt%). The peaks centered at wavenumber 1488cm-1, 1596cm-1 and 1612cm-1corresponding to the C=C stretching region of TDP were detected for the pure TDP at melting state and all its blends, indicating that TDP in the blends is in the amorphous state. The peak centered at wavenumber 1724 cm-1 corresponding to the

**3.2 FT-IR microscopy and mapping measurements** 

accumulated together to make a mapping mode spectrum.

Fig. 3. Infrared spectra in the carbonyl stretching region for binary P(BA-co-BT)/TDP=90/10 blend (a) and ternary P(BA-co-BT)/PEO/TDP=45/45/10 blend (b) resolved by curve-fitting program; Amor.: amorphous component; Crys.: crystalline component; Hydr.: hydrogenbonded component; Fitt.: cunve-fitted; Exp.: experimental spectrum.

Amor

Crys

P(BA-co-BT)/PEO/TDP 45/45/10 blend

Amor

P(BA-co-BT)/TDP 90/10 blend

(a)

(b)

Absorbance

Hydr

Absorbance

1800 1750 1700 1650

Wavenumber (cm-1)

1800 1750 1700 1650

Wavenumber (cm-1)

Fig. 3. Infrared spectra in the carbonyl stretching region for binary P(BA-co-BT)/TDP=90/10 blend (a) and ternary P(BA-co-BT)/PEO/TDP=45/45/10 blend (b) resolved by curve-fitting program; Amor.: amorphous component; Crys.: crystalline component; Hydr.: hydrogen-

bonded component; Fitt.: cunve-fitted; Exp.: experimental spectrum.

Hydr Crys

Fig. 4. Schematic illustration of the formation of P(BA-co-BT)/PEO/TDP compositional gradient film.

## **3.2 FT-IR microscopy and mapping measurements**

The films of polymer blends used for the measurements of FT-IR were prepared by casting the polymer solution on the surface of a silicon wafer and dried under vacuum condition for 2 days. The film used in this study was thin enough to obey the Lambert-Beer law (<0.6 absorbance units). FT-IR spectra were recorded on a Perkin-Elmer Spectrum 2000 spectrometer using a minimum of 64 co-added scans at a resolution of 4cm-1. Nitrogen was used to purge CO2 and gaseous water in the detector and sample compartments prior to and during the scans.

The sample for FT-IR mapping measurement of the film cross-section was prepared by slicing the film with microtome. FT-IR mapping measurements of the compositional gradient films were carried out by using a AIM-8800 FT-IR microscope (Shimazu Co.Ltd. Japan) equipped with a Diamond EX'Press. A square aperture of 30 х 30 μm was set on one side of the cross section and the aperture is automatically movable under a setting program. The 30 x 30 μm square aperture was displaced 50 times by a step of 10 μm. The sum of 64 scans with a resolution 4 cm-1 was averaged at each step. The spectra were then accumulated together to make a mapping mode spectrum.

#### **3.3 Results**

Figure 5 shows the FT-IR spectra of pure TDP, pure PEO, pure P(BA-co-BT), and their blends of P(BA-co-BT)/TDP92/8(wt%/wt%), PEO/ TDP92/8(wt%/wt%), and P(BA-co-BT)/PEO/TDP46/46/8(wt%/wt%wt%). The peaks centered at wavenumber 1488cm-1, 1596cm-1 and 1612cm-1corresponding to the C=C stretching region of TDP were detected for the pure TDP at melting state and all its blends, indicating that TDP in the blends is in the amorphous state. The peak centered at wavenumber 1724 cm-1 corresponding to the

Characterization of Compositional Gradient

Structure of Polymeric Materials by FTIR Technology 347

Fig. 6. Plots of the peak intensity of the P(BA-co-BT) C=O, PEO C-H, and TDP C=C

(b) 8 hours, and (c) 12 hours.

absorptions along the thickness direction of the resulted films after annealed for (a) 4 hours,

Fig. 5. FT-IR spectra of (a) pure TDP at 160oC, and (b) pure P(BA-co-BT), (c) 92/8 (wt%/wt%) P(BA-co-BT)/TDP binary blend (d) 46/46/8 (wt%/wt%/wt%) P(BA-co-BT)/PEO/TDP ternary blend (e) 92/8 (wt%/wt%) PEO/TDP binary blend, (f) pure PEO observed at room temperature.

carbonyl vibration region of P(BA-co-BT) and that at 2887cm-1 corresponding to the C-H stretching region of PEO were detected for the pure P(BA-co-BT) and PEO films, respectively. Both of these characteristic peaks also appeared in the ternary 46/46/8 (wt%/wt%/wt%) blend film. There are some characteristic bands in the C-H stretching region in the spectrum of pure P(BA-co-BT), but they are very weak and broad. Therefore, the two peaks at 1724cm-1 and 2887 cm-1 were used to analyze qualitatively the contents of the P(BA-co-BT) and PEO components, respectively, at the arbitrary position in the compositional gradient film.

The compositional change along the film thickness direction in the annealing process was investigated by FT-IR mapping measurements on the cross section of the resulted films after annealed for different times. In Figure 6, the intensities of the P(BA-co-BT) C=O, PEO C-H, and TDP C=C absorption peaks are plotted against the position in the film thickness direction. It reveals that the interdiffusion process begins from the interface between the P(BA-co-BT)/TDP and PEO/TDP blends, and develops towards the two outer surfaces of the film. Finally, a well-structured continuous compositional gradient along the film thickness direction was formed after 12 hours annealing.

Fig. 5. FT-IR spectra of (a) pure TDP at 160oC, and (b) pure P(BA-co-BT), (c) 92/8 (wt%/wt%) P(BA-co-BT)/TDP binary blend (d) 46/46/8 (wt%/wt%/wt%) P(BA-co-BT)/PEO/TDP ternary blend (e) 92/8 (wt%/wt%) PEO/TDP binary blend, (f) pure PEO

carbonyl vibration region of P(BA-co-BT) and that at 2887cm-1 corresponding to the C-H stretching region of PEO were detected for the pure P(BA-co-BT) and PEO films, respectively. Both of these characteristic peaks also appeared in the ternary 46/46/8 (wt%/wt%/wt%) blend film. There are some characteristic bands in the C-H stretching region in the spectrum of pure P(BA-co-BT), but they are very weak and broad. Therefore, the two peaks at 1724cm-1 and 2887 cm-1 were used to analyze qualitatively the contents of the P(BA-co-BT) and PEO components, respectively, at the arbitrary position in the

The compositional change along the film thickness direction in the annealing process was investigated by FT-IR mapping measurements on the cross section of the resulted films after annealed for different times. In Figure 6, the intensities of the P(BA-co-BT) C=O, PEO C-H, and TDP C=C absorption peaks are plotted against the position in the film thickness direction. It reveals that the interdiffusion process begins from the interface between the P(BA-co-BT)/TDP and PEO/TDP blends, and develops towards the two outer surfaces of the film. Finally, a well-structured continuous compositional gradient along the film

observed at room temperature.

compositional gradient film.

thickness direction was formed after 12 hours annealing.

Fig. 6. Plots of the peak intensity of the P(BA-co-BT) C=O, PEO C-H, and TDP C=C absorptions along the thickness direction of the resulted films after annealed for (a) 4 hours, (b) 8 hours, and (c) 12 hours.

Characterization of Compositional Gradient

**compositional gradient structure** 

1995, Yamada et al., 1990, 1993).

**4.1.1 Plasma-treatment of polystyrene dish** 

irradiation was performed for 15 seconds.

oC under vacuum condition for 5 hours before characterization.

**4.1.3 PAS-FTIR spectroscopy measurements** 

**4.1 Experiments** 

**4.1.2 Film preparation** 

that the gravity has no effect on the gradation process.

Structure of Polymeric Materials by FTIR Technology 349

side in the annealing process was the PEO/TDP or the P(BA-co-BT)/TDP blend, indicating

Plasma technology can add functional groups to a surface of organic and inorganic materials at the molecular level, changing surface chemistries for increased bond strength, wettability, permeability, and activating and changing surfaces from hydrophobic to hydrophilic without affecting the bulk properties. The compositional gradient structure in hyaluronic acid (HA) and thermal responsible poly(N- isopropylacrylamide) (PIPAAm) blend film was self-organized during solvent evaporation process on the oxygen-plasma treated polystyrene dish (PTPSD), while on the non-treated polystyrene dish (NTPSD) nearly homogenous blend film was formed at ambient temperature. HA is a naturally occurring linear polysaccharide widely distributed in body as components of the extracellular matrix (ECM) of connective tissues, and HA-based biomaterials have been recently utilized for a variety of clinical application and tissue engineering of skin, cartilage tissue, and bone based upon its specific properties, excellent biocompatibility and bioactivity (Kano et al., 1997, Zacchi et al., 1998, Aigner et al., 1998, Solchaga et al., 1999). PIPAAm is a synthetic polymer which has a sharp and reversible phase transition at ~32 °C and applicable in tissue engineering fields as a functional hydrogel and a cell sheet (Park et al., 2002, Yoshida et al.,

The plasma treatment was performed with a SWP-101EX (NISSIN Co, Ltd. Japan), using low-pressure region output power of the microwave oscillator at 2.0KW. Polystyrene dishes were set on the sample stage which was 15cm below the reactor. The oxygen discharge was utilized and the oxygen was filled at a rate of 500 cc/min and a pressure of 70 Pa. Plasma

1% HA solution was prepared by dissolving powder HA (weight average molecular weight 1680.000 by GPC, Life Core Biomedical Inc) in distilled water with stirring for 24 hours. 1% PIPAAM solution was prepared by diluting 15% PIPAAM (weight average molecular weight 220.000 by GPC, Kohjin Co. Ltd) aqueous solution by distilled water with stirring for 24 hours. After then 1% HA and 1% PIPAAm solutions were mixed together at the same weight ratios and stirred further for 24 hours before casting on the PTPSD and NTPSD at ambient and vacuum conditions, respectively. All resulted sample films were heated at 80

PAS-FTIR spectroscopy measurements were carried out on the JIR-SPX200 FT-IR spectrometer (JEOL Co. Ltd. Japan) equipped with a MTEC 300 photoacoustic cell (MTEC

**4. PAS-FTIR as a non-destructive, non-contact confirming method of** 

The results of FT-IR mapping measurements on the cross section of the resulted film are shown in Figure 7. It is clearly seen that the peak intensity of P(BA-co-BT) C=O vibration absorption at 1724cm-1 continuously increases along the thickness direction while that of the PEO C-H stretching absorption at 2887cm-1 decreases along the same direction. The weak bands centered at about 1600cm-1 and 1488cm-1 characteristic for TDP appear in all spectra and their peak intensities are almost constant over the whole thickness range, indicating that the TDP content is almost the same over the whole range in the thickness direction of the gradient film.

Fig. 7. Results of FT-IR mapping measurements on the cross section along the direction vertical to the sample surface.

The presence of a suitable content of TDP transformed the immiscible P(BA-co-BT)/PEO blend system into a miscible one through the intermolecular hydrogen bonding interactions, as experimentally demonstrated by DSC and FT-IR measurements (Hexig et al., 2004). In a miscible system, the favorable intermolecular interactions overwhelm the unfavorable intermolecular interactions and the free energy of the system decreases with mixing (Pines et al., 1972). As a result, the bilayered miscible polymer film shows a tendency towards homogeneous mixing. One more important factor is that the constant breaking and reforming of the hydrogen bond at a temperature above the glass transition (Coleman et al., 1991), especially at the high temperature used in this process provided the possibility and approach for the interdiffusion to occur. The optimal gradient structure can be obtained in the homogenizing process by controlling the interdiffusing time. The peak intensity of TDP C=C absorption is almost constant over the whole thickness range at all the annealing time, indicating that the motion of TDP is an isotropic thermal motion. The mapping measurements also revealed almost the same results for the films regardless that the upper side in the annealing process was the PEO/TDP or the P(BA-co-BT)/TDP blend, indicating that the gravity has no effect on the gradation process.

## **4. PAS-FTIR as a non-destructive, non-contact confirming method of compositional gradient structure**

Plasma technology can add functional groups to a surface of organic and inorganic materials at the molecular level, changing surface chemistries for increased bond strength, wettability, permeability, and activating and changing surfaces from hydrophobic to hydrophilic without affecting the bulk properties. The compositional gradient structure in hyaluronic acid (HA) and thermal responsible poly(N- isopropylacrylamide) (PIPAAm) blend film was self-organized during solvent evaporation process on the oxygen-plasma treated polystyrene dish (PTPSD), while on the non-treated polystyrene dish (NTPSD) nearly homogenous blend film was formed at ambient temperature. HA is a naturally occurring linear polysaccharide widely distributed in body as components of the extracellular matrix (ECM) of connective tissues, and HA-based biomaterials have been recently utilized for a variety of clinical application and tissue engineering of skin, cartilage tissue, and bone based upon its specific properties, excellent biocompatibility and bioactivity (Kano et al., 1997, Zacchi et al., 1998, Aigner et al., 1998, Solchaga et al., 1999). PIPAAm is a synthetic polymer which has a sharp and reversible phase transition at ~32 °C and applicable in tissue engineering fields as a functional hydrogel and a cell sheet (Park et al., 2002, Yoshida et al., 1995, Yamada et al., 1990, 1993).

## **4.1 Experiments**

348 Infrared Spectroscopy – Materials Science, Engineering and Technology

The results of FT-IR mapping measurements on the cross section of the resulted film are shown in Figure 7. It is clearly seen that the peak intensity of P(BA-co-BT) C=O vibration absorption at 1724cm-1 continuously increases along the thickness direction while that of the PEO C-H stretching absorption at 2887cm-1 decreases along the same direction. The weak bands centered at about 1600cm-1 and 1488cm-1 characteristic for TDP appear in all spectra and their peak intensities are almost constant over the whole thickness range, indicating that the TDP content is almost the same over the whole range in the thickness direction of the

Fig. 7. Results of FT-IR mapping measurements on the cross section along the direction

The presence of a suitable content of TDP transformed the immiscible P(BA-co-BT)/PEO blend system into a miscible one through the intermolecular hydrogen bonding interactions, as experimentally demonstrated by DSC and FT-IR measurements (Hexig et al., 2004). In a miscible system, the favorable intermolecular interactions overwhelm the unfavorable intermolecular interactions and the free energy of the system decreases with mixing (Pines et al., 1972). As a result, the bilayered miscible polymer film shows a tendency towards homogeneous mixing. One more important factor is that the constant breaking and reforming of the hydrogen bond at a temperature above the glass transition (Coleman et al., 1991), especially at the high temperature used in this process provided the possibility and approach for the interdiffusion to occur. The optimal gradient structure can be obtained in the homogenizing process by controlling the interdiffusing time. The peak intensity of TDP C=C absorption is almost constant over the whole thickness range at all the annealing time, indicating that the motion of TDP is an isotropic thermal motion. The mapping measurements also revealed almost the same results for the films regardless that the upper

gradient film.

vertical to the sample surface.

## **4.1.1 Plasma-treatment of polystyrene dish**

The plasma treatment was performed with a SWP-101EX (NISSIN Co, Ltd. Japan), using low-pressure region output power of the microwave oscillator at 2.0KW. Polystyrene dishes were set on the sample stage which was 15cm below the reactor. The oxygen discharge was utilized and the oxygen was filled at a rate of 500 cc/min and a pressure of 70 Pa. Plasma irradiation was performed for 15 seconds.

## **4.1.2 Film preparation**

1% HA solution was prepared by dissolving powder HA (weight average molecular weight 1680.000 by GPC, Life Core Biomedical Inc) in distilled water with stirring for 24 hours. 1% PIPAAM solution was prepared by diluting 15% PIPAAM (weight average molecular weight 220.000 by GPC, Kohjin Co. Ltd) aqueous solution by distilled water with stirring for 24 hours. After then 1% HA and 1% PIPAAm solutions were mixed together at the same weight ratios and stirred further for 24 hours before casting on the PTPSD and NTPSD at ambient and vacuum conditions, respectively. All resulted sample films were heated at 80 oC under vacuum condition for 5 hours before characterization.

## **4.1.3 PAS-FTIR spectroscopy measurements**

PAS-FTIR spectroscopy measurements were carried out on the JIR-SPX200 FT-IR spectrometer (JEOL Co. Ltd. Japan) equipped with a MTEC 300 photoacoustic cell (MTEC

Characterization of Compositional Gradient

(Hexig et al., 2010).

**5. Conclusions** 

of FGMs satisfactorily.

**6. References** 

Structure of Polymeric Materials by FTIR Technology 351

PIPAAm gradually change from the surfaces to the inside of the film for both films cast on NTPSD and PTPSD. For the film cast on PTPSD in vacuum condition, more significant compositional difference was observed between the dish side and the air side than the film cast on NTPSD in vacuum. In combination with XPS and EDX mapping measurements on the films cast on NTPSD and PTPSD at ambient and vacuum conditions, respectively, it was revealed that both the oxidized hydrophilic surface and evaporation rate of water molecules contribute to the formation of an ideal gradient structure in the HA/PIPAAm blend system

Developing a methodology for characterizing the gradient structure is with the same importance as developing a method to generate the gradient structure. Utilizing ATR-FTIR spectroscopy, we can confirm the compositional difference between the two surfaces of the compositional gradient materials, while FT-IR mapping measurements can give a information of the compositional change on the cross section. The step-scan PAS-FTIR is a non-destructive, non-contact method reveals the compositional difference corresponding to the varying sampling depth. Thus, with a reasonable combination of these FT-IR spectroscopy measurements, we can characterized and confirm the compositional gradient

[4] Amada, S., Ichikawa, Y., Munnekta, T., Munekata, T., Nagase, Y. & Shimizu, H.

[6] Suganuma, K., Okamoto, T., Shimada, M. & Koizumi, M. *J. Amer. Ceram. Soc*. 66, c117-

[11] G. Zuccarello, D. Scribner, R. Sands, and L. J. Buckley, *Advanced Materials*. 2002, 14,

[13] D. Ma, J. M. Lupton, R. Beavington, P.L. Burn, I. D. Samuel, *Adv, Funct, Mater*. 2002, 12,

[10] Ming Dao, Pei Gu, Akhilesh Maewal, and R. J. Asaro. Acta mater. 1997. 45. 3265

[1] Bruck, H. A., Evans, J. J. & Peterson, M. L.. *Exp. Mech.* 42. 361-371 (2002).

[2] Hench, L. L. & Polak, J. M. 295, 1014-1017 (2002). [3] Akaike, T. 35- 67 (Corona, Tokyo, Japan, 2005).

*Composites part B,* 28B, 13-20 (1997).

[8] Czubarow, P. & Seyferth, D. *J. Mater. Sci.* 32, 2121-2130 (1997). [9] Kawasaki, A. and Watanabe R., J. Japan Inst. Merals, 1987. 51. 525.

[12] M. Kryszewski, *Polymers for Advanced Technologies*. 1998, 9, 244.

[15] Y. Kano, S. Akiyama, H. Sano, *Polym. J*. 1997, 29, 158.

[14] Y. Aragi, M. Shimada, A. Uede, *Macromol. Chem. Phys*. 1996, 197, 2017.

[16] X. Xie, T. Xiao, Z. Zhang, A. Tanioka, *J. Colloid and Interface Sci.* 1998, 206, 189. [17] B. Hexig, H. Alata, N. Asakawa, Y. Inoue, *Sci. B Polym. Phys.* 2005, 43, 368. [18] B. Hexig, H. Alata, Y. Inoue, *J. Polym. Sci. B Polym. Phys.* 2005, 43, 3069. [19] B. Hexig, H. Alata, N. Asakawa, Y. Inoue, *Adv. Funct. Mater.* 2005, 15, 1630.

[5] Suresh, S. *Science* 292. 2447- 2451 (2001).

[7] Ishikawa, T. *et al*. *Nature* 416, 64-67(2002).

c118 (1983).

1261.

507.

photoacoustic Ames USA). Prior to the start of penetration experiment the cell was purged with helium for 30 seconds.

#### **4.2 Results**

The step-scan PAS-FTIR is a non-destructive, non-contact method with controllable sampling depth and need little or no sample preparation. Figure 8 shows the results of PAS-FTIR spectroscopy measurements on the films cast on NTPSD and PTPSD in vacuum condition. PAS-FTIR spectra of increasing shallow-sampling depth corresponding to different mirror velocities of 1.0, 2.0 and 5.0 mm/s indicate that the fractions of HA and

Fig. 8. Results of PAS-FTIR spectroscopy measurements on the films cast on NTPSD and PTPSD in vacuum condition.

PIPAAm gradually change from the surfaces to the inside of the film for both films cast on NTPSD and PTPSD. For the film cast on PTPSD in vacuum condition, more significant compositional difference was observed between the dish side and the air side than the film cast on NTPSD in vacuum. In combination with XPS and EDX mapping measurements on the films cast on NTPSD and PTPSD at ambient and vacuum conditions, respectively, it was revealed that both the oxidized hydrophilic surface and evaporation rate of water molecules contribute to the formation of an ideal gradient structure in the HA/PIPAAm blend system (Hexig et al., 2010).

## **5. Conclusions**

350 Infrared Spectroscopy – Materials Science, Engineering and Technology

photoacoustic Ames USA). Prior to the start of penetration experiment the cell was purged

The step-scan PAS-FTIR is a non-destructive, non-contact method with controllable sampling depth and need little or no sample preparation. Figure 8 shows the results of PAS-FTIR spectroscopy measurements on the films cast on NTPSD and PTPSD in vacuum condition. PAS-FTIR spectra of increasing shallow-sampling depth corresponding to different mirror velocities of 1.0, 2.0 and 5.0 mm/s indicate that the fractions of HA and

Fig. 8. Results of PAS-FTIR spectroscopy measurements on the films cast on NTPSD and

with helium for 30 seconds.

PTPSD in vacuum condition.

**4.2 Results** 

Developing a methodology for characterizing the gradient structure is with the same importance as developing a method to generate the gradient structure. Utilizing ATR-FTIR spectroscopy, we can confirm the compositional difference between the two surfaces of the compositional gradient materials, while FT-IR mapping measurements can give a information of the compositional change on the cross section. The step-scan PAS-FTIR is a non-destructive, non-contact method reveals the compositional difference corresponding to the varying sampling depth. Thus, with a reasonable combination of these FT-IR spectroscopy measurements, we can characterized and confirm the compositional gradient of FGMs satisfactorily.

## **6. References**


**18** 

*1Centre of Bioanalysis,* 

*Romania* 

**Fourier Transform Infrared** 

**Spectroscopy – Useful Analytical** 

**Tool for Non-Destructive Analysis** 

Georgiana-Ileana Truica1,2, Andreia Tache1,2 and Gabriel-Lucian Radu1,2

It is highly recognized that the use of Fourier Transform Infrared Spectrometry (FTIR) for chemical substances identification it is not a trivial task to be fulfilled by analytical chemists. The complexity of FTIR characterization comes mainly from the high degree of infrared absorption bands overlapping, that are difficult to be accurately ascribed, despite of the fact

Regardless all these difficulties, FTIR analysis became the main used technique when specific analytical topics have to be addressed, mainly when non-destructive analysis is needed. In this respect, according to our opinion, challenging analytical issues are raised in two important cases; the first one is that when historic (archaeological) or artistic materials have to be analyzed while the second issue came from the analysis of highly- specific biomaterials. Starting from these points, the present chapter is addressing, as case studies, both issues: the FTIR assessment of peculiar archaeological samples, amber samples, and, respectively the FTIR assessment of a developed bio-mimetic system based on human lipoproteins immobilization on conductive solid supports. In both cases the restricted amount of samples and the emergent call for non-destructive analyses create constraints in

Consequently, we consider that the versatility of FTIR techniques (especially variable angle reflectance- FTIR, VAR-FTIR and attenuated transmittance-FTIR, ATR-FTIR) in solving these

**2. Applicability of Fourier Transform Infrared technique in the assessment of** 

Archaeology is a science dealing mainly with the reconstruction of past events by the mean of a cautious credentials and artefacts collection. Historical archaeologists usually encounter

critical issues are presented and, moreover, supported with original experimental data.

that up to date computer-searchable databases of spectra are currently available.

**1. Introduction** 

the working procedure.

**archeological samples. Case study- Amber** 

Simona-Carmen Litescu1, Eugenia D. Teodor1,

*National Institute for Biological Sciences Bucharest 2Faculty of Applied Chemistry and Material Science,* 

*University Polytechnica of Bucharest* 


## **Fourier Transform Infrared Spectroscopy – Useful Analytical Tool for Non-Destructive Analysis**

Simona-Carmen Litescu1, Eugenia D. Teodor1, Georgiana-Ileana Truica1,2, Andreia Tache1,2 and Gabriel-Lucian Radu1,2 *1Centre of Bioanalysis, National Institute for Biological Sciences Bucharest 2Faculty of Applied Chemistry and Material Science, University Polytechnica of Bucharest Romania* 

## **1. Introduction**

352 Infrared Spectroscopy – Materials Science, Engineering and Technology

[20] B. Hexig, K. Isama, Y. Haishima, T. Tsuchiya, Y. Inoue, T. Akaike, *Journal of Biomaterials* 

[25] Y-Ho. Na, Y, He. N. Asakawa, N. Yoshie, Y. Inoue, Macromolecules, 2002, 35, 727-735.

[29] M. M. Coleman, J. F. Graf, P. C. Painter, *Specific Interaction and the Miscibility of Polymer* 

[35] Solchaga, L. A., Dennis, J. E., Goldberg, V. M. & Caplan, A. I. *J. Orthop. Res.* 17, 205-213

[38] Yamada, N. et al. Thermo-responsive polymeric surface; *Makromol Chem Rapid Commun.*

[27] T. Uesaka, K. Nakane, S. Maeda T. Ogihara, N. Ogata, polymer, 2000, 41, 8449.

[21] C. Allen, D. Maysinger, A. Eisenberg, *Colloids Surf*. B. 1999, 16, 3.

*Blends*, Technomic Publishing: Lancaster, PA, 1991. [30] B.Hexig, N. Asakawa, Y.Inoue, *J.Polym.Sci. Part B*, 2004, 42, 2971. [31] A. Pines, M. G. Gibby, and J. S. Waugh *J. Chem. Phys*. 1972. 56. 1776. [32] Kano, Y., Akiyama, S., Sano, H. & Yui, H. *Polym. J*. 29. 158-164 (1997).

[36] Park, J. U. & Tsuchiya, T. *J. Biomed. Mater. Res.* 60, 541-547 (2002).

11, 571-576 (1990). *J. Biomed. Mater. Res.* 27, 1243-1251 (1993).

[23] T. Miyoshi, K. Takegoshi, K. Hikichi, *Polymer*. 1996, 37, 11-18.

[22] E. E. Dormidontova, *Macromolecules.* 2002, 35, 987.

[26] Y. He, N. Asakawa, Y. Inoue, *Polym Int*. 2000, 49, 609.

[28] Z. B. Qiu, T. Ikehara, T. Nishi, *Polymer*. 2003, 44, 2799.

[33] Zacchi, V. *et al. J. Biomed. Mater. Res.* 40, 187-194 (1998). [34] Aigner, J. *et al. J. Biomed. Mater. Res.* 42, 172-181 (1998).

[37] Yoshida, R. *et al. Nature* 374, 240-242 (1995).

[24] H-L. Chen, S-F. Wang, *Polymer.* 2000, 41, 5157.

*Science* 2010, 21, 1957.

(1999).

It is highly recognized that the use of Fourier Transform Infrared Spectrometry (FTIR) for chemical substances identification it is not a trivial task to be fulfilled by analytical chemists. The complexity of FTIR characterization comes mainly from the high degree of infrared absorption bands overlapping, that are difficult to be accurately ascribed, despite of the fact that up to date computer-searchable databases of spectra are currently available.

Regardless all these difficulties, FTIR analysis became the main used technique when specific analytical topics have to be addressed, mainly when non-destructive analysis is needed. In this respect, according to our opinion, challenging analytical issues are raised in two important cases; the first one is that when historic (archaeological) or artistic materials have to be analyzed while the second issue came from the analysis of highly- specific biomaterials. Starting from these points, the present chapter is addressing, as case studies, both issues: the FTIR assessment of peculiar archaeological samples, amber samples, and, respectively the FTIR assessment of a developed bio-mimetic system based on human lipoproteins immobilization on conductive solid supports. In both cases the restricted amount of samples and the emergent call for non-destructive analyses create constraints in the working procedure.

Consequently, we consider that the versatility of FTIR techniques (especially variable angle reflectance- FTIR, VAR-FTIR and attenuated transmittance-FTIR, ATR-FTIR) in solving these critical issues are presented and, moreover, supported with original experimental data.

### **2. Applicability of Fourier Transform Infrared technique in the assessment of archeological samples. Case study- Amber**

Archaeology is a science dealing mainly with the reconstruction of past events by the mean of a cautious credentials and artefacts collection. Historical archaeologists usually encounter

Fourier Transform Infrared Spectroscopy –

**2.1 Assessment of amber origin based on FTIR analysis** 

and from Romanian sources. Further obtained results are detailed.

consequence, the amount of provided information increases.

Subsequent data acquisition the following observations were drawn:

**2.1.2 Geological amber FTIR assessment** 

2010b).

**2.1.1 FTIR method** 

C=C instauration.

Useful Analytical Tool for Non-Destructive Analysis 355

Local amber, Romanite, was also early used, as proved on some beads from the Middle Bronze Age from Pietroasa Mică, in a necropolis placed at the fringes of the Colţi area. In our studies two types of amber have been investigated: artefacts from the largest amber deposit discovered in Romania, in Cioclovina cave (over 3000 beads, most of them in amber) (Teodor ES et al., 2010) and artefacts from Roman Age; discovered in cemeteries dug systematically, for instance in Tomis, Callatis, Beroe, or Carsium cemeteries (Vîrgolici et al.

Taking into account the lack of referential with respect to ascription of amber origin, and considering as critical issue the transformation occurring on amber samples during historical ages strongly influenced by the storing conditions in the deposits where were discovered, in the attempt of establishing few definite FTIR criteria to be useful for amber origin proper assignment we started the analysis not directly with archaeological samples but first, with geological amber, of controlled and certified origin, both from Baltic sources

FTIR assays of amber were performed using the FTIR-VAR technique with a beam incidence angle of 45o, on a Bruker TENSOR 27 instrument, using the OPUS software version 6.0. The samples were used without any pre-treatment, as whole pieces fixed on a gold mirror, and all the spectra were registered versus a background of clean gold foil between 4000 and 600 cm-1. The spectral resolution was 4 cm-1, and the co-added scans 96, with an aperture of 4 nm. The FTIR-VAR technique is able to provide the same information as FTIR in terms of transmittance, with a slight loss in signal intensity, but with the advantage of preserving samples integrity. In addition, the bounds between atom plans are not destroyed, and, as

A large lot of samples from different controlled origins were analysed by FTIR-VAR, from Romania (Colti, Buzău County), from the Baltic region (Palanga and Kaliningrad), from Germany and Poland in order to settle a certain pattern to differentiate Romanian amber from other types of amber, especially the Baltic variety (the material was provided, mainly, from the National Geological Museum from Bucharest). In the case of unsure observations, to clarify some spectral zone from the fingerprint region (1300-900 cm-1), or the region 1800- 1450 cm-1, the transmittance spectra of the samples were used for a better evaluation. Spectra were analysed and assigned on the three wave-numbers domains of significance for amber, namely those between 3600-2000 cm -1, 1820-1350 cm-1 and the 1250-1045 cm-1 regions, which correlate with hydroxyl groups, carboxyl groups, carbonyl groups and with

a. There are no notable differences between Romanite spectra and Baltic spectra in the 3600-3000 cm-1 region mainly in case of geological samples. The noteworthy differences

on a list of so-called potential trace evidence, different items which require certain identification; the mentioned list may be formed by objects from various materials: paint, pottery, glass, bricks, liquid residues (oils, wines, and perfumes), fabrics, soil, metals etc. In order to rationalize on the evidences origin, source, transformation and historical age, the archaeologists need a deep analytical study, study that often has to be performed on irreplaceable pieces. Consequently, to accomplish these goals multiple analytical techniques, able to work as non-destructive techniques are used, such as: Fourier Transform Infrared Spectrometry (FTIR), scanning electron microscopy (SEM), polarized light microscopy (PLM) or X-photon spectrometry (XPS).

In our studies, in the amber analysis the main goal was to establish several criteria able to be useful in discrimination between different types of amber, the obtained FTIR information being used to ascribe the origin of the materials found in archaeological sites on the Romanian territory.

Amber is a fossil resin, extremely appreciated for its colour and beauty, and used as a gemstone from very early times. Natural amber deposits are found all around the world, the most famous being the ones from the Baltic sea, the Dominican republic, Sicily, Borneo, Spain, etc. Amber resources are documented in Romania since the late sixteenth century, around the village of Colţi (Buzău County), but they have certainly been exploited only starting with the modern age, after 1828, in the same area (Ghiurcă, 1999; Wollmann, 1996).

The existence of a local amber exploitation on Romanian territory aroused the question of geological origin of amber artefacts found by Romanian archaeologists over time.

During recent years, analysts used mostly FTIR, mass spectrometry and pyrolysis-gas chromatography-mass spectrometry for the analytical study of geological or archaeological amber from different territories (Beck, 1972; 1986; Boon et al., 1993; Lambert & Poinar, 2002; Shedrinski et al., 2004; Angelini & Bellintani, 2005; Guiliano et al., 2007; Pakutinskiene et al., 2007; Tonidandel et al., 2008; Teodor ED et al., 2009; Vîrgolici et al., 2010a). In 2005, Angelini and Bellintani reviewed the analytical techniques used for the differentiation of amber types, dedicating special attention to the most suitable methods for archaeological materials (non-destructive methods).

There are a few comparative analytical studies of Romanian geological amber, referred to as Rumanite or Romanite, which conclude that Romanian amber is geologically younger, but chemically almost identical, to Baltic amber (Stout et al., 2000).

Regarding archaeological amber from Romanian territory, it was basically unstudied since we started our research project in 2007. Some studies by Nikolaus Boroffka on prehistoric amber (Boroffka, 2006) refer to chemical analyses of amber with an assumed Romanian origin; in another paper, Boroffka (2001) discusses the results published by Banerjee (Banerjee et al. 1999) for two prehistoric beads from Romania, both considered of Baltic origin.

As a hint for the archaeological output, in order to rationalize on the importance of nondestructive analysis, it should be mentioned that the quantity and quality of amber artefacts from Romania is lower than those found in territories placed on the track of the so-called Amber Route (Teodor ES et al, 2010). Nevertheless, from our studies, the Baltic amber is encountered very early, in Neolithic, i.e. only three isolated beads (not published), and only in southern Romania, as a clue that Danube may be considered as a secondary Amber Trail.

## **2.1 Assessment of amber origin based on FTIR analysis**

Taking into account the lack of referential with respect to ascription of amber origin, and considering as critical issue the transformation occurring on amber samples during historical ages strongly influenced by the storing conditions in the deposits where were discovered, in the attempt of establishing few definite FTIR criteria to be useful for amber origin proper assignment we started the analysis not directly with archaeological samples but first, with geological amber, of controlled and certified origin, both from Baltic sources and from Romanian sources. Further obtained results are detailed.

#### **2.1.1 FTIR method**

2010b).

354 Infrared Spectroscopy – Materials Science, Engineering and Technology

on a list of so-called potential trace evidence, different items which require certain identification; the mentioned list may be formed by objects from various materials: paint, pottery, glass, bricks, liquid residues (oils, wines, and perfumes), fabrics, soil, metals etc. In order to rationalize on the evidences origin, source, transformation and historical age, the archaeologists need a deep analytical study, study that often has to be performed on irreplaceable pieces. Consequently, to accomplish these goals multiple analytical techniques, able to work as non-destructive techniques are used, such as: Fourier Transform Infrared Spectrometry (FTIR), scanning electron microscopy (SEM), polarized light microscopy

In our studies, in the amber analysis the main goal was to establish several criteria able to be useful in discrimination between different types of amber, the obtained FTIR information being used to ascribe the origin of the materials found in archaeological sites on the

Amber is a fossil resin, extremely appreciated for its colour and beauty, and used as a gemstone from very early times. Natural amber deposits are found all around the world, the most famous being the ones from the Baltic sea, the Dominican republic, Sicily, Borneo, Spain, etc. Amber resources are documented in Romania since the late sixteenth century, around the village of Colţi (Buzău County), but they have certainly been exploited only starting with the modern age, after 1828, in the same area (Ghiurcă, 1999; Wollmann, 1996). The existence of a local amber exploitation on Romanian territory aroused the question of

During recent years, analysts used mostly FTIR, mass spectrometry and pyrolysis-gas chromatography-mass spectrometry for the analytical study of geological or archaeological amber from different territories (Beck, 1972; 1986; Boon et al., 1993; Lambert & Poinar, 2002; Shedrinski et al., 2004; Angelini & Bellintani, 2005; Guiliano et al., 2007; Pakutinskiene et al., 2007; Tonidandel et al., 2008; Teodor ED et al., 2009; Vîrgolici et al., 2010a). In 2005, Angelini and Bellintani reviewed the analytical techniques used for the differentiation of amber types, dedicating special attention to the most suitable methods for archaeological materials

There are a few comparative analytical studies of Romanian geological amber, referred to as Rumanite or Romanite, which conclude that Romanian amber is geologically younger, but

Regarding archaeological amber from Romanian territory, it was basically unstudied since we started our research project in 2007. Some studies by Nikolaus Boroffka on prehistoric amber (Boroffka, 2006) refer to chemical analyses of amber with an assumed Romanian origin; in another paper, Boroffka (2001) discusses the results published by Banerjee (Banerjee et al. 1999)

As a hint for the archaeological output, in order to rationalize on the importance of nondestructive analysis, it should be mentioned that the quantity and quality of amber artefacts from Romania is lower than those found in territories placed on the track of the so-called Amber Route (Teodor ES et al, 2010). Nevertheless, from our studies, the Baltic amber is encountered very early, in Neolithic, i.e. only three isolated beads (not published), and only in southern Romania, as a clue that Danube may be considered as a secondary Amber Trail.

chemically almost identical, to Baltic amber (Stout et al., 2000).

for two prehistoric beads from Romania, both considered of Baltic origin.

geological origin of amber artefacts found by Romanian archaeologists over time.

(PLM) or X-photon spectrometry (XPS).

Romanian territory.

(non-destructive methods).

FTIR assays of amber were performed using the FTIR-VAR technique with a beam incidence angle of 45o, on a Bruker TENSOR 27 instrument, using the OPUS software version 6.0. The samples were used without any pre-treatment, as whole pieces fixed on a gold mirror, and all the spectra were registered versus a background of clean gold foil between 4000 and 600 cm-1. The spectral resolution was 4 cm-1, and the co-added scans 96, with an aperture of 4 nm. The FTIR-VAR technique is able to provide the same information as FTIR in terms of transmittance, with a slight loss in signal intensity, but with the advantage of preserving samples integrity. In addition, the bounds between atom plans are not destroyed, and, as consequence, the amount of provided information increases.

#### **2.1.2 Geological amber FTIR assessment**

A large lot of samples from different controlled origins were analysed by FTIR-VAR, from Romania (Colti, Buzău County), from the Baltic region (Palanga and Kaliningrad), from Germany and Poland in order to settle a certain pattern to differentiate Romanian amber from other types of amber, especially the Baltic variety (the material was provided, mainly, from the National Geological Museum from Bucharest). In the case of unsure observations, to clarify some spectral zone from the fingerprint region (1300-900 cm-1), or the region 1800- 1450 cm-1, the transmittance spectra of the samples were used for a better evaluation. Spectra were analysed and assigned on the three wave-numbers domains of significance for amber, namely those between 3600-2000 cm -1, 1820-1350 cm-1 and the 1250-1045 cm-1 regions, which correlate with hydroxyl groups, carboxyl groups, carbonyl groups and with C=C instauration.

Subsequent data acquisition the following observations were drawn:

a. There are no notable differences between Romanite spectra and Baltic spectra in the 3600-3000 cm-1 region mainly in case of geological samples. The noteworthy differences

Fourier Transform Infrared Spectroscopy –

amber where it occurs about 1640 cm-1).

for each duplicate recorded spectrum.

understanding of the noticed differences.

materials being the probable cause (Fig. 2).

from Romania.

(Figs. 2-3).

**2.1.3 Archaeological amber FTIR assessment** 

Useful Analytical Tool for Non-Destructive Analysis 357

d. The amber has an amorphous structure, and as consequence the reproducibility of determinations is affected by the heterogeneity of the sample, the obtained spectra depending on the analysed part of the sample. As a result of these observations, for each sample were registered two spectra, for different zones of sample, in order to observe if significant differences appear between the two, due to appearance/disappearance of certain absorption bands (specific vibration frequencies of interest). It was observed that difference between two spectra of the same sample appear only in the spectral region related to methyl, methylene, *etc.* (-CH3, -(CH2)n-, - CH, *etc*.) chains, namely 2962–2850 cm-1, and for the specific tensile bands of CH3 from 1375 cm-1. This led to the conclusion that the presence of functional groups (which are important because they correlate with the origin and the age of the sample, and giving the sample specificity) are equally distributed. Therefore it could be considered that the analytic information supplied by FTIR is acceptable, since the wavenumbers corresponding to carboxylic chains and hydroxy-carboxylic acids at 1547-1423 cm-1 for Romanite and, respectively, 1547-1300 cm-1 for Baltic amber are preserving their region

The differences appearing in the 1684-1642 cm-1 band of Baltic amber with respect to the 1547 cm-1 band of Romanite are determined by the shift of asymmetric vibration frequencies for the carboxylic type groups, as a function of the length of hydrocarbon chains. A strengthening-ageing of polymeric chains leads to shifts toward smaller wavenumbers. This can be observed also in 1046-900 cm-1 region, where the shifts toward the smaller wavenumbers are related to the better confirmation of the Baltic origin. The significant regions for each type of amber are summarised in Table 1, corresponding comment with respect to archaeological amber spectra being introduced at this point for a better

After analysis of geological specimens of amber, the next step was devoted to archaeological samples analysis. The investigated samples were originated from different times (Neolithic, Bronze Age, Roman Age, Byzhantine period, etc) found on different archaeological sites

Concerning the archaeological samples, the FTIR-VAR spectra pointed out some differences between the archaeological and geological material. Archaeological samples most generally presented a less intense signal in the 3600-3000 cm-1 region, which correlates with –OH groups and influence the H bonds (Fig. 2); the presence of –OH groups was confirmed by a signal of lower intensity for archaeological samples in the 746-736 cm -1 region, especially for Baltic amber, the relative dehydration of archaeological samples as compared to geological

The fingerprint zone is different in most of archaeological samples in comparison with geological ones. This region presents adsorption bands of higher intensity for the majority of archaeological samples and some shifts of the specific wavenumbers take place

due to intramolecular hydrogen bonds toward 1595 cm-1, if compared to the Baltic

appear (as in any IR technique) in the 'Baltic shoulder' region, 1250-1060 cm-1, and in the 1161-1155 cm-1 region. For the Baltic amber, the shoulder appears in the region 1275- 1155 cm-1, while that of Romanite has a different shape and is shifted to about 1045-1020 cm-1, as may be noticed from Fig. 1.

Fig. 1. The overlaid FTIR-VAR (reflectance) spectra, region 4000-600 cm-1 for geological reference Baltic amber (211) and reference Romanite (204)


Fig. 1. The overlaid FTIR-VAR (reflectance) spectra, region 4000-600 cm-1 for geological

b. An important difference in Variable Angle Reflectance spectra is registered for absorption bands from 900-600 cm-1 region (Fig. 1); in this region two species can be differentiated, depending on the age of the polymer. The absence of some characteristic bands indicates that the contraction-reticulation of the polymeric chain is finished (connected with the older age of fossil resin); the absence of a double peak at 667 cm-1 in the Baltic amber spectra and the appearance of the shifted shoulder toward 1045 cm-1, in the case of Romanite when compared to the Baltic shoulder, that occur at higher wavenumbers (demonstrated also by Fourier spectra de-convolution) provided evidence that Romanite was formed after Baltic amber (*i.e*., it is younger), as it may be

c. There are vibration frequency shifts determined by the degree of ethers and esters formation, correlated to the number of functional groups from the chain and showing evidence on the intramolecular bounds, which appear frequently in amber type resins with younger ages (for example the OH group, specific to Romanite, is slightly shifted

reference Baltic amber (211) and reference Romanite (204)

observed in Fig. 1.

cm-1, as may be noticed from Fig. 1.

appear (as in any IR technique) in the 'Baltic shoulder' region, 1250-1060 cm-1, and in the 1161-1155 cm-1 region. For the Baltic amber, the shoulder appears in the region 1275- 1155 cm-1, while that of Romanite has a different shape and is shifted to about 1045-1020 due to intramolecular hydrogen bonds toward 1595 cm-1, if compared to the Baltic amber where it occurs about 1640 cm-1).

d. The amber has an amorphous structure, and as consequence the reproducibility of determinations is affected by the heterogeneity of the sample, the obtained spectra depending on the analysed part of the sample. As a result of these observations, for each sample were registered two spectra, for different zones of sample, in order to observe if significant differences appear between the two, due to appearance/disappearance of certain absorption bands (specific vibration frequencies of interest). It was observed that difference between two spectra of the same sample appear only in the spectral region related to methyl, methylene, *etc.* (-CH3, -(CH2)n-, - CH, *etc*.) chains, namely 2962–2850 cm-1, and for the specific tensile bands of CH3 from 1375 cm-1. This led to the conclusion that the presence of functional groups (which are important because they correlate with the origin and the age of the sample, and giving the sample specificity) are equally distributed. Therefore it could be considered that the analytic information supplied by FTIR is acceptable, since the wavenumbers corresponding to carboxylic chains and hydroxy-carboxylic acids at 1547-1423 cm-1 for Romanite and, respectively, 1547-1300 cm-1 for Baltic amber are preserving their region for each duplicate recorded spectrum.

The differences appearing in the 1684-1642 cm-1 band of Baltic amber with respect to the 1547 cm-1 band of Romanite are determined by the shift of asymmetric vibration frequencies for the carboxylic type groups, as a function of the length of hydrocarbon chains. A strengthening-ageing of polymeric chains leads to shifts toward smaller wavenumbers. This can be observed also in 1046-900 cm-1 region, where the shifts toward the smaller wavenumbers are related to the better confirmation of the Baltic origin. The significant regions for each type of amber are summarised in Table 1, corresponding comment with respect to archaeological amber spectra being introduced at this point for a better understanding of the noticed differences.

## **2.1.3 Archaeological amber FTIR assessment**

After analysis of geological specimens of amber, the next step was devoted to archaeological samples analysis. The investigated samples were originated from different times (Neolithic, Bronze Age, Roman Age, Byzhantine period, etc) found on different archaeological sites from Romania.

Concerning the archaeological samples, the FTIR-VAR spectra pointed out some differences between the archaeological and geological material. Archaeological samples most generally presented a less intense signal in the 3600-3000 cm-1 region, which correlates with –OH groups and influence the H bonds (Fig. 2); the presence of –OH groups was confirmed by a signal of lower intensity for archaeological samples in the 746-736 cm -1 region, especially for Baltic amber, the relative dehydration of archaeological samples as compared to geological materials being the probable cause (Fig. 2).

The fingerprint zone is different in most of archaeological samples in comparison with geological ones. This region presents adsorption bands of higher intensity for the majority of archaeological samples and some shifts of the specific wavenumbers take place (Figs. 2-3).

Fourier Transform Infrared Spectroscopy –

Useful Analytical Tool for Non-Destructive Analysis 359

Fig. 2. The overlaid FTIR-VAR (reflectance) spectra, region 4000-600 cm-1 for geological reference Baltic amber (211), reference Romanite (204), and an archaeological sample from

Another difference in VAR spectra of geological origin is registered for absorption bands from 800-600 cm-1 region: 622-621 cm-1 are ascribed to C-O, C=O out of plane bands; 746-736 cm-1 are the confirmation of OH and H bonds; 676-666 cm-1 are the confirmation of CH2 bonds. The absence of a double peak at 667 cm-1 in the Baltic amber spectra indicates that the contraction-reticulation of the polymeric chain is complete (an aspect related to the older age of the fossil resin, in our case the Baltic species). This region is not so well defined in

Nufăru (746)-Byzanthine period- asigned by us to Baltic origin

archaeological samples, especially for Baltic amber (Figs. 2 and 3).


Table 1. The main characteristics of spectral domains in Baltic amber and Romanite and comparison between geological and archaeological samples)

OH and bound intramolecular OH) from hydroxy-acids; observed wavenumbers shifts characteristic of amorphous

out of plane vibration of bonded C-H-C. Plane vibration of H bonds (to O). Assignment confirmed by the peaks from 620-750 cm-1 domain. 1600-1604 –COOunsaturated from carboxylic acids;

usually occurs as wide band at 1595 cm-1, shifts and appears as shoulder toward 1617 **Comments regarding the differences between geological and archaeological spectra** 

Less intense signals in this region for archaeological samples and slight shifts

More intense signals in archaeological samples

Not always present in archaeological samples

More intense signals in archaeological samples,

Less intense signals in this region for archaeological samples and slight shifts

More intense signals in archaeological samples,

archaeological samples

More intense signals in archaeological samples,

Not so well defined in archaeological samples

shifts

shifts

shifts

**domain (cm-1) Signal Assignment** 

structures.

3893-3200 OH frequencies, alcoholic (phenolic free

1684-1600 C=C; C=O frequencies of bond vibrations;

1617 OH group, specific to Romanite, that

or respectively 1625, due to H intramolecular bonds.

1547-1423 Specific for carboxylic chains and hydroxy-

1046-doublet Specific peak OH & C=O, slightly shifted due to intramolecular bonds. 750-620 Out of plane vibrations of C=O; combined

3893-3200 Same assignment as above, the differences

1684-1600 Same assignment as above, the differences

1547-1300 Same functional groups, carboxylic chains and hydroxyl-carboxylic acids;

1046/1001 Very well defined peak of OH & C=O;

to Romanite

comparison between geological and archaeological samples)

750-620 Single peak, sharp, intense, specific for C-O,

bonds with those specific to C-C=O

appear in intensity and slight shifts, due to the different number of intramolecular bonds with respect to those of Romanite;

appear in intensity and slight shifts, due to the different number of intramolecular bonds with respect to those of Romanite

compared to 1046-doublet of Romanite, has different shape and is shifted toward smaller wavenumbers in Baltic amber

region slightly different in shape compared

Table 1. The main characteristics of spectral domains in Baltic amber and Romanite and

1155-1275 Baltic shoulder Not always present in

carboxylic acids;

**Sample of controlled origin (geological)** 

Romanian amber (Colti, Buzău County)

Baltic amber (Bitterfeld, Germany)

**Wavenumber** 

Fig. 2. The overlaid FTIR-VAR (reflectance) spectra, region 4000-600 cm-1 for geological reference Baltic amber (211), reference Romanite (204), and an archaeological sample from Nufăru (746)-Byzanthine period- asigned by us to Baltic origin

Another difference in VAR spectra of geological origin is registered for absorption bands from 800-600 cm-1 region: 622-621 cm-1 are ascribed to C-O, C=O out of plane bands; 746-736 cm-1 are the confirmation of OH and H bonds; 676-666 cm-1 are the confirmation of CH2 bonds. The absence of a double peak at 667 cm-1 in the Baltic amber spectra indicates that the contraction-reticulation of the polymeric chain is complete (an aspect related to the older age of the fossil resin, in our case the Baltic species). This region is not so well defined in archaeological samples, especially for Baltic amber (Figs. 2 and 3).

Fourier Transform Infrared Spectroscopy –

Rosia Montană, statuette (6 samples)

Dobrogea region (Tomis, Callatis, Beroe,

Pitesti, one bead (2 samples) Neolithic

Grădistea-Cotlogeni, one bead

Cioclovina hoard, 12 beads

Carsium, Noviodunum) 25 beads (85 samples)

(43 samples)

Nufăru

Useful Analytical Tool for Non-Destructive Analysis 361

**Sample origin Age FTIR-VAR assignment** 

Late Bronze

Roman

Table 2. The main categories of archaeological samples analyzed by FTIR-VAR and the

Fig. 4. The overlaid FTIR-VAR (reflectance) spectra, region 4000-600 cm-1 for geological reference Baltic amber (211), reference Romanite (204), and an archaeological sample from

Cioclovina (456)-Bronze period- assigned by us to Romanian origin

One pectoral cross, 2 beads (5 samples) Byzantine Baltic amber

assignment of amber origin in comparison with geological samples

Baltic amber (unpublished)

Baltic amber

10 beads Romanite 2 beads Baltic amber (Teodor ES et al., 2010)

16 beads Baltic amber 9 beads Romanite (Vîrgolici et al., 2010b)

Fig. 3. The overlaid FTIR-VAR (reflectance) spectra, region 4000-600 cm-1 for geological reference Baltic amber (211), reference Romanite (204), and an archaeological sample from Pitesti (1076) –neolithic period- assigned by us to Baltic origin

Based on the relative comparison with the geological reference material (both Baltic and Romanian amber, see Table 1) and according to the FTIR-VAR bands of archaeological samples from the region 1820-1350 cm-1, 1275-1020 cm-1 and 900-600 cm-1, the classification of several archaeological samples is summarised in Table 2. Some spectra obtained for the samples from Nufăru (Roman Age), Cioclovina (Bronze Age) and a Neolithic bead are presented in Figs. 2, 3 and 4.

Fig. 3. The overlaid FTIR-VAR (reflectance) spectra, region 4000-600 cm-1 for geological reference Baltic amber (211), reference Romanite (204), and an archaeological sample from

Based on the relative comparison with the geological reference material (both Baltic and Romanian amber, see Table 1) and according to the FTIR-VAR bands of archaeological samples from the region 1820-1350 cm-1, 1275-1020 cm-1 and 900-600 cm-1, the classification of several archaeological samples is summarised in Table 2. Some spectra obtained for the samples from Nufăru (Roman Age), Cioclovina (Bronze Age) and a Neolithic bead are

Pitesti (1076) –neolithic period- assigned by us to Baltic origin

presented in Figs. 2, 3 and 4.


Table 2. The main categories of archaeological samples analyzed by FTIR-VAR and the assignment of amber origin in comparison with geological samples

Fig. 4. The overlaid FTIR-VAR (reflectance) spectra, region 4000-600 cm-1 for geological reference Baltic amber (211), reference Romanite (204), and an archaeological sample from Cioclovina (456)-Bronze period- assigned by us to Romanian origin

Fourier Transform Infrared Spectroscopy –

**Wavenumber domain (cm-1)** 

cm-1, 98 scans being acquired for each spectrum, with 2.2 Hz.

to oxidative changes induced on cellular membrane by ROS.

2800-3000 CH2 – symmetric and asymmetric

1740 C=O – stretching vibration for ester

1630-1680 C=O – stretching vibration for

1541 N-H – deformation

1466 CH2 – deformation vibration for

1170-1245 PO2- – asymmetric stretching

1063-1100 PO2- – symmetric stretching

Table 3. Main IR absorption bands of low-density lipoprotein

**3.1.2 FTIR assessment of native and oxidised low-density lipoprotein** 

well defined, no matter the type of the sample, either free or deposed LDL.

stretching vibration, CH3 – asymmetric stretching vibration

group of lipids

carbonyl group for protein, amide I

(scissoring)vibration for amine group of protein, amide II

methylene group

vibration for phosphate groups

vibration for phosphate groups CO-O-C vibration for esters groups

Useful Analytical Tool for Non-Destructive Analysis 363

bromide (KBr) pellet, the background spectrum being recorded against KBr pellet. In reflectance mode the spectral range was 4000 – 600 cm-1, the background being recorded for an unmodified polycrystalline gold sheet after that being recorded the samples' spectrum. The optimum reflectance angle was 45 degrees, the aperture 6 cm, the spectra resolution 4

As mentioned in the previous section, the first FTIR studies were performed to evaluate the suitable deposition of the low-density lipoprotein on the gold support in order to preserve the lipoprotein structural characteristics. This is an important experimental issue, because the protein structure and surface charging are the most important structural features necessary to be maintained to ensure feasible "ex-vivo" analytical information with respect

The main regions corresponding to FTIR absorption bands and specific for native LDL are given in table 3. Absorption bands from 2800–3000 cm-1 corresponding to methyl groups are

**Signal assignment Comments and observations** 

(<sup>s</sup>

2927 cm-1 (asCH2), 2854 cm-1 (<sup>s</sup>

2954 cm-1 (asCH3) and 2866 cm-1


1656 cm-1 band correspond to amide I helix structure being the result of 60% vibration of C=O group and 20% vibration of N-H

Is due to 60 % vibration of N-H bound and the rest from -CO-NHfunctional group



Studies performed on incidence angles different of 45o proved that ester groups are oriented preferentially at 75o, the intensity of the corresponding band at 1099cm-1 increasing suddenly

CH3) are specific for lipid chains

CH2),

These results based on comparison of FTIR-VAR spectra about origin of amber artefacts were clarified and confirmed by other non-destructive analysis (Raman, statistical analysis) and supported by historical context, in some cases. In other cases, the assignment of samples to Baltic amber or local (Romanian) amber contributed to understanding of archaeological (historical) context.

FTIR-VAR is a reliable tool for non-destructive investigating of amber. The spectra are more complex than FTIR-transmittance spectra, the obtained signal are rather highly difficult to ascribe and the variability of results is high, but coupled with statistical analysis and corroborated with other techniques (Raman spectroscopy, X-rays fluorescence) is a consistent method to diagnose the origin of amber in archaeological artifacts.

## **3. Applicability of Fourier Transform Infrared technique in the assessment of bio-mimicking systems. Case study- Human low density lipoproteins**

In order to sustain the assertion regarding the FTIR versatility as non-destructive analytical technique able to provide valuable information with respect to noteworthy molecular changes, the second exemplification is based on assessment of significant bio-markers used to indicate oxidative modifications occurring on cellular membrane. At the level of living organisms the cellular membranes are the main targets for reactive oxygen species (ROS) and reactive nitrogen species (RNS). Both ROS and RNS induce lipid peroxidation (LPO) of unsaturated fatty acids from membrane phospholipids. At the cellular level, reactive species are responsible for changes in the membrane properties, inducing modification on cell permeability and enzyme activities (Ahsan et al., 2003).

In our experiments we developed a bio-mimetic to be applied in assessment of lipoperoxidation processes using as significant oxidative substrate a component of the cellular membrane, low-density lipoproteins. The system development was based on system deposing a thin lipoproteic layer on the surface of conductive solid support, usually gold sheet, and quantification of the lipoprotein peroxidation degree by electrochemical measurements, the intensity of the registered intensity current at the lipoperoxides specific peak potential being proportional with the amount of formed lipoperoxides. At this point have to be mentioned the fact that we employed the VAR-FTIR in analysis of structural modifications occurring to low-densitylipoprotein as result of the oxidative changes induced by free radicals (ROS) attack.

## **3.1.1 FTIR method**

Starting from the important experimental issue which states that an appropriate development of a bio-mimetic system by immobilization of a compound to a support preserves the structure of the immobilized compound, the first FTIR study was performed on low-density lipoprotein layer deposed on gold support, the results being compared to those obtained for un-bounded lipoprotein. For the deposed layer the reflectance mode was employed, while for free lipoprotein the transmittance mode was used. FTIR spectra were recorded at room temperature using a Bruker Tensor 27 Fourier Transform spectrometer. The spectra were collected and ascribed using Opus software. In transmittance mode the spectral range was 4000 – 400 cm-1, the aperture 4 cm, the spectra resolution 4 cm-1 and 98 scans being acquired for each spectrum, 20 Hz. Samples were pressed into a potassium

These results based on comparison of FTIR-VAR spectra about origin of amber artefacts were clarified and confirmed by other non-destructive analysis (Raman, statistical analysis) and supported by historical context, in some cases. In other cases, the assignment of samples to Baltic amber or local (Romanian) amber contributed to understanding of archaeological

FTIR-VAR is a reliable tool for non-destructive investigating of amber. The spectra are more complex than FTIR-transmittance spectra, the obtained signal are rather highly difficult to ascribe and the variability of results is high, but coupled with statistical analysis and corroborated with other techniques (Raman spectroscopy, X-rays fluorescence) is a

**3. Applicability of Fourier Transform Infrared technique in the assessment of** 

In order to sustain the assertion regarding the FTIR versatility as non-destructive analytical technique able to provide valuable information with respect to noteworthy molecular changes, the second exemplification is based on assessment of significant bio-markers used to indicate oxidative modifications occurring on cellular membrane. At the level of living organisms the cellular membranes are the main targets for reactive oxygen species (ROS) and reactive nitrogen species (RNS). Both ROS and RNS induce lipid peroxidation (LPO) of unsaturated fatty acids from membrane phospholipids. At the cellular level, reactive species are responsible for changes in the membrane properties, inducing modification on cell

In our experiments we developed a bio-mimetic to be applied in assessment of lipoperoxidation processes using as significant oxidative substrate a component of the cellular membrane, low-density lipoproteins. The system development was based on system deposing a thin lipoproteic layer on the surface of conductive solid support, usually gold sheet, and quantification of the lipoprotein peroxidation degree by electrochemical measurements, the intensity of the registered intensity current at the lipoperoxides specific peak potential being proportional with the amount of formed lipoperoxides. At this point have to be mentioned the fact that we employed the VAR-FTIR in analysis of structural modifications occurring to low-densitylipoprotein as result of the oxidative changes induced

Starting from the important experimental issue which states that an appropriate development of a bio-mimetic system by immobilization of a compound to a support preserves the structure of the immobilized compound, the first FTIR study was performed on low-density lipoprotein layer deposed on gold support, the results being compared to those obtained for un-bounded lipoprotein. For the deposed layer the reflectance mode was employed, while for free lipoprotein the transmittance mode was used. FTIR spectra were recorded at room temperature using a Bruker Tensor 27 Fourier Transform spectrometer. The spectra were collected and ascribed using Opus software. In transmittance mode the spectral range was 4000 – 400 cm-1, the aperture 4 cm, the spectra resolution 4 cm-1 and 98 scans being acquired for each spectrum, 20 Hz. Samples were pressed into a potassium

consistent method to diagnose the origin of amber in archaeological artifacts.

permeability and enzyme activities (Ahsan et al., 2003).

by free radicals (ROS) attack.

**3.1.1 FTIR method** 

**bio-mimicking systems. Case study- Human low density lipoproteins** 

(historical) context.

bromide (KBr) pellet, the background spectrum being recorded against KBr pellet. In reflectance mode the spectral range was 4000 – 600 cm-1, the background being recorded for an unmodified polycrystalline gold sheet after that being recorded the samples' spectrum. The optimum reflectance angle was 45 degrees, the aperture 6 cm, the spectra resolution 4 cm-1, 98 scans being acquired for each spectrum, with 2.2 Hz.

## **3.1.2 FTIR assessment of native and oxidised low-density lipoprotein**

As mentioned in the previous section, the first FTIR studies were performed to evaluate the suitable deposition of the low-density lipoprotein on the gold support in order to preserve the lipoprotein structural characteristics. This is an important experimental issue, because the protein structure and surface charging are the most important structural features necessary to be maintained to ensure feasible "ex-vivo" analytical information with respect to oxidative changes induced on cellular membrane by ROS.

The main regions corresponding to FTIR absorption bands and specific for native LDL are given in table 3. Absorption bands from 2800–3000 cm-1 corresponding to methyl groups are well defined, no matter the type of the sample, either free or deposed LDL.



Fourier Transform Infrared Spectroscopy –

C=O groups.

components of the cellular membrane.

Useful Analytical Tool for Non-Destructive Analysis 365

Fig. 6. FTIR spectra of LDL on Au support (variable angle reflectance, incidence angle 45o)

intensity and shifted as wavenumber, thus proving the lipoperoxides formation.

The performed studies proved that in the presence of free radicals (Fig. 7), the IR absorption bands corresponding to ester groups from lipid residues at 1740 cm-1 were changed in

Another proof of lipoperoxides existence was the presence of HO absorption bands from 3600–3700 cm-1 and 917 cm-1. In detail, the signal ascription are as follows: the band from 3278 cm-1 increasing in intensity and shoulders rise at 3334 cm-1, 3304 cm-1, 3220 cm-1 and 3197 cm-1 assigned to O-H and N-H vibrations. The bands from 3072 cm-1 are corresponding to hidroxil and amine groups involved in hydrogen bonds. In the same time, it is important to mention that several changes arose around 1717 cm-1 indicating the formation of new

The changes observed on FTIR spectra of deposed LDL correlated perfectly with electrochemical and MALDI information (Litescu et al, 2011), proving the usefulness of nondestructive FTIR analysis in ascribing the oxidative modification occurring on lipoproteic

As observable from figures 5 and 6, the amide specific band absorptions for proteins, amide I band around 1654 cm-1 and amide II band around 1541 cm-1 (Firth et al. 2008; Banuelos et al. 1995) are not changed when LDL was deposited on the gold support. This observation is important because it proved that the secondary structure of protein is preserved subsequent deposition therefore it can be concluded that the deposition on solid support did not affect theLDL functionality, and, consequently, that deposed LDL is expected to react with free radicals according to the same pathway as free LDL. Moreover, it should be mentioned that this argument is consistent with the data published by Paker (Paker, 1991) where it is mentioned that LDL *ex vivo* peroxidation pathway is similar as *in vivo* peroxidation pathway.

Fig. 5. FTIR spectra of LDL free, on transmittance (KBr pelleted)

When lipoprotein are subjected to the attack of peroxyl radicals, thermally induced from aqueous solutions of azo-initiator 2,2'-azobis (2methylpropionamidine).dihydrochloride (AAPH), the lipoperoxidation occurs according to a radical pathway (Litescu et al, 2002; Tache et al., 2011) leading to lipo-peroxides formation on the LDL layer. In our experiments the FTIR analysis was performed first on the LDL deposed on the solid support, then the LDL was subjected to free radicals attack for 10 minutes, allowed to dry on inert atmosphere and after that the oxidised LDL layer was once more assessed by FTIR.

As observable from figures 5 and 6, the amide specific band absorptions for proteins, amide I band around 1654 cm-1 and amide II band around 1541 cm-1 (Firth et al. 2008; Banuelos et al. 1995) are not changed when LDL was deposited on the gold support. This observation is important because it proved that the secondary structure of protein is preserved subsequent deposition therefore it can be concluded that the deposition on solid support did not affect theLDL functionality, and, consequently, that deposed LDL is expected to react with free radicals according to the same pathway as free LDL. Moreover, it should be mentioned that this argument is consistent with the data published by Paker (Paker, 1991) where it is mentioned that LDL *ex vivo* peroxidation pathway is similar as *in vivo* peroxidation

Fig. 5. FTIR spectra of LDL free, on transmittance (KBr pelleted)

and after that the oxidised LDL layer was once more assessed by FTIR.

When lipoprotein are subjected to the attack of peroxyl radicals, thermally induced from aqueous solutions of azo-initiator 2,2'-azobis (2methylpropionamidine).dihydrochloride (AAPH), the lipoperoxidation occurs according to a radical pathway (Litescu et al, 2002; Tache et al., 2011) leading to lipo-peroxides formation on the LDL layer. In our experiments the FTIR analysis was performed first on the LDL deposed on the solid support, then the LDL was subjected to free radicals attack for 10 minutes, allowed to dry on inert atmosphere

pathway.

Fig. 6. FTIR spectra of LDL on Au support (variable angle reflectance, incidence angle 45o)

The performed studies proved that in the presence of free radicals (Fig. 7), the IR absorption bands corresponding to ester groups from lipid residues at 1740 cm-1 were changed in intensity and shifted as wavenumber, thus proving the lipoperoxides formation.

Another proof of lipoperoxides existence was the presence of HO absorption bands from 3600–3700 cm-1 and 917 cm-1. In detail, the signal ascription are as follows: the band from 3278 cm-1 increasing in intensity and shoulders rise at 3334 cm-1, 3304 cm-1, 3220 cm-1 and 3197 cm-1 assigned to O-H and N-H vibrations. The bands from 3072 cm-1 are corresponding to hidroxil and amine groups involved in hydrogen bonds. In the same time, it is important to mention that several changes arose around 1717 cm-1 indicating the formation of new C=O groups.

The changes observed on FTIR spectra of deposed LDL correlated perfectly with electrochemical and MALDI information (Litescu et al, 2011), proving the usefulness of nondestructive FTIR analysis in ascribing the oxidative modification occurring on lipoproteic components of the cellular membrane.

Fourier Transform Infrared Spectroscopy –

when coupled with statistical analysis.

200, ISSN: 0570-4928

8090102654 9788090102651

389-407, ISSN: 1582-0688

*Archaeometry* 47, 441-54, ISSN: 0003-813X

**5. Acknowledgment** 

232522/2009.

**6. References** 

Useful Analytical Tool for Non-Destructive Analysis 367

In the same time, it should be emphasized that, in our opinion, FTIR-VAR technique use as non-destructive way to analyse archaeological artefacts is strongly recommended, especially

This work was partially supported by EU FP7-SME project Sensbiosyn, contract no

Angelini I. & Bellintani P. (2005). Archaeological ambers from northern Italy: an FTIR–

Ahsan, H., Ali, A. & Ali, R.(2003). Oxygen free radicals and systemic autoimmunity. *Clinical* 

Banuelos, S., Arrondo J. L. R., Goni F. M.& Pifat G.(1995). Surface-core relationships in

Banerjee A., Ghiurcă V., Langer B. & Wilhelm M. (1999). Determination of the provenance of

Beck C. W. (1986). Spectroscopic investigations of amber. *Applied Spectroscopy Review* 22, 57–

Boon J. J., Tom, A. & Pureveen, J. (1993). Microgram scale pyrolysis mass spectrometric and

Boroffka N. (2001). Bemerkugen zu einigen Bersteinfunden aus Rumänien. *Archäologisches* 

Boroffka N.( 2006). Resursele minerale din România şi stadiul actual al cercetărilor privind

Firth, C. A., Crone E. M., Flavall E. A., Roake J. A. & Gieseg S. P. (2008) Macrophage

Ghiurcă V. (1999). Chihlimbarul şi alte resurse gemologice din judeţul Buzău. *Mousaios* 5,

Guiliano M., Asia L., Onoratini G. & Mille G. (2007). Applications of diamond crystal ATR

Lambert, J. B. & Poinar, G. O. (2002). Amber: the organic gemstone. *Accounts of Chemical* 

mediated protein hydroperoxide formation and lipid oxidation in low density lipoprotein are inhibited by the inflammation marker 7,8-dihydroneopterin.

FTIR spectroscopy to the characterization of ambers. *Spectrochim. Acta A Mol.* 

*and . Experimental Immunology* 131, 398–404, ISSN: 1365-2249

Beck C. W. (1972). Amber in Archaeology. *Archaeology* 23, 7-11, ISSN: 0003-8113

*Biology & Chemistry*, 270, 9192–9196, ISSN 0021-9258

*Korrespondenzblatt* 31, 3, 395-409, ISSN: 0342-734X

*Biomol. Spectrosc*. 67, 1407-11, ISSN: 1386-1425

*Research* 35, 628–36, ISSN: 0001-4842

mineritul prehistoric. *Apulum* 43, 71-94, ISSN: 1013-428X

*Biochimica Biophysica Acta* 1783, 1095–101. ISSN 1570-9639

DRIFT study of provenance by comparison with the geological amber database*.* 

human low density lipoproteins as studied by infrared spectroscopy. *Journal of* 

two archaeological amber beads from Romania by FTIR- and Solid-State-Carbon-13 NMR spectroscopy. *Archäologisches Korrespondenzblatt* 29, 593-606, ISSN: 0342-734X

pyrolysis gas chromatographic characterisation of geological and archaeological amber and resin samples, in *Amber in archaeology: proceedings of the Second International Conference on Amber in Archaeology, Liblice, 1990* (eds. C. W. Beck and J. Bouzek), 9–27, Institute of Archaeology, Czech Academy of Sciences, Praha, ISBN:

Fig. 7. FTIR spectra of native peroxidised LDL deposed on Au support

## **4. Conclusions**

The use of FTIR, especially as variable angle reflectance technique, proved to be of crucial importance in two main types of applications where the critical issues are the importance to conserve the sample integrity and the limited amount of available sample.

Two examples of FTIR non-destructive analysis employment in two very different research domains, archaeology and modified surfaces are supporting the highest utility of the technique when the analysis is dealing with valuable samples, sometimes of patrimony value, which have to remain un-affected subsequent analysis, as it was the amber samples case study, or when the analysis have to provide arguments on sharp and specific structural changes, as it was the case of lipoprotein oxidation study.

It could be concluded that FTIR proved its efficiency in evaluation of the oxidative modifications on the level of one of the main components of the cellular membrane, oxidative process which plays, generally, as key-event in several age –related and degenerative pathologies.

In the same time, it should be emphasized that, in our opinion, FTIR-VAR technique use as non-destructive way to analyse archaeological artefacts is strongly recommended, especially when coupled with statistical analysis.

## **5. Acknowledgment**

This work was partially supported by EU FP7-SME project Sensbiosyn, contract no 232522/2009.

## **6. References**

366 Infrared Spectroscopy – Materials Science, Engineering and Technology

Fig. 7. FTIR spectra of native peroxidised LDL deposed on Au support

conserve the sample integrity and the limited amount of available sample.

changes, as it was the case of lipoprotein oxidation study.

The use of FTIR, especially as variable angle reflectance technique, proved to be of crucial importance in two main types of applications where the critical issues are the importance to

Two examples of FTIR non-destructive analysis employment in two very different research domains, archaeology and modified surfaces are supporting the highest utility of the technique when the analysis is dealing with valuable samples, sometimes of patrimony value, which have to remain un-affected subsequent analysis, as it was the amber samples case study, or when the analysis have to provide arguments on sharp and specific structural

It could be concluded that FTIR proved its efficiency in evaluation of the oxidative modifications on the level of one of the main components of the cellular membrane, oxidative process which plays, generally, as key-event in several age –related and

**4. Conclusions** 

degenerative pathologies.


**19** 

*Spain* 

**Infrared Spectroscopy in the Analysis** 

**of Building and Construction Materials** 

In the characterization of building and construction materials, the most frequently analytical tool performed have been X-ray diffraction but also, thermal analysis and microscopic techniques. Nowadays, infrared and other spectroscopic techniques have become as a useful, non-destructive and easy technique to study the phase composition of initial but also the evolved materials due to their exposure to the climatic conditions. Moreover, by using this tool is possible the detection of crystalline but also the amorphous phases very frequently developed on certain cementitious materials, mainly at early ages. The infrared spectroscopy is used both to gather information about the structure of compounds and as

The infrared spectra are quick and easy to achieve and refers to the spectrum region between the visible and microwave regions. In theory, infrared radiation is absorbed by molecules and converted into energy of molecular vibration; when the radiant energy matches the energy of a specific molecular vibration, absorption occurs. The frequencies at which a molecule absorbs radiation give information on the groups present in the molecule. As an approximation, the energy of a molecule can be separated into three additive components associated with the motion of the electrons in the molecule, the vibration of the constituent atoms, and the rotation of the molecule as a whole. The absorption in the infrared region arises predominantly from excitation of molecular vibrations. Then, if a molecule is placed in an electromagnetic field, a transfer of energy from the field to the

�� � ℎ� Where �� is the difference in energy between two quantized states, ℎ is the Planck's constant and � is the frequency of the light. Then the molecule "absorbs" Δ� when it is excited from ��to �� and "emits" Δ� when it reverts form �� to ��. The infrared absorption spectra originate in photons in the infrared region that are absorbed by transitions between

analytical tool to assess in qualitative and quantitative analysis of mixtures.

molecule will occur when Bohr's frequency condition is satisfied.

two vibrational levels of the molecule in the electronic ground state.

**1. Introduction** 

Lucia Fernández-Carrasco1, D. Torrens-Martín1, L.M. Morales1 and Sagrario Martínez-Ramírez2

*In memorial of Prof. Tomás Vázquez* 

*1Escola Tècnica Superior d'Arquitectura (UPC), LiTA, Barcelona 2Instituto de Estructura de la Materia (CSIC), IEM-CSIC, Madrid* 


## **Infrared Spectroscopy in the Analysis of Building and Construction Materials**

Lucia Fernández-Carrasco1, D. Torrens-Martín1, L.M. Morales1 and Sagrario Martínez-Ramírez2 *1Escola Tècnica Superior d'Arquitectura (UPC), LiTA, Barcelona 2Instituto de Estructura de la Materia (CSIC), IEM-CSIC, Madrid Spain* 

*In memorial of Prof. Tomás Vázquez* 

#### **1. Introduction**

368 Infrared Spectroscopy – Materials Science, Engineering and Technology

Litescu, SC, Cioffi, N, Sabbatini L, Radu G.L. (2002), Study of Phenol-Like Compounds

Litescu SC, Eremia SAV, Diaconu M., Tache A., Radu G.L (2011) Biosensor applications on

Pakutinskiene I., Kiuberis J., Bezdicka P., Senvaitiene J. & Kareiva A. (2007). Analytical

Shedrinsky A. M., Wampler T. P. & Chugunov K. V. (2004). The examination of amber beads

Stout E. C., Beck C. W. & Anderson, K. B. (2000). Identification of rumanite (Romanian

TacheA, Cotrone S, Litescu SC, Cioffi N., Torsi L.; Sabbatini L & Radu GL (2011)

Teodor E.D., Liţescu S. C., Neacşu A., Truică G. & Albu C. (2009). Analytical methods to

Teodor E.S., Teodor E.D., Vîrgolici M., Manea M.M., Truica G. & Litescu S. C. (2010). Non-

Tonidandel L., Ragazzi E., Roghi G. & Traldi P. (2008). Mass spectrometry in the

Vîrgolici M., Ponta C., Manea M., Neguţ D., Cutrubinis M., Moise I., Şuvăilă R., Teodor E. S.,

Vîrgolici M., Petroviciu I., Teodor E. D., Liţescu S. C., Manea M. M., Ponta C., Niculescu G.,

Wollmann V. (1996). Der Bernsteinbergbau von Colti, in *Bernstein - Tränen der Götter. Katalog* 

memorial site. *J. Anal. Appl. Pyrolysis* 71, 69–81, ISSN: 0165-2370

*Electroanalysis,* 14(12), 858-865, ISSN 1040-0397

*Clinical Nutrition*, 53, 1050S-1055S, ISSN 0002-9165

*Spectrosc.* 52, 287- 94, ISSN: 1205-6685

*Minerals* 27, 665–78, ISSN: 0342-1791

*Sci.* 37, 2386-96, ISSN: 0305-4403

*Spectrom*. 22, 630-38, ISSN: 0951-4198

Origins, *J. Chrom. A* 1217, 1977–1987, ISSN: 0021- 9673

(Roman age), *Rev. Roum. Chim.* 55*,* 349-55, ISSN: 0035-3930

*Cent. Eur. J. Chem.* 7, 560-68, ISSN: 1895-1066

760, ISSN 00032719

Antioxidative Behavior on Low- Density Lipoprotein Gold Modified Electrode,

Assessment of Reactive Oxygen Species and Antioxidants, in *Environmental Biosensors*, Ed. Vernon Sommerset, Intech, ISBN: 978-953-307-486-3, Riejda Croatia Paker, L., (1991). Protective role of vitamin E in biological systems, *American Journal of* 

characterization of Baltic amber by FTIR, XRD and SEM. *Canadian J. Anal. Sci.* 

from the collection of the state hermitage museum found in Arzhan-2 burial

amber) as thermally altered succinite (Baltic amber). *Physics and Chemistry of* 

Spectrochemical Characterization of Thin Layers of Lipoprotein Self-Assembled Films on Solid Supports Under Oxidation Process, *Analytical Letters,* 44 ( 5), 747 —

differentiate Romanian amber and Baltic amber for archaeological applications,

destructive analysis of amber artefacts from Prehistoric Cioclovina Hoard. *J. Arch.* 

characterization of ambers I. Studies of amber samples of different origin and ages by laser desorption ionization, atmospheric pressure chemical ionization and atmospheric pressure photoionization mass spectrometry. *Rapid Commun. Mass* 

Sârbu C. & Medvedovici A. (2010a). Thermal Desorption/Gas Chromatography/Mass Spectrometry Approach for Characterization of the Volatile Fraction from Amber Specimens: a Possibility of Tracking Geological

Sarbu C. & Andrei Medvedovici A. (2010b). TD/CGC/MS and FT-IR characterization of archaeological amber artefacts from Romanian collections

*des Ausstellung des Deutchen Bergbau-Museums Bochum in Zusammenarbeit mit dem Ostpreußischen Landesmuseum Lüneburg und dem Siebenbürgischen Museum Gundelsheim*, M. Ganzelewski, R. Slotta (eds), 369-377, Bochum, ISBN: 3-921533-57-0

In the characterization of building and construction materials, the most frequently analytical tool performed have been X-ray diffraction but also, thermal analysis and microscopic techniques. Nowadays, infrared and other spectroscopic techniques have become as a useful, non-destructive and easy technique to study the phase composition of initial but also the evolved materials due to their exposure to the climatic conditions. Moreover, by using this tool is possible the detection of crystalline but also the amorphous phases very frequently developed on certain cementitious materials, mainly at early ages. The infrared spectroscopy is used both to gather information about the structure of compounds and as analytical tool to assess in qualitative and quantitative analysis of mixtures.

The infrared spectra are quick and easy to achieve and refers to the spectrum region between the visible and microwave regions. In theory, infrared radiation is absorbed by molecules and converted into energy of molecular vibration; when the radiant energy matches the energy of a specific molecular vibration, absorption occurs. The frequencies at which a molecule absorbs radiation give information on the groups present in the molecule. As an approximation, the energy of a molecule can be separated into three additive components associated with the motion of the electrons in the molecule, the vibration of the constituent atoms, and the rotation of the molecule as a whole. The absorption in the infrared region arises predominantly from excitation of molecular vibrations. Then, if a molecule is placed in an electromagnetic field, a transfer of energy from the field to the molecule will occur when Bohr's frequency condition is satisfied.

#### �� � ℎ�

Where �� is the difference in energy between two quantized states, ℎ is the Planck's constant and � is the frequency of the light. Then the molecule "absorbs" Δ� when it is excited from ��to �� and "emits" Δ� when it reverts form �� to ��. The infrared absorption spectra originate in photons in the infrared region that are absorbed by transitions between two vibrational levels of the molecule in the electronic ground state.

Infrared Spectroscopy in the Analysis of Building and Construction Materials 371

Fig. 1. General rules in the interpretation of building cementing material IR spectra.

features of the crystal lattice.

shoulder at 538 cm-1.

groups. The Ca-O bands appear at lower frequencies.

As a resume, the vibrations can be divided in stretching and bending: vibrations can involve either a change in bond length (stretching) or bond angle (bending); some bonds can stretch in-phase (symmetrical stretching) or out-of-phase (asymmetric stretching). The main structure elements in building the crystal lattice of silicates are tetrahedral SiO4 groups, with may be either isolated as in the orthosilicates or connected with one another by common O atoms as in building an Si2O7 group from two connected tetraheda [Matossi]. Connection of SiO4 groups so as to form a ring of tetrahedral occurs in the crystal of alite (C3S) or belite (C2S). The infrared spectra of all silicates [Matossi] contain two reflection maxima near 1000 and 500 cm-1, which have been interpreted as a two active frequencies of a tetrahedral point group. In addition to these, there may occur other maxima corresponding to other particular

The spectrum of the main constituent of OPC, ܥ3ܵ, shows two regions dominated by the internal modes of SiO44- tetrahedral units, with to broad absorption bands centred between 890 and 955 cm-1, solved in to maxima near to 870 and 940 cm-1 due to the symmetric and antisymmetric stretching of Si-O bonds within tetrahedral SiO4 groups, 1 and 3, respectively. Another absorption band of medium intensity appears close to 525 cm-1 and a lower intense band sited near to 450 cm-1 due to the symmetric and antisymmetric bending of the O-Si-O bonds, 2 and 4, respectively (see details of maxims in Table 1 and Figure 2). The other calcium silicate phase spectra, ܥ2ܵ, exhibits strong bands in the area 1000-800 cm-1 with maximums at 990 and 840 cm-1 due the stretching Si-O bond of the silicon tetrahedron and the bending vibration absorption band appear at lower frequencies, 520 cm-1 and a

The ܥ3ܣ-cubic tricalcium aluminate polymorph spectra (Figure 3), shows a well-defined spectra with two dominant absorption areas with very broad bands. The first ones appear in the area between 950-650 cm-1 and the second ones appearing between 500-380 cm-1, respectively. The main observed maxima appear near to 900, 865, 820, 780, 720 and 705 cm-1 of AlO4-tetrahedral groups, and close to 520, 510, 460 and 414 cm-1 due to AlO6-octahedral

The grey colour of Portland cements is due to the presence of the names ferrite phases; in absence of elements other than calcium, aluminium, iron and oxygen, calcium

The application of infrared spectroscopy to the inorganic compounds started as a more frequent technique during the 60's with Lawson. This author made a first attempt to compile the work done in the relatively new field-Inorganic Infrared Spectroscopy since 1952 where 1171 references were reported. Farmer, in 1964, studied the silicates and Nakamoto in relation to the coordinated compounds prepared a helpful atlas of these compounds. Afremow (1966) presented for an important research of inorganic pigments and extenders in the mid-infrared region from 1500 cm-1 to 200 cm-1. The study of surface chemistry and the nature of surface functional groups was also advanced by Basila (1968).

In the first decade of infrared research on the study of Portland cements, Vázquez (1969), was a lead the way in the study by infrared spectroscopy the main present compounds in the Portland cement but also later, made some research about the carbonation processes of calcium aluminate cements. Also, the hydration of Portland cement and its constituents was developed by Bensted (1974).

After that initial period, several reports have been done in the study on cementitious materials by infrared. More recent studies in relation with the calcium aluminates cements were reported by Vázquez (1993). Later on, different papers have present some characterization of materials and evolution over several exposition conditions using the infrared spectroscopy as a complementary technique join to mainly XRD and SEM analytical tools.

Without doubt, the infrared spectroscopy has not been really used in the qualitative and quantitative analysis of these materials; the main uses have rather been in identification of compounds and few structural studies. The main objective of this chapter will be to present a revision of infrared spectra useful in the study of the building and construction materials, mainly cements, from the point of view of characterization.

## **2. Characterization of cementitious systems by infrared spectroscopy**

As a general rule, as it is easier to bend a bond than to stretch or compress it, in the spectra the stretching frequencies are higher than the corresponding bending frequencies; bonds to hydrogen have higher stretching frequencies than those to heavier atoms; and double bonds have higher stretching frequencies than single bonds (Figure 1).

## **2.1 Portland cement**

The ordinary Portland cement is made by firing raw materials - limestone, clay minerals, sand and iron minerals- at around ������ in a rotary kiln. At this temperature a series of chemical reaction take place and the clinker synthesized. Clinker is cooled, mixed with setting regulators (e.g. gypsum) grounded to a fine powder to obtain the cement. The common phases present in the cement clinkers are: alite (���� � ���2, �3�1), belite (���� � ���2, �2�1), tricalcium aluminate (���� � ��2�3, �3�1), and tetracalcium aluminate ferrite (���� � ��2�3���2�3, �4��1). One typical composition of cement consists of: �3� = 55-60% (wt); �2� = 15-20% (wt); �3� = 5-10% (wt); �4�� = 5-8% (wt) and ��̅ �2 = 2.6% (wt). In this chapter, synthetic silicates and aluminates phases have been used to identify infrared vibrations bands previous to study the more complicated commercial cement.

<sup>1</sup> Cement chemistry nomenclature is used: � � ���� �� � ���2� �� � ��2�3� �� � ��2�3� �� � �2����̅� ��<sup>3</sup>

The application of infrared spectroscopy to the inorganic compounds started as a more frequent technique during the 60's with Lawson. This author made a first attempt to compile the work done in the relatively new field-Inorganic Infrared Spectroscopy since 1952 where 1171 references were reported. Farmer, in 1964, studied the silicates and Nakamoto in relation to the coordinated compounds prepared a helpful atlas of these compounds. Afremow (1966) presented for an important research of inorganic pigments and extenders in the mid-infrared region from 1500 cm-1 to 200 cm-1. The study of surface chemistry and the nature of surface functional groups was also advanced by Basila (1968). In the first decade of infrared research on the study of Portland cements, Vázquez (1969), was a lead the way in the study by infrared spectroscopy the main present compounds in the Portland cement but also later, made some research about the carbonation processes of calcium aluminate cements. Also, the hydration of Portland cement and its constituents was

After that initial period, several reports have been done in the study on cementitious materials by infrared. More recent studies in relation with the calcium aluminates cements were reported by Vázquez (1993). Later on, different papers have present some characterization of materials and evolution over several exposition conditions using the infrared spectroscopy as a complementary technique join to mainly XRD and SEM analytical

Without doubt, the infrared spectroscopy has not been really used in the qualitative and quantitative analysis of these materials; the main uses have rather been in identification of compounds and few structural studies. The main objective of this chapter will be to present a revision of infrared spectra useful in the study of the building and construction materials,

As a general rule, as it is easier to bend a bond than to stretch or compress it, in the spectra the stretching frequencies are higher than the corresponding bending frequencies; bonds to hydrogen have higher stretching frequencies than those to heavier atoms; and double bonds

The ordinary Portland cement is made by firing raw materials - limestone, clay minerals, sand and iron minerals- at around ������ in a rotary kiln. At this temperature a series of chemical reaction take place and the clinker synthesized. Clinker is cooled, mixed with setting regulators (e.g. gypsum) grounded to a fine powder to obtain the cement. The common phases present in the cement clinkers are: alite (���� � ���2, �3�1), belite (���� � ���2, �2�1), tricalcium aluminate (���� � ��2�3, �3�1), and tetracalcium aluminate ferrite (���� � ��2�3���2�3, �4��1). One typical composition of cement consists of: �3� = 55-60%

chapter, synthetic silicates and aluminates phases have been used to identify infrared

<sup>1</sup> Cement chemistry nomenclature is used: � � ���� �� � ���2� �� � ��2�3� �� � ��2�3� �� � �2����̅� ��<sup>3</sup>

�2 = 2.6% (wt). In this

**2. Characterization of cementitious systems by infrared spectroscopy** 

mainly cements, from the point of view of characterization.

have higher stretching frequencies than single bonds (Figure 1).

(wt); �2� = 15-20% (wt); �3� = 5-10% (wt); �4�� = 5-8% (wt) and ��̅

vibrations bands previous to study the more complicated commercial cement.

developed by Bensted (1974).

**2.1 Portland cement** 

tools.

Fig. 1. General rules in the interpretation of building cementing material IR spectra.

As a resume, the vibrations can be divided in stretching and bending: vibrations can involve either a change in bond length (stretching) or bond angle (bending); some bonds can stretch in-phase (symmetrical stretching) or out-of-phase (asymmetric stretching). The main structure elements in building the crystal lattice of silicates are tetrahedral SiO4 groups, with may be either isolated as in the orthosilicates or connected with one another by common O atoms as in building an Si2O7 group from two connected tetraheda [Matossi]. Connection of SiO4 groups so as to form a ring of tetrahedral occurs in the crystal of alite (C3S) or belite (C2S). The infrared spectra of all silicates [Matossi] contain two reflection maxima near 1000 and 500 cm-1, which have been interpreted as a two active frequencies of a tetrahedral point group. In addition to these, there may occur other maxima corresponding to other particular features of the crystal lattice.

The spectrum of the main constituent of OPC, ܥ3ܵ, shows two regions dominated by the internal modes of SiO44- tetrahedral units, with to broad absorption bands centred between 890 and 955 cm-1, solved in to maxima near to 870 and 940 cm-1 due to the symmetric and antisymmetric stretching of Si-O bonds within tetrahedral SiO4 groups, 1 and 3, respectively. Another absorption band of medium intensity appears close to 525 cm-1 and a lower intense band sited near to 450 cm-1 due to the symmetric and antisymmetric bending of the O-Si-O bonds, 2 and 4, respectively (see details of maxims in Table 1 and Figure 2). The other calcium silicate phase spectra, ܥ2ܵ, exhibits strong bands in the area 1000-800 cm-1 with maximums at 990 and 840 cm-1 due the stretching Si-O bond of the silicon tetrahedron and the bending vibration absorption band appear at lower frequencies, 520 cm-1 and a shoulder at 538 cm-1.

The ܥ3ܣ-cubic tricalcium aluminate polymorph spectra (Figure 3), shows a well-defined spectra with two dominant absorption areas with very broad bands. The first ones appear in the area between 950-650 cm-1 and the second ones appearing between 500-380 cm-1, respectively. The main observed maxima appear near to 900, 865, 820, 780, 720 and 705 cm-1 of AlO4-tetrahedral groups, and close to 520, 510, 460 and 414 cm-1 due to AlO6-octahedral groups. The Ca-O bands appear at lower frequencies.

The grey colour of Portland cements is due to the presence of the names ferrite phases; in absence of elements other than calcium, aluminium, iron and oxygen, calcium

Infrared Spectroscopy in the Analysis of Building and Construction Materials 373

2000 1800 1600 1400 1200 1000 800 600 400 W a v e n u m b e rs (cm - 1 )

cm-1, for the gypsum, bassanite, and anhydrite, respectively. They are �3 antisymmetric stretch vibration modes of ��� tetrahedra. The gypsum, bassanite, and anhydrite present two absorption bands (669, 604 cm-1), three **(**660, 629, 600 cm-1) and three (677, 615, 600 cm-1), anti-symmetric bending vibrations, respectively. The peaks at 595 cm-1 in gypsum and 594 cm-1 in bassanite split into two peaks (610 and 591 cm-1) in anhydrite, which indicates a lowering of symmetry in anhydrite. The shift of frequency from 677 cm-1 in anhydrite to 660 cm-1 in bassanite indicates that the sulfate ions in bassanite are linked with water molecules by hydrogen bonding, because in general the hydrogen bonding will lower the frequency of the absorption band. In addition, there is a very weak peak at 1140 cm-1 in gypsum, 1150 cm-1 in bassanite, and 1150 cm-1 in anhydrite, which should be the ν1 symmetric stretch vibration modes of SO4 tetrahedral. The Table 2 present the characteristics absorption bands of sulfate compounds. For gypsum and bassanite presence it will be also possible to analyze 4000-3000 cm-1 region were 1 O-H absorptions can be observed (3553 and 3399 cm-1 for gypsum and

gypsum 1140 1117 669, 604 2500 - 1900 3553, 3399 1686, 1618

The calcium aluminate cement (���) was developed as a solution to the sulphates attack in OPC, and was patented in France in 1908. The ���s are cements consisting predominantly of hydraulic calcium aluminates: mainly monocalcium aluminate, ��, but also contains

According to Tarte, in the interpretation of IR spectra of inorganic aluminates, the characteristic frequency ranges are "condensed" AlO4 tetrahedral in the 900-700 cm-1, "isolated" AlO4 tetrahedral 800-650 cm-1, "condensed" AlO6 octahedral in the 680-500 cm-1,

anhydrite 1150 1120 677, 615, 600 2500 - 1900 --- ---

<sup>1098</sup>660, 629, 600 2500 - 1900 3611, 3557 1618

**OVERTONES OH-STRECHT OH-BEND** 

Fig. 4. IR Spectrum of ordinary Portland cement.

3611 and 3557 cm-1 for bassanite).

Bassanite 1150 1117,

**2.2 Calcium aluminate cement** 

minor amounts of �����, ��� and ����.

**FUNDAMENTALS** 

**ν<sup>1</sup> ν<sup>3</sup> ν<sup>4</sup>**

Table 2. Absorption bands of sulfate compounds (cm-1)

**Sulfates** 


Table 1. Characteristic absorbance bands for cement Portland phases

Fig. 2. Infrared spectra of pure C3S (left) and C2S (right).

aluminoferrite forms a solid solution series of formula ������������)��� for all values of x in the range 0-0.7, compositions with x > 0.7 do not exist at ordinary pressures. The spectrum of ���� presents as significant absorption bands the sited between 800-830 cm-1 with maxima close to 720 cm-1 due to 1 [(Fe,Al)O45-]; moreover, a broad and less intense band with several maximums between 620 and 670 cm-1 is also present (Figure 3).

Fig. 3. Infrared spectra of ��� (left) and ���� (right).

In a real cement the main phases are alite (C3S-base solid solutions i.e. MgO, Na2O) and belite (C2S-base solid solutions i.e.- Al2O3, Fe2O3). The presence of these impurities prompts a change in the crystalline structure of the silicate phases that may cause modifications in the Infrared spectra compared to the pure phases. Figure 4 present the infrared spectra of the ordinary Portland cement. In this spectrum, it is possible to identify different vibrations bands from the calcium silicates, calcium aluminates, and gypsum, the last one added as setting regulator.

The gypsum can lose part of the structure water and the sulfates can be present as bassanite and/or anhydrite. In the IR spectras, the strongest peak is presented at 1102, 1111, and 1094

Fig. 4. IR Spectrum of ordinary Portland cement.

��� 900, 865, 820, 780, 720, 705 520, 510, 460, 414

aluminoferrite forms a solid solution series of formula ������������)��� for all values of x in the range 0-0.7, compositions with x > 0.7 do not exist at ordinary pressures. The spectrum of ���� presents as significant absorption bands the sited between 800-830 cm-1 with maxima close to 720 cm-1 due to 1 [(Fe,Al)O45-]; moreover, a broad and less intense

In a real cement the main phases are alite (C3S-base solid solutions i.e. MgO, Na2O) and belite (C2S-base solid solutions i.e.- Al2O3, Fe2O3). The presence of these impurities prompts a change in the crystalline structure of the silicate phases that may cause modifications in the Infrared spectra compared to the pure phases. Figure 4 present the infrared spectra of the ordinary Portland cement. In this spectrum, it is possible to identify different vibrations bands from the calcium silicates, calcium aluminates, and gypsum, the last one added as

The gypsum can lose part of the structure water and the sulfates can be present as bassanite and/or anhydrite. In the IR spectras, the strongest peak is presented at 1102, 1111, and 1094

band with several maximums between 620 and 670 cm-1 is also present (Figure 3).

**phase Fundamental vibrations (cm-1)** 

Table 1. Characteristic absorbance bands for cement Portland phases

���� 700 - 500

Fig. 2. Infrared spectra of pure C3S (left) and C2S (right).

Fig. 3. Infrared spectra of ��� (left) and ���� (right).

setting regulator.

��� 935 521 ����� 991, 879, 847 509

> cm-1, for the gypsum, bassanite, and anhydrite, respectively. They are �3 antisymmetric stretch vibration modes of ��� tetrahedra. The gypsum, bassanite, and anhydrite present two absorption bands (669, 604 cm-1), three **(**660, 629, 600 cm-1) and three (677, 615, 600 cm-1), anti-symmetric bending vibrations, respectively. The peaks at 595 cm-1 in gypsum and 594 cm-1 in bassanite split into two peaks (610 and 591 cm-1) in anhydrite, which indicates a lowering of symmetry in anhydrite. The shift of frequency from 677 cm-1 in anhydrite to 660 cm-1 in bassanite indicates that the sulfate ions in bassanite are linked with water molecules by hydrogen bonding, because in general the hydrogen bonding will lower the frequency of the absorption band. In addition, there is a very weak peak at 1140 cm-1 in gypsum, 1150 cm-1 in bassanite, and 1150 cm-1 in anhydrite, which should be the ν1 symmetric stretch vibration modes of SO4 tetrahedral. The Table 2 present the characteristics absorption bands of sulfate compounds. For gypsum and bassanite presence it will be also possible to analyze 4000-3000 cm-1 region were 1 O-H absorptions can be observed (3553 and 3399 cm-1 for gypsum and 3611 and 3557 cm-1 for bassanite).


Table 2. Absorption bands of sulfate compounds (cm-1)

#### **2.2 Calcium aluminate cement**

The calcium aluminate cement (���) was developed as a solution to the sulphates attack in OPC, and was patented in France in 1908. The ���s are cements consisting predominantly of hydraulic calcium aluminates: mainly monocalcium aluminate, ��, but also contains minor amounts of �����, ��� and ����.

According to Tarte, in the interpretation of IR spectra of inorganic aluminates, the characteristic frequency ranges are "condensed" AlO4 tetrahedral in the 900-700 cm-1, "isolated" AlO4 tetrahedral 800-650 cm-1, "condensed" AlO6 octahedral in the 680-500 cm-1,

Infrared Spectroscopy in the Analysis of Building and Construction Materials 375

Fig. 7. IR spectra of commercial calcium aluminate cement.

**2.3 Calcium sulfoaluminate cement** 

The Figure 8 displays the IR spectra of yeelimite.

**phase Fundamental vibrations (cm-1)** 

Table 3. Characteristic absorbance bands for calcium aluminate cement phases.

�� 840, 805, 780, 720 680, 640, 570, 540, 450, 420

����� 850, 780, 610, 575, 460, 410 ��� 945, 920, 860, 840, 810, 745, 680, 660, 640, 575, 540, 440, 422 ���� 920, 720, 710, 1020, 973 650, 530, 480, 420

From the sustainability point of view new cement production has been developed in the past decades. One of these new cements is calcium sulphoaluminate (���) that was first developed in China in1980´s. Industrial production requires essentially gypsum, bauxite and limestone as raw materials, which are burnt at 1,300ºC in a conventional rotary kiln. These starting materials lead to a final clinker based on the quinary system CaO–SiO2– Al2O3–Fe2O3–SO3 and formed by three main minerals: tetracalcium trialuminate sulphate or yeelimite (�����̅)*;* dicalcium silicate or belite (���) and calcium sulphate or anhydrite (��̅). Minor phases such as C3A, C4AF, C12A7 and (C2AS) can also be present. The infrared spectra of main mineral phase of calcium sulphoaluminate cement can be described as follow: yeelimite has two absorption bands due to vibrational modes of sulphate [SO4]2– groups at 1110 cm-1, a very intense absorption band due to silicate groups near to 800 cm-1, the third band at 620 cm-1 is due to vibrational modes of [AlO4]5– tetrahedra; ii) belite, anhydrite, C12A7 and C2AS spectras have been described previously. Then, the infrared spectrum of the CSA cement presents the most intense bands located at 1110 and 800 cm–1, in the region where stretching vibrations of [SO4]2– groups lie. A broadened signal appears between 900 and 800 cm–1, centered at 857 cm–1. This feature is strongly asymmetrical: this is probably the result of the convolution of the two bands of C2S, that appear unresolved or as a consequence of lower crystal perfection caused by the presence of foreign ions in the lattice or because of the small particle size of minerals of CSA clinker. But it is also possible to highlight the presence of the three anhydrite bands at 677, 615, and 600 cm-1, respectively.

"isolated" AlO6 octahedral 530-400 cm-1. In the spectra of the main present phase of CAC, the CA, the most relevant signals are presented in the two region 850-750 cm-1 and 750-500 cm-1 due to the mentioned groups, with maxima near to 840, 805, 780 and 720 cm-1 of AlO4 tetrahedral and close to 680, 640, 570, 540, 450 and 420 cm-1 of AlO6 groups (Figure 5).

The infrared spectrum of C12A7 contains absorption bands mainly in two regions: a very broad absorption region between 680-900 cm-1 of tetrahedral groups and another area in the 650-400 cm-1 range with very sharp and intense bands due to octahedral groups. The maxima appear close to 850, 780, 610, 575, 460 and 410 cm-1 (Figure 5).

Fig. 5. Infrared spectra of ܣܥ) left) and ܥଵଶܣ) right).

The minor phases present in the CAC are the CA2 aluminate with the two absorption in the areas 950-700 cm-1 and 690-410 cm-1 with maxima at 945, 920, 860, 840, 810 and 745 cm-1 of AlO4 groups and near to 680, 660, 640, 575, 540, 440 and 422 cm-1 of AlO6 groups. The C2AS mineral presents the AlO4 vibration area between 920 and 720 cm-1 while the AlO6 groups give absorption bands between 720-400 cm-1. At higher frequencies, in this spectra appear the signals due to the Si-O vibrations, 1020 and 973 cm-1. The IR spectra of CA2 and C2AS are presented in Figure 6.

Fig. 6. Infrared spectra of ܣܥଶ (left) and ܥଶܣܵ) right).

Then, the most relevant signals on the FTIR spectrum for CAC are the absorption bands in the region between 850 and 650 cm-1 – the bands at around 840, 805 and 780 cm-1 – attributed to AlO4 groups; the bands between 750 and 400 cm-1 – with bands at about 720, 685, 640 and 570 cm-1, ascribed to AlO6 groups; and the bands at under 400 cm-1 owing to Ca-O bonds [15, 16]. The Figure 7 shows the IR spectra of the commercial cement. The Table 3 presents the characteristics absorption bands of CAC mineral compounds.

Fig. 7. IR spectra of commercial calcium aluminate cement.


Table 3. Characteristic absorbance bands for calcium aluminate cement phases.

#### **2.3 Calcium sulfoaluminate cement**

374 Infrared Spectroscopy – Materials Science, Engineering and Technology

"isolated" AlO6 octahedral 530-400 cm-1. In the spectra of the main present phase of CAC, the CA, the most relevant signals are presented in the two region 850-750 cm-1 and 750-500 cm-1 due to the mentioned groups, with maxima near to 840, 805, 780 and 720 cm-1 of AlO4 tetrahedral and close to 680, 640, 570, 540, 450 and 420 cm-1 of AlO6 groups

The infrared spectrum of C12A7 contains absorption bands mainly in two regions: a very broad absorption region between 680-900 cm-1 of tetrahedral groups and another area in the 650-400 cm-1 range with very sharp and intense bands due to octahedral groups. The

The minor phases present in the CAC are the CA2 aluminate with the two absorption in the areas 950-700 cm-1 and 690-410 cm-1 with maxima at 945, 920, 860, 840, 810 and 745 cm-1 of AlO4 groups and near to 680, 660, 640, 575, 540, 440 and 422 cm-1 of AlO6 groups. The C2AS mineral presents the AlO4 vibration area between 920 and 720 cm-1 while the AlO6 groups give absorption bands between 720-400 cm-1. At higher frequencies, in this spectra appear the signals due to the Si-O vibrations, 1020 and 973 cm-1. The IR spectra of CA2 and C2AS are

Then, the most relevant signals on the FTIR spectrum for CAC are the absorption bands in the region between 850 and 650 cm-1 – the bands at around 840, 805 and 780 cm-1 – attributed to AlO4 groups; the bands between 750 and 400 cm-1 – with bands at about 720, 685, 640 and 570 cm-1, ascribed to AlO6 groups; and the bands at under 400 cm-1 owing to Ca-O bonds [15, 16]. The Figure 7 shows the IR spectra of the commercial cement. The Table 3 presents

maxima appear close to 850, 780, 610, 575, 460 and 410 cm-1 (Figure 5).

Fig. 5. Infrared spectra of ܣܥ) left) and ܥଵଶܣ) right).

Fig. 6. Infrared spectra of ܣܥଶ (left) and ܥଶܣܵ) right).

the characteristics absorption bands of CAC mineral compounds.

(Figure 5).

presented in Figure 6.

From the sustainability point of view new cement production has been developed in the past decades. One of these new cements is calcium sulphoaluminate (���) that was first developed in China in1980´s. Industrial production requires essentially gypsum, bauxite and limestone as raw materials, which are burnt at 1,300ºC in a conventional rotary kiln. These starting materials lead to a final clinker based on the quinary system CaO–SiO2– Al2O3–Fe2O3–SO3 and formed by three main minerals: tetracalcium trialuminate sulphate or yeelimite (�����̅)*;* dicalcium silicate or belite (���) and calcium sulphate or anhydrite (��̅). Minor phases such as C3A, C4AF, C12A7 and (C2AS) can also be present. The infrared spectra of main mineral phase of calcium sulphoaluminate cement can be described as follow: yeelimite has two absorption bands due to vibrational modes of sulphate [SO4]2– groups at 1110 cm-1, a very intense absorption band due to silicate groups near to 800 cm-1, the third band at 620 cm-1 is due to vibrational modes of [AlO4]5– tetrahedra; ii) belite, anhydrite, C12A7 and C2AS spectras have been described previously. Then, the infrared spectrum of the CSA cement presents the most intense bands located at 1110 and 800 cm–1, in the region where stretching vibrations of [SO4]2– groups lie. A broadened signal appears between 900 and 800 cm–1, centered at 857 cm–1. This feature is strongly asymmetrical: this is probably the result of the convolution of the two bands of C2S, that appear unresolved or as a consequence of lower crystal perfection caused by the presence of foreign ions in the lattice or because of the small particle size of minerals of CSA clinker. But it is also possible to highlight the presence of the three anhydrite bands at 677, 615, and 600 cm-1, respectively. The Figure 8 displays the IR spectra of yeelimite.

Infrared Spectroscopy in the Analysis of Building and Construction Materials 377

Ca6[Al(OH)6]2(SO4)326H2O presents a very strong anti-symmetrical stretching frequency of the sulphate ion (3 SO4) centred towards 1120 cm-1; this band is indicative of relative isolation of this ion in the hexagonal prism structure. The water absorption bands appear in the region 1600-1700 cm-1 (1640 and 1675 cm-1 2 H2O) and above 3000 cm-1 (3420 due to <sup>1</sup> H2O and 3635 cm-1 from OHfree). The presence of aluminate bands are near to 550 cm-1 ( AlO6) due to stretching Al-O groups, and 855 cm-1 (Al-O-H bending). The Figure 10 shows

Fig. 10. Structure model of ettringite (according to Dr. J. Neubauer/University Erlangen/

the second one at 353 cm-1assigned to Ca-O lattice vibrations (Figure 11).

Fig. 11. Infrared spectra of gel C-S-H (left) and portlandite (right)

The other crystalline phase present in cement hydration, *portlandite, Ca(OH)2*, shows two prominent sharp peaks, the first one at 3645 cm-1due to the presence of OH stretching and

While crystalline materials give sharp well-defined bands and the glasses give broad, poorly defined bands, the C-S-H samples lie between these two extremes. The distribution function, which describes the line shape of the bands, is strongly dependent upon the

the structure and the infrared spectra of ettringite compound.

Germany) (left), infrared spectra of ettringite (right).

Fig. 8. Infrared spectra of yeelimite.

#### **3. Infrared analysis of hydrated cementitious materials**

The hydration of Portland cement give rise both, amorphous phase calcium silicate hydrated (C-S-H gel) and two crystalline phases, ettringite (Ca6Al2(OH)12·(SO4)3·26H2O) and portlandite. The C-S-H gel is the primary binding phase in Portland cement but poorly crystalline. The Figure 9 presents the infrared spectra of a hydrated commercial Portland cement.

Fig. 9. IR spectra of hydrated Portland cement.

The ettringite (C6A�̅ H32) is the first developed phase due to the reaction of aluminates with sulfates of Portland cement. According to the structure model by Taylor [76], the crystals are based on columns of cations of the composition {Ca3[Al(OH)6]·12 H2O}3+. In there, the Al(OH)63--octahedral are bound up with the edgesharing CaO8-polyhedra, that means each aluminum-ion, bound into the crystal, is connected to Ca2+-ions, with which they share OHions. The intervening channels contain the SO4 2--tetrahedral and the remaining H2O molecules (fig. 2). The H2O molecules are partly bound very loose into the ettringite structure. According to Bensted, the infrared spectra of ettringite C3A3C�̅H32 or

2000 1500 1000 500 W a ve n u m b e rs (cm - 1 )

The hydration of Portland cement give rise both, amorphous phase calcium silicate hydrated (C-S-H gel) and two crystalline phases, ettringite (Ca6Al2(OH)12·(SO4)3·26H2O) and portlandite. The C-S-H gel is the primary binding phase in Portland cement but poorly crystalline. The Figure 9 presents the infrared spectra of a hydrated commercial Portland

H32) is the first developed phase due to the reaction of aluminates with

sulfates of Portland cement. According to the structure model by Taylor [76], the crystals are based on columns of cations of the composition {Ca3[Al(OH)6]·12 H2O}3+. In there, the Al(OH)63--octahedral are bound up with the edgesharing CaO8-polyhedra, that means each aluminum-ion, bound into the crystal, is connected to Ca2+-ions, with which they share OHions. The intervening channels contain the SO42--tetrahedral and the remaining H2O molecules (fig. 2). The H2O molecules are partly bound very loose into the ettringite structure. According to Bensted, the infrared spectra of ettringite C3A3C�̅H32 or

**3. Infrared analysis of hydrated cementitious materials** 

Fig. 8. Infrared spectra of yeelimite.

Fig. 9. IR spectra of hydrated Portland cement.

The ettringite (C6A�̅

cement.

Ca6[Al(OH)6]2(SO4)326H2O presents a very strong anti-symmetrical stretching frequency of the sulphate ion (3 SO4) centred towards 1120 cm-1; this band is indicative of relative isolation of this ion in the hexagonal prism structure. The water absorption bands appear in the region 1600-1700 cm-1 (1640 and 1675 cm-1 2 H2O) and above 3000 cm-1 (3420 due to <sup>1</sup> H2O and 3635 cm-1 from OHfree). The presence of aluminate bands are near to 550 cm-1 ( AlO6) due to stretching Al-O groups, and 855 cm-1 (Al-O-H bending). The Figure 10 shows the structure and the infrared spectra of ettringite compound.

Fig. 10. Structure model of ettringite (according to Dr. J. Neubauer/University Erlangen/ Germany) (left), infrared spectra of ettringite (right).

The other crystalline phase present in cement hydration, *portlandite, Ca(OH)2*, shows two prominent sharp peaks, the first one at 3645 cm-1due to the presence of OH stretching and the second one at 353 cm-1assigned to Ca-O lattice vibrations (Figure 11).

Fig. 11. Infrared spectra of gel C-S-H (left) and portlandite (right)

While crystalline materials give sharp well-defined bands and the glasses give broad, poorly defined bands, the C-S-H samples lie between these two extremes. The distribution function, which describes the line shape of the bands, is strongly dependent upon the

Infrared Spectroscopy in the Analysis of Building and Construction Materials 379

area of 3000-3700 cm-1 of the various polymorphs of the aluminium hydroxide (Fig. 13), the spectroscopy of the calcium aluminate cement does not specify clearly the difference between the three forms of the Al(0H)3. However and as guidance that can help in the interpretation, the bayerite has a band in 3550 cm-1, a shoulder in 3430 cm-1 and one other shoulder in 3660 cm-1; these are not observed in the spectrum of gibbsite. Table 4 shows the bands of greater interest in the three polymorphisms, according to some authors

4000 3500 3000 2500 2000 1500 1000 500

W a v e n u m b e rs (c m - 1 )

Absorption bands (cm-1) references

1025, 969 Fernández

1024, 975 Frederickson

1060, 1030, 823, 770, 461 Van der Marel

Fig. 12. Infrared spectra of main CAC hydrates, CAH10 and C3AH6.

(Fernández, Frederickson, Van der Marel).

Fig. 13. Aluminium hydroxide, gibbsite, infrared spectra.

Gibbsite 3620, 3524, 3468, 3395

Bayerite 3360, 3620, 3540, 3420, 3401, 3454, 3533

Nordstrandite 3660, 3558, 3521, 3490, 3455, 3380, 3360

Table 4. Characteristic frequencies of aluminium hydroxides.

Transmittance (%)

distribution of bond angles and bond lengths within common environments, and the broad half-width of the absorption bands of the C-S-H samples reflect their low symmetry and crystallinity. This distribution is assumed to be symmetry for glasses; hence, any asymmetry of the shape of the bands is due to a superimposition of several symmetrically shaped bands. The infrared spectra of synthetic *C-S-H gel* samples show a broad band in the 3800-3000 cm-1 region attributed to OH stretching vibrations of water molecules with maxima close to 3420 and 3626 cm-1, 1428 and 666 cm-1(Figure 11). According to Martinez-Ramirez, depending on the C/S ratio of the C-S-H gel the frequency of the maximum can be different.

With respect to the *CACs*, the normal CAC hydration with water gives up to the development of hydrated calcium aluminates, CAH10 at low temperatures but C2AH8 and C3AH6 at intermediate and high temperatures joint to AH3 according to the following reactions:

CA + 10H CAH10 (1)

2CA + 11H C2AH8 + AH3 (2)

$$\text{\textbulletCA} \dagger \text{12} \,\text{H} \xrightarrow{\text{\textbullet}} \text{\textbullet}\_{3} \text{AH}\_{6} \dagger \text{2} \,\text{AH}\_{3} \tag{3}$$

High early strength, good chemical resistance and high temperature resistance of calcium aluminate cement (CAC) products had encouraged the use of CAC in certain applications. However, conversion of hexagonal phases, CAH10 and/or C2AH8 to cubic C3AH6 and AH3 in hydrated CAC cement under certain temperature conditions has been the major consequence in limiting its use to special applications. The presence of a minor amount of C2ASH8 (strätlingite) in CAC at later ages may be responsible of some strength recovery after conversion process.

The IR spectra of CAH10 have a very broad and intense band due to hydroxyl vibration in the 3400-3550 cm-1 region, with maxima near to 3500 cm-1. A very weak band at 1650 cm-1 is associated to the H-O-H deformation vibration. The 1200-400 cm-1 region is a very poor resolution area due to the complexity and associated vibrations sometimes indicating a low crystalline grade; but some absorption bands at 1,024, 774 (shoulder) 699 and a doublet close to 573-528 cm-1. The IR spectra of the CAH10 phase are depicted in Figure 12. *The β-C2AH8 hexagonal phase presents in the* 3400-3700 cm-1 region absorption bands at 3,465 and 3,625 cm-1 due to OH vibrations of the molecular water. In the 1100-400 cm-1 region there is a very complex vibration area with difficulties in the interpretation.

The *C3AH6* and the gibbsite are the stable phases in this system. The C3AH6 structure can be described as [Al(OH)6]2- octahedrals connected by Ca2+ cations. The IR spectra presents a very intense OH-free band at 3670 cm-1. This compound do not presents water molecular in the structure so, in the area between 3,400 and 3,600 cm-1, there is not the presence of the deformation H-O-H band. Others fundamental bands due to the stretching and bending vibrations of the Al-O in the octahedral AlO6 groups, appear at 802, 525 and 412 cm-1 (Figure 12).

Different AH3 polymorphs can be identified by FTIR (Table 4). Although the differences in the strength of OH bond are reflected mainly by the numerous absorption maxima in the

distribution of bond angles and bond lengths within common environments, and the broad half-width of the absorption bands of the C-S-H samples reflect their low symmetry and crystallinity. This distribution is assumed to be symmetry for glasses; hence, any asymmetry of the shape of the bands is due to a superimposition of several symmetrically shaped bands. The infrared spectra of synthetic *C-S-H gel* samples show a broad band in the 3800-3000 cm-1 region attributed to OH stretching vibrations of water molecules with maxima close to 3420 and 3626 cm-1, 1428 and 666 cm-1(Figure 11). According to Martinez-Ramirez, depending on the C/S ratio of the C-S-H gel the frequency of the maximum can

With respect to the *CACs*, the normal CAC hydration with water gives up to the development of hydrated calcium aluminates, CAH10 at low temperatures but C2AH8 and C3AH6 at intermediate and high temperatures joint to AH3 according to the following

High early strength, good chemical resistance and high temperature resistance of calcium aluminate cement (CAC) products had encouraged the use of CAC in certain applications. However, conversion of hexagonal phases, CAH10 and/or C2AH8 to cubic C3AH6 and AH3 in hydrated CAC cement under certain temperature conditions has been the major consequence in limiting its use to special applications. The presence of a minor amount of C2ASH8 (strätlingite) in CAC at later ages may be responsible of some strength recovery

The IR spectra of CAH10 have a very broad and intense band due to hydroxyl vibration in the 3400-3550 cm-1 region, with maxima near to 3500 cm-1. A very weak band at 1650 cm-1 is associated to the H-O-H deformation vibration. The 1200-400 cm-1 region is a very poor resolution area due to the complexity and associated vibrations sometimes indicating a low crystalline grade; but some absorption bands at 1,024, 774 (shoulder) 699 and a doublet close to 573-528 cm-1. The IR spectra of the CAH10 phase are depicted in Figure 12. *The β-C2AH8 hexagonal phase presents in the* 3400-3700 cm-1 region absorption bands at 3,465 and 3,625 cm-1 due to OH vibrations of the molecular water. In the 1100-400 cm-1 region there is a very

The *C3AH6* and the gibbsite are the stable phases in this system. The C3AH6 structure can be described as [Al(OH)6]2- octahedrals connected by Ca2+ cations. The IR spectra presents a very intense OH-free band at 3670 cm-1. This compound do not presents water molecular in the structure so, in the area between 3,400 and 3,600 cm-1, there is not the presence of the deformation H-O-H band. Others fundamental bands due to the stretching and bending vibrations of the Al-O in the octahedral AlO6 groups, appear at 802, 525 and 412 cm-1 (Figure

Different AH3 polymorphs can be identified by FTIR (Table 4). Although the differences in the strength of OH bond are reflected mainly by the numerous absorption maxima in the

complex vibration area with difficulties in the interpretation.

CA + 10H CAH10 (1)

2CA + 11H C2AH8 + AH3 (2)

3CA +12 H C3AH6 + 2AH3 (3)

be different.

reactions:

after conversion process.

12).

Fig. 12. Infrared spectra of main CAC hydrates, CAH10 and C3AH6.

area of 3000-3700 cm-1 of the various polymorphs of the aluminium hydroxide (Fig. 13), the spectroscopy of the calcium aluminate cement does not specify clearly the difference between the three forms of the Al(0H)3. However and as guidance that can help in the interpretation, the bayerite has a band in 3550 cm-1, a shoulder in 3430 cm-1 and one other shoulder in 3660 cm-1; these are not observed in the spectrum of gibbsite. Table 4 shows the bands of greater interest in the three polymorphisms, according to some authors (Fernández, Frederickson, Van der Marel).

Fig. 13. Aluminium hydroxide, gibbsite, infrared spectra.


Table 4. Characteristic frequencies of aluminium hydroxides.

Infrared Spectroscopy in the Analysis of Building and Construction Materials 381

Special thanks to Tomás Vázquez to teach and pass on to us his enthusiasm for the infrared spectroscopy techniques. The authors also thank the support by the MICINN (Ministerio de

A. S. Povaennykh "The use of infrared spectra for the determination of minerals" American

Afremow L. "High resolution spectra of inorganic pigment and extenders in the mid-

Bensted J, Varna SP. Some applications of IR and Raman. Spectroscopy in cement chemistry,

Farmer, V. C. (Ed.) The Infrared Spectra of Minerals. Mineralogical Society, London, 1974. Fernandez, L., Vazquez, T. *"Aplicación de la espectroscopia infrarroja al estudio de cemento* 

Ghosh, S. N. *"Infrared and Raman spectral studies in cement and concrete"* Cement and Concrete

Ghosh, S. N., Chatterjee, A. K. *"Absorption and reflection infrared spectra of major cement minerals, ckinkers and cements"* Journal of Materials Science 9(10):1577-1584, 1974. Hughes, T.L., Methven, C.M., Jones, T.G.J., Pelham, S.E., Fletcher, P., Hall, C. *"Determining* 

Lawson, K. E. "Infrared Absorption of Inorganic Substances", Reinhold Publishing Corp.,

Matossi, F. "Vibration Frequencies and Binding Forces in Some Silicate Groups" The Journal

Nakamoto, K. Infrared Spectra of Inorganic and Coordination Compounds. Wiley, New

Tarte, P. *"Infrared spectra of inorganic aluminates and characteristic vibrational frequencies of AlO4*

Tarte, P. Etude infra-rouge des orthosilicates et des ortho-germanates. Une nouvelle method

d'interpretation des spectres. Spectrochim. Acta, I 8, 467 -483, 1962. Taylor, H. F. W.: Cement Chemistry. Reedwood Books, Trowbridge, 2nd Edition, 1997. Van der Marel and Beutelspacher, H.: Atlas of Infrared Spectroscopy of Clay Minerals and

*tetrahedra and AlO6 octahedra"* Spectrochimica Acta: Part A Molecular Spectroscopy

Nyquist, R. P. and R. O. Kagel "Infrared spectra of inorganic compounds (3800 - 45 cm-1).

*cement composition by Fourier Transform infrared spectroscopy"* Advanced Cement

*aluminoso"* Materiales de Construcción 46(241):53-65, 1996.

Frederickson, L.D.: Analytcal Chemistry, vol. 26, p. 1883, die., 1954.

infrared region from 1500 cm-1 to 200 cm-1" Journal of paint technology vol. 28

Part III: Hydration of Portland cement and its constituents. *Cement Technology*,

**5. Acknowledgments** 

Academic Press, New York., 1971.

issue 34 (1966).

(5):440-450, 1974.

New York, 1961.

23A(7):2127-2143, 1967.

York., 1970.

Research 10(6):771-782 (1980).

Based Materials 2(3):91-104, 1995.

of Chemical Physics Vol. 17, n 8, 1949.

Schaefer, C., Matossi, F., Wirtz, K., Zeits. F. Physik 89, 210, 1934.

Their Admixtures. Ed. ELSEVIER. New York. 1976.

**6. References** 

Ciencia e Innovación) with the BIA00767- 2008 project.

Mineralogist, Volume 63, pages 956-959, 1978.

Basila M. R. "Infrared Spectra of Adsorbed Molecules", 1968.

#### **4. Carbonated compounds**

The infrared spectroscopy is very sensitive to detect the presence of carbonates. The bands more features and more valid for its identification are those indicated in Table 5. The *Calcium carbonate* phases formed after portlandite carbonation, calcite, aragonite or/and vaterite; although vaterite is the least thermodynamically stable of the three crystalline calcium carbonate polymorphs. Indeed, vaterite has been observed following exposure of C-S-H gels to carbon dioxide (accelerated carbonation). The formation of vaterite may occur upon carbonation of pastes with high lime contents, and is favoured by the presence of imperfectly crystalline portlandite. The observed absorptions bands for calcium carbonate phases are due to the planar CO3 -2 ion. There are four vibrational modes in the free CO3 - 2 ion: i) the symmetric stretching, 1[CO3]; ii) the out-of-plane bend, ν2 [CO3]; iii) the asymmetric stretch, *v*3[CO3]; and iv) the split in-plane bending vibrations 4[CO3]; and Ca-O lattice vibrations. Depending on the calcium carbonate polymorph the vibration of the bands appears at different wavenumber. Figure 14 shows the spectra of calcite, vaterite and aragonite.

Fig. 14. Infrared spectra of calcite, vaterite and aragonite.


Table 5. Calcium carbonate polymorphs infrared bands

## **5. Acknowledgments**

Special thanks to Tomás Vázquez to teach and pass on to us his enthusiasm for the infrared spectroscopy techniques. The authors also thank the support by the MICINN (Ministerio de Ciencia e Innovación) with the BIA00767- 2008 project.

## **6. References**

380 Infrared Spectroscopy – Materials Science, Engineering and Technology

The infrared spectroscopy is very sensitive to detect the presence of carbonates. The bands more features and more valid for its identification are those indicated in Table 5. The *Calcium carbonate* phases formed after portlandite carbonation, calcite, aragonite or/and vaterite; although vaterite is the least thermodynamically stable of the three crystalline calcium carbonate polymorphs. Indeed, vaterite has been observed following exposure of C-S-H gels to carbon dioxide (accelerated carbonation). The formation of vaterite may occur upon carbonation of pastes with high lime contents, and is favoured by the presence of imperfectly crystalline portlandite. The observed absorptions bands for calcium carbonate

2 ion: i) the symmetric stretching, 1[CO3]; ii) the out-of-plane bend, ν2 [CO3]; iii) the asymmetric stretch, *v*3[CO3]; and iv) the split in-plane bending vibrations 4[CO3]; and Ca-O lattice vibrations. Depending on the calcium carbonate polymorph the vibration of the bands appears at different wavenumber. Figure 14 shows the spectra of calcite, vaterite and

calcite vaterite aragonite

<sup>3</sup> 1420 1482 1492-1404

<sup>4</sup> 875, 848 856 877

2 712 713, 700 744



**4. Carbonated compounds** 

phases are due to the planar CO3

Fig. 14. Infrared spectra of calcite, vaterite and aragonite.

Table 5. Calcium carbonate polymorphs infrared bands

<sup>1</sup> 1063 1085

aragonite.


**20** 

*Brazil* 

Daniel A. Macedo et al.\*

**Infrared Spectroscopy Techniques in the** 

*Material Science and Engineering Post Graduate Program, UFRN* 

**Characterization of SOFC Functional Ceramics** 

Infrared spectroscopy is certainly one of the most important analytical techniques available nowadays for scientists. One of the greatest advantages of infrared spectroscopy is that virtually any sample in any physical state can be analyzed. The technique is based on the vibrations of atoms of a molecule. An infrared spectrum is obtained by passing infrared radiation through a sample and determining what fraction of the incident radiation is absorbed at a particular energy. The energy at which any peak in an absorption spectrum appears corresponds to the frequency of a vibration of a part of a sample molecule (Stuart,

Fourier transform infrared spectroscopy (FTIR) has been widely used for *in situ* analysis of adsorbed species and surface reactions. Infrared spectroscopy techniques have being used for the characterization of solid oxide fuel cells (SOFCs). FTIR is utilized to identify the structure of the SOFC electrode and electrolyte surface (Resini et al., 2009; Guo et al., 2010). Liu and co-workers (Liu et al., 2002) were pioneers in the *in situ* surface characterization by

The development of high-performance electrode and electrolyte materials for SOFC is an important step towards reducing the fuel cell operation temperature to the low and intermediate range (500 – 700 ºC). As the operating temperature is reduced, many cell parts, such as the auxiliary components can be easily and cost-efficiently produced. To meet long operational lifetime, material compatibility and thermomechanical resistance would be less critical as the range of possibilities for lower temperature increases. To that end, recent research at UFRN, Natal, Brazil has successfully focused on novel synthesis processes based on microwave-assisted combustion and modified polymeric precursor methods in order to synthesize high performance cobaltite-based composite cathodes for low-intermediary-

Moisés R. Cesário2, Graziele L. Souza3, Beatriz Cela4, Carlos A. Paskocimas1, Antonio E. Martinelli1,

**1. Introduction** 

FTIR under SOFC operating conditions.

Dulce M. A. Melo1,2 and Rubens M. Nascimento1

*2Chemistry Pos Graduate Program, UFRN, Brazil 3Chemical Engineering Department, UFRN, Brazil* 

*1Material Science and Engineering Post Graduate Program, UFRN, Brazil* 

*4Forschungszentrum Jülich GmbH, Central Department of Technology (ZAT), Germany* 

temperature SOFCs.

 \*

2004).

Vázquez, T. "Espectroscopía Infraroja de algunos compuestos de interés en la química del cemento". Cuadernos de Investigación del Instituto Eduardo torroja de la Construcción y del Cemento, 1969.

## **Infrared Spectroscopy Techniques in the Characterization of SOFC Functional Ceramics**

Daniel A. Macedo et al.\* *Material Science and Engineering Post Graduate Program, UFRN Brazil* 

## **1. Introduction**

382 Infrared Spectroscopy – Materials Science, Engineering and Technology

Vázquez, T. "Espectroscopía Infraroja de algunos compuestos de interés en la química del

Construcción y del Cemento, 1969.

cemento". Cuadernos de Investigación del Instituto Eduardo torroja de la

Infrared spectroscopy is certainly one of the most important analytical techniques available nowadays for scientists. One of the greatest advantages of infrared spectroscopy is that virtually any sample in any physical state can be analyzed. The technique is based on the vibrations of atoms of a molecule. An infrared spectrum is obtained by passing infrared radiation through a sample and determining what fraction of the incident radiation is absorbed at a particular energy. The energy at which any peak in an absorption spectrum appears corresponds to the frequency of a vibration of a part of a sample molecule (Stuart, 2004).

Fourier transform infrared spectroscopy (FTIR) has been widely used for *in situ* analysis of adsorbed species and surface reactions. Infrared spectroscopy techniques have being used for the characterization of solid oxide fuel cells (SOFCs). FTIR is utilized to identify the structure of the SOFC electrode and electrolyte surface (Resini et al., 2009; Guo et al., 2010). Liu and co-workers (Liu et al., 2002) were pioneers in the *in situ* surface characterization by FTIR under SOFC operating conditions.

The development of high-performance electrode and electrolyte materials for SOFC is an important step towards reducing the fuel cell operation temperature to the low and intermediate range (500 – 700 ºC). As the operating temperature is reduced, many cell parts, such as the auxiliary components can be easily and cost-efficiently produced. To meet long operational lifetime, material compatibility and thermomechanical resistance would be less critical as the range of possibilities for lower temperature increases. To that end, recent research at UFRN, Natal, Brazil has successfully focused on novel synthesis processes based on microwave-assisted combustion and modified polymeric precursor methods in order to synthesize high performance cobaltite-based composite cathodes for low-intermediarytemperature SOFCs.

<sup>\*</sup> Moisés R. Cesário2, Graziele L. Souza3, Beatriz Cela4, Carlos A. Paskocimas1, Antonio E. Martinelli1, Dulce M. A. Melo1,2 and Rubens M. Nascimento1

*<sup>1</sup>Material Science and Engineering Post Graduate Program, UFRN, Brazil* 

*<sup>2</sup>Chemistry Pos Graduate Program, UFRN, Brazil 3Chemical Engineering Department, UFRN, Brazil* 

*<sup>4</sup>Forschungszentrum Jülich GmbH, Central Department of Technology (ZAT), Germany* 

Infrared Spectroscopy Techniques in the Characterization of SOFC Functional Ceramics 385

Fig. 1. A typical SOFC single cell with its basic components (anode, cathode and electrolyte)

ohmic losses and electrode polarization losses associated with thermally activated processes of both ionic transport and electrode reactions. In special, the electrochemical activity of the cathode dramatically deteriorates with decreasing temperature for typical strontium-doped lanthanum manganite (LSM) based electrodes (Singhal 2000; Steele, 2000; Steele & Heinzel,

Thus, to achieve acceptable performance of IT-SOFCs, reducing the electrolyte resistance and electrode polarization losses are two key points. Losses attributed to electrolytes can be minimized by decreasing the thickness of electrolyte layers or substituting traditional YSZ by other electrolytes, such as doped ceria and apatite-like materials. The overall electrochemical limitation would then be ruled by electrode polarization losses. Due to the higher activation energy and lower reaction kinetics for oxygen reduction at the cathode compared with those of the fuel oxidation at the anode, the polarization loss from the air electrode (cathode) that limits the overall cell performance (Ivers-Tiffee et al., 2001; Tsai & Barnett, 1997). Therefore, the development of new cathode materials with high electrocatalytic activity for oxygen reduction becomes a critical issue for the development of

The development of high performance cathodes is based on reducing its thickness from hundreds to few micrometers associated with the addition of an electrolyte material in the formation of composite cathodes. These are interesting approaches that can improve electrode performance for the reduction of oxygen. The latter allows the extension of the triple phase boundaries (TPB) from the electrolyte/cathode interface deep into the bulk of the electrode, permitting electrochemical reactions to take place within the electrode. The

and the mechanism of energy production.

**2.1 Development of cathode materials** 

2001).

IT-SOFCs.

This book chapter reviews how powerful infrared spectroscopy techniques play a fundamental role in the novel synthesis approaches developed during the last 5 years (2007 – 2011) by this research group. Synthesized high performance cathode powders had their features investigated using different materials characterization techniques, such as Fourier Transform Infrared Spectroscopy (FTIR), FAR and MID-Infrared spectroscopy.

## **2. Solid oxide fuel cells and its components**

Fuel cells are highly efficient power generation devices which convert chemical energy of gaseous fuels (hydrogen, fossil fuels and ethanol among others) directly into electric power in a silent and environmentally friendly way. These electrochemical devices are promising alternatives to traditional mobile and stationary power sources, such as internal combustion engines and coal burning power plants. Among the various types of fuel cells, solid oxide fuel cells (SOFCs) have advantages linked to high energy conversion efficiency and excellent fuel flexibility because of their high operating temperature compared to other types of fuel cells. Fig. 1 depicts a typical SOFC single cell produced by our research group. It consists of three basic components: two porous electrodes (anode and cathode) and a solid electrolyte. Each one of these components must fulfil specific performance requirements such as microstructural stability during preparation and operation; chemical and physical compatibility, i.e., similar thermal expansion coefficients; adequate porosity and catalytic activity to achieve the highest performance (Minh & Takahashi, 1995; Singhal, 2000).

The SOFC electrolyte material, typically yttria stabilized zirconia (YSZ) or rare earth doped ceria, must be an electronic insulating but ion-conducting ceramic that allows only protons or oxygen ions to pass through. Furthermore, the electrolyte material must be dense to separate the air and fuel, chemically and structurally stable over a wide range of partial pressures of oxygen and temperatures. The cathode (air electrode) material, typically lanthanum manganites and cobaltites, has to be electrocatalyst for oxygen reduction into oxide ions. When an oxygen ionic conducting oxide is adopted as electrolyte, these ions diffuse through the material to the anode (fuel electrode), driven by the differences in oxygen chemical potential between fuel and air constituents of the cell, where they electrochemically oxidize fuels such as hydrogen, methane, and hydrocarbons. The released electrons flow through an external circuit to the cathode to complete the circuit. The complete mechanism of electric power production in a SOFC is also shown in Fig. 1.

SOFCs have attracted significant attention in the last 20-30 years due to their high efficiency, fuel flexibility and environmental advantages (Minh, 2004; Molenda et al., 2007; McIntosh & Gorte, 2004). However, typical SOFCs operate at 1000 ºC. Elevated operating temperatures introduce a series of difficulties such as sintering of the electrodes and high reactivity between cell components. For these reasons, there is a considerable research interest in reducing the operating temperature of these devices down to the range between 500 and 800 °C, which characterizes intermediate temperature solid oxide fuel cells (IT-SOFCs) or even lower, which would imply the use of inexpensive metallic materials, rapid start-up and shut-down, minimization of thermal degradation and reactions between cell components, and longer operational lifetime. Furthermore, as the operation temperature is reduced, system reliability increases, increasing the possibility of using SOFCs for a wide variety of applications, including residential and automotive devices. On the other hand, reduced operating temperatures reduces the overall electrochemical performance due to increased

This book chapter reviews how powerful infrared spectroscopy techniques play a fundamental role in the novel synthesis approaches developed during the last 5 years (2007 – 2011) by this research group. Synthesized high performance cathode powders had their features investigated using different materials characterization techniques, such as Fourier

Fuel cells are highly efficient power generation devices which convert chemical energy of gaseous fuels (hydrogen, fossil fuels and ethanol among others) directly into electric power in a silent and environmentally friendly way. These electrochemical devices are promising alternatives to traditional mobile and stationary power sources, such as internal combustion engines and coal burning power plants. Among the various types of fuel cells, solid oxide fuel cells (SOFCs) have advantages linked to high energy conversion efficiency and excellent fuel flexibility because of their high operating temperature compared to other types of fuel cells. Fig. 1 depicts a typical SOFC single cell produced by our research group. It consists of three basic components: two porous electrodes (anode and cathode) and a solid electrolyte. Each one of these components must fulfil specific performance requirements such as microstructural stability during preparation and operation; chemical and physical compatibility, i.e., similar thermal expansion coefficients; adequate porosity and catalytic

activity to achieve the highest performance (Minh & Takahashi, 1995; Singhal, 2000).

complete mechanism of electric power production in a SOFC is also shown in Fig. 1.

SOFCs have attracted significant attention in the last 20-30 years due to their high efficiency, fuel flexibility and environmental advantages (Minh, 2004; Molenda et al., 2007; McIntosh & Gorte, 2004). However, typical SOFCs operate at 1000 ºC. Elevated operating temperatures introduce a series of difficulties such as sintering of the electrodes and high reactivity between cell components. For these reasons, there is a considerable research interest in reducing the operating temperature of these devices down to the range between 500 and 800 °C, which characterizes intermediate temperature solid oxide fuel cells (IT-SOFCs) or even lower, which would imply the use of inexpensive metallic materials, rapid start-up and shut-down, minimization of thermal degradation and reactions between cell components, and longer operational lifetime. Furthermore, as the operation temperature is reduced, system reliability increases, increasing the possibility of using SOFCs for a wide variety of applications, including residential and automotive devices. On the other hand, reduced operating temperatures reduces the overall electrochemical performance due to increased

The SOFC electrolyte material, typically yttria stabilized zirconia (YSZ) or rare earth doped ceria, must be an electronic insulating but ion-conducting ceramic that allows only protons or oxygen ions to pass through. Furthermore, the electrolyte material must be dense to separate the air and fuel, chemically and structurally stable over a wide range of partial pressures of oxygen and temperatures. The cathode (air electrode) material, typically lanthanum manganites and cobaltites, has to be electrocatalyst for oxygen reduction into oxide ions. When an oxygen ionic conducting oxide is adopted as electrolyte, these ions diffuse through the material to the anode (fuel electrode), driven by the differences in oxygen chemical potential between fuel and air constituents of the cell, where they electrochemically oxidize fuels such as hydrogen, methane, and hydrocarbons. The released electrons flow through an external circuit to the cathode to complete the circuit. The

Transform Infrared Spectroscopy (FTIR), FAR and MID-Infrared spectroscopy.

**2. Solid oxide fuel cells and its components** 

Fig. 1. A typical SOFC single cell with its basic components (anode, cathode and electrolyte) and the mechanism of energy production.

ohmic losses and electrode polarization losses associated with thermally activated processes of both ionic transport and electrode reactions. In special, the electrochemical activity of the cathode dramatically deteriorates with decreasing temperature for typical strontium-doped lanthanum manganite (LSM) based electrodes (Singhal 2000; Steele, 2000; Steele & Heinzel, 2001).

Thus, to achieve acceptable performance of IT-SOFCs, reducing the electrolyte resistance and electrode polarization losses are two key points. Losses attributed to electrolytes can be minimized by decreasing the thickness of electrolyte layers or substituting traditional YSZ by other electrolytes, such as doped ceria and apatite-like materials. The overall electrochemical limitation would then be ruled by electrode polarization losses. Due to the higher activation energy and lower reaction kinetics for oxygen reduction at the cathode compared with those of the fuel oxidation at the anode, the polarization loss from the air electrode (cathode) that limits the overall cell performance (Ivers-Tiffee et al., 2001; Tsai & Barnett, 1997). Therefore, the development of new cathode materials with high electrocatalytic activity for oxygen reduction becomes a critical issue for the development of IT-SOFCs.

#### **2.1 Development of cathode materials**

The development of high performance cathodes is based on reducing its thickness from hundreds to few micrometers associated with the addition of an electrolyte material in the formation of composite cathodes. These are interesting approaches that can improve electrode performance for the reduction of oxygen. The latter allows the extension of the triple phase boundaries (TPB) from the electrolyte/cathode interface deep into the bulk of the electrode, permitting electrochemical reactions to take place within the electrode. The

Infrared Spectroscopy Techniques in the Characterization of SOFC Functional Ceramics 387

preparation of LSM-SDC films from powders synthesized using commercial gelatin have been encountered. This chapter presents recent results in obtaining cathodes based on LSM and SDC powders obtained by the modified Pechini method using commercial gelatin as polymerizing agent. LSM-SDC films were prepared using different amounts of ethyl cellulose as pore former. Among the many methods which can be used to produce SOFC cathodes, slurry spin coating was selected due to its fast processing time and high uniformity over the surface. In this technique, a high spin rotation quickly lays the slurry on the substrate which dries in short times, hindering particle agglomeration, which also contributes to the uniform distribution of particles along the green film. The characteristics of not only LSM, SDC and LSCF powders, but also LSM-SDC spin-coated and LSCF-SDC

screen-printed films have been investigated by different techniques.

spectroscopy to characterize working SOFC anodes.

electrochemical modeling analyses (Guo et al., 2010).

**4. Experimental** 

**3. Infrared spectroscopy techniques applied in the SOFC research** 

The fundamental studies of the catalytic reactions mechanisms that occur near the threephase boundary in the anode of a solid oxide fuel cell still deserving attention and investigation. Many in situ spectroscopies such as Raman and infrared spectroscopy are routinely used in catalysis research to characterize surface intermediates and reaction mechanisms. It is very difficult to apply in situ spectroscopy techniques to an operating SOFC anode (Atkinson et al., 2004). Recently research groups (Liu et al., 2002; Guo et al., 2010) presented their methodologies on the possible ways to apply the infrared emission

Guo and co-workers (Guo et al., 2010) worked recently developing a novel experimental technique to measure in situ surface deformation and temperature on the anode surface of a SOFC button cell, along with cell electrochemical performance under operating conditions. An adaptation of a SOFC button cell test apparatus was integrated with a Sagnac interferometric optical setup and IR thermometer. This optical technique was capable of in situ, noncontact, electrode surface deformation, and temperature measurement under SOFC operating conditions. The surface deformation measurement sensitivity is half-wavelength and is immune to temperature uctuation and environmental vibration. The experimental data can be used for the validation and further development of SOFC structural and

Other application of infrared spectroscopy techniques, for example FTIR, is for functional ceramic, electrodes and electrolyte, characterization. Through analyze of the absorption bands in low wave number region the oxygen-metal linkage are associated; the enlarged bands are attributed to the stretching vibration of hydrogen-bonded OH groups present. Moreover, the residual carbon from the synthesis not eliminated in the thermal treatment processes, can also be observed in some bands, for example associated to carboxylate anion

During the last years the research of the group on SOFC development has been focused on the synthesis of LaSrMnO3 (LSM), SmCeO2 (SDC) and LaSrCoFeO3 (LSCF) powders by two different methods. LSM and SDC powders have been prepared by the modified Pechini

(COO-) stretching and the C-O groups or to stretching vibration of the C-O bonds.

**4.1 Preparation and characterization of functional materials for SOFC** 

preparation of composite cathodes by the combination of strontium-doped lanthanum manganite (LSM) with yttria stabilized zirconia (YSZ) or even samarium oxide-doped ceria (SDC) has been intensively studied in the last years. Whereas rare earth doped ceria, both samarium (SDC) and gadolinium doped ceria (CGO), have higher ionic conductivity than YSZ. Thus, composite cathodes such as LSM-SDC or LSM-CGO have better electrochemical performance than LSM-YSZ, and are particularly indicated for IT-SOFCs (Zhang et al., 2007; Chen et al., 2008; Wang et al., 2007).

## **2.1.1 Lanthanum manganite and cobaltite based cathodes**

Lanthanum manganites (LaMnO3), particularly strontium-doped lanthanum manganites (La1−xSrxMnO3 - LSM), are widely used as cathode materials in SOFCs operating at high temperatures (800 – 1000 °C), which is the temperature range where yttria-stabilized zirconia (YSZ) is commonly used as electrolyte. Doping Sr into LaMnO3 considerably increases electrical conductivity because of the increased number of holes. LSM cathodes present high stability, electro-catalytic activity for O2 reduction at high temperatures and thermal expansion coefficient reasonably similar to YSZ electrolyte (Escobedo et al., 2008; Jiang, 2003; Minh, 1993). Even thought LSM cathodes have been successfully used as functional materials in high temperature SOFCs, they have been replaced for more efficient lanthanum strontium cobaltite ferrites (LSCF). LSMs are poor ionic conductors and the electrochemical reactions are limited to the region close to triple phase boundaries (TPB). On the other hand, LSCF is a mixed electronic/ionic conductor (MEIC) with appreciable ionic conductivity. The exchange of oxygen ions occurs at the electrode surface with the diffusion of oxygen through the mixed conductor (Liu et al., 2007; Shao et al., 2009).

Due to low ionic conductivity and high activation energy to oxygen dissociation at low temperatures, cathodes made of LSM and LSCF are normally combined with the electrolyte material forming a composite cathode. The composite cathodes containing ceria are more attractive to low and intermediate temperatures (500 – 700 °C) than the ones containing YSZ, which has lower ionic conductivity in such temperatures (Yang et al., 2007; Xu et al., 2006; Xu et al., 2005; Chen et al., 2007; Zhang et al., 2008; Lin et al., 2008).

Among the extensive number of chemical synthesis routes available for the preparation of LSM and LSCF powders, the polymeric precursor method (Pechini method), solid state reaction, spray pyrolysis and combustion synthesis have been successful used (Conceição et al., 2009; Cela et al., 2009; Grossin & Noudem, 2004; Guo et al., 2006; Macedo et al., 2009; Conceição et al., 2011). In spite of the proved efficiency of the polymeric precursor method to produce monophasic nanometric powders, alternative synthesis methods mainly using microwave technology have being implemented in an attempt to minimize energy consumption and total powder production time (Liu et al., 2007). An alternative method for preparing nanoparticles with particle sizes of order of nanometers using low annealing temperatures has been developed using commercial gelatin as a polymerizing agent. This method has been named by the authors as *soft chemical route* (Medeiros et al., 2004; Maia et al., 2006).

Regarding the development of LSM based composite cathodes, several works have reported the preparation of LSM-SDC cathodes from the powders obtained by different synthesis methods (Ye et al., 2007; Chen et al., 2007; Xu et al., 2009). However, no reports on the preparation of LSM-SDC films from powders synthesized using commercial gelatin have been encountered. This chapter presents recent results in obtaining cathodes based on LSM and SDC powders obtained by the modified Pechini method using commercial gelatin as polymerizing agent. LSM-SDC films were prepared using different amounts of ethyl cellulose as pore former. Among the many methods which can be used to produce SOFC cathodes, slurry spin coating was selected due to its fast processing time and high uniformity over the surface. In this technique, a high spin rotation quickly lays the slurry on the substrate which dries in short times, hindering particle agglomeration, which also contributes to the uniform distribution of particles along the green film. The characteristics of not only LSM, SDC and LSCF powders, but also LSM-SDC spin-coated and LSCF-SDC screen-printed films have been investigated by different techniques.

## **3. Infrared spectroscopy techniques applied in the SOFC research**

The fundamental studies of the catalytic reactions mechanisms that occur near the threephase boundary in the anode of a solid oxide fuel cell still deserving attention and investigation. Many in situ spectroscopies such as Raman and infrared spectroscopy are routinely used in catalysis research to characterize surface intermediates and reaction mechanisms. It is very difficult to apply in situ spectroscopy techniques to an operating SOFC anode (Atkinson et al., 2004). Recently research groups (Liu et al., 2002; Guo et al., 2010) presented their methodologies on the possible ways to apply the infrared emission spectroscopy to characterize working SOFC anodes.

Guo and co-workers (Guo et al., 2010) worked recently developing a novel experimental technique to measure in situ surface deformation and temperature on the anode surface of a SOFC button cell, along with cell electrochemical performance under operating conditions. An adaptation of a SOFC button cell test apparatus was integrated with a Sagnac interferometric optical setup and IR thermometer. This optical technique was capable of in situ, noncontact, electrode surface deformation, and temperature measurement under SOFC operating conditions. The surface deformation measurement sensitivity is half-wavelength and is immune to temperature uctuation and environmental vibration. The experimental data can be used for the validation and further development of SOFC structural and electrochemical modeling analyses (Guo et al., 2010).

Other application of infrared spectroscopy techniques, for example FTIR, is for functional ceramic, electrodes and electrolyte, characterization. Through analyze of the absorption bands in low wave number region the oxygen-metal linkage are associated; the enlarged bands are attributed to the stretching vibration of hydrogen-bonded OH groups present. Moreover, the residual carbon from the synthesis not eliminated in the thermal treatment processes, can also be observed in some bands, for example associated to carboxylate anion (COO-) stretching and the C-O groups or to stretching vibration of the C-O bonds.

## **4. Experimental**

386 Infrared Spectroscopy – Materials Science, Engineering and Technology

preparation of composite cathodes by the combination of strontium-doped lanthanum manganite (LSM) with yttria stabilized zirconia (YSZ) or even samarium oxide-doped ceria (SDC) has been intensively studied in the last years. Whereas rare earth doped ceria, both samarium (SDC) and gadolinium doped ceria (CGO), have higher ionic conductivity than YSZ. Thus, composite cathodes such as LSM-SDC or LSM-CGO have better electrochemical performance than LSM-YSZ, and are particularly indicated for IT-SOFCs (Zhang et al., 2007;

Lanthanum manganites (LaMnO3), particularly strontium-doped lanthanum manganites (La1−xSrxMnO3 - LSM), are widely used as cathode materials in SOFCs operating at high temperatures (800 – 1000 °C), which is the temperature range where yttria-stabilized zirconia (YSZ) is commonly used as electrolyte. Doping Sr into LaMnO3 considerably increases electrical conductivity because of the increased number of holes. LSM cathodes present high stability, electro-catalytic activity for O2 reduction at high temperatures and thermal expansion coefficient reasonably similar to YSZ electrolyte (Escobedo et al., 2008; Jiang, 2003; Minh, 1993). Even thought LSM cathodes have been successfully used as functional materials in high temperature SOFCs, they have been replaced for more efficient lanthanum strontium cobaltite ferrites (LSCF). LSMs are poor ionic conductors and the electrochemical reactions are limited to the region close to triple phase boundaries (TPB). On the other hand, LSCF is a mixed electronic/ionic conductor (MEIC) with appreciable ionic conductivity. The exchange of oxygen ions occurs at the electrode surface with the diffusion

Due to low ionic conductivity and high activation energy to oxygen dissociation at low temperatures, cathodes made of LSM and LSCF are normally combined with the electrolyte material forming a composite cathode. The composite cathodes containing ceria are more attractive to low and intermediate temperatures (500 – 700 °C) than the ones containing YSZ, which has lower ionic conductivity in such temperatures (Yang et al., 2007; Xu et al., 2006;

Among the extensive number of chemical synthesis routes available for the preparation of LSM and LSCF powders, the polymeric precursor method (Pechini method), solid state reaction, spray pyrolysis and combustion synthesis have been successful used (Conceição et al., 2009; Cela et al., 2009; Grossin & Noudem, 2004; Guo et al., 2006; Macedo et al., 2009; Conceição et al., 2011). In spite of the proved efficiency of the polymeric precursor method to produce monophasic nanometric powders, alternative synthesis methods mainly using microwave technology have being implemented in an attempt to minimize energy consumption and total powder production time (Liu et al., 2007). An alternative method for preparing nanoparticles with particle sizes of order of nanometers using low annealing temperatures has been developed using commercial gelatin as a polymerizing agent. This method has been named by the authors as *soft chemical route* (Medeiros et al., 2004; Maia et

Regarding the development of LSM based composite cathodes, several works have reported the preparation of LSM-SDC cathodes from the powders obtained by different synthesis methods (Ye et al., 2007; Chen et al., 2007; Xu et al., 2009). However, no reports on the

Chen et al., 2008; Wang et al., 2007).

al., 2006).

**2.1.1 Lanthanum manganite and cobaltite based cathodes** 

of oxygen through the mixed conductor (Liu et al., 2007; Shao et al., 2009).

Xu et al., 2005; Chen et al., 2007; Zhang et al., 2008; Lin et al., 2008).

#### **4.1 Preparation and characterization of functional materials for SOFC**

During the last years the research of the group on SOFC development has been focused on the synthesis of LaSrMnO3 (LSM), SmCeO2 (SDC) and LaSrCoFeO3 (LSCF) powders by two different methods. LSM and SDC powders have been prepared by the modified Pechini

Infrared Spectroscopy Techniques in the Characterization of SOFC Functional Ceramics 389

Fig. 2. Synthesis mechanism of the modified Pechini method.

scaling this up to a polypeptide (a protein chain), each of the peptide bonds will be broken in exactly the same way. This means that a mixture of the amino acids that makes up the protein will form, although in the form of their positive ions because of the presence of hydrogen ions from the citric acid. The presence of positive ions (not shown in Fig.2) is due to the fact that an amino acid has both a basic amine group and an acidic carboxylic acid group. There is an internal transference of a hydrogen ion from the -COOH group to the -NH2 group to grant both a negative and positive charge to the ion. This is called a zwitterion. This is how amino acids exist even in the solid state. After being dissolved in water, the gelatin forms a simple solution which also contains this ion. A zwitterion is a compound with no overall electrical

method using gelatin as a polymerizing agent. LSCF powder has been synthesized by microwave-assisted combustion. The electrochemical performance of a single electrolytesupported SOFC containing LSCF-SDC composite cathode film has also been investigated.

The raw materials used to synthesize LSM and SDC powders by the modified Pechini method, in which gelatin replaces ethylene glycol, were lanthanum, strontium, manganese (VETEC, Brazil), samarium and cerium (Sigma-Aldrich, Germany) nitrates. For synthesizing LSM, a solution of manganese citrate was firstly prepared from manganese nitrate and citric acid with a molar ratio of 1:3 (metal/citric acid) under stirring for 2 h at 70 °C. Stoichiometric amounts of the other citrates were added one by one at 1 hour intervals. The temperature was slowly increased up to 80 ºC and then gelatin was added at a weight ratio of 40:60 (gelatin/citric acid). The solution was stirred on a hot plate, which allowed the temperature to be controlled until a polymeric resin was formed. This resin was pre-calcined at 300 °C for 2 h, forming a black solid mass which was grounded into a powder and calcined at 500, 700 or 900 °C for 4 h to obtain the LSM perovskite structure. SDC powder was obtained following the same procedure but calcined between 700 and 900 ºC.

To further understand the synthesis mechanism used in this work, which uses gelatin instead of ethylene glycol, it is necessary to review the traditional polymeric precursor method proposed by Pechini (Pechini, 1967). The Pechini method involves the formation of stable metal-chelate complexes with certain alpha-hydroxycarboxyl acids, such as citric acid, and polyesterification in the presence of a polyhydroxy alcohol, such as ethylene glycol, to form a polymeric resin. The metal cations are homogeneously distributed in the polymeric resin, which is then calcined to yield the desired oxides. The most common materials used as source of cations are nitrate salts since they can be fully removed at low temperatures (400 – 500 ºC). The synthesis mechanism of the modified Pechini method used in this work can be explained in three basic steps, as shown in Fig. 2. It stands out by its simplicity and low cost, using only citric acid, gelatin and metal nitrates as reagents.

In the first step, a solution of metallic citrate is prepared from the mixture of deionized water, citric acid and the salts of each metal ion. The citric acid acts as a chelating agent to remove metal ions from the solution. Chelation is the ability that a chemical substance has to form a ring-like structure with a metal ion, resulting in a compound with different chemical properties compared to the original metal, which prevents it from following different chemical routes. The chelating agent acts as a crayfish which traps the metal in its claws. When metal ions are in solution, they are surrounded by water molecules which avoid the establishment of new bonds. Thus, the chelating agent replaces such water molecules (bonds) and forms a ring-like structure, resulting in the phenomenon known as chelation. The number of binding sites that can form coordination bonds in the metal determines the number of rings formed.

The second step of this synthesis mechanism is the gelatin hydrolysis. Gelatin is a natural polymer, composed of a mixture of high molecular weight polypeptides (proteins) obtained by controlled hydrolysis of collagen fiber. Therefore, it can be used for cheap aqueous polymeric precursor synthesis in order to attain nanocrystalline powders. As illustrated in Fig. 2, the gelatin hydrolysis corresponds to the breaking of peptide bonds in the presence of water. For easier understanding, the second step reaction, in this illustration, is taking place in a low molecular weight peptide. An extra hydrogen ion is necessary to react with the - NH2 group on the left-hand end of the peptide, the one not involved in the peptide bond. By

method using gelatin as a polymerizing agent. LSCF powder has been synthesized by microwave-assisted combustion. The electrochemical performance of a single electrolytesupported SOFC containing LSCF-SDC composite cathode film has also been investigated. The raw materials used to synthesize LSM and SDC powders by the modified Pechini method, in which gelatin replaces ethylene glycol, were lanthanum, strontium, manganese (VETEC, Brazil), samarium and cerium (Sigma-Aldrich, Germany) nitrates. For synthesizing LSM, a solution of manganese citrate was firstly prepared from manganese nitrate and citric acid with a molar ratio of 1:3 (metal/citric acid) under stirring for 2 h at 70 °C. Stoichiometric amounts of the other citrates were added one by one at 1 hour intervals. The temperature was slowly increased up to 80 ºC and then gelatin was added at a weight ratio of 40:60 (gelatin/citric acid). The solution was stirred on a hot plate, which allowed the temperature to be controlled until a polymeric resin was formed. This resin was pre-calcined at 300 °C for 2 h, forming a black solid mass which was grounded into a powder and calcined at 500, 700 or 900 °C for 4 h to obtain the LSM perovskite structure. SDC powder

was obtained following the same procedure but calcined between 700 and 900 ºC.

low cost, using only citric acid, gelatin and metal nitrates as reagents.

number of rings formed.

To further understand the synthesis mechanism used in this work, which uses gelatin instead of ethylene glycol, it is necessary to review the traditional polymeric precursor method proposed by Pechini (Pechini, 1967). The Pechini method involves the formation of stable metal-chelate complexes with certain alpha-hydroxycarboxyl acids, such as citric acid, and polyesterification in the presence of a polyhydroxy alcohol, such as ethylene glycol, to form a polymeric resin. The metal cations are homogeneously distributed in the polymeric resin, which is then calcined to yield the desired oxides. The most common materials used as source of cations are nitrate salts since they can be fully removed at low temperatures (400 – 500 ºC). The synthesis mechanism of the modified Pechini method used in this work can be explained in three basic steps, as shown in Fig. 2. It stands out by its simplicity and

In the first step, a solution of metallic citrate is prepared from the mixture of deionized water, citric acid and the salts of each metal ion. The citric acid acts as a chelating agent to remove metal ions from the solution. Chelation is the ability that a chemical substance has to form a ring-like structure with a metal ion, resulting in a compound with different chemical properties compared to the original metal, which prevents it from following different chemical routes. The chelating agent acts as a crayfish which traps the metal in its claws. When metal ions are in solution, they are surrounded by water molecules which avoid the establishment of new bonds. Thus, the chelating agent replaces such water molecules (bonds) and forms a ring-like structure, resulting in the phenomenon known as chelation. The number of binding sites that can form coordination bonds in the metal determines the

The second step of this synthesis mechanism is the gelatin hydrolysis. Gelatin is a natural polymer, composed of a mixture of high molecular weight polypeptides (proteins) obtained by controlled hydrolysis of collagen fiber. Therefore, it can be used for cheap aqueous polymeric precursor synthesis in order to attain nanocrystalline powders. As illustrated in Fig. 2, the gelatin hydrolysis corresponds to the breaking of peptide bonds in the presence of water. For easier understanding, the second step reaction, in this illustration, is taking place in a low molecular weight peptide. An extra hydrogen ion is necessary to react with the - NH2 group on the left-hand end of the peptide, the one not involved in the peptide bond. By

Fig. 2. Synthesis mechanism of the modified Pechini method.

scaling this up to a polypeptide (a protein chain), each of the peptide bonds will be broken in exactly the same way. This means that a mixture of the amino acids that makes up the protein will form, although in the form of their positive ions because of the presence of hydrogen ions from the citric acid. The presence of positive ions (not shown in Fig.2) is due to the fact that an amino acid has both a basic amine group and an acidic carboxylic acid group. There is an internal transference of a hydrogen ion from the -COOH group to the -NH2 group to grant both a negative and positive charge to the ion. This is called a zwitterion. This is how amino acids exist even in the solid state. After being dissolved in water, the gelatin forms a simple solution which also contains this ion. A zwitterion is a compound with no overall electrical

Infrared Spectroscopy Techniques in the Characterization of SOFC Functional Ceramics 391

A preliminary performance test was carried out to qualitatively evaluate the prepared LSCF-SDC composite cathode. A single SOFC supported on commercial, 200µm thick, YSZ electrolyte (Kerafol, Germany) with Ni-YSZ anode and LSCF-SDC composite cathode screen printed was used. To assemble the single cell, first the anode suspension was printed and the half cell was sintered at 1300 °C for 4 h. The anode was reduced during cell operation. Later on, the LSCF-SDC cathode was printed on the other electrolyte side and sintered at 950°C for 4 h to avoid undesirable reactions. An in-house test station was used in the performance evaluation. The single cell was tested using dry hydrogen as fuel and oxygen as oxidant, in the ratio of 40ml/min H2 and 40ml/min O2. The current–voltage characteristic of the cell was measured using linear sweep voltammetry (LSV) over a temperature range of 800 to 950 °C. The microstructure of the cathode/electrolyte interface after the performance

XRD patterns of the calcined LSM powders can be observed in Fig. 3. The as-synthesized material depicts the main diffraction peaks characteristic of the rhombohedral structure with space group *R* 3 *c* (Konysheva et al., 2009) and the respective JCPDS (Joint Committee on Powder Diffraction Standards) chart number 53-0058. Secondary phases were found and identified as SrCO3 and Mn3O4. The presence of SrCO3 was identified at 500 and 700 ºC by comparison with JCPDS pattern number 84-1778. The formation of strontium carbonate is due to the reaction between SrO and CO2 produced during organic compound decomposition. Mn3O4 is present only in the powder calcined at 900 °C and was identified by comparison with pattern JCPDS 89-4837. It probably came from the change of oxidation state of the Mn ion. Vargas and co-workers (Vargas et al., 2008) have reported that

10 20 30 40 50 60 70 80

2 (degree)

+ +

@

\*

+

\*

\*

\*

\* \*

La0.8Sr0.2MnO3 SrCO3 Mn3 O4

@@

Fig. 3. XRD patterns of LSM powders calcined at different temperatures.

900 ºC 700 ºC 500 ºC

Intensity (a.u.)

350 ºC

\* \*

\*

**4.2 Preparation and characterization of a SOFC single cell** 

**5.1 Characterization of functional materials for SOFC** 

test was examined by SEM.

**5. Results and discussion** 

charge, but which contains separate parts, positively and negatively charged. As the pH decreases by adding an acid to a solution of an amino acid, the -COO- side of the zwitterion picks up a hydrogen ion, forming the desirable compound.

In the third step, proteins act as amides and react with the free hydroxyl of the citric acid. The hydroxyl of the citric acid reacts like the hydroxyl of an alcohol in a standard esterification process. During the reaction of the amide with the citric acid, the substitution of the amide hydroxyl by an alkoxy radical (-OR) takes place. At the end of the process, the oxygen in the –OH group of the citric acid remains in the chain whereas the hydroxyl oxygen in the amino acids is eliminated in the form of water. This mechanism results in long chains containing metal cations tightly bond and evenly distributed.

To synthesize LSCF powders, an aqueous solution of La(NO3)3, Sr(NO3)2, Co(NO3)2, and Fe(NO3)3 (VETEC, Brazil) and urea as fuel was heated up to 100 ºC under stirring for 10 minutes. Afterwards, the becker was placed in a microwave oven set to 810 W and 2.45 GHz. Self-ignition takes place in about 1 minute. The resulting powder was then calcined at 900 ºC for 4 h in order to remove carbon residues remaining in the ash and to convert the powder to the desired LSCF phase with a well-defined crystalline perovskite structure.

LSM, SDC and LSCF powders were characterized by XRD using a Shimadzu XDR-7000 diffractometer and scanning electron microscopy (SEM-SSX 550, Shimadzu). Infrared spectra were also recorded with FTIR (IR Prestige-21, Shimadzu) in the 400 – 4600 cm−1 spectral range. Specific surface area measurements were performed only for the LSM powders. An infrared reflectance spectrum of a LSM pellet prepared from a powder calcined at 900 °C was recorded with a Fourier-transform spectrometer (Bomem DA 8-02) equipped with a fixed-angle specular reflectance accessory (external incidence angle of 11.5°).

#### **4.1.1 Preparation and characterization of composite cathodes**

Composite powders consisted of 50 wt.% cathode (LSM or LSCF) and 50 wt.% SDC were prepared by mixing in a ball mill for 24 h in order to deposit LSM-SDC and LSCF-SDC composite films by spin coating (LSM-SDC) and screen printing (LSCF-SDC). Prior to spin coating deposition, ceramic suspensions of LSM-SDC composite powders were prepared using ethanol and different amounts of ethyl cellulose as pore-forming material. Ethyl cellulose was added in the weight ratios of 4, 8 and 10 wt.% with respect to the total solid weight. With the purpose of studying the influence of the milling process on the particle size distribution, another mixture was prepared without milling and without the addition of ethyl cellulose. Before coating, the ceramic suspensions were ultrasonically treated and then deposited by spin coating onto YSZ substrates. Commercially available YSZ powder (Tosoh Corporation, Japan) was compressed into pellets (13 mm in diameter) under uniaxial pressure (74 MPa) and then sintered in air at 1450 °C for 4 h to increase its mechanical strength for application as ceramic substrate. In the deposition process, 25 layers were applied using initial and final rotation speeds of 500 rpm for 15 seconds and 5000 rpm for 30 seconds, respectively. The composite films were attained after sintering at 1150 °C for 4 h. LSCF-SDC cathodes were screen-printed onto YSZ electrolytes and sintered at 950 °C for 4 h. The morphological characterization of the surface and cross section of LSM-SDC composite cathodes was performed using a field emission gun scanning electron microscope (FEG-SEM, Zeiss-Supra 35) and a scanning electron microscope (SEM-SSX 550, Shimadzu).

#### **4.2 Preparation and characterization of a SOFC single cell**

A preliminary performance test was carried out to qualitatively evaluate the prepared LSCF-SDC composite cathode. A single SOFC supported on commercial, 200µm thick, YSZ electrolyte (Kerafol, Germany) with Ni-YSZ anode and LSCF-SDC composite cathode screen printed was used. To assemble the single cell, first the anode suspension was printed and the half cell was sintered at 1300 °C for 4 h. The anode was reduced during cell operation. Later on, the LSCF-SDC cathode was printed on the other electrolyte side and sintered at 950°C for 4 h to avoid undesirable reactions. An in-house test station was used in the performance evaluation. The single cell was tested using dry hydrogen as fuel and oxygen as oxidant, in the ratio of 40ml/min H2 and 40ml/min O2. The current–voltage characteristic of the cell was measured using linear sweep voltammetry (LSV) over a temperature range of 800 to 950 °C. The microstructure of the cathode/electrolyte interface after the performance test was examined by SEM.

#### **5. Results and discussion**

390 Infrared Spectroscopy – Materials Science, Engineering and Technology

charge, but which contains separate parts, positively and negatively charged. As the pH decreases by adding an acid to a solution of an amino acid, the -COO- side of the zwitterion

In the third step, proteins act as amides and react with the free hydroxyl of the citric acid. The hydroxyl of the citric acid reacts like the hydroxyl of an alcohol in a standard esterification process. During the reaction of the amide with the citric acid, the substitution of the amide hydroxyl by an alkoxy radical (-OR) takes place. At the end of the process, the oxygen in the –OH group of the citric acid remains in the chain whereas the hydroxyl oxygen in the amino acids is eliminated in the form of water. This mechanism results in long

To synthesize LSCF powders, an aqueous solution of La(NO3)3, Sr(NO3)2, Co(NO3)2, and Fe(NO3)3 (VETEC, Brazil) and urea as fuel was heated up to 100 ºC under stirring for 10 minutes. Afterwards, the becker was placed in a microwave oven set to 810 W and 2.45 GHz. Self-ignition takes place in about 1 minute. The resulting powder was then calcined at 900 ºC for 4 h in order to remove carbon residues remaining in the ash and to convert the powder

LSM, SDC and LSCF powders were characterized by XRD using a Shimadzu XDR-7000 diffractometer and scanning electron microscopy (SEM-SSX 550, Shimadzu). Infrared spectra were also recorded with FTIR (IR Prestige-21, Shimadzu) in the 400 – 4600 cm−1 spectral range. Specific surface area measurements were performed only for the LSM powders. An infrared reflectance spectrum of a LSM pellet prepared from a powder calcined at 900 °C was recorded with a Fourier-transform spectrometer (Bomem DA 8-02) equipped with a fixed-angle

Composite powders consisted of 50 wt.% cathode (LSM or LSCF) and 50 wt.% SDC were prepared by mixing in a ball mill for 24 h in order to deposit LSM-SDC and LSCF-SDC composite films by spin coating (LSM-SDC) and screen printing (LSCF-SDC). Prior to spin coating deposition, ceramic suspensions of LSM-SDC composite powders were prepared using ethanol and different amounts of ethyl cellulose as pore-forming material. Ethyl cellulose was added in the weight ratios of 4, 8 and 10 wt.% with respect to the total solid weight. With the purpose of studying the influence of the milling process on the particle size distribution, another mixture was prepared without milling and without the addition of ethyl cellulose. Before coating, the ceramic suspensions were ultrasonically treated and then deposited by spin coating onto YSZ substrates. Commercially available YSZ powder (Tosoh Corporation, Japan) was compressed into pellets (13 mm in diameter) under uniaxial pressure (74 MPa) and then sintered in air at 1450 °C for 4 h to increase its mechanical strength for application as ceramic substrate. In the deposition process, 25 layers were applied using initial and final rotation speeds of 500 rpm for 15 seconds and 5000 rpm for 30 seconds, respectively. The composite films were attained after sintering at 1150 °C for 4 h. LSCF-SDC cathodes were screen-printed onto YSZ electrolytes and sintered at 950 °C for 4 h. The morphological characterization of the surface and cross section of LSM-SDC composite cathodes was performed using a field emission gun scanning electron microscope (FEG-SEM, Zeiss-Supra 35) and a scanning electron microscope (SEM-SSX 550, Shimadzu).

picks up a hydrogen ion, forming the desirable compound.

chains containing metal cations tightly bond and evenly distributed.

specular reflectance accessory (external incidence angle of 11.5°).

**4.1.1 Preparation and characterization of composite cathodes** 

to the desired LSCF phase with a well-defined crystalline perovskite structure.

#### **5.1 Characterization of functional materials for SOFC**

XRD patterns of the calcined LSM powders can be observed in Fig. 3. The as-synthesized material depicts the main diffraction peaks characteristic of the rhombohedral structure with space group *R* 3 *c* (Konysheva et al., 2009) and the respective JCPDS (Joint Committee on Powder Diffraction Standards) chart number 53-0058. Secondary phases were found and identified as SrCO3 and Mn3O4. The presence of SrCO3 was identified at 500 and 700 ºC by comparison with JCPDS pattern number 84-1778. The formation of strontium carbonate is due to the reaction between SrO and CO2 produced during organic compound decomposition. Mn3O4 is present only in the powder calcined at 900 °C and was identified by comparison with pattern JCPDS 89-4837. It probably came from the change of oxidation state of the Mn ion. Vargas and co-workers (Vargas et al., 2008) have reported that

Fig. 3. XRD patterns of LSM powders calcined at different temperatures.

Infrared Spectroscopy Techniques in the Characterization of SOFC Functional Ceramics 393

Fig. 4. SEM micrographs of LSM powders calcined at: (a) 500 ºC, (b) 700 ºC and (c) 900 ºC for

4 h. FEG-SEM high magnification image of powder calcined at 900 °C (d).

Fig. 5. FTIR spectra of LSM powders calcined at different temperatures.

The infrared reflectance spectrum in the FAR and MID ranges is presented in Fig. 6. The almost continuous decrease of the reflectivity as a function of the wave number shows that this spectrum is dominated by a conduction mechanism (Gosnet et al., 2008). Moreover, first

temperatures higher than 1000 ºC are necessary to remove all the carbon present in the LSM powders. Lower calcination temperatures result in free carbon or as carbonates from the degradation of citrates.

Table 1 lists the specific surface area (SBET), average crystallite size (DXRD), average particle size (dBET), microstrain and the dBET/DXRD ratio for LSM powders calcined at 700 and 900 °C. There is a clear dependency between the average crystallite size and the calcination temperature. The thermal treatment temperature is directly proportional to the crystallite size, because of the crystallization process that occurs in high temperatures. Therefore, the specific surface area decreases with the onset of particle sintering. High particle sizes promote lower surface area to volume ratio, which results in small deformations in the crystalline structure parameters.


\*Calculated from specific surface area

\*Theoretical density: LSM= 6.521 g/cm3

\*\*Calculated by Rietveld refinement

Table 1. Specific surface area (SBET), average particle size (dBET), average crystallite size (DXRD), microstrain and crystalline degree of LSM powders.

The dBET/DXRD ratio, which indicates the degree of particle agglomeration, increases with the calcination temperature. In a previous study (Cela et al., 2009), dBET/DXRD equal to 12 was established for the LSM powder synthesized with ethylene glycol as the polymerizing agent and calcined at 900 °C. This result shows that the substitution of ethylene glycol for gelatin reduces powder agglomeration, improving the potential to prepare ceramic suspensions to be deposited as cathode films with good adherence to the substrate.

LSM powders were observed by scanning electron microscopy (Fig. 4). SEM micrographs revealed the presence of soft particles agglomerates. Fig. 4(d) displays a FEG-SEM high magnification image of the LSM powder calcined at 900 °C. As it can be seen, this powder reveals strong agglomeration of particles with an estimated size lower than 100 nm. Results from SEM analyses are in good agreement with specific surface area measurements.

In the FTIR spectra of the LSM powders calcined between 500 and 900 ºC, displayed in Fig. 5, a band located at 603 cm−1 is attributed to M – O (metal – oxygen) stretching, and is characteristic of the perovskite structure. On the other hand, bands corresponding to carbon bonds, particularly carboxyl groups are visible in 1000 – 2500 cm−1 interval. The two absorption bands at 1460 cm−1 and 2360 cm−1 are due to C=O vibration and can be related to traces of carbonate. In the range of 3300 – 3670 cm−1 a band corresponding to O – H bond can be attributed to adsorbed water due to the contact of the sample with the environment. With this information and the XRD analyses, it became clear that the carbonate content of the synthesized powders decreases and even vanishes at higher calcination temperatures.

temperatures higher than 1000 ºC are necessary to remove all the carbon present in the LSM powders. Lower calcination temperatures result in free carbon or as carbonates from the

Table 1 lists the specific surface area (SBET), average crystallite size (DXRD), average particle size (dBET), microstrain and the dBET/DXRD ratio for LSM powders calcined at 700 and 900 °C. There is a clear dependency between the average crystallite size and the calcination temperature. The thermal treatment temperature is directly proportional to the crystallite size, because of the crystallization process that occurs in high temperatures. Therefore, the specific surface area decreases with the onset of particle sintering. High particle sizes promote lower surface area to volume ratio, which results in small deformations in the

> dBET (nm)\*

Table 1. Specific surface area (SBET), average particle size (dBET), average crystallite size

The dBET/DXRD ratio, which indicates the degree of particle agglomeration, increases with the calcination temperature. In a previous study (Cela et al., 2009), dBET/DXRD equal to 12 was established for the LSM powder synthesized with ethylene glycol as the polymerizing agent and calcined at 900 °C. This result shows that the substitution of ethylene glycol for gelatin reduces powder agglomeration, improving the potential to prepare ceramic suspensions to

LSM powders were observed by scanning electron microscopy (Fig. 4). SEM micrographs revealed the presence of soft particles agglomerates. Fig. 4(d) displays a FEG-SEM high magnification image of the LSM powder calcined at 900 °C. As it can be seen, this powder reveals strong agglomeration of particles with an estimated size lower than 100 nm. Results from SEM analyses are in good agreement with specific surface area

In the FTIR spectra of the LSM powders calcined between 500 and 900 ºC, displayed in Fig. 5, a band located at 603 cm−1 is attributed to M – O (metal – oxygen) stretching, and is characteristic of the perovskite structure. On the other hand, bands corresponding to carbon bonds, particularly carboxyl groups are visible in 1000 – 2500 cm−1 interval. The two absorption bands at 1460 cm−1 and 2360 cm−1 are due to C=O vibration and can be related to traces of carbonate. In the range of 3300 – 3670 cm−1 a band corresponding to O – H bond can be attributed to adsorbed water due to the contact of the sample with the environment. With this information and the XRD analyses, it became clear that the carbonate content of the synthesized powders decreases and even vanishes at higher

700 18.79 48.9 22.80 0.004869 2.1 900 8.46 108.7 25.21 0.003917 4.3

DXRD (nm)\*\* Microstrain

(%) dBET/DXRD

(m2/g)

(DXRD), microstrain and crystalline degree of LSM powders.

be deposited as cathode films with good adherence to the substrate.

degradation of citrates.

crystalline structure parameters.

\*Calculated from specific surface area \*Theoretical density: LSM= 6.521 g/cm3 \*\*Calculated by Rietveld refinement

measurements.

calcination temperatures.

Calcination temperature (°C) SBET

Fig. 4. SEM micrographs of LSM powders calcined at: (a) 500 ºC, (b) 700 ºC and (c) 900 ºC for 4 h. FEG-SEM high magnification image of powder calcined at 900 °C (d).

Fig. 5. FTIR spectra of LSM powders calcined at different temperatures.

The infrared reflectance spectrum in the FAR and MID ranges is presented in Fig. 6. The almost continuous decrease of the reflectivity as a function of the wave number shows that this spectrum is dominated by a conduction mechanism (Gosnet et al., 2008). Moreover, first

Infrared Spectroscopy Techniques in the Characterization of SOFC Functional Ceramics 395

Fig. 7. XRD patterns of SDC powders calcined at different temperatures (a) and FEG-SEM

image of the SDC powder calcined at 900 °C (b).

Fig. 6. Infrared reflectance spectrum in the FAR and MID ranges for LSM powder calcined at 900 °C.

order phonon features characteristic of the crystal lattice are clearly superposed on the decreasing reflectivity curve below 700 cm-1. The reflectivity line shape of these phonons is rather characteristic of conducting materials and denotes possibly electron-phonon coupling in the LSM powder calcined at 900 °C, as also observed in La0.7Ca0.3MnO3 ceramics (Kim et al., 1996). On the other hand, the two modes above 1000 cm-1 are likely to be defect modes from impurities due to unreacted materials, probably Mn3O4, as observed by XRD.

The XRD patterns of SDC powders, as synthesized by the modified Pechini method and after calcination from 700 to 900 ºC, are shown in Fig. 7. XRD results revealed no peaks corresponding to secondary phases, i.e., there are no obvious peaks from phases other than SDC (JCPDS 75-0158) until the detection limit of the X-ray diffraction. The Rietveld refinement of the diffraction data using Maud program indicates that SDC powders exhibit cubic structures with space group Fm-3m and crystallite sizes in the range of 7-38 nm. Typical particle morphology of the SDC powder calcined at 900 °C is shown in Fig. 7(b), from which it can be seen that the particles are nearly spherical and strongly agglomerated, as expected for nanoparticles synthesized by the polymeric precursor method. FEG-SEM images also showed that the spherical particles are lower than 50 nm.

The FTIR spectra of SDC powders synthesized by the modified Pechini method and calcined at different temperatures are shown in Fig. 8. The peaks at around 3400 and 1640 cm-1 correspond to the H-O stretching and H-O-H bending vibration, respectively. They indicate that water or hydroxyl groups still existed not only in the as-prepared, but also in all calcined samples. A low intensity band at 1380 cm-1 in the as-prepared powder might indicate the physical adsorption of CO2 or be caused by residual NO3- (Chen et al., 2006). However, this band is eliminated after heat treatment. Therefore, although a synthesis temperature as low as 350 °C was sufficient to produce the SDC phase, powders with higher purity were obtained by calcining above 700 °C.

Fig. 6. Infrared reflectance spectrum in the FAR and MID ranges for LSM powder calcined at

order phonon features characteristic of the crystal lattice are clearly superposed on the decreasing reflectivity curve below 700 cm-1. The reflectivity line shape of these phonons is rather characteristic of conducting materials and denotes possibly electron-phonon coupling in the LSM powder calcined at 900 °C, as also observed in La0.7Ca0.3MnO3 ceramics (Kim et al., 1996). On the other hand, the two modes above 1000 cm-1 are likely to be defect modes from impurities due to unreacted materials, probably Mn3O4, as observed

The XRD patterns of SDC powders, as synthesized by the modified Pechini method and after calcination from 700 to 900 ºC, are shown in Fig. 7. XRD results revealed no peaks corresponding to secondary phases, i.e., there are no obvious peaks from phases other than SDC (JCPDS 75-0158) until the detection limit of the X-ray diffraction. The Rietveld refinement of the diffraction data using Maud program indicates that SDC powders exhibit cubic structures with space group Fm-3m and crystallite sizes in the range of 7-38 nm. Typical particle morphology of the SDC powder calcined at 900 °C is shown in Fig. 7(b), from which it can be seen that the particles are nearly spherical and strongly agglomerated, as expected for nanoparticles synthesized by the polymeric precursor method. FEG-SEM

The FTIR spectra of SDC powders synthesized by the modified Pechini method and calcined at different temperatures are shown in Fig. 8. The peaks at around 3400 and 1640 cm-1 correspond to the H-O stretching and H-O-H bending vibration, respectively. They indicate that water or hydroxyl groups still existed not only in the as-prepared, but also in all calcined samples. A low intensity band at 1380 cm-1 in the as-prepared powder might indicate the physical adsorption of CO2 or be caused by residual NO3- (Chen et al., 2006). However, this band is eliminated after heat treatment. Therefore, although a synthesis temperature as low as 350 °C was sufficient to produce the SDC phase, powders with higher

images also showed that the spherical particles are lower than 50 nm.

purity were obtained by calcining above 700 °C.

900 °C.

by XRD.

Fig. 7. XRD patterns of SDC powders calcined at different temperatures (a) and FEG-SEM image of the SDC powder calcined at 900 °C (b).

Infrared Spectroscopy Techniques in the Characterization of SOFC Functional Ceramics 397

#

\* \*

\* \* \* \*

2-, respectively. Based on these

\*La0.6Sr0.4Co0.2Fe0.8O3

+

La2 CoO4 SrCO3

10 20 30 40 50 60 70 80 # # + ++ +

2 (degree)

may dramatically deteriorate the cathode performance for SOFCs due to its poisoning effect on the cathode surface, further calcination at 900 °C under air for 4 h was carried out in order to eliminate the SrCO3 impurity, driving its decomposition to CO2. Accompanied by the formation of SrCO3, another K2NiF4-structured phase with many small peaks was identified in the as-prepared LSCF powder. This phase was ascribed as La2CoO4, based on JCPDS card 34–1296. It is important to mention that it has good electronic and oxygen ionic conductivity and activity for oxygen reduction, being therefore, expected to have negligible effect on the cathode performance of LSCF (Borovskikh et al., 2003; Vashook et al., 2000). In any case, it is

Fig. 10 shows SEM images of the LSCF powder after calcination at 900 °C. As expected for ceramic materials synthesized by the combustion method, the powder presents a porous microstructure consisting of small uniform particles. The porous structure of this material

Fig. 11 shows the FTIR spectra for the as-prepared and calcined LSCF powders. The strong absorption bands in the low wave number region (~ 595 cm-1) are associated with metaloxygen (M-O) bonds characteristic of the perovskite structure. The similar absorption intensity of this band confirms that the perovskite-type structure is almost completely obtained after the combustion reaction, as observed by X-ray diffraction. The band at 1620 cm-1 corresponds to the stretching vibration frequency of coordinated H2O (δHOH). The bands related to absorbed water (3400 cm-1) were also present in both samples. Before calcination at 900 °C for 4 h, the LSCF powder exhibited strong absorption bands at 1381 cm-1 and 1454 cm-1, which indicate the


results, it is possible to conclude that during LSCF synthesis, some nitrates can react with carbon dioxide to form carbonaceous phases which decompose after further calcination at

\*

+

Fig. 9. XRD patterns of LSCF powder: (a) as-prepared and (b) calcined at 900 °C.

observed that both secondary phases decomposed completely after calcination.

can be attributed to significant gas evolution during the combustion reaction.

900 °C for 4 h. Therefore, FTIR results are consistent with XRD analyses.

\*

(b)

Intensity (a.u.)

characteristic vibration of NO3

(a)

\*

Fig. 8. FTIR spectra of SDC powders at different temperatures.

Fig. 9 shows the XRD patterns of LSCF powder both as-prepared and calcined at 900 °C for 4 h. Prior to calcination and after only 1 minute of microwave irradiation, the powder already exhibited mainly the perovskite-type structure. In addition, some peaks of deleterious phases (LaCoO4 and SrCO3) were detected before calcination. The formation of SrCO3 can be attributed to the reaction of Sr(NO3)2 and CO2. Compared with the as-prepared powder, the LSCF powder after calcination displayed higher crystallization. Since carbonaceous phases

**4000 3500 3000 2500 2000 1500 1000 500**

Wavenumber (cm-1

Fig. 9 shows the XRD patterns of LSCF powder both as-prepared and calcined at 900 °C for 4 h. Prior to calcination and after only 1 minute of microwave irradiation, the powder already exhibited mainly the perovskite-type structure. In addition, some peaks of deleterious phases (LaCoO4 and SrCO3) were detected before calcination. The formation of SrCO3 can be attributed to the reaction of Sr(NO3)2 and CO2. Compared with the as-prepared powder, the LSCF powder after calcination displayed higher crystallization. Since carbonaceous phases

)

Transmittance (%)

<sup>50</sup> 350 ºC

Fig. 8. FTIR spectra of SDC powders at different temperatures.

700 ºC

900 ºC

800 ºC

Fig. 9. XRD patterns of LSCF powder: (a) as-prepared and (b) calcined at 900 °C.

may dramatically deteriorate the cathode performance for SOFCs due to its poisoning effect on the cathode surface, further calcination at 900 °C under air for 4 h was carried out in order to eliminate the SrCO3 impurity, driving its decomposition to CO2. Accompanied by the formation of SrCO3, another K2NiF4-structured phase with many small peaks was identified in the as-prepared LSCF powder. This phase was ascribed as La2CoO4, based on JCPDS card 34–1296. It is important to mention that it has good electronic and oxygen ionic conductivity and activity for oxygen reduction, being therefore, expected to have negligible effect on the cathode performance of LSCF (Borovskikh et al., 2003; Vashook et al., 2000). In any case, it is observed that both secondary phases decomposed completely after calcination.

Fig. 10 shows SEM images of the LSCF powder after calcination at 900 °C. As expected for ceramic materials synthesized by the combustion method, the powder presents a porous microstructure consisting of small uniform particles. The porous structure of this material can be attributed to significant gas evolution during the combustion reaction.

Fig. 11 shows the FTIR spectra for the as-prepared and calcined LSCF powders. The strong absorption bands in the low wave number region (~ 595 cm-1) are associated with metaloxygen (M-O) bonds characteristic of the perovskite structure. The similar absorption intensity of this band confirms that the perovskite-type structure is almost completely obtained after the combustion reaction, as observed by X-ray diffraction. The band at 1620 cm-1 corresponds to the stretching vibration frequency of coordinated H2O (δHOH). The bands related to absorbed water (3400 cm-1) were also present in both samples. Before calcination at 900 °C for 4 h, the LSCF powder exhibited strong absorption bands at 1381 cm-1 and 1454 cm-1, which indicate the characteristic vibration of NO3 -1 and the stretching of the CO3 2-, respectively. Based on these results, it is possible to conclude that during LSCF synthesis, some nitrates can react with carbon dioxide to form carbonaceous phases which decompose after further calcination at 900 °C for 4 h. Therefore, FTIR results are consistent with XRD analyses.

Infrared Spectroscopy Techniques in the Characterization of SOFC Functional Ceramics 399

Fig. 12. FEG-SEM images of the surface of LSM-SDC composite cathodes containing different amounts of ethyl cellulose: (a) 0 wt.%, (b) 4 wt.%, (c) 8 wt.%, (d) 10 wt.%.

cathodes are currently under way.

cathodes containing different amounts of ethyl cellulose. In these images, it is possible to observe the effects of the milling process and the addition of ethyl cellulose in the final microstructure of composite cathodes. The film obtained without addition of ethyl cellulose and without prior milling treatment of powders (Fig. 12a) depicted particle sizes of about 200 nm. On the other hand, in the films obtained with the powders previously milled for 24 h (Figure 12b-d) it can be observed that both particle agglomeration and pore size decrease as the ethyl cellulose content increases. Moreover, the particle size is lower than 200 nm. Typically, if the slurry contains 10 wt.% ethyl cellulose (Fig. 12d), the film exhibits a highly porous surface morphology and uniform pore structure, which are essential conditions for obtaining high-performance cathodes. The electrochemical evaluation of these composite

Fig. 13 shows cross-sectional SEM images of the LSM-SDC cathodes containing 0 and 10 wt.% ethyl cellulose after milling LSM and SDC powders. The film without addition of ethyl cellulose (Fig. 13a) is tightly adhered to the substrate and shows few pores. However, the film containing 10 wt.% ethyl cellulose (~ 10 μm thick) is also well adhered to the YSZ substrate and exhibits higher pore uniformity than that in Fig. 13a. Such a porous microstructure fulfils the need for a SOFC cathode, which possess high active surface area,

while permitting rapid diffusion of oxygen through the porous cathode film.

Fig. 10. SEM images of LSCF powder calcined at 900 °C: (a) porous structure and (b) typical small particles.

Fig. 11. FTIR spectra for the as-prepared and calcined LSCF powders.

#### **5.1.1 Characterization of composite cathodes**

LSM, SDC and LSCF powders synthesized by the above mentioned methods were used to prepare LSM-SDC and LSCF-SDC composite films onto YSZ substrates. A previous study performed by Ye and co-workers (Ye et al., 2007) demonstrated that the amount of ethyl cellulose in LSM-SDC slurries is restricted by cracking of the surface of the films. These authors observed that cracking took place when the amount of pore former reached 15 wt.% and is caused by the large amount of organics which evaporates during sintering. Therefore, in the present study, LSM-SDC slurries with a maximum of 10 wt.% ethyl cellulose were used. Fig. 12 shows FEG-SEM images of the porous structure of LSM-SDC composite

Fig. 10. SEM images of LSCF powder calcined at 900 °C: (a) porous structure and (b) typical

4000 3500 3000 2500 2000 1500 1000 500

Wavenumber (cm-1

LSM, SDC and LSCF powders synthesized by the above mentioned methods were used to prepare LSM-SDC and LSCF-SDC composite films onto YSZ substrates. A previous study performed by Ye and co-workers (Ye et al., 2007) demonstrated that the amount of ethyl cellulose in LSM-SDC slurries is restricted by cracking of the surface of the films. These authors observed that cracking took place when the amount of pore former reached 15 wt.% and is caused by the large amount of organics which evaporates during sintering. Therefore, in the present study, LSM-SDC slurries with a maximum of 10 wt.% ethyl cellulose were used. Fig. 12 shows FEG-SEM images of the porous structure of LSM-SDC composite

3400 O-H

Fig. 11. FTIR spectra for the as-prepared and calcined LSCF powders.

**5.1.1 Characterization of composite cathodes** 

 as-prepared calcined

HOH

)

1454 CO3 2-

> 1381 NO3 -

595 M-O

small particles.

10

15

20

25

30

Transmittance (%)

35

40

45

Fig. 12. FEG-SEM images of the surface of LSM-SDC composite cathodes containing different amounts of ethyl cellulose: (a) 0 wt.%, (b) 4 wt.%, (c) 8 wt.%, (d) 10 wt.%.

cathodes containing different amounts of ethyl cellulose. In these images, it is possible to observe the effects of the milling process and the addition of ethyl cellulose in the final microstructure of composite cathodes. The film obtained without addition of ethyl cellulose and without prior milling treatment of powders (Fig. 12a) depicted particle sizes of about 200 nm. On the other hand, in the films obtained with the powders previously milled for 24 h (Figure 12b-d) it can be observed that both particle agglomeration and pore size decrease as the ethyl cellulose content increases. Moreover, the particle size is lower than 200 nm. Typically, if the slurry contains 10 wt.% ethyl cellulose (Fig. 12d), the film exhibits a highly porous surface morphology and uniform pore structure, which are essential conditions for obtaining high-performance cathodes. The electrochemical evaluation of these composite cathodes are currently under way.

Fig. 13 shows cross-sectional SEM images of the LSM-SDC cathodes containing 0 and 10 wt.% ethyl cellulose after milling LSM and SDC powders. The film without addition of ethyl cellulose (Fig. 13a) is tightly adhered to the substrate and shows few pores. However, the film containing 10 wt.% ethyl cellulose (~ 10 μm thick) is also well adhered to the YSZ substrate and exhibits higher pore uniformity than that in Fig. 13a. Such a porous microstructure fulfils the need for a SOFC cathode, which possess high active surface area, while permitting rapid diffusion of oxygen through the porous cathode film.

Infrared Spectroscopy Techniques in the Characterization of SOFC Functional Ceramics 401

thick YSZ electrolyte. Nonetheless, voltage and power density values cannot be compared to literature data because of the thick (200 µm) electrolyte used as cell support, which is responsible for high ohmic polarization. The improvement in cell performance is attributed to the good catalytic activity of the novel LSCF-SDC composite film prepared from the

Results showed that LSM, SDC and LSCF powders were successfully synthesized by modified Pechini and microwave-assisted combustion methods. The milling process of the LSM-SDC composite powders and the addition of ethyl cellulose to the slurries seem to reduce particle agglomeration and pore size, that could improve the triple phase boundary (TPB) and, consequently, the electrochemical efficiency of the cathode. According to the results gathered herein, LSM-SDC films containing 10 wt.% ethyl cellulose (maximum concentration in order to avoid surface cracks) exhibited uniform and continuous pore structure and excellent adhesion to the YSZ substrate. The composite films produced presented porous microstructures desirable for the electrochemical reduction of the oxidant. Homogeneity and agglomeration absence or phase segregation were also characteristics

FTIR technique was successfully applied for powder characterization. Results were in agreement with XRD data, demonstrating that the FTIR is also a powerful tool for evaluating the crystalline structure evolution after a chemical synthesis. Moreover, the data could also indicate that water or hydroxyl groups still existed for some calcinated samples as well as

The authors acknowledge PPGCEM-UFRN, PPGQ-UFRN, PRH-ANP 14, CAPES (PRÓ-ENGENHARIAS and PDEE program – BEX 6775/10-1), CAPES-PROCAD, and CNPq

Atkinson, A.; Barnett, S.; Gorte, R. J.; Irvine, J. T. S.; McEvoy, A. J.; Mogensen, M.; Singhal, S.

Borovskikh, L.; Mazo, G.; Kemnitz, E. (2003). Reactivity of oxygen of complex cobaltates

Cela, B.; Macedo, D. A.; Souza, G. L.; Nascimento, R. M.; Martinelli, A. E.; Paskocimas, C. A.

Chen, K.; Lü, Z.; Ai, N.; Chen, X.; Hu, J.; Huang, X.; Su, W. (2007). Effect of SDC-impregnated

*of Power Sources*, Vol. 167, No. 1, (1 May 2007), pp. (84-89), ISSN 0378-7753

C.; Vohs J. (2004). Advanced anodes for high-temperature fuel cells. *Nature* 

La1−xSrxCoO3−δ and LaSrCoO4. *Solid State Sciences*, Vol. 5, No. 3, (23 March 2003),

(2009). Strontium-doped lanthanum manganite synthesis for solid oxide fuel cells cathode. *Journal of New Materials for Electrochemical Systems*, Vol. 12, No. 2-3, (1 April

LSM cathodes on the performance of anode-supported YSZ films for SOFCs. *Journal* 

physical adsorption of CO2, giving a qualitative information of purity of the samples.

(477294/2006-5, 502238/2007-0 and 554576/2010-4) for their financial support.

*Materials*, Vol. 3, pp. (17-27), ISSN 1476-1122

pp. (409–17), ISSN 1293-2558

2009), pp. (109-113), ISSN 1480-2422

powders obtained by different synthesis methods.

found in the composite films prepared.

**7. Acknowledgments** 

**8. References** 

**6. Conclusions** 

Fig. 13. Cross-sectional SEM images of LSM-SDC cathodes containing different amounts of ethyl cellulose: (a) 0 wt.% and (b) 10 wt.%.

#### **5.2 Characterization of a SOFC single cell**

A SEM micrograph of the cathode/electrolyte interface and preliminary results on the electrochemical activity of YSZ electrolyte-supported SOFCs containing Ni-YSZ anode and a LSCF-SDC composite cathode are shown in Fig. 14. As it can be seen in Fig. 14(a), the composite film not only has good adhesion to the electrolyte, but also possesses a porous microstructure which is required for the oxidant electrochemical reduction. It indicates that such a composite film can have a good performance as SOFC cathode. By the LSV technique, qualitative information about electrochemical activity of this SOFC was acquired. The power density curves (Fig. 14b) revealed that maximum power densities were 19, 26, 36 and 46 mW/cm2 at 800, 850, 900 and 950 ºC. It is possible to compare these first results with literature data and safely state that the LSCF-SDC cathode composite is qualitatively better than other plain standard materials or cathode composites already reported. It should also be mentioned that the result obtained at 800 ºC is similar to that reported by Muccillo et al (Muccillo et al., 2006) for a SOFC single cell with LSM-YSZ cathode, Ni-YSZ anode and 70 µm

Fig. 14. Electrolyte-supported SOFC: (a) SEM micrograph of the cathode (LSCF-SDC)/electrolyte (YSZ) interface, (b) I-V and power density curves at different temperatures.

thick YSZ electrolyte. Nonetheless, voltage and power density values cannot be compared to literature data because of the thick (200 µm) electrolyte used as cell support, which is responsible for high ohmic polarization. The improvement in cell performance is attributed to the good catalytic activity of the novel LSCF-SDC composite film prepared from the powders obtained by different synthesis methods.

## **6. Conclusions**

400 Infrared Spectroscopy – Materials Science, Engineering and Technology

Fig. 13. Cross-sectional SEM images of LSM-SDC cathodes containing different amounts of

A SEM micrograph of the cathode/electrolyte interface and preliminary results on the electrochemical activity of YSZ electrolyte-supported SOFCs containing Ni-YSZ anode and a LSCF-SDC composite cathode are shown in Fig. 14. As it can be seen in Fig. 14(a), the composite film not only has good adhesion to the electrolyte, but also possesses a porous microstructure which is required for the oxidant electrochemical reduction. It indicates that such a composite film can have a good performance as SOFC cathode. By the LSV technique, qualitative information about electrochemical activity of this SOFC was acquired. The power density curves (Fig. 14b) revealed that maximum power densities were 19, 26, 36 and 46 mW/cm2 at 800, 850, 900 and 950 ºC. It is possible to compare these first results with literature data and safely state that the LSCF-SDC cathode composite is qualitatively better than other plain standard materials or cathode composites already reported. It should also be mentioned that the result obtained at 800 ºC is similar to that reported by Muccillo et al (Muccillo et al., 2006) for a SOFC single cell with LSM-YSZ cathode, Ni-YSZ anode and 70 µm

Fig. 14. Electrolyte-supported SOFC: (a) SEM micrograph of the cathode (LSCF-SDC)/electrolyte (YSZ) interface, (b) I-V and power density curves at different

ethyl cellulose: (a) 0 wt.% and (b) 10 wt.%.

**5.2 Characterization of a SOFC single cell** 

temperatures.

Results showed that LSM, SDC and LSCF powders were successfully synthesized by modified Pechini and microwave-assisted combustion methods. The milling process of the LSM-SDC composite powders and the addition of ethyl cellulose to the slurries seem to reduce particle agglomeration and pore size, that could improve the triple phase boundary (TPB) and, consequently, the electrochemical efficiency of the cathode. According to the results gathered herein, LSM-SDC films containing 10 wt.% ethyl cellulose (maximum concentration in order to avoid surface cracks) exhibited uniform and continuous pore structure and excellent adhesion to the YSZ substrate. The composite films produced presented porous microstructures desirable for the electrochemical reduction of the oxidant. Homogeneity and agglomeration absence or phase segregation were also characteristics found in the composite films prepared.

FTIR technique was successfully applied for powder characterization. Results were in agreement with XRD data, demonstrating that the FTIR is also a powerful tool for evaluating the crystalline structure evolution after a chemical synthesis. Moreover, the data could also indicate that water or hydroxyl groups still existed for some calcinated samples as well as physical adsorption of CO2, giving a qualitative information of purity of the samples.

## **7. Acknowledgments**

The authors acknowledge PPGCEM-UFRN, PPGQ-UFRN, PRH-ANP 14, CAPES (PRÓ-ENGENHARIAS and PDEE program – BEX 6775/10-1), CAPES-PROCAD, and CNPq (477294/2006-5, 502238/2007-0 and 554576/2010-4) for their financial support.

## **8. References**


Infrared Spectroscopy Techniques in the Characterization of SOFC Functional Ceramics 403

Liu, S.; Qian, X.; Xiao. (2007). Synthesis and characterization of La0.8Sr0.2Co0.5Fe0.5O3±<sup>δ</sup>

*Technology*, Vol. 44, No. 3, (3 October 2007), pp. (187-193), ISSN 0928-0707 Liu, X.; Faguy, P. W.; Liu, M. (2002). In Situ Potential-Dependent FTIR Emission

Macedo, D. A.; Cela, B.; Martinelli, A. E.; Nascimento, R. M.; Melo, D. M. A.; Paskocimas, C.

Maia, A. O. G.; Meneses, C. T.; Menezes, A. S.; Flores, W. H.; Melo, D. M. A.; Sasaki, J. M.

McIntosh S.; Gorte R. J. (2004). Direct Hydrocarbon Solid Oxide Fuel Cells. *Chemical Reviews*, Vol. 104, No. 10, (21 July 2004), pp. (4845-4866), ISSN 0009-2665 Medeiros, A. M. L.; Miranda, M. A. R.; Menezes, A. S.; Jardim, P. M.; Silva, L. R. D.; Gouveia,

Minh, N. Q. & Takahashi T (1995). Science and Technology of Ceramic Fuel Cell, Elsevier

Minh, N. Q. (1993). Ceramic fuel cells. *Journal of the American Ceramic Society*, Vol. 76, No.

Minh, N. Q. (2004). Solid oxide fuel cell technology—features and applications. *Solid State Ionics*, Vol. 174, No. 1-4, (29 October 2004), pp. (271-277), ISSN 0167-2738 Molenda, J.; Swierczek, K.; Zajac, W. (2007). Functional materials for the IT-SOFC. *Journal of Power Sources*, Vol. 173, No. 2, (15 November 2007), pp. (657-670), ISSN 0378-7753 Muccillo, R.; Muccillo, E. N. S.; Fonseca, F. C.; França, Y. V.; Porfírio, T. C.; Florio, D. Z.;

*of Power Sources*, Vol. 156, No. 2, (1 June 2006), pp. (455-460), ISSN 0378-7753 Pechini, M. P. (1967). Method of preparing lead and alkaline earth titanates and niobates and

Resini, C.; Venkov, T.; Hadjiivanov, K.; Presto S.; Riani, P.; Marazza, R.; Ramis, G.; Busca, G.

Singhal, S. C. (2000). Advances in solid oxide fuel cell technology. *Solid State Ionics*, Vol. 135,

Berton, M. A. C.; Garcia, M. C. (2006). Development and testing of anodesupported solid oxide fuel cells with slurry-coated electrolyte and cathode. *Journal* 

coating method using the same to form a capacitor. US Patent No. 3,330,697 (11 July

(2009). An FTIR study of the dispersed Ni species on Ni-YSZ catalysts. *Applied Catalysis A: General,* Vol. 353, No. 1, (January 2009), pp. (137-143), ISSN 0926-860X Shao, J.; Tao, Y.; Wang, J.; Xu, C.; Wang, W. G. (2009). Investigation of precursors in the

preparation of nanostructured La0.6Sr0.4Co0.2Fe0.8O3−δ via a modified combined complexing method. *Journal of Alloys and Compounds,* Vol. 484, No. 1-2, (18

Science, ISBN 0-444-89568-X, Amsterdam, The Netherlands.

September 2006), pp. (3729-3733), ISSN 0022-3093

563, (March 1993), pp. (563–588), ISSN 1551-2916

September 2009), pp. (263-267), ISSN 0925-8388

No. 1-4, (1 November 2000), pp. (305-313), ISSN 0167-2738

1298), ISSN 1945-7111 (online)

pp. (399-406), ISSN 1422-6375

2422

1967)

nanopowders by microwave assisted sol–gel route, *Journal of Sol-Gel Science and* 

Spectroscopy. *Journal of The Electrochemical Society*, Vol. 149, No. 10, 2002, pp. (1293-

A.; Rabelo, A. A. (2009) Synthesis, processing and characterization of ZrO2-8Y2O3, ZrO2-8CeO2 and La0.78Sr0.22MnO3 powders. *Journal of New Materials for Electrochemical Systems*, Vol. 12, No. 2-3, (1 April 2009), pp. (103-108), ISSN 1480-

(2006). Synthesis and X-ray structural characterization of NiO nanoparticles obtained through gelatin. *Journal of Non-Crystalline Solids*, Vol. 352, No. 32-35, (15

S. T.; Sasaki, J. M. (2004). Synthesis and Characterization of Cr2O3 Nanoparticles Obtained by Gelatin. *Journal of Metastable and Nanocrystalline Materials*, Vol. 20-21,


Chen, M.; Kim, B. H.; Xu, Q.; Nam, O. J.; Ko, J. H. (2008). Synthesis and performances of Ni–

Chen, W.; Li, F.; Yu, J. (2006). Combustion synthesis and characterization of nanocrystalline

Conceição, L.; Silva, A. M.; Ribeiro, N. F. P.; Souza, M. M. V. M. (2011). Combustion

Conceição, L.; Silva, R. C. B.; Ribeiro, N. F. P.; Souza, M. M. V. M. (2009). Influence of the

Escobedo, C. A. C.; Saldaña, J. M.; Miró, A. M. B.; Jesús, F. S. (2008). Determination of

*Power Sources*, Vol. 180, No. 1, (15 May 2008), pp. (209-214), ISSN 0378-7753 Gosnet, A. M. H.; Koubaa, M.; Santander-Syro, A. F.; Lobo, R. P. S. M.; Lecoeur, P.; Mercey,

Grossin, D.; Noudem, J. G. (2004). Synthesis of fine La0.8Sr0.2MnO3 powder by different ways. *Solid State Sciences*, Vol. 6, No. 9, (September 2004), pp. 939-944, ISSN 1293-2558 Guo, H.; Iqbal, G.; Kang, B. S. (2010). Development of an In Situ Surface Deformation and

Guo, R. S.; Wei, Q. T.; Li, H. L.; Wang, F. H. (2006). Synthesis and properties of

Ivers-Tiffee, E.; Weber, A.; Herbstritt, D. (2001). Materials and technologies for SOFC-

Jiang, S. P. (2003). Issues on development of (La,Sr)MnO3 cathode for solid oxide fuel cells.

Kim, K. H.; Gu, J. Y.; Choi, H. S.; Park, J. W.; Noh, T. W. (1996), Frequency Shifts of the

Konysheva, E.; Irvine, J. T. S.; Besmehn, A. (2009). Crystal structure, thermochemical

Lin, H.; Ding, C.; Sato, K.; Tsutai, Y.; Ohtaki, H.; Iguchi, M.; Wada, C.; Hashida, T. (2008).

15, (November 2008), pp. (2947-2953), ISSN 0955-2219

No. 1, (January 2006), pp. (57–62), ISSN 167-577X

(December 2009), pp. (1417-1423), ISSN 1044-5803

314), ISSN 0025-5408

5), ISSN 1098-0121

0378-7753

2010), pp. (55-62), ISSN 1744-7402

2006), pp. 261-265, ISSN 0167-577X

2001), pp. (1805-1811), ISSN 0955-2219

pp. (778-783), ISSN 0167-2738

August 1996), pp. (1877 – 1880), ISSN 0031-9007

February 2008), pp. (73-76), ISSN 0921-5107

SDC cermets for IT-SOFC anode. *Journal of the European Ceramic Society,* Vol. 28, No

CeO2-based powders via ethylene glycol–nitrate process. *Material Letters*, Vol. 60,

synthesis of La0.7Sr0.3Co0.5Fe0.5O3 (LSCF) porous materials for application as cathode in IT-SOFC. *Materials Research Bulletin*, Vol. 46, No. 2, (February 2011), pp. (308-

synthesis method on the porosity, microstructure and electrical properties of La0.7Sr0.3MnO3 cathode materials. *Materials Characterization,* Vol. 60, No. 12,

strontium and lanthanum zirconates in YPSZ–LSM mixtures for SOFC. *Journal of* 

B. (2008). Metallic nature of strained thin single-crystal La2/3Sr1/3MnO3 films. *Physical Review B*, Vol. 78, No. 11, (September 2008), Article Number 115118, pp. (1-

Temperature Measurement Technique for a Solid Oxide Fuel Cell Button Cell. *International Journal of Applied Ceramic Technology*, Vol. 7, No. 1, (January/February

La0.7Sr0.3MnO3 cathode by gel combustion. *Materials Letters*, Vol. 60, No. 2, (January

components. *Journal of the European Ceramic Society*, Vol. 21, No. 10-11, (22 August

*Journal of Power Sources,* Vol. 124, No. 2, (24 November 2003), pp. (390-402), ISSN

Internal Phonon Modes in La0.7Ca0.3MnO3. *Physical Review Letters*, Vol. 77, No. 9, (26

stability, electrical and magnetic properties of the two-phase composites in the La0.8Sr0.2MnO3 ± δ–CeO2 system. *Solid State Ionics*, Vol. 180, No. 11-13, (22 June 2009),

Preparation of SDC electrolyte thin films on dense and porous substrates by modified sol–gel route. *Materials Sciences and Engineering B*, Vol. 148, No. 1-3, (25


**21** 

*México* 

**Infrared Spectroscopy of** 

*Universidad Autónoma de Ciudad Juárez, Instituto de Ingeniería y Tecnología,* 

Perla E. García Casillas,

*Ciudad Juárez Chihuahua* 

**Functionalized Magnetic Nanoparticles** 

Claudia A. Rodriguez Gonzalez and Carlos A. Martínez Pérez

Nanotechnology development has allowed that nanomaterials can be used in biomedical applications, and nanometer sized objects can interact with biological entities like cells, virus, protein, enzyme, etc. For this reason, many research projects has been focused in the development of nanosystems, nanoparticles and nanodevices for this applications. This area is relatively new, according to the ISI web of knowledge, the publications of the nanoparticles for biomedical applications started on 2000 year, and since that time they have increased exponentially (Figure 1). The nanoparticles (NPs) used for biomedical purposes generally include zero-dimensional nanospheres and one-dimensional nanowires and

Fig. 1. Trends of Nanoparticles (NPs) in Biomedical application, information extracted from

NPs in Biomedical applications

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Figure 2, extracted from ISI Web of Knowledge, shows the publications related to gold, silver, iron and magnetic nanoparticles for biomedical application. Magnetic nanoparticles

**1. Introduction** 

nanotubes.

**Publication number**

ISI Web of Knowledge.


## **Infrared Spectroscopy of Functionalized Magnetic Nanoparticles**

Perla E. García Casillas, Claudia A. Rodriguez Gonzalez and Carlos A. Martínez Pérez *Universidad Autónoma de Ciudad Juárez, Instituto de Ingeniería y Tecnología, Ciudad Juárez Chihuahua México* 

#### **1. Introduction**

404 Infrared Spectroscopy – Materials Science, Engineering and Technology

Steele B. C. H.; Heinzel, A. (2001). Materials for fuel-cell technologies. *Nature*, Vol. 414, (15

Steele, B. C. H. (2000). Appraisal of Ce1−yGdyO2−y/2 electrolytes for IT-SOFC operation at

Stuart, B. H (2004). Infrared Spectroscopy: Fundamentals and Applications (Analytical Techniques in the Sciences (AnTs) \*), John Wiley and Son's Ltd. England. Tsai, T.; Barnett, S. A. (1997). Effect of LSM-YSZ cathode on thin-electrolyte solid oxide fuel

Vargas, R. A.; Chiba, R. ; Andreoli, M.; Seo, E. S. M. (2008). Synthesis and characterization of

Wang, K.; Ran, R.; Shao, Z. (2007). Methane-fueled IT-SOFCs with facile in situ inorganic

*Power Sources*, Vol. 170, No. 2, (10 July 2007), pp. (251-258), ISSN 0378-7753 Xu, X.; Cao, C.; Xia, C.; Peng, D. (2009). Electrochemical performance of LSM–SDC

Xu, X.; Jiang, Z.; Fan, X.; Xia, C. (2006). LSM–SDC electrodes fabricated with an ion-

Yang, K.; Shen, J. H.; Yang, K. Y.; Hung, I. M.; Fung, K. Z.; Wang, M. C. (2007).

Ye, F., Wang, Z.; Weng W.; Cheng, K.; Song, C.; Du, P.; Sheng, G.; Han, G. (2007). Spin-

Zhang, L.; Zhao, F.; Peng, R.; Xia, C. (2008). Effect of firing temperature on the performance

Vol. 177, No. 19-25, (15 October 2006), pp. (2113-2117), ISSN 0167-2738 Xu, X.; Xia, C.; Xiao, G.; Peng, D. (2005). Fabrication and performance of functionally graded

500°C. *Solid State Ionics*, Vol. 129, No. 1-4, (April 2000), pp. (95-110), ISSN 0167-2738

cell performance. *Solid State Ionics*, Vol. 93, No. 3-4, (January 1997), pp. (207-217),

La1-xSrxMnO3± δ and La1-xSrxCo1-yFeyO3-δ used as cathode in solid oxide fuel cells. *Cerâmica*, Vol. 54, No. 331, (July - September 2008), pp. (366-372), ISSN 0366-6913 Vashook, V. V.; Ullmann, H.; Olshevskaya, O. P.; Kulik, V. P.; Lukashevich, V. E.;

Kokhanovskij, L. V. (2000). Composition and electrical conductivity of some cobaltates of the type La2−xSrxCoO4.5−x/2±δ. *Solid State Ionics*, Vol. 138, No. 1-2, (4

templating synthesized mesoporous Sm0.2Ce0.8O1.9 as catalytic layer. *Journal of* 

electrodes prepared with ion-impregnated LSM. *Ceramics International*, Vol. 35, No.

impregnating process for SOFCs with doped ceria electrolytes. *Solid State Ionics*,

cathodes for IT-SOFCs based on doped ceria electrolytes. *Solid State Ionics*, Vol. 176,

Characterization of the yttria-stabilized zirconia thin film electrophoretic deposited on La0.8Sr0.2MnO3 substrate. *Journal of Alloys and Compounds,* Vol. 436, No. 1-2, (14

coating derived LSM-SDC films with uniform pore structure. *Thin Solid Films*, Vol.

of LSM–SDC cathodes prepared with an ion-impregnation method. *Solid State Ionics*, Vol. 179, No. 27-32, (30 September 2008), pp. (1553-1556), ISSN 0167-2738 Zhang, Y.; Huang, X.; Lu, Z.; Liu, Z.; Ge, X.; Xu, J.; Xin, X.; Sha, X.; Su, X. (2007). Ni–

Sm0.2Ce0.8O1.9 anode-supported YSZ electrolyte film and its application in solid oxide fuel cells. *Journal of Alloys and Compounds*, Vol. 428, No. 1-2, (31 January 2007),

November 2001), pp. (345-352), ISSN 0028-0836

December 2000), pp. (99–104), ISSN 0167-2738

6, (August 2009), pp. (2213-2218), ISSN 0272-8842

June 2007), pp. (351-357), ISSN 0925-8388

pp. (302-306), ISSN 0925-8388

No. 17-18, (31 May 2005), pp. (1513-1520), ISSN 0167-2738

516, No. 16, (30 June 2008), pp. (5206-5209), ISSN 0040-6090

ISSN 0167-2738

Nanotechnology development has allowed that nanomaterials can be used in biomedical applications, and nanometer sized objects can interact with biological entities like cells, virus, protein, enzyme, etc. For this reason, many research projects has been focused in the development of nanosystems, nanoparticles and nanodevices for this applications. This area is relatively new, according to the ISI web of knowledge, the publications of the nanoparticles for biomedical applications started on 2000 year, and since that time they have increased exponentially (Figure 1). The nanoparticles (NPs) used for biomedical purposes generally include zero-dimensional nanospheres and one-dimensional nanowires and nanotubes.

Fig. 1. Trends of Nanoparticles (NPs) in Biomedical application, information extracted from ISI Web of Knowledge.

Figure 2, extracted from ISI Web of Knowledge, shows the publications related to gold, silver, iron and magnetic nanoparticles for biomedical application. Magnetic nanoparticles

Infrared Spectroscopy of Functionalized Magnetic Nanoparticles 407

thermal decomposition, etc. With this variety of methods particles with different morphologies such as spheres, rods, wires and tubes has been obtained (Palla et al, 1999; Joralemon et al, 2005; Terrazas et al, 2010). However, the coprecipitation method is still the most popular method for their simplicity and easy way to manipulate the size and morphology of the particles by the use of templates, besides being a method whose raw

Recent publications have been emphasized on the particle size control of magnetite nanoparticles, because under some critical value, the material exhibits a superparamagnetic behavior, this means that there is no hysteresis in the magnetization curves, which implies that the retentivity and coercivity are close to zero (Cullity and Graham, 2009). Biomedical applications involve strict requirements on particle size, and it can be by using a chemical coprecipitation through the control of nucleation and growth process. Magnetite nanoparticles obtained by chemical coprecipitation method are produced by the precipitation of divalent (Fe+2) and trivalent (Fe+3) iron salts in an alkaline medium. The size and the number of nucleus are influenced by the alkaline medium and the addition velocity, which results in a nucleation and growing process; a fast nucleation will form high concentration of nuclei and small particles, while a slow nucleation will form low nuclei

materials are relatively inexpensive.

concentration generating larger nanoparticles (figure 3).

Fig. 3. Schematic representation of nucleation and growth kinetics.

overall chemical equation (Cornell et al, 2003; Gnanaprakash et al, 2007):

As mentioned above, chemical co-precipitation consists in the precipitation of divalent (Fe+2) and trivalent (Fe+3) iron salts in an alkaline medium, maintaining a molar ratio of 1:2, by using ammonium hydroxide, ammonia or some other alkaline solution to increase pH reaction that is required to magnetite formation. Commonly the addition of alkaline solution to divalent and trivalent iron solution is made slowly, drop by drop (titration) under vigorous agitation using a magnetic agitator. The initial solution of divalent and trivalent iron cations had acidic pH and after the titration is close to 12, a black precipitate is formed which indicates that the reaction has been completed. The chemical reaction that takes place during magnetite formation from iron salts solutions by increasing the pH can be represented in the following

**2.1 Chemical coprecipitation** 

Fig. 2. Trends of silver, gold, iron and magnetic nanoparticles used in Biomedical applications, information extracted from ISI Web of Knowledge

are at the forefront as the most promising materials for clinical diagnostic and therapeutic applications. Magnetic nanoparticles (MNPs) are widely used for labeling and manipulating biomolecules, targeting drugs and genes, magnetic resonance imaging, as well as hyperthermia treatment (Varadan et al., 2008; Cornell and Schwertmann, 2003). Magnetite Nanoparticles (Fe3O4) are the most used magnetic material for biomedical applications because they have a high enough saturation magnetization to allow its manipulation with an external field, superparamagnetic behavior and ability to bond with different molecules to surface functionalized (Cullity and Graham, 2009; Neuberger et al., 2005). In biomedical applications, the characteristics of the magnetite nanoparticles have a significant advantage when they interact with biological molecules, therefore, many methods of synthesis have been developed in order to control surface morphology, particle size, particle distribution, and chemical stability among others (An-Hui et al., 2007;. Gao and Gu, 2009).

#### **2. Magnetic nanoparticles**

Many magnetic nanoparticles such as magnetite, strontium and cobalt ferrites, lantaniumzinc ferrites, niquel, iron and some compounds with a rare earth like SmCo5 have been development (Pankhurst et al., 2003). Magnetite, Fe3O4 is the magnetic material most used in biomedical application due to its several interesting properties such as great chemical stability, low toxicity, and its magnetic saturation mentioned above for being manipulated with an external field, biocompatibility and the heating ability in presence of a field, which made it an interesting candidate for hyperthermia treatment (Sun et al., 2004) for this reason in recent years, much effort has been focused in the design and controlled synthesis of this material with certain shape and particle size. Many methods for synthesis of magnetite nanoparticles have been developed like co-precipitation, microemulsion, sol-gel, sputtering,

Fig. 2. Trends of silver, gold, iron and magnetic nanoparticles used in Biomedical

are at the forefront as the most promising materials for clinical diagnostic and therapeutic applications. Magnetic nanoparticles (MNPs) are widely used for labeling and manipulating biomolecules, targeting drugs and genes, magnetic resonance imaging, as well as hyperthermia treatment (Varadan et al., 2008; Cornell and Schwertmann, 2003). Magnetite Nanoparticles (Fe3O4) are the most used magnetic material for biomedical applications because they have a high enough saturation magnetization to allow its manipulation with an external field, superparamagnetic behavior and ability to bond with different molecules to surface functionalized (Cullity and Graham, 2009; Neuberger et al., 2005). In biomedical applications, the characteristics of the magnetite nanoparticles have a significant advantage when they interact with biological molecules, therefore, many methods of synthesis have been developed in order to control surface morphology, particle size, particle distribution, and chemical stability among others (An-Hui et al.,

2004 2005 2006 2007 2008 2009 2010 2011 2012

Many magnetic nanoparticles such as magnetite, strontium and cobalt ferrites, lantaniumzinc ferrites, niquel, iron and some compounds with a rare earth like SmCo5 have been development (Pankhurst et al., 2003). Magnetite, Fe3O4 is the magnetic material most used in biomedical application due to its several interesting properties such as great chemical stability, low toxicity, and its magnetic saturation mentioned above for being manipulated with an external field, biocompatibility and the heating ability in presence of a field, which made it an interesting candidate for hyperthermia treatment (Sun et al., 2004) for this reason in recent years, much effort has been focused in the design and controlled synthesis of this material with certain shape and particle size. Many methods for synthesis of magnetite nanoparticles have been developed like co-precipitation, microemulsion, sol-gel, sputtering,

applications, information extracted from ISI Web of Knowledge

Silver Gold Iron

magnetic

2007;. Gao and Gu, 2009).

0

50

100

150

**Publication number**

200

250

300

**2. Magnetic nanoparticles** 

thermal decomposition, etc. With this variety of methods particles with different morphologies such as spheres, rods, wires and tubes has been obtained (Palla et al, 1999; Joralemon et al, 2005; Terrazas et al, 2010). However, the coprecipitation method is still the most popular method for their simplicity and easy way to manipulate the size and morphology of the particles by the use of templates, besides being a method whose raw materials are relatively inexpensive.

Recent publications have been emphasized on the particle size control of magnetite nanoparticles, because under some critical value, the material exhibits a superparamagnetic behavior, this means that there is no hysteresis in the magnetization curves, which implies that the retentivity and coercivity are close to zero (Cullity and Graham, 2009). Biomedical applications involve strict requirements on particle size, and it can be by using a chemical coprecipitation through the control of nucleation and growth process. Magnetite nanoparticles obtained by chemical coprecipitation method are produced by the precipitation of divalent (Fe+2) and trivalent (Fe+3) iron salts in an alkaline medium. The size and the number of nucleus are influenced by the alkaline medium and the addition velocity, which results in a nucleation and growing process; a fast nucleation will form high concentration of nuclei and small particles, while a slow nucleation will form low nuclei concentration generating larger nanoparticles (figure 3).

Fig. 3. Schematic representation of nucleation and growth kinetics.

#### **2.1 Chemical coprecipitation**

As mentioned above, chemical co-precipitation consists in the precipitation of divalent (Fe+2) and trivalent (Fe+3) iron salts in an alkaline medium, maintaining a molar ratio of 1:2, by using ammonium hydroxide, ammonia or some other alkaline solution to increase pH reaction that is required to magnetite formation. Commonly the addition of alkaline solution to divalent and trivalent iron solution is made slowly, drop by drop (titration) under vigorous agitation using a magnetic agitator. The initial solution of divalent and trivalent iron cations had acidic pH and after the titration is close to 12, a black precipitate is formed which indicates that the reaction has been completed. The chemical reaction that takes place during magnetite formation from iron salts solutions by increasing the pH can be represented in the following overall chemical equation (Cornell et al, 2003; Gnanaprakash et al, 2007):

Infrared Spectroscopy of Functionalized Magnetic Nanoparticles 409

Fig. 4. Difference between conventional coprecipitation (a) and coprecipitation with fast

Fe+2 y Fe+3 solution

 NH3 + H2O NH4+ + OH- (6) By the use of urea in the chemical coprecipitation with reflux and aging conditions method, the speed of pH increment is considerably slower than the one observed in the slow injection and rapid injection co-precipitation methods. The condition of slow and uniform pH increment and the reacting time of 20 hours under favorable conditions for magnetite formation, affect the nucleation and particle growth process making it very slow, which is reflected in the formation of bigger particles than those promoted in slow injection and

(a) (b)

NH4OH

Fe+2 y Fe+3 solution is added quickly

The difference of these methodologies lies in the particle size and its distribution as is shown in figure 5. Magnetite obtained by coprecipitation with reflux and aging shows higher particle size and a wide particle size distribution. According to these results, magnetite obtained by coprecipitation with fast injection shows the smallest average particle size and the closest particle size distribution; differing from magnetite obtained by common

The magnetite obtained using these three methodologies have a spherical particle shape with a superparamagnetic behavior, and their saturation magnetization is influenced by the particle size. Values of 55.9, 64.3 and 78.2 emu/g were obtained from nanoparticles with an average particle size of 16 nm, 27 nm and 200 nm respectively. The saturation magnetization increases when the particle sizes are larger. Hysteresis loops of the synthesized

Magnetite structure is an inverse spinel with a face center cubic unit based on 32 O-2- ions with a regularly cubic close packed along the [111] direction. There are eight formula units per unit cell. Magnetite differs from other iron oxides in that it contains both divalent and trivalent iron. Its formula is written as Fe III[Fe II Fe III]O4 and the brackets denote octahedral sites, tetrahedral Fe spins are directed antiparallel to octahedral Fe 3+ and Fe 2+ spins so

injection (b).

rapid injection methods.

NH4OH added drop by drop

coprecipitation, Table 1.

nanoparticles are shown in figure 6.

$$2\text{Fe}^{\ast 3} + \text{Fe}^{\ast 2} + 8\text{OH}^{\ast 1} \xrightarrow{\rightharpoonup} \text{FeO}\text{Fe}\_2\text{O}\_3 + 4\text{H}\_2\text{O} \tag{1}$$

In general, the solubility of trivalent iron oxide (Fe+3) is smaller than the one observed on divalent iron oxides (Fe+2). The trivalent iron hydrolyzes and forms hydroxide species. The hydrolysis can be induced by heating up the solution. The complete hydrolysis corresponds to the formation of a trivalent iron oxide-hydroxide and it is represented according to the following chemical reaction:

$$(\text{Fe(H}\_2\text{O)}\_6)^{\*3} \longrightarrow \text{FeCOH} + 3\text{H}^+ + 4\text{H}\_2\text{O} \tag{2}$$

The divalent iron cation in solution (Fe+2) reacts to form the divalent iron oxide in basic conditions (presence of hydroxyl ion OH- ), which is presented in equation 3:

$$\text{Fe}^{\bullet} \quad \text{+} \quad \text{2OH} \cdot \xrightarrow{} \text{Fe(OH)}\_{2} \tag{3}$$

Under the reaction conditions, divalent iron hydroxide and trivalent iron oxide-hydroxide species were likely to be formed. This being established, it is suggested that the following chemical reaction mechanism occurred: trivalent iron cation hydrolyzes forming (FeOOH) as pH increases; under alkaline conditions divalent iron cation forms Fe(OH)2. Both chemical species reacted to each other at pH values of around 10 to 11, forming magnetite according to equation 4:

$$\text{2FeOOH} + \text{Fe(OH)}\_2 \xrightarrow{\text{---} \longrightarrow \text{ Fe}\_3\text{O}\_4 + \text{ 2H}\_2\text{O}} \tag{4}$$

#### **2.2 Chemical coprecipitation with fast injection**

The chemical coprecipitation with fast injection differs from conventional coprecipitation in the speed at pH of the reaction pH solution is increased; in order to favor magnetite formation abruptly. Divalent and trivalent iron salt solutions have an initial pH of 0 to 1. On the conventional coprecipitation method, the pH of the solution is increased by the addition of an alkaline solution drop by drop, which is considered slow speed; while on the rapid injection method, the pH of the solution is increased by adding the salt solution directly to ammonium hydroxide solution, speed to be considered rapid and explosive. The difference between both methods is schematically shown in figure 4.

#### **2.3 Chemical coprecipitation with reflux and aging conditions**

A trivalent iron solution is placed into a bowl flask and heated up to 80 ºC under refluxing conditions for a period of time of 2 hours. A precipitate is formed and separated from supernatant. Trivalent iron cation is hydrolyzed due to an increment of temperature promoting the hydrolysis and forming a trivalent iron oxide-hydroxide (FeOOH). After the 2 hours of hydrolysis reaction, a yellowish precipitate is obtained.

Another solution is prepared with divalent iron and urea. This solution is mixed with the previous precipitate and heated up to 90ºC-96ºC for 20 hours under refluxing conditions. The required pH condition is obtained through the slow decomposition of the urea when the temperature increases above 90ºC, this condition will increase the pH uniformly favoring a more slower nucleation in the solution (Terrazas et al., 2010).

$$\text{(NH}\_2\text{)}\_2\text{CO} + \text{H}\_2\text{O} \quad \begin{array}{c} \longrightarrow \text{ 2NH}\_3 \text{ + CO}\_2 \end{array} \tag{5}$$

 2Fe+3 + Fe+2 + 8OH-1 FeO.Fe2O3 + 4H2O (1) In general, the solubility of trivalent iron oxide (Fe+3) is smaller than the one observed on divalent iron oxides (Fe+2). The trivalent iron hydrolyzes and forms hydroxide species. The hydrolysis can be induced by heating up the solution. The complete hydrolysis corresponds to the formation of a trivalent iron oxide-hydroxide and it is represented according to the

 (Fe(H2O)6)+3 FeOOH + 3H+ + 4H2O (2) The divalent iron cation in solution (Fe+2) reacts to form the divalent iron oxide in basic

Under the reaction conditions, divalent iron hydroxide and trivalent iron oxide-hydroxide species were likely to be formed. This being established, it is suggested that the following chemical reaction mechanism occurred: trivalent iron cation hydrolyzes forming (FeOOH) as pH increases; under alkaline conditions divalent iron cation forms Fe(OH)2. Both chemical species reacted to each other at pH values of around 10 to 11, forming magnetite

The chemical coprecipitation with fast injection differs from conventional coprecipitation in the speed at pH of the reaction pH solution is increased; in order to favor magnetite formation abruptly. Divalent and trivalent iron salt solutions have an initial pH of 0 to 1. On the conventional coprecipitation method, the pH of the solution is increased by the addition of an alkaline solution drop by drop, which is considered slow speed; while on the rapid injection method, the pH of the solution is increased by adding the salt solution directly to ammonium hydroxide solution, speed to be considered rapid and explosive. The difference

A trivalent iron solution is placed into a bowl flask and heated up to 80 ºC under refluxing conditions for a period of time of 2 hours. A precipitate is formed and separated from supernatant. Trivalent iron cation is hydrolyzed due to an increment of temperature promoting the hydrolysis and forming a trivalent iron oxide-hydroxide (FeOOH). After the

Another solution is prepared with divalent iron and urea. This solution is mixed with the previous precipitate and heated up to 90ºC-96ºC for 20 hours under refluxing conditions. The required pH condition is obtained through the slow decomposition of the urea when the temperature increases above 90ºC, this condition will increase the pH uniformly

(NH2)2CO + H2O 2NH3 + CO2 (5)

), which is presented in equation 3:

2FeOOH + Fe(OH)2 Fe3O4 + 2H2O (4)

Fe(OH)2 (3)

following chemical reaction:

according to equation 4:

conditions (presence of hydroxyl ion OH-

Fe+2 + 2OH-

**2.2 Chemical coprecipitation with fast injection** 

between both methods is schematically shown in figure 4.

**2.3 Chemical coprecipitation with reflux and aging conditions** 

2 hours of hydrolysis reaction, a yellowish precipitate is obtained.

favoring a more slower nucleation in the solution (Terrazas et al., 2010).

Fig. 4. Difference between conventional coprecipitation (a) and coprecipitation with fast injection (b).

$$\text{NH}\_3 \quad \text{+} \quad \text{H}\_2\text{O} \quad \begin{array}{c} \longrightarrow \\ \longrightarrow \end{array} \text{NH}\_4\text{\*} \quad \text{+} \quad \text{OH} \cdot \tag{6}$$

By the use of urea in the chemical coprecipitation with reflux and aging conditions method, the speed of pH increment is considerably slower than the one observed in the slow injection and rapid injection co-precipitation methods. The condition of slow and uniform pH increment and the reacting time of 20 hours under favorable conditions for magnetite formation, affect the nucleation and particle growth process making it very slow, which is reflected in the formation of bigger particles than those promoted in slow injection and rapid injection methods.

The difference of these methodologies lies in the particle size and its distribution as is shown in figure 5. Magnetite obtained by coprecipitation with reflux and aging shows higher particle size and a wide particle size distribution. According to these results, magnetite obtained by coprecipitation with fast injection shows the smallest average particle size and the closest particle size distribution; differing from magnetite obtained by common coprecipitation, Table 1.

The magnetite obtained using these three methodologies have a spherical particle shape with a superparamagnetic behavior, and their saturation magnetization is influenced by the particle size. Values of 55.9, 64.3 and 78.2 emu/g were obtained from nanoparticles with an average particle size of 16 nm, 27 nm and 200 nm respectively. The saturation magnetization increases when the particle sizes are larger. Hysteresis loops of the synthesized nanoparticles are shown in figure 6.

Magnetite structure is an inverse spinel with a face center cubic unit based on 32 O-2- ions with a regularly cubic close packed along the [111] direction. There are eight formula units per unit cell. Magnetite differs from other iron oxides in that it contains both divalent and trivalent iron. Its formula is written as Fe III[Fe II Fe III]O4 and the brackets denote octahedral sites, tetrahedral Fe spins are directed antiparallel to octahedral Fe 3+ and Fe 2+ spins so

Infrared Spectroscopy of Functionalized Magnetic Nanoparticles 411

b a c


Magnetic field (Oe)

Fig. 6. Hysteresis loops of magnetite obtained by common coprecipitation (a), with a fast

that the Fe 3+ moments cancel, leaving a spontaneous magnetization equivalent to one Fe 2+ moment per molecule, eight tetrahedral sites are occupied by trivalent iron, and the divalent and trivalent cations occupies the sixteen octahedral sites (Cornell and Schwertmann, 2003). X-ray diffraction is the most widely used technique to determine the crystalline structure of a material. However in the case of magnetite it can be confusing because the magnetite has the same crystalline structure that maghemite, but this one has a interstitial voids, therefore by using XRD is not conclusive. The difference between the two materials is that some of the interstitial atomic positions of the maghemite are not fully occupied, and consecuently having atomic holes. In the case of magnetite, the infrared spectroscopy is very useful because this technique arises as a result of divalent and trivalent cations interaction with electromagnetic radiation, this interaction involves excitation for vibration or rotation of molecules in their ground electronic state, and they are associated with stretching deformation of the interatomic bonds and bending deformation of the interbond angles.

Infrared spectra of the magnetite shows the chareacteristic bands at 590 and 450cm-1 approximately due to the Fe-O bond in tetrahedrical and octahedrical positions. Figure 7 shows the infrared spectra of the magnetite with a different particle size, the band at 600 cm-1 approximately is broadening when the particle size decreases. According to Nasrazadani (1993) this effect indicates an increment of cation vacancy in the lattice, this behavior corroborated with the decreased value of the lattice parameter, which is shown in table 2. These values are minor than the lattice parameter of defect free magnetite (8.396 A ), this small reduction is assumed to be due to the prevalence of a small amount of cation


deficiency.

injection (b) and aging and reflux(c)

FTIR spectroscopy provides a fast mean of identification.




0

20

Magnetization (emu/g)

40

60

80

Fig. 5. Comparison of magnetite particle sizes and their distribution obtained by common coprecipitation (•), coprecipitation with fast injection (▲) and coprecipitation with reflux and aging (■)


Table 1. Average particle size of magnetite obtained by different methodologies

Fig. 5. Comparison of magnetite particle sizes and their distribution obtained by common coprecipitation (•), coprecipitation with fast injection (▲) and coprecipitation with reflux

injection

27.6 16.2 206.9

8.2 4.4 58.9

Table 1. Average particle size of magnetite obtained by different methodologies

Coprecipitation with fast

0 50 100 150 200 250 300 350 400 450

**Particle size (nm)**

Coprecipitation with aging and reflux

and aging (■)

**Frecuency**

Average particle size (nm)

Standard Deviation (nm)

Common coprecipitation

Fig. 6. Hysteresis loops of magnetite obtained by common coprecipitation (a), with a fast injection (b) and aging and reflux(c)

that the Fe 3+ moments cancel, leaving a spontaneous magnetization equivalent to one Fe 2+ moment per molecule, eight tetrahedral sites are occupied by trivalent iron, and the divalent and trivalent cations occupies the sixteen octahedral sites (Cornell and Schwertmann, 2003). X-ray diffraction is the most widely used technique to determine the crystalline structure of a material. However in the case of magnetite it can be confusing because the magnetite has the same crystalline structure that maghemite, but this one has a interstitial voids, therefore by using XRD is not conclusive. The difference between the two materials is that some of the interstitial atomic positions of the maghemite are not fully occupied, and consecuently having atomic holes. In the case of magnetite, the infrared spectroscopy is very useful because this technique arises as a result of divalent and trivalent cations interaction with electromagnetic radiation, this interaction involves excitation for vibration or rotation of molecules in their ground electronic state, and they are associated with stretching deformation of the interatomic bonds and bending deformation of the interbond angles. FTIR spectroscopy provides a fast mean of identification.

Infrared spectra of the magnetite shows the chareacteristic bands at 590 and 450cm-1 approximately due to the Fe-O bond in tetrahedrical and octahedrical positions. Figure 7 shows the infrared spectra of the magnetite with a different particle size, the band at 600 cm-1 approximately is broadening when the particle size decreases. According to Nasrazadani (1993) this effect indicates an increment of cation vacancy in the lattice, this behavior corroborated with the decreased value of the lattice parameter, which is shown in table 2. These values are minor than the lattice parameter of defect free magnetite (8.396 A ), this small reduction is assumed to be due to the prevalence of a small amount of cation deficiency.

Infrared Spectroscopy of Functionalized Magnetic Nanoparticles 413

**Co0.16 Fe2.84O4**

**Co0.08 Fe2.92O4**

**Co0.04 Fe2.96O4**

 **% Transmitance**

**Co0.02 Fe2.98O4**

Fig. 8. Infrared spectra of cobalt doped magnetite.

**3. Functionalization of magnetic nanoparticles** 

most used coating materials for magnetite nanoparticles.

**Fe3O4**

4000 3500 3000 2500 2000 1500 1000 700 600 500 400 **Wavenumber (cm-1**)

One of the most important aspects of the nanoparticles for biomedical applications is the surface preparation of the nanoparticles in order to improve their biocompatibility with biological entities and provide chemical stability. The nanoparticles surfaces can be modified with a biocompatible or/and biodegradable polymeric coating. The polymer can be natural such as chitosan (C6H13NO5), collagen, folic acid (C19H19N7O6) or synthetic as dextran (H(C6H10O5)xOH), tetraethyl orthosilicate (SiC8H20O4), N-(2-aminoethyl-3 aminopropyl) trimethoxysilane (C8H22N2O3Si), poly-lactic-co-glycolic acid ( PLGA), polyethylene glycol *(*C2nH4n+2On+1), etc. This surface modification needs to have a functional groups like: carboxyl (-COOH), hydroxyl (-OH), amine (-NH2), etc, with the capability to bond with a biological molecules. Table 3 shows a summary of recent publications of the

One of the most used techniques to ensure that the functionalization of magnetic nanoparticles has occurred, is the Fourier infrared spectroscopy (FTIR) because of its simplicity and availability. This technique provides the information about the excitation of vibration or rotation of molecules in their ground electronic State. In magnetite structure, these vibration, are associated with the stretching deformation of the interatomic bond of the iron with other molecules. Magnetite with a silica (figure x, MS sample) shell is confirmed by the characteristic adsorption band at 1090 cm-1 due to silane group presence. When a aminosilane is used like a coating, the spectra (figure, MA sample) show the band at 2943 cm-1 due to the stretching of C-H from methyl group (-CH2, -CH3), the band at 1072 cm-1 is due to the Si-O bond and the bands at 3309 and 1654 cm-1 are attributed to the amine

Fig. 7. Infrared spectra of the magnetite obtained by common coprecipitation (a), with a fast injection (b) and aging and reflux(c).


Table 2. Variation of the magnetite lattice cell with a determined particle size

Substitution of the cation on magnetic structures has been studied in order to improve the magnetic properties and FTIR spectroscopy is one of the techniques used in this kind of studies. In magnetite structure, the divalent iron is totally or partially replaced for strontium, cobalt, copper, nickel, manganese, cadmium, aluminum and gadolinium (Brabers et al., 1998). Figure 8 shows the infrared spectra of magnetite doped with cobalt; this cation occupies octahedral sites without changing inverse spinel crystal structure of magnetite.

\*KBr Bands **a**

\*1090

\*1072

441

**b**

438

451

**C**

606 575

606

606

Fig. 7. Infrared spectra of the magnetite obtained by common coprecipitation (a), with a fast

2000 1800 1600 1400 1200 1000 800 600 400

Wavenumber (cm-1

\*1090

)

Sample Method of Synthesis Particle size (nm) Lattice cell (A)

Substitution of the cation on magnetic structures has been studied in order to improve the magnetic properties and FTIR spectroscopy is one of the techniques used in this kind of studies. In magnetite structure, the divalent iron is totally or partially replaced for strontium, cobalt, copper, nickel, manganese, cadmium, aluminum and gadolinium (Brabers et al., 1998). Figure 8 shows the infrared spectra of magnetite doped with cobalt; this cation occupies octahedral sites without changing inverse spinel crystal structure of

Magnetite Coprecipiation with aging and reflux 206.9 ± 58.9 8.34468

Magnetite Common coprecipitation 27.6 ± 8.2 8. 34270

Magnetite Coprecipitation with fast injection 16.2 ± 4.4 8. 33475

Table 2. Variation of the magnetite lattice cell with a determined particle size

injection (b) and aging and reflux(c).

% de Transmitance

\*1633

\*1626

\*1626

magnetite.

Fig. 8. Infrared spectra of cobalt doped magnetite.

## **3. Functionalization of magnetic nanoparticles**

One of the most important aspects of the nanoparticles for biomedical applications is the surface preparation of the nanoparticles in order to improve their biocompatibility with biological entities and provide chemical stability. The nanoparticles surfaces can be modified with a biocompatible or/and biodegradable polymeric coating. The polymer can be natural such as chitosan (C6H13NO5), collagen, folic acid (C19H19N7O6) or synthetic as dextran (H(C6H10O5)xOH), tetraethyl orthosilicate (SiC8H20O4), N-(2-aminoethyl-3 aminopropyl) trimethoxysilane (C8H22N2O3Si), poly-lactic-co-glycolic acid ( PLGA), polyethylene glycol *(*C2nH4n+2On+1), etc. This surface modification needs to have a functional groups like: carboxyl (-COOH), hydroxyl (-OH), amine (-NH2), etc, with the capability to bond with a biological molecules. Table 3 shows a summary of recent publications of the most used coating materials for magnetite nanoparticles.

One of the most used techniques to ensure that the functionalization of magnetic nanoparticles has occurred, is the Fourier infrared spectroscopy (FTIR) because of its simplicity and availability. This technique provides the information about the excitation of vibration or rotation of molecules in their ground electronic State. In magnetite structure, these vibration, are associated with the stretching deformation of the interatomic bond of the iron with other molecules. Magnetite with a silica (figure x, MS sample) shell is confirmed by the characteristic adsorption band at 1090 cm-1 due to silane group presence. When a aminosilane is used like a coating, the spectra (figure, MA sample) show the band at 2943 cm-1 due to the stretching of C-H from methyl group (-CH2, -CH3), the band at 1072 cm-1 is due to the Si-O bond and the bands at 3309 and 1654 cm-1 are attributed to the amine

Infrared Spectroscopy of Functionalized Magnetic Nanoparticles 415

Fig. 9. FT spectra of magnetite (M), magnetite-aminosilane (MA) and magnetite-silica-

4000 3500 3000 2500 2000 1500 1000

**C- H**

Wavenumber (cm(-1))

+ **3 CH3O H Fe Fe Fe**

**O**

**O** 

O

Si

O

O

Si

O

O

O Si

O

O

Si

O

O

Fe Fe

**O – Si – (CH2 )3 – NH - (C H2)2 – N H2 +**

**Fe-O**

**3 C H3O H**

NH2

+

**Si-O-Si**

**Si-O**

**Si-O**

**N-H**

aminosilane (MSA) obtained by reflux and aging method.

**CH3O – Si – (C H2)3 – N H - (C H2)2 – NH2**

**CH3O** 

**CH3O** 

**CH3O** 

**Magenetite -Aminosilane shells**

**CH3O** 

**Magnetie-Silica-aminosilane shells**

+

**O - Si - OH**

M

**Fe OH**

**Fe Fe**

**Fe OH Fe OH**

**O - Si - OH** 

% de Transmitance

MA

MSA

**N-H**

MS

Scheme 1. Suggested mechanism of coated magnetite nanoparticles

**C H3O – Si – (C H2)3 – N H - (CH 2)2 – NH2**


Table 3. Materials used in functionalization of magnetite nanoparticles.

group (-NH2). A sample with a double coating silica-aminosilane (Figure 9, MSA sample) shows the band of both materials, and a new band is shown at 802 cm-1 due to Si-O-Si bond. Using this information, a suggested mechanism of coated particles can be proposed (Scheme 1) in magnetite-aminosilane shell, the silicon is bonded with the iron through the deprotonation of magnetite; when a silica shell is added before the aminosilane groups, the silicon is bonded in the same way with the magnetite and the silicon bonded with aminosilane trough S-O-S bond.

delivery, Metal separation in waste

imagines

10-50 NMR imagines, Gen delivery

10-50 NMR imagines, Drug delivery,

250 Tissue engineering, Cell targeting

separation

10-50 o NMR imagines, Entities separation

**Cellulose** 20-50 Drug delivery Huixia et al, 2011

delivery

Table 3. Materials used in functionalization of magnetite nanoparticles.

**Starch** 10-20 Cell Separation Dong-Hyun et al, 2009

group (-NH2). A sample with a double coating silica-aminosilane (Figure 9, MSA sample) shows the band of both materials, and a new band is shown at 802 cm-1 due to Si-O-Si bond. Using this information, a suggested mechanism of coated particles can be proposed (Scheme 1) in magnetite-aminosilane shell, the silicon is bonded with the iron through the deprotonation of magnetite; when a silica shell is added before the aminosilane groups, the silicon is bonded in the same way with the magnetite and the silicon bonded with

Metal separation

engineering, Drug delivery, hyperthermia

**Application Reference** 

10-20 Drug delivery Young-Lee et al, 2006

Schweoger et al, 2011; Chen et al, 2011; Del Campo et al, 2001; Ajay and grupta, 2005

Liu et al, 2011; Quin et al, 2011; Catherine et al, 2003; Zhang et al

Kami et al, 2011; Phadatare et al,

Pardoe et al, 2001; Morteza et al,

Schliehe et al, 2011; Chih-Hang et al, 2011; Mu andFeng,2001; Yoshida and. Babensee, 2006;

McCarthy et al, 2011; Ramirez et

Madhumita et al, 2011; Ammar et al, 2004; Andreva et al, 2006;

Coroto et al, 2011; Arami et al, 2011; Del campo et al, 2001;

2011; Zang et al, 2008;

,2007

2009.

al, 2003

Gao et al, 2010;

Gaihre et al, 2009

**Material Particle size** 

**Polyethylene glycol (PET)** 

**(PVA)** 

**Polyvinyl** 

**Methyl** 

**Polyvinyl alcohol** 

**Pyrrolidone (PVP)**

**polymethacrylate** 

aminosilane trough S-O-S bond.

**poly-lactic-coglycolic acid (PLGA)** 

**(nm)** 

**Silica** 20-300 DNA separation, Drug

**Dextran** 10-200 Drug delivery NMR

**Polystyrene (PS)** 10-20 NMR imagines DNA

**Polypyrrole** 20-100 Protein separation,

**Chitosan** 20-100 o Cell targeting, Tissue

**Gelatin** 50-100 DNA separation, drug

Fig. 9. FT spectra of magnetite (M), magnetite-aminosilane (MA) and magnetite-silicaaminosilane (MSA) obtained by reflux and aging method.

#### **Magenetite -Aminosilane shells**

Scheme 1. Suggested mechanism of coated magnetite nanoparticles

Infrared Spectroscopy of Functionalized Magnetic Nanoparticles 417

On the other hand adipic acid is a materials that has not been widely studied like a coating shell on magnetic nanoparticles; however, these materials have been widely used in drug delivery systems. In figure 11, the spectrum of the magnetite coated with this polymer shows the bands at 1415 y 1550cm-1 due to the symmetric and asymmetric carboxylate ion (COO-) and approximately at 600cm-1 the band due to the Fe-O bond in octahedral sites of

One of the most important aspects of nanoparticles in biomedical applications is their surface functionalization in order to improve their biocompatibility with biological entities, and Fourier infrared spectroscopy (FTIR) is very useful technique that provides information about iron oxides in their ground electronic state, and when this material is bonding with a polymeric coating provides information about mechanism of functionalized magnetic nanoparticles. This technique is widely used in characterization nanoparticles due to its simplicity and availability. In magnetite structure it provides information about the excitation of vibration or rotation of the trivalent and divalent iron cations and allows

An-Hui Lu, Salabas E. L., Schüth Ferdi. (2007). Magnetic Nanoparticles: Synthesis,

Ammar Azioune, Amel Ben Slimane, Lobnat Ait Hamou, Anne Pleuvy, Mohamed M.

Babita Gaihre, Myung Seob Khil, Douk Rae Lee, Hak Yong Kim. (2009). Gelatin-coated

Berry Catherine, C, Stephen Wells , Stuart Charles, Adam S.G. Curtis. (2003) Dextran and

Brabers V. A. M., Walz F.and Kronmüller H. (1998) Impurity effects upon the Verwey

Chen Yuwei and Wang Jianlong (2011). Preparation and characterization of magnetic

Chih-Hang Chu, Yu-Chao Wang, Hsin-Ying Huang, Li-Chen Wu and Chung-Shi Yang.

transition in magnetite, *Phys. Rev*. B, Volume 58, pp 14163–14166.

*occurrences, and uses*, wiley publisher, ISBN 3-527-30274-3.

Protection, Functionalization, and Application, Angewandte Chemie-International,

Chehimi, Christian Perruchot, and Steven P. Armes.( 2004). Synthesis and Characterization of Active Ester-Functionalized Polypyrrole−Silica Nanoparticles: Application to the Covalent Attachment of Proteins, *Langmuir*, Volume 20, Issue 8,

magnetic iron oxide nanoparticles as carrier system: Drug loading and *in vitro* drug release study, *International Journal of Pharmaceutics*, Volume 365, Issues 1-2, pp 180-

albumin derivatised iron oxide nanoparticles: influence on fibroblasts in vitro,

chitosan nanoparticles and its application for Cu(II) removal, *Chemical Engineering* 

(2011). Ultrafine PEG-coated poly(lactic-*co*-glycolic acid) nanoparticles formulated by hydrophobic surfactant-assisted one-pot synthesis for biomedical applications, *Nanotechnology*, Volume 22, Number 18, DOI:10.1088/0957-4484/22/18/185601. Cornell R. M. and Schwertmann Udo (2003). *The iron oxides: Structure, properties, reactions,* 

knowing the occupied sites when the divalent iron is replaced with other cations.

Volume 46, Issue 8, pp 1222-1244.

*Biomaterials*, Volume 24, Issue 25, pp 4551-4557.

*Journal*, Volume 168, Issue 1, 15,pp 286-292.

the magnetite.

**4. Conclusions** 

**5. References** 

pp 3350–3356.

189.

Chitosan is a natural polymeric material widely used in biomedical applications as coating in magnetic nanoparticles for biomedical applications. The spectra of this material, Figure 10 show bands at 1400 cm-1 due to C-O of the primary OH groups; at 1600cm-1 due to the N-H; at 2943cm-1 due to the stretching of C-H from methyl group (- CH2, -CH3) and the band at 1100cm-1 of the hydroxyl group of the piranosic ring of the chitosan beside the Fe-O bond due to octahedral sites of the magnetite. The band at 2250 cm-1 is due to carbon dioxide air.

Fig. 10. FT spectra of magnetite coated with a chitosan shell.

Fig. 11. FT spectra of magnetite coated with an adipic acid shell.

On the other hand adipic acid is a materials that has not been widely studied like a coating shell on magnetic nanoparticles; however, these materials have been widely used in drug delivery systems. In figure 11, the spectrum of the magnetite coated with this polymer shows the bands at 1415 y 1550cm-1 due to the symmetric and asymmetric carboxylate ion (COO-) and approximately at 600cm-1 the band due to the Fe-O bond in octahedral sites of the magnetite.

## **4. Conclusions**

416 Infrared Spectroscopy – Materials Science, Engineering and Technology

Chitosan is a natural polymeric material widely used in biomedical applications as coating in magnetic nanoparticles for biomedical applications. The spectra of this material, Figure 10 show bands at 1400 cm-1 due to C-O of the primary OH groups; at 1600cm-1 due to the N-H; at 2943cm-1 due to the stretching of C-H from methyl group (- CH2, -CH3) and the band at 1100cm-1 of the hydroxyl group of the piranosic ring of the chitosan beside the Fe-O bond due to octahedral sites of the magnetite. The band at 2250

4000 3500 3000 2500 2000 1500 1000

Wavenumber (cm-1

)

Vs V COO-

as COO-

cm-1 is due to carbon dioxide air.

% Transmitance

MA

Fig. 11. FT spectra of magnetite coated with an adipic acid shell.

Fig. 10. FT spectra of magnetite coated with a chitosan shell.

One of the most important aspects of nanoparticles in biomedical applications is their surface functionalization in order to improve their biocompatibility with biological entities, and Fourier infrared spectroscopy (FTIR) is very useful technique that provides information about iron oxides in their ground electronic state, and when this material is bonding with a polymeric coating provides information about mechanism of functionalized magnetic nanoparticles. This technique is widely used in characterization nanoparticles due to its simplicity and availability. In magnetite structure it provides information about the excitation of vibration or rotation of the trivalent and divalent iron cations and allows knowing the occupied sites when the divalent iron is replaced with other cations.

## **5. References**


Infrared Spectroscopy of Functionalized Magnetic Nanoparticles 419

Joralemon MJ and Rachel KOR. (2005). Shell click-crosslinked (SCC) nanoparticles: a new

Liu G., Hong R.Y., Guo L.,. Li Y.G, Li H.Z. (2011)., Preparation, characterization and MRI

Madhumita Bhaumik, Arjun Maity, V.V. Srinivasu, Maurice S. Onyango. (2011) Enhanced

Morteza Mahmoudi, Abdolreza Simchi and Mohammad Imani. (2009). Cytotoxicity of

Mu L and Feng S.S.(2001). Fabrication, characterization and in vitro release of paclitaxel

Nasrazadani S. adn Raman A. (1993). The application of infrares spectroscopy to the study

Neuberger Tobias, Schöpf Bernhard, Hofmann Heinrich, Hofmann Margarete and von

Qiang Gao, Dan Luo, Jun Ding, Yu-Qi Feng. (2010). Rapid magnetic solid-phase extraction

Phadatare M.R., Khot V.M., Salunkhe A.B., Thorat N.D., Pawar S.H. (2011). Studies on

Nanoparticles, *J. Phys. Chem. C*, volume 113, Issue 22, pp 9573–9580. Mutsumi Yoshida and Julia E. Babensee (2006). Differential effects of agarose and

*Materials Research Part A*, Volume 79A, Issue 2, pp 393–408.

Volume 127, Issue 48,pp 16892–16899.

*Release*,Volume 76, Issue 3, pp 239-254.

*Magn Mater.,Volume* 293, Issue 1,pp 483-496.

*Res*., Volume 2, Issue 1,pp 215–221.

21, 14219-14225.

1355-1365.

172.

*Surface Science*, Volume 257, Issue 15, pp 6711-6717.

methodology for synthesis and orthogonal functionalization, *J. Am. Chem. Soc.*,

application of carboxymethyl dextran coated magnetic nanoparticles, *Applied* 

removal of Cr(VI) from aqueous solution using polypyrrole/Fe3O4 magnetic nanocomposite*, Journal of Hazardous Materials*, Volume 190, Issues 1-3, pp 381-390. McCarthy Joseph E, Prina-Mello Adriele, Rakovich Tatsiana, Volkov Yuri and K. Gun'ko

Yurii. (2011) Fabrication and characterization of multimodal magnetic - fluorescent polystyrene nanowires as selective cell imaging probes, *J. Mater. Chem*., DOI: 2011,

Uncoated and Polyvinyl Alcohol Coated Superparamagnetic Iron Oxide

poly(lactic-*co*-glycolic acid) on dendritic cell maturation*, Journal of Biomedical* 

(Taxol®) loaded poly (lactic-co-glycolic acid) microspheres prepared by spray drying technique with lipid/cholesterol emulsifiers, *Journal of Controlled* 

of rust systems: Study of cation deficiency in magnetite produced during its transformation to maghemite and hematite, *Corrosion Science,* Volume 34, No. 8, pp.

Rechenberg Brigitte. (2005). Superparamagnetic nanoparticles for biomedical applications: Possibilities and limitations of a new drug delivery system, *J. Magn.* 

based on magnetite/silica/poly(methacrylic acid–co–ethylene glycol dimethacrylate) composite microspheres for the determination of sulfonamide in milk samples, *Journal of Chromatography A*, Volume 1217, Issue 35, pp 5602-5609. Palla B.J.,. Shah D.O, Garcia Casillas P., Matutes-Aquino J.(1999) Preparation of

nanoparticles of barium ferrite from precipitation in microemulsions, *J. Nanoparticle* 

polyethylene glycol coating on NiFe2O4 nanoparticles for biomedical applications, *Journal of Magnetism and Magnetic Materials*, doi:10.1016/j.jmmm.2011.09.020. Pankhurst Q A, Connolly J, Jones S K and Dobson J. (2003). Applications of magnetic

nanoparticles in biomedicine J. Phys. D: Appl. Phys, Volume 36, Issue 13,pp 167-


Cullity Bernard Dennis and Graham Chad D. (2009), *Introduction to magnetic materials*, wiley

Daisuke Kami, Shogo Takeda,Yoko Itakura, Satoshi Gojo, Masatoshi Watanabe and Masashi

Daria V. Andreeva. Dmitry A. Gorin, Dmitry G. Shchukin, Gleb B. Sukhorukov.(2006).

Di Corato Riccardo,. Bigall Nadja C, Ragusa Andrea, k DorfsDir, Genovese Alessandro ,

Gao J, Gu H. (2009). Multifunctional magnetic nanoparticles: design, synthesis, and biomedical applications, *Acc. Chem. Res.*, Volume 42, Issue 8,pp 1097–1107. Gnanaprakash G., Mahadevan S., Jayakumar T., Kalyanasundaram P., Philip J., Baldev R.

Gupta Ajay Kumar and Mona Gupta.(2005). Synthesis and surface engineering of iron oxide

Hamed Arami, Zachary Stephen, Omid Veiseh and Miqin Zhang. (2011). Chitosan-Coated

Heath Pardoe, Wanida Chua-anusorna, Timothy G. St. Pierre, Jon Dobson. (2001). Structural

Huan Qin, Dong Xu and Sihua Yang. (2011). Dextran-coated Fe3O4 magnetic nanoparticles

*Macromolecular Rapid Communications*, Volume 27, Issue 12,pp 931–936. De Campos Angela, M, Alejandro Sánchez, Maŕa J. Alonso. (2001). Chitosan nanoparticles:

Toyoda.(2011). Application of Magnetic Nanoparticles to Gene Delivery *Int. J. Mol.* 

Magnetic Microcapsules with Low Permeable Polypyrrole Skin Layer,

a new vehicle for the improvement of the delivery of drugs to the ocular surface. Application to cyclosporin A, *International Journal of Pharmaceutics,* Volume 224,

Marotta Roberto, Manna Liberato, and Pellegrino Teresa (2011). Multifunctional Nanobeads Based on Quantum Dots and Magnetic Nanoparticles: Synthesis and Cancer Cell Targeting and Sorting, *ACS Nano*, Volume 5, Issue 2, pp 1109–1121. Dong-Hyun Kim, Kyoung-Nam Kim, Kwang-Mahn Kim, Yong-Keun Lee (2009). Targeting

to carcinoma cells with chitosan- and starch-coated magnetic nanoparticles for magnetic hyperthermia*, Journal of Biomedical Materials Research Part A,* Volume

(2007). Effect of initial pH and temeprature of iron salt solutions on formation of magnetite nanoparticles. *Materials chemistry and Physics* Volume 103, Issue 1,pp168-

nanoparticles for biomedical applications, *Biomaterials*, Volume 26, Issue 18, pp

Iron Oxide Nanoparticles for Molecular Imaging and Drug Delivery, Chitosan for biomaterials I, *Advances in Polymer Science*, Volume 243/2011, 163-

Zhong. (2011). Biosynthesis of spherical Fe3O4/bacterial cellulose nanocomposites as adsorbents for heavy metal ions, *Carbohydrate Polymers*, Volume 86, Issue 4, pp

and magnetic properties of nanoscale iron oxide particles synthesized in the presence of dextran or polyvinyl alcohol, *Journal of Magnetism and Magnetic* 

as a contrast agent in thermoacoustic tomography for hepatocellular carcinoma detection, *Journal of Physics: Conference Series*, Volume 277, Number 1,

TongWan, Yuanyuan Jia, Hongjiang Yang, Jing Li, Lin Yan, Cheng

publisher, ISBN 978-0-471-47741-9.

*Sci* , Volume 12*,* Issue *6*,pp 3705-3722.

Issues 1-2, 14 August 2001, pp 159-168.

88. Issue 1, pp 1–11.

175.

Huixia Zhu, Shiru Jia,

3995-4021.

1558-1564.

184, DOI: 10.1007/12\_2011\_121.

*Materials*, Volume 225, Issues 1-2, pp 41-46.

doi:10.1088/1742-6596/277/1/012028.


**22** 

Tarik Chafik

*Morocco* 

*University Abdelmalek Essaadi,* 

**Determination of Adsorption Characteristics** 

**Gas Phase FTIR Spectroscopy Flow Analysis** 

Adsorption is an important operation for gas mixture separation and purification (Yang et al.1987). Thus, measurement of adsorption capacity is crucial for correct designing and operating of adsorption facilities. The heat of adsorption is also an important adsorption characteristic usually needed if adsorbent regeneration and/or adsorbate recovery is targeted (Do, 1998). Accurate knowledge of this thermodynamic parameter is of interest because the heat effects involved during adsorption and/or desorption may induce significant local warming to start combustion of either the adsorbate or the adsorbent itself,

In case of physical adsorption, the involved heat effects might be determined with the isosteric method based on Clausius–Clapeyron equation using adsorption isotherms data at various temperatures (Sircar et al.1999). However, it is to be noted that the isosteric method is limited only to physisorption processes involving relatively low heat of adsorption and for equilibrium reached within accessible range of pressure and temperature (Ranke et al.2002). Under equilibrium, the physically adsorbed species are trapped in different potential wells predisposed within the accessible porosity. During desorption, the species adsorbed in more easily accessible porosity desorbs first. As for those adsorbed in the less accessible porosity, one has to supply required thermal energy for their transfer into gas phase. This energy could be estimated by TPD method using a relatively inexpensive and

This chapter is devoted to the description of an easy and efficient method based on the application of gas phase Flow FTIR spectroscopy analysis for determination of adsorption characteristics of volatile organic compounds. As adsorbent beds are usually operated under dynamic conditions, the adopted analytical approach is based on gas phase composition monitoring at reactor outlet during adsorption/ desorption experiments carried out under dynamic regime. This method permits further simultaneous detection of new IR bands that may originate from adsorbate dissociation during adsorption or

particularly, when activated carbon is used (EPA, 1998).

simple experiment to set up and to run (Cvetanovic et al.1967).

**1. Introduction** 

desorption.

**of Volatile Organic Compounds Using** 

*Laboratory of Chemical Engineering and Resources Valorisation (FST/L01),* 

*Faculty of Sciences and Techniques of Tangier, Tangier* 


## **Determination of Adsorption Characteristics of Volatile Organic Compounds Using Gas Phase FTIR Spectroscopy Flow Analysis**

Tarik Chafik

*Laboratory of Chemical Engineering and Resources Valorisation (FST/L01), University Abdelmalek Essaadi, Faculty of Sciences and Techniques of Tangier, Tangier Morocco* 

### **1. Introduction**

420 Infrared Spectroscopy – Materials Science, Engineering and Technology

Ramírez Liliana P. and Landfester Katharina. (2003). Magnetic Polystyrene Nanoparticles

Schweiger Christoph, Pietzonka Clemens, Heverhagen Johannes, Kissel Thomas.(2011).

Sidhu P.S., Gilkes R.J,.Posner A.M. (1978). The synthesis and some properties of Co, Ni, Zn,

Sun Yong-kang , Ming Ma, Zhang Yu, Ning Gu.(2004) Synthesis of nanometer-size

Terrazas Reza R., Martinez Perez C. A., Rodriguez Gonzalez C.A. Romero Monreal H. T

Varadan V. K, Linfeng Chen and Jining Xie. (2008). *Nanomedicine: design and applications of magnetic nanomaterials, nanosensors*, wiley publisher, ISBN 978-0-470-03351-7.

Zhang J., Rana S., Srivastava R.S., Misra R.D.K. (2008). On the chemical synthesis and drug

magnetite nanoparticles, *Acta Biomaterialia*, Volume 4, Issue 1, pp 40-48.

*Macromolecular Chemistry and Physics*, Volume 204, Issue 1, pp 22–31. Schliehe Christopher, Schliehe Constanze, Thiry Marc, I. Tromsdorf Ulrich, Hentschel

Joachim, Weller Horst, Marcus Groettrup..

Issues 1-2, 15,pp 130-137.

1046.

Young Lee Ha, Hyun Lim Nak,

Volume 4, Issue 3, pp 429-435.

*Controlled Release*, Volume 151, Issue 3, pp 278-285.

*Engineering Aspects*, Volume 245, Issues 1-3, pp 15-19.

*Langmuir*, Volume 23*,* Issue 11, pp 6342–6351.

with a High Magnetite Content Obtained by Miniemulsion Processes†,

nanocrystals into PLGA microsphere vaccines enables their intracellular localization in dendritic cells by electron and fluorescence microscopy, *Journal of* 

Novel magnetic iron oxide nanoparticles coated with poly(ethylene imine)-*g*poly(ethylene glycol) for potental biomedical application: Synthesis, stability, cytotoxicity and MR imaging, *International Journal of Pharmaceutics*, Volume 408,

Cu, Mn and Cd substituted magnetites*, Journal of Inorganic and Nuclear Chemistry*,

maghemite particles from magnetite, *Colloids and Surfaces A: Physicochemical and* 

amd Garcia Casillas P. E.(2010). Effect of the polymeric coating over Fe3O4 particles used for magnetic separation, *Cent. Eur. J. Che, 2010*, Volume 8, Issue 5, pp 1041-

Hai Bang Lee, Sun Hang Cho. (2006). Preparation and magnetic resonance imaging effect of polyvinylpyrrolidone-coated iron oxide nanoparticles, *Journal of Biomedical Materials Research Part B: Applied Biomaterials*, Volume 79B, Issue, pp 142–150. Zhang J. L., Srivastava R. S, and. Misra R. D. K. (2007). Core−Shell Magnetite Nanoparticles

Surface Encapsulated with Smart Stimuli-Responsive Polymer: Synthesis, Characterization, and LCST of Viable Drug-Targeting Delivery System,

delivery response of folate receptor-activated, polyethylene glycol-functionalized

. Seo Jin A, Soon Hong Yuk, Byung Kook Kwak, Gilson Khang,

(2011). Microencapsulation of inorganic

Adsorption is an important operation for gas mixture separation and purification (Yang et al.1987). Thus, measurement of adsorption capacity is crucial for correct designing and operating of adsorption facilities. The heat of adsorption is also an important adsorption characteristic usually needed if adsorbent regeneration and/or adsorbate recovery is targeted (Do, 1998). Accurate knowledge of this thermodynamic parameter is of interest because the heat effects involved during adsorption and/or desorption may induce significant local warming to start combustion of either the adsorbate or the adsorbent itself, particularly, when activated carbon is used (EPA, 1998).

In case of physical adsorption, the involved heat effects might be determined with the isosteric method based on Clausius–Clapeyron equation using adsorption isotherms data at various temperatures (Sircar et al.1999). However, it is to be noted that the isosteric method is limited only to physisorption processes involving relatively low heat of adsorption and for equilibrium reached within accessible range of pressure and temperature (Ranke et al.2002). Under equilibrium, the physically adsorbed species are trapped in different potential wells predisposed within the accessible porosity. During desorption, the species adsorbed in more easily accessible porosity desorbs first. As for those adsorbed in the less accessible porosity, one has to supply required thermal energy for their transfer into gas phase. This energy could be estimated by TPD method using a relatively inexpensive and simple experiment to set up and to run (Cvetanovic et al.1967).

This chapter is devoted to the description of an easy and efficient method based on the application of gas phase Flow FTIR spectroscopy analysis for determination of adsorption characteristics of volatile organic compounds. As adsorbent beds are usually operated under dynamic conditions, the adopted analytical approach is based on gas phase composition monitoring at reactor outlet during adsorption/ desorption experiments carried out under dynamic regime. This method permits further simultaneous detection of new IR bands that may originate from adsorbate dissociation during adsorption or desorption.

Determination of Adsorption Characteristics of

concentration in N2 (molar fraction % )

**2.3 Determination of adsorbed and desorbed amounts** 

accuracy was cheeked over 3 experiments and represented as mean value.

Fig. 1. Calibration curve o-xylene IR band area integral (2600-3200 cm-1) versus its

The sample pre-treatment as well as adsorption and desorption experiments were performed with a flow rate of 100 cm3 min-1 that was passed through a quartz reactor (Utype) containing 1 g of bentonite meshes. The sample was first pre-treated under N2 flow at 473 K for 30 minutes than adsorption was carried out using the model mixture flow until saturation was reached in order to obtain breakthrough curves. The gas mixture was switched again to pure N2 flow, to proceed with isothermal desorption until o-xylene concentration at the reactor outlet reached zero. This step was followed by a subsequent linear heating in order to perform Temperature Programmed Desorption (TPD) experiment. The monitoring of *o*-xylene IR bands during adsorption at 300 K shows gradual increase of IR bands until equilibrium was reached, which corresponds to adsorbent saturation (Fig2, part A). Following this step, the gas mixture was switched to pure nitrogen flow (0.36 % o-xylene/N2N2), to perform isothermal desorption giving rise to gradual decrease of IR

Volatile Organic Compounds Using Gas Phase FTIR Spectroscopy Flow Analysis 423

The o-xylene concentrations at reactor outlet during adsorption and desorption were monitored with a FTIR spectrometer (Jasco 410, resolution 4 cm-1), using a Pyrex gas cell equipped with CaF2 windows. The quantitative analysis is facilitated by FTIR instrumentation and programs that allows high frequency spectra acquisition and manipulation (substraction, multiplication, smoothing). This operation is particularly easy when there are no IR bands overlapping and the application of Beer–Lambert law permits relating IR bands area to concentration. In the present work, the quantitative treatment was achieved by integrating o-xylene IR bands located between 2600 and 3200 cm-1. Preliminary calibration with *o*xylene/N2 mixtures of known composition was carried out using reactor by pass, in order to correlate bands area with concentration. Figure 1 shows the calibration curve representing the integral of *o*-xylene IR bands area between 2600 and 3200 cm-1 as function of concentration. The FTIR response was found to produce linear plot in the studied concentration range and its

Although, Infra Red (IR) spectroscopy is generally used for compounds identification based on measurement of IR radiation resulting from molecular vibration perturbation. The application of Beer's law allows quantitative exploitation of IR spectra using preliminary calibration with known composition. Therefore, the technique was found to be adequate for transient signals monitoring and quantification owing to faster response to instantaneous change in IR bands positions and intensities due to rapid spectra acquisitions allowed by FTIR instrumentation (Chafik et al, 1998). In addition, the high transparency of gases and the low background values achieved, allows detection of lower species concentration in the gas stream. Further increase of analytical sensitivity might be obtained with multiple-pass gas cell.

In the present study, we show an example of the application of this experimental methodology to the investigation of adsorption characteristic of local bentonite clay. Adsorption / desorption experiment were performed at laboratory scale with modelcontaminated gaseous stream containing o-xylene that has been selected as representative VOC because it is environmentally relevant regarding industrial concern. The work aimed to help further development of low cost materials involved in environmental engineering control.

## **2. Experimental**

#### **2.1 Adsorbent**

The clay tested in the present work comes from deposits located in the north of Morocco (Nador area). The results corresponding to its textural and mineralogical composition have been published (Harti et al, 2007). The clay was found to be a mixture of phases, namely, opal, montmorillonite, kaolinite, muscovite, topaz, rutile, calcite, dolomite, suggesting a bentonite type clay. Textural studies perfomed with N2 adsorption desorption at 77 K have shown a BET specific surface area of 79 m2 g-1, and a negligible micropore volume of 0.002 cm3 g-1 as compared with total pore volume of 0.205 cm3 g-1. The pore size distribution obtained following BJH method revealed the presence of a wide pore size distribution in the mesopore range with a significant contribution of pores widths between 8 and 50 nm.

### **2.2 Adsorption and desorption experiments**

Adsorption/desorption experiments were performed under dynamic conditions at atmospheric pressure using the experimental apparatus as reported elsewhere (Zaitan et al.2005). The bentonite adsorptive properties were investigated with respect to O-xylene vapour. Prior to adsorption, a model mixture with a given concentration of o-xylene vapour in nitrogen flow, was prepared by means of a saturator connected to a condenser that was immersed in a thermostatically controlled bath. This temperature was carefully checked so as to maintain constant *o*-xylene vapour pressure and consequently keep the *o*-xylene concentration unchanged. The resulting concentration is expressed as molar fraction (or partial pressure; P/P0), where *P* is the vapour pressure of *o*-xylene obtained from Antoine equation and *P*0 the atmospheric pressure taken as 760 Torr. Hence, fixing condenser temperature between -8 and 40°C permitted obtaining a value of concentration at reactor inlet (*Ci*n) in a range of 700 - 11000 ppm (Zaitan et al, 2006).

#### **2.3 Determination of adsorbed and desorbed amounts**

422 Infrared Spectroscopy – Materials Science, Engineering and Technology

Although, Infra Red (IR) spectroscopy is generally used for compounds identification based on measurement of IR radiation resulting from molecular vibration perturbation. The application of Beer's law allows quantitative exploitation of IR spectra using preliminary calibration with known composition. Therefore, the technique was found to be adequate for transient signals monitoring and quantification owing to faster response to instantaneous change in IR bands positions and intensities due to rapid spectra acquisitions allowed by FTIR instrumentation (Chafik et al, 1998). In addition, the high transparency of gases and the low background values achieved, allows detection of lower species concentration in the gas stream. Further increase of analytical sensitivity might be obtained with multiple-pass

In the present study, we show an example of the application of this experimental methodology to the investigation of adsorption characteristic of local bentonite clay. Adsorption / desorption experiment were performed at laboratory scale with modelcontaminated gaseous stream containing o-xylene that has been selected as representative VOC because it is environmentally relevant regarding industrial concern. The work aimed to help further development of low cost materials involved in environmental engineering

The clay tested in the present work comes from deposits located in the north of Morocco (Nador area). The results corresponding to its textural and mineralogical composition have been published (Harti et al, 2007). The clay was found to be a mixture of phases, namely, opal, montmorillonite, kaolinite, muscovite, topaz, rutile, calcite, dolomite, suggesting a bentonite type clay. Textural studies perfomed with N2 adsorption desorption at 77 K have shown a BET specific surface area of 79 m2 g-1, and a negligible micropore volume of 0.002 cm3 g-1 as compared with total pore volume of 0.205 cm3 g-1. The pore size distribution obtained following BJH method revealed the presence of a wide pore size distribution in the mesopore range with a significant contribution of pores

Adsorption/desorption experiments were performed under dynamic conditions at atmospheric pressure using the experimental apparatus as reported elsewhere (Zaitan et al.2005). The bentonite adsorptive properties were investigated with respect to O-xylene vapour. Prior to adsorption, a model mixture with a given concentration of o-xylene vapour in nitrogen flow, was prepared by means of a saturator connected to a condenser that was immersed in a thermostatically controlled bath. This temperature was carefully checked so as to maintain constant *o*-xylene vapour pressure and consequently keep the *o*-xylene concentration unchanged. The resulting concentration is expressed as molar fraction (or partial pressure; P/P0), where *P* is the vapour pressure of *o*-xylene obtained from Antoine equation and *P*0 the atmospheric pressure taken as 760 Torr. Hence, fixing condenser temperature between -8 and 40°C permitted obtaining a value of concentration at reactor

gas cell.

control.

**2. Experimental 2.1 Adsorbent** 

widths between 8 and 50 nm.

**2.2 Adsorption and desorption experiments** 

inlet (*Ci*n) in a range of 700 - 11000 ppm (Zaitan et al, 2006).

The o-xylene concentrations at reactor outlet during adsorption and desorption were monitored with a FTIR spectrometer (Jasco 410, resolution 4 cm-1), using a Pyrex gas cell equipped with CaF2 windows. The quantitative analysis is facilitated by FTIR instrumentation and programs that allows high frequency spectra acquisition and manipulation (substraction, multiplication, smoothing). This operation is particularly easy when there are no IR bands overlapping and the application of Beer–Lambert law permits relating IR bands area to concentration. In the present work, the quantitative treatment was achieved by integrating o-xylene IR bands located between 2600 and 3200 cm-1. Preliminary calibration with *o*xylene/N2 mixtures of known composition was carried out using reactor by pass, in order to correlate bands area with concentration. Figure 1 shows the calibration curve representing the integral of *o*-xylene IR bands area between 2600 and 3200 cm-1 as function of concentration. The FTIR response was found to produce linear plot in the studied concentration range and its accuracy was cheeked over 3 experiments and represented as mean value.

Fig. 1. Calibration curve o-xylene IR band area integral (2600-3200 cm-1) versus its concentration in N2 (molar fraction % )

The sample pre-treatment as well as adsorption and desorption experiments were performed with a flow rate of 100 cm3 min-1 that was passed through a quartz reactor (Utype) containing 1 g of bentonite meshes. The sample was first pre-treated under N2 flow at 473 K for 30 minutes than adsorption was carried out using the model mixture flow until saturation was reached in order to obtain breakthrough curves. The gas mixture was switched again to pure N2 flow, to proceed with isothermal desorption until o-xylene concentration at the reactor outlet reached zero. This step was followed by a subsequent linear heating in order to perform Temperature Programmed Desorption (TPD) experiment.

The monitoring of *o*-xylene IR bands during adsorption at 300 K shows gradual increase of IR bands until equilibrium was reached, which corresponds to adsorbent saturation (Fig2, part A). Following this step, the gas mixture was switched to pure nitrogen flow (0.36 % o-xylene/N2N2), to perform isothermal desorption giving rise to gradual decrease of IR

Determination of Adsorption Characteristics of

Desorption with Programmed Temperature).

―·―·without adsorbent

heating rate of 5K/min


Volatile Organic Compounds Using Gas Phase FTIR Spectroscopy Flow Analysis 425

bands with time, as shown in part B. The sample was, than, linearly heated up to 473K to carry out TPD experiment under N2 flow (part C of Figure 2). It is to be noted, that the recorded FTIR spectra do not reveal any new species formation originating from o-xylene transformation due to catalytic activity. This information provided by the use of FTIR

The evolution of *o*-xylene concentration at the reactor outlet (*C*out) obtained from FTIR spectra, permitted the monitoring of adsorbent loading as function of time during

Figure 3 shows the profile of the variation of *o*-xylene concentration in the gas flow at reactor outlet, represented as relative values (*C*out/*Ci*n), during the aforementioned cycle of successive steps (adsorption at 300K until saturation followed by isothermal desorption than

Fig. 3. Profile of the variation of *o*-xylene concentration in the gas flow at reactor outlet, represented as relative values (*C*out/*Ci*n), during a cycle of successive steps; adsorption performed with a mixture of 0.36% xylene in N2 at 300K until saturation followed by isothermal desorption than Temperature Programmed Desorption carried out with a linear

It is to be noted that the adsorption/ desorption experiments carried out in the present work with samples of 1g of bentonite compressed as small meshes contained in a quartz microreactor (type U with an internal volume ≈ 1 cm3). The experiments carried out under atmospheric pressure (760 torr) using total flow rate of 100 cm3/min resulted in a residence time <1 sec. Under these conditions the flow pass through the sample in mode 'plug flow' and since small amounts are involved in adsrorption or desorption experiments, the reactor

analysis technique can be considered as an advantage of the experimental approach.

adsorption (i.e. breakthrough curve) and desorption processes.

Fig. 2. Evolution of o-xylene IR bands during the following successive experiments: A; Isothermal adsorption at 300K, B; isothermal desoprtion and TPD under N2 at linear heating rate of 5K/min.

bands with time, as shown in part B. The sample was, than, linearly heated up to 473K to carry out TPD experiment under N2 flow (part C of Figure 2). It is to be noted, that the recorded FTIR spectra do not reveal any new species formation originating from o-xylene transformation due to catalytic activity. This information provided by the use of FTIR analysis technique can be considered as an advantage of the experimental approach.

The evolution of *o*-xylene concentration at the reactor outlet (*C*out) obtained from FTIR spectra, permitted the monitoring of adsorbent loading as function of time during adsorption (i.e. breakthrough curve) and desorption processes.

Figure 3 shows the profile of the variation of *o*-xylene concentration in the gas flow at reactor outlet, represented as relative values (*C*out/*Ci*n), during the aforementioned cycle of successive steps (adsorption at 300K until saturation followed by isothermal desorption than Desorption with Programmed Temperature).

―·―·without adsorbent

424 Infrared Spectroscopy – Materials Science, Engineering and Technology

Fig. 2. Evolution of o-xylene IR bands during the following successive experiments: A; Isothermal adsorption at 300K, B; isothermal desoprtion and TPD under N2 at linear

heating rate of 5K/min.


Fig. 3. Profile of the variation of *o*-xylene concentration in the gas flow at reactor outlet, represented as relative values (*C*out/*Ci*n), during a cycle of successive steps; adsorption performed with a mixture of 0.36% xylene in N2 at 300K until saturation followed by isothermal desorption than Temperature Programmed Desorption carried out with a linear heating rate of 5K/min

It is to be noted that the adsorption/ desorption experiments carried out in the present work with samples of 1g of bentonite compressed as small meshes contained in a quartz microreactor (type U with an internal volume ≈ 1 cm3). The experiments carried out under atmospheric pressure (760 torr) using total flow rate of 100 cm3/min resulted in a residence time <1 sec. Under these conditions the flow pass through the sample in mode 'plug flow' and since small amounts are involved in adsrorption or desorption experiments, the reactor

Determination of Adsorption Characteristics of

enough for recovering its initial performances.

if collecting and reusing VOCs is targeted.

Freundlich equation deviates at pressures higher than 4 torr.

according to Clausius-Clapeyron equation (Rouquerol, 1999):

**2.4 Isosteric heat of adsorption** 

and temperature at equilibrium.

Volatile Organic Compounds Using Gas Phase FTIR Spectroscopy Flow Analysis 427

Therefore, the data corresponding to total adsorbed amount (nads) were found to fit the mass balance equation (nads ≈ n1ads + n2ads ) with a precision around 2%, corresponding to the accuracy of the used analytical methodology. It is to be noted that the nads values do not suffer major changes during successive cycles of adsorption desorption experiments carried out with the same benonite samples and the precision remains at worst around 5%. However, for bentonite sample used for more than 3 successive cycles of adsorption/desorption experiments, a treatment under N2 flow at 473 K for 30 min, is

Another important information given by TPD experiment concerns the desorption performance indicated by the temperature at the peak maximum. The Tm value of 353 K shown by Figure 3, is slightly, lower than the Tm value of 368 K given by TPD peak obtained

The obtained results reveals adsorptive properties of bentonite clay such as larger reversibly adsorbed fraction, lower temperature for complete thermal desorption and absence of catalytic activity that might be of interest for adsorbent regeneration and adsorbate recovery. Even though its lower BET surface area, Bentonite's potential use as adsorbent material deserves to be investigated because adsorbent with higher specific surface area; such as active carbons is not usually the best (Brasseur et al, 2004). Thus, investigation of adsorption energies is also needed for selection of efficient adsorbent material, particularly,

The amounts adsorbed at equilibrium corresponding to adsorption capacity obtained from breakthrough curves, were determined for different adsorption temperatures and for different xylene pressures, yields to isotherms shown in Figure 4. The isotherms are represented in the form of *N* f P where N is the adsorbed amount per adsorbent weight at equilibrium and P the O-xylene partial pressure in the mixture flow. The experimental isotherms were modeled with Langmuir and Freundlich equations using a nonlinear regression method (MathCAD software). The corresponding fitting curves are shown as solid and dashed lines, respectively, for Langmuir and Freundlich models (Fig. 4). Apparently, the experimental data were well represented by Langmuir model while

The adsorption isotherms presented in Figure 4 were used for determination of isosteric heat (Qst) of adsorption by extrapolation at different temperatures and for a given coverage

> st ln <sup>Q</sup> <sup>1</sup> *<sup>P</sup> <sup>R</sup>*

Where R is the perfect gas constant, P and T correspond, respectively; to partial pressure

 

*T*

(4)

(where tb and tc correspond to the starting and the ending of TPD curve).

using the same heating rate for SiO2 Degussa (200 m2g-1)(Zaitan et al 2005).

could be considered operating under differential mode as Continuous flow Stirred Tank Reactor (CSTR). This approach permit creation of appropriate and accurate conditions that avoid limitation due to mass and heat transfer within the adsorbent particles in the reactor bed (Zaitan et al, 2008). Consequently, the experiments could be considered as occurring under isothermal conditions in which the pressure drop as well as the variation in fluid velocity between the reactor inlet and outlet were considered negligible. Thus, the adsorbed amount can be calculated using the mass balance equation (Ruthven, 1984).

$$\mathbf{m\_{ads}} = \frac{\mathbf{FC\_{in}}}{\mathbf{m}} \left[ \mathbf{t\_a} - \int\_0^{\mathbf{t\_a}} \frac{\mathbf{C\_{out}}}{\mathbf{C\_{in}}} \mathbf{dt} \right] \tag{1}$$

Where nads is the adsorbed mole number of toluene at saturation (adsorption capacity in mol*/*g , Cin and Cout the molar fractions of toluene at reactor inlet and outlet, m is the absorbent mass, ta the saturation time and F the gaseous molar flow rate.

The adsorbed amount measured from breakthrough curve is generally used to indicate adsorbent performance in terms of a given constituent removal from a flowing stream. In the present study, the adsorption capacity is obtained by integration of the breakthrough curve, according to the equation (1) and considering the curve corresponding to the reactor response in the absence of solid. Accordingly, the numerical integration of the breakthrough curve using MathCAD software yields to a total adsorbed amount of 420 µmol g–1, for adsorption carried out with a flow containing 0.36% *o*-xylene in N2 (Fig. 3 part A). This amount is lower than the adsorption capacity of 1958 µmol g–1 reported in our previous work for SiO2 Degussa (200 m2g-1) (Zaitan et al 2005). In comparison with the literature, substantially higher values of 4666, 2800, 1800 and 2050 µmol g-1 are reported for xylene adsorption, respectively, for activated carbon AC40 (1300 m2/g)( Benkhedda et al, 2000) and zeolithes (Al-Meso 100 (915 m2/g) UL-ZSM5-100-2 (840 m2/g), UL-ZSM5-100-6 (780 m2/g) (Huang et al, 2000)

The numerical integration of the desorption curve (part B of Figure 3) using equation (2) and considering the curve in the absence of solid, permitted quantification of loosely adsorbed fraction (n1ads) of 370 µmol.g–1 released during isothermal desorption.

$$m\_{1ads} = \frac{\text{FC}\_{\text{in}}}{\text{m}} \left[ \left. \frac{^{\text{t}\_{\text{j}}}\text{C}\_{\text{out}}}{\text{C}\_{\text{in}}} \text{dt} \right| \right] \tag{2}$$

Where ta and tb correspond to starting and ending time of isothermal desorption.

It is to be noted that although the *o*-xylene desorption curve reached zero, the calculated weakly adsorbed fraction represents 88% of the total adsorbed amount. The remaining oxylene adsorbed amount corresponds to the more strongly adsorbed fraction. The removal of the latter fraction requires thermal treatment with linear heating rate according to TPD method. The integration of the corresponding curve, shown in part C of Figure 3 using equation (3), permits obtaining an irreversible amount of 60 µmol.g–1.

$$\mathbf{n}\_{2\text{ads}} = \frac{\mathbf{F}\mathbf{C}\_{\text{in}}}{\mathbf{m}} \cdot \left| \int\_{\mathbf{t}\_b}^{\mathbf{t}\_c} \frac{\mathbf{C}\_{\text{out}}}{\mathbf{C}\_{\text{in}}} \mathbf{dt} \right| \tag{3}$$

(where tb and tc correspond to the starting and the ending of TPD curve).

Therefore, the data corresponding to total adsorbed amount (nads) were found to fit the mass balance equation (nads ≈ n1ads + n2ads ) with a precision around 2%, corresponding to the accuracy of the used analytical methodology. It is to be noted that the nads values do not suffer major changes during successive cycles of adsorption desorption experiments carried out with the same benonite samples and the precision remains at worst around 5%. However, for bentonite sample used for more than 3 successive cycles of adsorption/desorption experiments, a treatment under N2 flow at 473 K for 30 min, is enough for recovering its initial performances.

Another important information given by TPD experiment concerns the desorption performance indicated by the temperature at the peak maximum. The Tm value of 353 K shown by Figure 3, is slightly, lower than the Tm value of 368 K given by TPD peak obtained using the same heating rate for SiO2 Degussa (200 m2g-1)(Zaitan et al 2005).

The obtained results reveals adsorptive properties of bentonite clay such as larger reversibly adsorbed fraction, lower temperature for complete thermal desorption and absence of catalytic activity that might be of interest for adsorbent regeneration and adsorbate recovery. Even though its lower BET surface area, Bentonite's potential use as adsorbent material deserves to be investigated because adsorbent with higher specific surface area; such as active carbons is not usually the best (Brasseur et al, 2004). Thus, investigation of adsorption energies is also needed for selection of efficient adsorbent material, particularly, if collecting and reusing VOCs is targeted.

## **2.4 Isosteric heat of adsorption**

426 Infrared Spectroscopy – Materials Science, Engineering and Technology

could be considered operating under differential mode as Continuous flow Stirred Tank Reactor (CSTR). This approach permit creation of appropriate and accurate conditions that avoid limitation due to mass and heat transfer within the adsorbent particles in the reactor bed (Zaitan et al, 2008). Consequently, the experiments could be considered as occurring under isothermal conditions in which the pressure drop as well as the variation in fluid velocity between the reactor inlet and outlet were considered negligible. Thus, the adsorbed

at

0 in

(1)

(2)

(3)

in out ads <sup>a</sup>

FC <sup>C</sup> n t dt m C

Where nads is the adsorbed mole number of toluene at saturation (adsorption capacity in mol*/*g , Cin and Cout the molar fractions of toluene at reactor inlet and outlet, m is the

The adsorbed amount measured from breakthrough curve is generally used to indicate adsorbent performance in terms of a given constituent removal from a flowing stream. In the present study, the adsorption capacity is obtained by integration of the breakthrough curve, according to the equation (1) and considering the curve corresponding to the reactor response in the absence of solid. Accordingly, the numerical integration of the breakthrough curve using MathCAD software yields to a total adsorbed amount of 420 µmol g–1, for adsorption carried out with a flow containing 0.36% *o*-xylene in N2 (Fig. 3 part A). This amount is lower than the adsorption capacity of 1958 µmol g–1 reported in our previous work for SiO2 Degussa (200 m2g-1) (Zaitan et al 2005). In comparison with the literature, substantially higher values of 4666, 2800, 1800 and 2050 µmol g-1 are reported for xylene adsorption, respectively, for activated carbon AC40 (1300 m2/g)( Benkhedda et al, 2000) and zeolithes (Al-Meso 100 (915

amount can be calculated using the mass balance equation (Ruthven, 1984).

absorbent mass, ta the saturation time and F the gaseous molar flow rate.

m2/g) UL-ZSM5-100-2 (840 m2/g), UL-ZSM5-100-6 (780 m2/g) (Huang et al, 2000)

fraction (n1ads) of 370 µmol.g–1 released during isothermal desorption.

equation (3), permits obtaining an irreversible amount of 60 µmol.g–1.

*n ads*

Where ta and tb correspond to starting and ending time of isothermal desorption.

The numerical integration of the desorption curve (part B of Figure 3) using equation (2) and considering the curve in the absence of solid, permitted quantification of loosely adsorbed

b

t in out <sup>1</sup>

a

b

t in

tc

in out 2ads

FC <sup>C</sup> n . dt m C

It is to be noted that although the *o*-xylene desorption curve reached zero, the calculated weakly adsorbed fraction represents 88% of the total adsorbed amount. The remaining oxylene adsorbed amount corresponds to the more strongly adsorbed fraction. The removal of the latter fraction requires thermal treatment with linear heating rate according to TPD method. The integration of the corresponding curve, shown in part C of Figure 3 using

t in FC <sup>C</sup> dt m C

The amounts adsorbed at equilibrium corresponding to adsorption capacity obtained from breakthrough curves, were determined for different adsorption temperatures and for different xylene pressures, yields to isotherms shown in Figure 4. The isotherms are represented in the form of *N* f P where N is the adsorbed amount per adsorbent weight

at equilibrium and P the O-xylene partial pressure in the mixture flow. The experimental isotherms were modeled with Langmuir and Freundlich equations using a nonlinear regression method (MathCAD software). The corresponding fitting curves are shown as solid and dashed lines, respectively, for Langmuir and Freundlich models (Fig. 4). Apparently, the experimental data were well represented by Langmuir model while Freundlich equation deviates at pressures higher than 4 torr.

The adsorption isotherms presented in Figure 4 were used for determination of isosteric heat (Qst) of adsorption by extrapolation at different temperatures and for a given coverage according to Clausius-Clapeyron equation (Rouquerol, 1999):

$$\mathbf{Q}\_{\rm st} = -R \left( \frac{\partial \ln P}{\partial \left( \mathbf{J}\_T' \right)} \right) \tag{4}$$

Where R is the perfect gas constant, P and T correspond, respectively; to partial pressure and temperature at equilibrium.

Determination of Adsorption Characteristics of

physisorption adsorption.

Volatile Organic Compounds Using Gas Phase FTIR Spectroscopy Flow Analysis 429

isosteres correspond to straight lines, which depends strongly on the accuracy of experimental conditions. So far, a temperature measurement error of 2 K may yield to an uncertainty of 8

For temperatures ranging from 300 to 363 K, values of isosteric heat of adsorption in range of 40 to 44 kJ/mol were obtained for o-xylene loading between 100 and 250 µmol/g corresponding to a coverage between 0.05≤θ≤0.125 (θ=nads/nads m,where nads is the adsorbed amount and nads m is the adsorbed amount at monolayer). Nevertheless, these values do not exceed the heat of vaporization (55 kJ/mole) (CRC Handbook, 1985) indicating a weak

In the following section, the part C of the cycle (Figure 3), corresponding to TPD experiment carried out after adsorbent saturation followed by isothermal desorption, will be used for the determination of the desorption energy. The TPD method proposed by Cvetanovic and Amenomiya (Cvetanovic et al., 1967) was also validated under various experimental conditions by several authors (Yang et al.1999; Joly et al.2000; Kanervo et al.2006, Yoshimoto et al.2007). The method allows an easier estimation of the binding energy between the adsorbate and the adsorbent by plotting the desorption rate of the adsorbate as function of temperature. Although, a criticism is reported concerning the assumption related with experimental conditions permitting to achieve TPD measurement free of the influence of

The analysis of TPD curves collected at different linear heating rates (*β*), is based on exploitation of the shift of temperature at desorption peak maximum (Tm) as a function of *β*. Hence the activation energy for desorption Ed is extracted from the slope of the line obtained

 2lnTm – lnβ = Ed/RTm + constant (5) Where Tm is the desorption temperature at peak maximum (K), *β* is the linear rate of temperature rise (K/min), Ed is the desorption energy (kJ/mol) and R is the perfect gas

Therefore, in order to provide reliable data on the desorption energy, the TPD curves have to be well defined, with clearly detectable *T*m positions. This was the case of our experiment

Table 1 gives the *T*m positions obtained with different heating rates *β* during the TPD (part C of the cycle). For this purpose additional experiments set of adsorption and desorption cycles at 300K were carried out with the same adsorption pressure of 2.72 Torr of O-xylene (molar concentration of 0.36% in N2) but with the TPD experiment performed using

Β (K/min) 4 5 7 10 12 Tm (K) 350 353 356 361 364

Table 1. Linear heating rate β used for TPD experiments and its corresponding Temperature

kJ/mol for the isosteric heat calculation (Bülow et al, 2002 and Dulaurent et al 2000).

**2.5 Estimation of the desorption energy with TPD method**

diffusion and readsorption (Gorte et al.1996).

from the following equation:

as shown in part C of Figure 3.

at Peak maximum Tm (K)

different linear heating rate *β* (curves not shown).

constant (kJ/molK).

\_\_\_\_Langmuir equation ------- Freundlich model

Fig. 4. Experimental and Modeled adsorption isotherms of O-xylene at different temperatures

This approach do not requires an assumption on model fitting with experimental data even though we have shown that Langmuir equation described well the adsorption process in the studied T, P ranges. Thus, (Qst) values were extracted from the slops of isosteres (Figure 5) representing the plot of ln P = <sup>1</sup> *<sup>f</sup> <sup>T</sup>* . One has to check very carefully whether the resulting

●, 206; ■, 133; ×, 100 µmol g-1

Fig. 5. Plot of isostere corresponding to ln(P) versus1/T corresponding to the following adsorbed amounts

Fig. 4. Experimental and Modeled adsorption isotherms of O-xylene at different

 

This approach do not requires an assumption on model fitting with experimental data even though we have shown that Langmuir equation described well the adsorption process in the studied T, P ranges. Thus, (Qst) values were extracted from the slops of isosteres (Figure 5)

Fig. 5. Plot of isostere corresponding to ln(P) versus1/T corresponding to the following

. One has to check very carefully whether the resulting

\_\_\_\_Langmuir equation ------- Freundlich model

representing the plot of ln P = <sup>1</sup> *<sup>f</sup> <sup>T</sup>*

●, 206; ■, 133; ×, 100 µmol g-1

adsorbed amounts

temperatures

isosteres correspond to straight lines, which depends strongly on the accuracy of experimental conditions. So far, a temperature measurement error of 2 K may yield to an uncertainty of 8 kJ/mol for the isosteric heat calculation (Bülow et al, 2002 and Dulaurent et al 2000).

For temperatures ranging from 300 to 363 K, values of isosteric heat of adsorption in range of 40 to 44 kJ/mol were obtained for o-xylene loading between 100 and 250 µmol/g corresponding to a coverage between 0.05≤θ≤0.125 (θ=nads/nads m,where nads is the adsorbed amount and nads m is the adsorbed amount at monolayer). Nevertheless, these values do not exceed the heat of vaporization (55 kJ/mole) (CRC Handbook, 1985) indicating a weak physisorption adsorption.

### **2.5 Estimation of the desorption energy with TPD method**

In the following section, the part C of the cycle (Figure 3), corresponding to TPD experiment carried out after adsorbent saturation followed by isothermal desorption, will be used for the determination of the desorption energy. The TPD method proposed by Cvetanovic and Amenomiya (Cvetanovic et al., 1967) was also validated under various experimental conditions by several authors (Yang et al.1999; Joly et al.2000; Kanervo et al.2006, Yoshimoto et al.2007). The method allows an easier estimation of the binding energy between the adsorbate and the adsorbent by plotting the desorption rate of the adsorbate as function of temperature. Although, a criticism is reported concerning the assumption related with experimental conditions permitting to achieve TPD measurement free of the influence of diffusion and readsorption (Gorte et al.1996).

The analysis of TPD curves collected at different linear heating rates (*β*), is based on exploitation of the shift of temperature at desorption peak maximum (Tm) as a function of *β*. Hence the activation energy for desorption Ed is extracted from the slope of the line obtained from the following equation:

$$\text{2In}\,\text{T}\_{\text{m}} - \text{ln}\,\theta = \text{E}\_{\text{d}}/\text{RT}\_{\text{m}} + \text{constant} \tag{5}$$

Where Tm is the desorption temperature at peak maximum (K), *β* is the linear rate of temperature rise (K/min), Ed is the desorption energy (kJ/mol) and R is the perfect gas constant (kJ/molK).

Therefore, in order to provide reliable data on the desorption energy, the TPD curves have to be well defined, with clearly detectable *T*m positions. This was the case of our experiment as shown in part C of Figure 3.

Table 1 gives the *T*m positions obtained with different heating rates *β* during the TPD (part C of the cycle). For this purpose additional experiments set of adsorption and desorption cycles at 300K were carried out with the same adsorption pressure of 2.72 Torr of O-xylene (molar concentration of 0.36% in N2) but with the TPD experiment performed using different linear heating rate *β* (curves not shown).


Table 1. Linear heating rate β used for TPD experiments and its corresponding Temperature at Peak maximum Tm (K)

Determination of Adsorption Characteristics of

**3. Conclusion** 

facilities.

**4. References** 

(3) 411

196,157–172

42.

al, 1997, Cardona-Martinez and Dumesic, 1997).

Volatile Organic Compounds Using Gas Phase FTIR Spectroscopy Flow Analysis 431

differential heat of adsorption provided by microcalorimetric methods (Simonot-Grange et

The use of FTIR spectroscopy was found to permit accurate quantitative analysis that allows determination of adsorbent saturation loading from breakthrough curves and differentiation between weakly and strongly physisorption. The experimental methodology allows, also, the possibility of simultaneous detection of further adsorbate dissociation through appearance of new IR bands. The data obtained using this analytical approach have been used to derivate useful thermodynamic parameters related to the heat involved during adsorption and/or desorption processes. It was shown that significant difference might be obtained for the values of the heat of adsorption depending on to the nature of the involved porosity. This aspect need to be considered for designing and operating of adsorption

Benkhedda, J., Jaubert, J. N., Barth, D., Perrin, L., Bailly, M., J. Chem. Thermodyn. (2000) 32

Brasseur, A.; Gambin, A.; Laudet, A.; Marien, J.; Pirard, J.P. Elaboration of new formulations to remove micropollutants in MSWI flue gas. Chemosphere 2004, 56, 745-756. Bülow, M.; Shen, D.; Jale, S. Measurement of sorption equilibria under isosteric

Cardona-Martinez, N.; Dumesic, J.A.; Microcalorimetric measurements of differential heats of adsorption on reactive catalyst surfaces. Thermochimica Acta. 1997, 1,3, 43-53 . Chafik T., Dulaurent O., Gass J.L., and Bianchi D.,(1998), Heat of adsorption of carbon

Cvetanovic, R. J. and Amenomiya, Y. (1967) Application of a Temperature-Programmed

Pines, H. and Weisz, P. B. Eds.; Academic Press: New York, Vol. 17, p. 103.

Do, D. D. Adsorption Analysis: Equilibria and kinetics; Series on Chemical Engineering 2;

Dulaurent, O.; Bianchi, D. Adsorption isobars for CO on a Pt/Al2O3 catalyst at high

EPA,( May 1998) Chemical Emergency Preparedness and Prevention Office. *Fire hazard from* 

Joly, J.P. and Perrard, A. (2000) Determination of the heat of adsorption of ammonia on

Harti, S.; Cifredo; G.; Gatica J.M; Vidal H, and Chafik, T. Physicochemical characterization

Spectroscopy at high temperatures, J. Catal., 179(2) 503

CRC Handbook, 65 th ed.; CRC Press Inc, Boca Ratton, FL, 1985

model. Appl. Catal. A: General 2000, 196, 271-280

monoliths. Applied Clay Science (2007) 36, 287–296

*carbon adsorption deodorizing systems. EPA 550-F-97-002-e.* 

Imperial College Press: London, 1998.

conditionsThe principles, advantages and limitations. Appl. Surf. Sci. 2002,

monoxide on Pt/Rh/CeO2/Al2O3 Three way catalyst using in-situ Infrared

Desorption Technique to Catalyst Studies. In Advances in Catalysis; Eley, D. D.,

temperatures using FTIR spectroscopy: isosteric heat of adsorption and adsorption

zeolites from temperature-programmed desorption experiments. Langmuir 1538–

and adsorptive properties of some Moroccan clay minerals extruded as lab-scale

As observed in Figure 6, a linear and positive relationship was observed between 2lnTm – ln β and 1/*T*m, and a value for the heat of desorption of 75 kJ/mol is obtained from the slope of this plot. This value is higher than those obtained using isosteric method. As stated above, the analytical procedure adopted in the present work permitted quantification of weakly and strongly adsorbed fractions occurring within the predisposed porosity of the clay. Moreover, previous work on textural characterisation revealed that the studied clay is mainly mesoporous and presents wider pore sizes distribution with a significant contribution of pores widths around 9 nm and 40nm (Harti et al, 2007). It is known that the efficiency of an adsorbent depends on its pore structure and the size of adsorbate molecules which affect their diffusion and interaction within the pores. Thus, adsorption in larger pores is associated with lower heats of adsorption corresponding to the weakly adsorbed fraction (n1ads) which represents about 88% of the total adsorbed. While the remaining adsorbed fraction corresponds to the more strongly adsorbed physisorbed species trapped in deeper potentials of predisposed micropores and narrowest mesopores. The resulting heat of adsorption values might be enhanced by "confinement effect", particularly, in case of physisorption of molecules disposing with size similar to those of adsorbent pores (Yoshimoto et al.2007). As shown in the present work, these species will need supplying thermal energy for their transfer to the gas phase, corresponding to the desorption energy which is equal to the heat of adsorption in case of physisorption.

Fig. 6. Cvetanovic curve obtained with TPD experiments performed with β and Tm values presented in table 2 for an adsorption test performed with a flow containing 3600 ppmv of xylene

On the other hand, it is to be noted that the use of Clausius Clapeyron for the estimation of the heat of adsorption, is efficient and accurate when the adsorption process is essentially reversible. This approach is no longer valid for stronger adsorption. As shown in the present work, significant difference might be obtained in the values of the heat of physisorption depending on the nature of the involved porosity. Thus, the analytical method must be able to quantify different adsorbed fractions (i.e. weakly and strongly pysisorption) or else, at best, only an average of the heats of adsorption is obtained, as it is the case of integral or differential heat of adsorption provided by microcalorimetric methods (Simonot-Grange et al, 1997, Cardona-Martinez and Dumesic, 1997).

## **3. Conclusion**

430 Infrared Spectroscopy – Materials Science, Engineering and Technology

As observed in Figure 6, a linear and positive relationship was observed between 2lnTm – ln β and 1/*T*m, and a value for the heat of desorption of 75 kJ/mol is obtained from the slope of this plot. This value is higher than those obtained using isosteric method. As stated above, the analytical procedure adopted in the present work permitted quantification of weakly and strongly adsorbed fractions occurring within the predisposed porosity of the clay. Moreover, previous work on textural characterisation revealed that the studied clay is mainly mesoporous and presents wider pore sizes distribution with a significant contribution of pores widths around 9 nm and 40nm (Harti et al, 2007). It is known that the efficiency of an adsorbent depends on its pore structure and the size of adsorbate molecules which affect their diffusion and interaction within the pores. Thus, adsorption in larger pores is associated with lower heats of adsorption corresponding to the weakly adsorbed fraction (n1ads) which represents about 88% of the total adsorbed. While the remaining adsorbed fraction corresponds to the more strongly adsorbed physisorbed species trapped in deeper potentials of predisposed micropores and narrowest mesopores. The resulting heat of adsorption values might be enhanced by "confinement effect", particularly, in case of physisorption of molecules disposing with size similar to those of adsorbent pores (Yoshimoto et al.2007). As shown in the present work, these species will need supplying thermal energy for their transfer to the gas phase, corresponding to the desorption energy

Fig. 6. Cvetanovic curve obtained with TPD experiments performed with β and Tm values presented in table 2 for an adsorption test performed with a flow containing 3600 ppmv of

On the other hand, it is to be noted that the use of Clausius Clapeyron for the estimation of the heat of adsorption, is efficient and accurate when the adsorption process is essentially reversible. This approach is no longer valid for stronger adsorption. As shown in the present work, significant difference might be obtained in the values of the heat of physisorption depending on the nature of the involved porosity. Thus, the analytical method must be able to quantify different adsorbed fractions (i.e. weakly and strongly pysisorption) or else, at best, only an average of the heats of adsorption is obtained, as it is the case of integral or

which is equal to the heat of adsorption in case of physisorption.

xylene

The use of FTIR spectroscopy was found to permit accurate quantitative analysis that allows determination of adsorbent saturation loading from breakthrough curves and differentiation between weakly and strongly physisorption. The experimental methodology allows, also, the possibility of simultaneous detection of further adsorbate dissociation through appearance of new IR bands. The data obtained using this analytical approach have been used to derivate useful thermodynamic parameters related to the heat involved during adsorption and/or desorption processes. It was shown that significant difference might be obtained for the values of the heat of adsorption depending on to the nature of the involved porosity. This aspect need to be considered for designing and operating of adsorption facilities.

## **4. References**


**23** 

*University of Stellenbosch* 

*South Africa* 

**Identification of Rocket Motor** 

N. Hamp, J.H. Knoetze, C. Aldrich\* and C. Marais

**Characteristics from Infrared Emission Spectra** 

The prediction of infrared (IR) emission spectra from the exhaust gases of rocket plumes finds numerous applications in the strategic identification of rockets. These rocket fingerprints could be classified, thus allowing for the distinction between friend and foe. Likewise, the plume radiation intensity could also be reduced for stealth purposes, where accurate prediction of the spectra could be used to determine whether rockets have the required stealth characteristics during their design phase already. This would reduce the

The challenge of predicting the plume radiance is describing the thermodynamic combustion process within the rocket chamber, the plume structure and the rocket plume chemical composition. The factors guiding these processes are the rocket motor design parameters, as well as the rocket motor fuel chemistry. In addition, environmental conditions have a

Previously, attempts were made to model the middle IR band emission spectra (2 to 5.5 μm) from the rocket fuel chemistry and the physical properties during combustion by making use of techniques such as quantum mechanics and computational fluid dynamics. These methods proved to be too time consuming and the accuracies of the predictions were not

More recently, Roodt (1998) was the first to show that the IR spectra could be modelled with a multilayer perceptron neural network using the elemental composition and other physical properties of the rocket motor fuel as input. Although these models were successful, there were some indications that they were not optimal and in this investigation the use of multilayer perceptrons similar to the ones used by Roodt (1998), as well as linear partial least squares (PLS) and neural network PLS (with and without weight updating) are considered.

In addition, the modelling problem is considered in terms of a forward mapping, i.e. prediction of the emission spectra of the rockets from their design parameters, as well as a reverse mapping, where the rocket design parameters are predicted from the middle-IR

significant impact on the plume structure and the plume chemical composition.

high manufacturing and testing costs involved in later stages.

**1. Introduction** 

acceptable (Roodt, 1998).

Corresponding Author

 \*

spectral absorbances of the rocket plume.


## **Identification of Rocket Motor Characteristics from Infrared Emission Spectra**

N. Hamp, J.H. Knoetze, C. Aldrich\* and C. Marais *University of Stellenbosch South Africa* 

## **1. Introduction**

432 Infrared Spectroscopy – Materials Science, Engineering and Technology

Huang, Q., Vinh-Thang, H., Melekian, A., Eic, M., Trong-On, D., Kaliaguine, S., Micropor.

Kanervo, J.M., Keskitalo, T.J., Slioor, R.I. and Krause, A.O.I.(2006) Temperature-

Gorte, R. J. (1996) Temperature-programmed desorption for oxide catalysts. Catalysis

Simonot-Grange, M.H.; Bertrand, O.; Pilverdier, E.; Bellat, J.P.; Paulin, C. Differential

Sircar, S. , Mohr, R., Ristic, C. and Rao, M. B. (1999) Isosteric Heat of Adsorption: Theory

Ranke, W. and Josephy, Y. (2002) Determination of Adsorption Energies and Kinetic Parameters by Isosteric Methods. *Phys. Chem. Chem. Phys. 4* ,2483–2498. Rouquerol, F. , Rouquerol, J. and K.S.W. Sing, *Adsorption by Powders and Porous Solids*, Academic Press (Eds.), Harcourt Brace & Company, London, (1999) pp. 47–49. Yang R.T., Long Q., Padin J., Takahashi A., Takahashi T., Adsorbents for Dioxins (1999) A

Yoshimoto R., Hara K.,Okumura K., Katada N., M. Niwa M., (2007) Analysis of Toluene

Yang R.T., Long Q., Padin J., Takahashi A., Takahashi T., Adsorbents for Dioxins (1999) A

Zaitan H., Chafik T., FTIR Determination of Adsorption Characteristics for Volatile Organic

Zaitan, H.; Feronnato, C.; Bianchi, D.; Achak, O.; Chafik, T. Etude des propriétés texturales

Zaitan H., Bianchi D., Ouafae Achak O., and Chafik T., A comparative study of the

information for porous catalysts, Journal of Catalysis 382–393

programmed desorption as a tool to extract quantitative kinetic or energetic

calorimetric enthalpies of adsorption of p-xylene and m-xylene on Y faujasites at

New Technique for Sorbent Screening for Low-Volatile Organics. *Ind. Eng. Chem.* 

Adsorption on Na-Form Zeolite with a Temperature-Programmed Desorption

New Technique for Sorbent Screening for Low-Volatile Organics. *Ind. Eng. Chem.* 

Compounds Removal on Diatomite Mineral Compared to Commercial Silica (2005)

et adsorbantes d'une diatomite marocaine: Application au traitement D'air chargé d'un polluant de type composé organique volatil. Ann. Chim. Sci. Mat. 2006, 31(2),

adsorption and desorption of o-xylene onto bentonite clay and alumina. Journal of

Mesopor. Mater., 87 (2006) 224

25°C. J. Therm. Anal. 1997, 48, 741-754.

Method. *J. Phys. Chem. 111*, 1474-1479.

Hazardous Materials 153 (2008) 852–859

and Experiment. *J. Phys. Chem. B 103* ,6539–6546.

Today 405-414.

*Res. 38*, 2726-2731.

*Res. 38*, 2726-2731.

183-196.

*C.R. Chimie 8*, 1701–1708.

The prediction of infrared (IR) emission spectra from the exhaust gases of rocket plumes finds numerous applications in the strategic identification of rockets. These rocket fingerprints could be classified, thus allowing for the distinction between friend and foe. Likewise, the plume radiation intensity could also be reduced for stealth purposes, where accurate prediction of the spectra could be used to determine whether rockets have the required stealth characteristics during their design phase already. This would reduce the high manufacturing and testing costs involved in later stages.

The challenge of predicting the plume radiance is describing the thermodynamic combustion process within the rocket chamber, the plume structure and the rocket plume chemical composition. The factors guiding these processes are the rocket motor design parameters, as well as the rocket motor fuel chemistry. In addition, environmental conditions have a significant impact on the plume structure and the plume chemical composition.

Previously, attempts were made to model the middle IR band emission spectra (2 to 5.5 μm) from the rocket fuel chemistry and the physical properties during combustion by making use of techniques such as quantum mechanics and computational fluid dynamics. These methods proved to be too time consuming and the accuracies of the predictions were not acceptable (Roodt, 1998).

More recently, Roodt (1998) was the first to show that the IR spectra could be modelled with a multilayer perceptron neural network using the elemental composition and other physical properties of the rocket motor fuel as input. Although these models were successful, there were some indications that they were not optimal and in this investigation the use of multilayer perceptrons similar to the ones used by Roodt (1998), as well as linear partial least squares (PLS) and neural network PLS (with and without weight updating) are considered.

In addition, the modelling problem is considered in terms of a forward mapping, i.e. prediction of the emission spectra of the rockets from their design parameters, as well as a reverse mapping, where the rocket design parameters are predicted from the middle-IR spectral absorbances of the rocket plume.

<sup>\*</sup> Corresponding Author

Identification of Rocket Motor Characteristics from Infrared Emission Spectra 435

This is a useful expression, since **T** can be expressed in terms of **W**, the PLS input weights, without having to break down **X** into its residuals for each latent dimension. A matrix of linear inner model regression parameters on the diagonal and zero-values off the diagonal,

where **B**PLS is the m × l matrix of overall regression coefficients which converges to multiple

Artificial neural networks (ANNs) are a non-linear function mapping technique that was initially developed to imitate the brain from both a structural and computational perspective. Its parallel architecture is primarily responsible for its computational power. The multilayer

A multilayer perceptron neural network (Bishop, 1995; Haykin, 1999) consists of an input and an output layer of nodes, which may be separated by one or more layers of hidden nodes (see figure 1 below). Each node links to another node with a weighted connection, *ω*. Considering a network with a single hidden layer, where the hidden and output layers are denoted by superscripts (1) and (2) respectively, then for r = 1, 2, …, H hidden nodes the

(�)��� �� ����

�

���

(�)�������

(�) =����(�) (10)

(�)

�

� � ���(���) � (9)

� ��

(�) (8)

φis a

**B** can now be defined. Equations (5) and (4) can further be used to obtain

perceptron network architecture is probably the most popular and is used here.

(�)�=����

�(�) = tanh(�) <sup>=</sup> � � ���(���)

��

�

���

Here Ω(1) is the H × m matrix of weights (ωrk) in the hidden layer, ω represents a vector of weights for a single node and β is a bias value associated with each node. The function

The advantage of this form of the function is that its derivative is simple to calculate, i.e.

weights of the network are updated within the neural network PLS algorithm below.

This derivative form becomes useful when calculating the Jacobian matrix used when the

The performance of an ANN is measured by the root-mean-square error (*RMSE*) which is also the function to be minimised. The Levenberg-Marquardt optimization algorithm (Marquardt, 1963) and resilient propagation algorithm (RPROP) (Riedmiller & Braun, 1993)

nonlinear functional relationship is represented by equation (8):

(�)� �(�)� ��

sigmoidal activation function, typically of the form:

were used to train the neural networks in this study.

linear regression coefficients for h = m.

��� ���� �(�)� ��

**2.2 Multilayer perceptron neural networks** 

� = �(���) (6)

�� = ���� = ����� = ����� (7)

## **2. Partial least squares (PLS) and neural network models**

#### **2.1 Linear PLS**

The advantage of PLS lies in the fact that a multivariate regression problem can be decomposed into a number of uncorrelated univariate or SISO (single input, single output) data mappings. This is especially useful when the available data are sparse, such as when dealing with relatively small sets of samples across many highly correlated input, as well as output variables.

The linear PLS algorithm has various forms like the one given by Lorber et al. (1987). The nonlinear iterative partial least squares (NIPALS) algorithm for training PLS models was pioneered by H. Wold (1966). The NIPALS algorithm may be computationally less efficient, but it is well understood and serves as the basis for nonlinear neural network PLS algorithms. The objective of the NIPALS algorithm is to project input and output matrices **X** and **Y**  (consisting of rows corresponding to data sample points i = 1, 2, … n**)** onto a subset of latent variables, **T** and **U**, which are referred to as the input and output scores, respectively. The dimensionalities of variable spaces of **X** and **Y** are denoted by k = 1, 2, … m and j = 1, 2, … l respectively. The output scores can then be fitted to the input scores by linear least squares regression in order to obtain the so-called inner linear relationship coefficients, ba for a = 1, 2, … h:

$$\mathbf{u}\_a = \mathbf{t}\_a b\_a + \mathbf{e}\_a \tag{1}$$

Here the h primary latent dimensions explaining most of the model variance are retained. The decompositions of **X** and **Y** can be defined using the loading vectors **p** and **q** such that PLS outer models become:

$$X = \sum\_{a=1}^{h} \mathbf{t}\_a \mathbf{p}\_a^T + F \tag{2}$$

$$Y = \sum\_{a=1}^{h} \hat{u}\_a q\_a^T + E \tag{3}$$

The matrices, **F** and **E** are the resulting residual matrices when a model with h ≤ min(n,m) latent dimensions is used for the approximation of **X** and the prediction of **Y**. The remaining latent dimensions usually explain random noise that may be present in the data. The predicted scores of **u** are calculated using the inner model

$$\mathfrak{u}\_a + \mathfrak{t}\_a b\_a \tag{4}$$

The linear projections constituting the NIPALS algorithm (see Appendix A) are described in Baffi et al. (1999a), where it is further shown that the n × h score matrix, **T** can be related to the input matrix, **X** by

$$T = XR\tag{5}$$

where **R** is obtained from

$$\mathcal{R} = \mathcal{W}(\mathbf{P}^T \mathbf{W}) \tag{6}$$

This is a useful expression, since **T** can be expressed in terms of **W**, the PLS input weights, without having to break down **X** into its residuals for each latent dimension. A matrix of linear inner model regression parameters on the diagonal and zero-values off the diagonal, **B** can now be defined. Equations (5) and (4) can further be used to obtain

$$
\tilde{Y} = \tilde{U}Q^T = XRBQ^T = XB\_{PLS} \tag{7}
$$

where **B**PLS is the m × l matrix of overall regression coefficients which converges to multiple linear regression coefficients for h = m.

#### **2.2 Multilayer perceptron neural networks**

434 Infrared Spectroscopy – Materials Science, Engineering and Technology

The advantage of PLS lies in the fact that a multivariate regression problem can be decomposed into a number of uncorrelated univariate or SISO (single input, single output) data mappings. This is especially useful when the available data are sparse, such as when dealing with relatively small sets of samples across many highly correlated input, as well as

The linear PLS algorithm has various forms like the one given by Lorber et al. (1987). The nonlinear iterative partial least squares (NIPALS) algorithm for training PLS models was pioneered by H. Wold (1966). The NIPALS algorithm may be computationally less efficient, but it is well understood and serves as the basis for nonlinear neural network PLS algorithms. The objective of the NIPALS algorithm is to project input and output matrices **X** and **Y**  (consisting of rows corresponding to data sample points i = 1, 2, … n**)** onto a subset of latent variables, **T** and **U**, which are referred to as the input and output scores, respectively. The dimensionalities of variable spaces of **X** and **Y** are denoted by k = 1, 2, … m and j = 1, 2, … l respectively. The output scores can then be fitted to the input scores by linear least squares regression in order to obtain the so-called inner linear relationship coefficients, ba for

Here the h primary latent dimensions explaining most of the model variance are retained. The decompositions of **X** and **Y** can be defined using the loading vectors **p** and **q** such that

்

ൌ

ୀଵ

ෝ ൌ

ୀଵ

predicted scores of **u** are calculated using the inner model

The matrices, **F** and **E** are the resulting residual matrices when a model with h ≤ min(n,m) latent dimensions is used for the approximation of **X** and the prediction of **Y**. The remaining latent dimensions usually explain random noise that may be present in the data. The

The linear projections constituting the NIPALS algorithm (see Appendix A) are described in Baffi et al. (1999a), where it is further shown that the n × h score matrix, **T** can be related to

(1) ܾ ൌ

(4) ܾ ෝ

(5) ൌ

(2)

(3)

**2. Partial least squares (PLS) and neural network models** 

**2.1 Linear PLS** 

output variables.

a = 1, 2, … h:

PLS outer models become:

the input matrix, **X** by

where **R** is obtained from

Artificial neural networks (ANNs) are a non-linear function mapping technique that was initially developed to imitate the brain from both a structural and computational perspective. Its parallel architecture is primarily responsible for its computational power. The multilayer perceptron network architecture is probably the most popular and is used here.

A multilayer perceptron neural network (Bishop, 1995; Haykin, 1999) consists of an input and an output layer of nodes, which may be separated by one or more layers of hidden nodes (see figure 1 below). Each node links to another node with a weighted connection, *ω*.

Considering a network with a single hidden layer, where the hidden and output layers are denoted by superscripts (1) and (2) respectively, then for r = 1, 2, …, H hidden nodes the nonlinear functional relationship is represented by equation (8):

$$f\_{lj}\left(\mathbf{x}\_{l\cdot},\mathbf{0}^{(1)},\varpi\_{j}^{(2)},\mathbf{\mathcal{J}}^{(1)},\boldsymbol{\mathcal{J}}\_{j}^{(2)}\right) = \sum\_{r=1}^{H} \varpi\_{lr}^{(2)} \phi\_{lr} \left\{ \sum\_{k=1}^{m} \left(\varpi\_{rk}^{(1)} \mathbf{x}\_{lk}\right) + \boldsymbol{\mathcal{J}}\_{r}^{(1)} \right\} + \boldsymbol{\mathcal{J}}\_{j}^{(2)} \tag{8}$$

Here Ω(1) is the H × m matrix of weights (ωrk) in the hidden layer, ω represents a vector of weights for a single node and β is a bias value associated with each node. The function φ is a sigmoidal activation function, typically of the form:

$$\left|\phi(\mathbf{z}) = \tanh(\mathbf{z})\right| = \begin{array}{c} 1 - \exp(-2\mathbf{z}) \\ \end{array} \Big|\, 1 + \exp(-2\mathbf{z}) \tag{9}$$

The advantage of this form of the function is that its derivative is simple to calculate, i.e.

$$\phi'(\mathbf{z}) = 1 - \phi^2(\mathbf{z}) \tag{10}$$

This derivative form becomes useful when calculating the Jacobian matrix used when the weights of the network are updated within the neural network PLS algorithm below.

The performance of an ANN is measured by the root-mean-square error (*RMSE*) which is also the function to be minimised. The Levenberg-Marquardt optimization algorithm (Marquardt, 1963) and resilient propagation algorithm (RPROP) (Riedmiller & Braun, 1993) were used to train the neural networks in this study.

Identification of Rocket Motor Characteristics from Infrared Emission Spectra 437

Fig. 2. Diagram illustrating the NNPLS algorithm wherein data are transformed to latent scores, then neural networks used to learn the scores (adapted from Qin & McAvoy, 1992).

in equation (4) with a feed-forward multilayer perceptron neural network, such that

improve the weight updating procedure originally suggested by S. Wold et al. (1998).

**3. Experimental data** 

nozzle diameter to the nozzle throat diameter (EC).

A neural network has the advantage that it is a universal approximator and the inner PLS model is therefore not limited to some predefined functional form. In Qin & McAvoy (1992) the neural network PLS (NNPLS) algorithm is introduced by replacing the linear inner relationship

The NIPALS algorithm now replaces the inner linear regression coefficient calculation (Appendix A, step x) by a neural network training step. The use of a nonlinear function as inner PLS relationship influences both the inner and outer mappings of the PLS algorithm. If the inner mapping is highly nonlinear, this approach may no longer be acceptable. This problem was addressed by S. Wold et al. (1998) by updating the PLS weights, **w** using a complicated, nonintuitive Taylor series linearization method. More recently, Baffi et al. (1999b) proposed an error-based (EB) input weight (**w**) updating procedure using a Taylor series expansion to

The data set of rocket motor features consisted of 14 elemental rocket propellant compositions and 4 rocket motor design parameters. The elemental compositions were molar values calculated from a 100 kg basis and included the elements C, H, O, N, Al, K, F, Cu, Pb, S, Cl, Si, Ti and Fe. The design parameters consisted of the nozzle throat temperature (TC), pressure (PC), nozzle diameter (DT) and the expansion ratio of the outlet

ෝ ൌ ݂൫ǡ ࣚሺଵሻǡ ࣚሺଶሻǡ ࢼሺଵሻǡ ߚሺଶሻ൯ (13)

Fig. 1. Multi-layered perceptron neural network with one hidden layer.

$$RMSE = \sqrt{\overbrace{\sum\_{l=1}^{n} SSE\_{l}}} \Big/\_{nl} \tag{11}$$

*n* refers to the training vector number (i.e. observation) and *SSEi* is the sum-square error of the *i*th training vector for all *l* output nodes:

$$SSE\_l = \sum\_{j=1}^{l} \left( \mathbf{y}\_{true,j}^l - \mathbf{y}\_{pred,j}^l \right)^2 \tag{12}$$

The weight matrices are initially randomised. A subset of the input dataset is applied to the network input nodes and the outputs of the hidden and output nodes are calculated. The *SSE* is calculated as in equation 12 upon which the weight matrices are updated using the optimisation framework. The procedure is repeated for the remaining input dataset to calculate the *RMSE* which completes a single iteration. A number of these iterations are necessary to minimise the *RMSE*.

#### **2.3 Neural network PLS**

When applying linear PLS to nonlinear problems, it may not be sensible to discard the minor latent dimensions, as they may contain valuable information with regard to the mapping. It may therefore be advantageous to derive a nonlinear relationship for the PLS inner model. This can be accomplished by use of a multilayer perceptron neural network such as described above and illustrated in figure 2.

Fig. 1. Multi-layered perceptron neural network with one hidden layer.

the *i*th training vector for all *l* output nodes:

necessary to minimise the *RMSE*.

such as described above and illustrated in figure 2.

**2.3 Neural network PLS** 

���� � �<sup>∑</sup> ���� � ���

���� � ���������

���

�

*n* refers to the training vector number (i.e. observation) and *SSEi* is the sum-square error of

The weight matrices are initially randomised. A subset of the input dataset is applied to the network input nodes and the outputs of the hidden and output nodes are calculated. The *SSE* is calculated as in equation 12 upon which the weight matrices are updated using the optimisation framework. The procedure is repeated for the remaining input dataset to calculate the *RMSE* which completes a single iteration. A number of these iterations are

When applying linear PLS to nonlinear problems, it may not be sensible to discard the minor latent dimensions, as they may contain valuable information with regard to the mapping. It may therefore be advantageous to derive a nonlinear relationship for the PLS inner model. This can be accomplished by use of a multilayer perceptron neural network

� � �������

��

� � �

� (11)

(12)

Fig. 2. Diagram illustrating the NNPLS algorithm wherein data are transformed to latent scores, then neural networks used to learn the scores (adapted from Qin & McAvoy, 1992).

A neural network has the advantage that it is a universal approximator and the inner PLS model is therefore not limited to some predefined functional form. In Qin & McAvoy (1992) the neural network PLS (NNPLS) algorithm is introduced by replacing the linear inner relationship in equation (4) with a feed-forward multilayer perceptron neural network, such that

$$\mathfrak{u}\_a = f\left(\mathfrak{t}\_a, \varpi^{(1)}, \varpi^{(2)}, \mathfrak{f}^{(1)}, \mathfrak{f}^{(2)}\right) \tag{13}$$

The NIPALS algorithm now replaces the inner linear regression coefficient calculation (Appendix A, step x) by a neural network training step. The use of a nonlinear function as inner PLS relationship influences both the inner and outer mappings of the PLS algorithm. If the inner mapping is highly nonlinear, this approach may no longer be acceptable. This problem was addressed by S. Wold et al. (1998) by updating the PLS weights, **w** using a complicated, nonintuitive Taylor series linearization method. More recently, Baffi et al. (1999b) proposed an error-based (EB) input weight (**w**) updating procedure using a Taylor series expansion to improve the weight updating procedure originally suggested by S. Wold et al. (1998).

#### **3. Experimental data**

The data set of rocket motor features consisted of 14 elemental rocket propellant compositions and 4 rocket motor design parameters. The elemental compositions were molar values calculated from a 100 kg basis and included the elements C, H, O, N, Al, K, F, Cu, Pb, S, Cl, Si, Ti and Fe. The design parameters consisted of the nozzle throat temperature (TC), pressure (PC), nozzle diameter (DT) and the expansion ratio of the outlet nozzle diameter to the nozzle throat diameter (EC).

Identification of Rocket Motor Characteristics from Infrared Emission Spectra 439

A correlation map of the rocket motor design features in figure 4 shows that there is very

1 C

Cl

Si

Ti

Fe

T<sup>c</sup>

P<sup>c</sup>

E<sup>c</sup>

DT

S

Fig. 4. A map of correlation factors of the rocket motor design parameters and chemistry to investigate the presence of potentially redundant correlated information in the underlying

All data in the forward mappings (prediction of emission spectra) were mean-centred and

Although the complete data set consisted of 417 measured IR spectra, it covered only 18 different rockets, i.e. it contained 399 replicates. These replicates were not used in the validation of the models. Instead, leave-one-out cross-validation (Hjorth, 1994) was used to assess the quality of the models, i.e. the set of n (= 18) independent samples was split into n-1 training samples, while the nth point was reserved for model validation. The trainingvalidation split was repeated n times until each data point had been omitted once for validation. A validation set of n predictions on the 'unseen' data was therefore derived from all the available data and a predicted residual estimate sum of squares (PRESS) was

݁ൌ ܵܵܧܴܲ

ୀଵ

ଶ 

(14)





0

0.2

0.4

0.6

0.8

structure.

**4. Construction of models** 

H O N Al K F Cu Pb S Cl Si Ti Fe Tc Pc Ec DT

C

H

O

N

Al

K

F

Cu

calculated on the validation set.

**4.1 Model validation** 

scaled to unit variance during the training of the models.

little correlation between the variables representing the rocket motor parameters.

Pb

The data set of IR emission spectra consisted of radiometer absorbance values at 146 different wavelengths in the middle-IR band (2 to 5.5 μm).

Two types of rocket motor propellants were used, namely a composite (C) and a doublebase (DB) type. The C-type propellants consisted of heterogeneous grains where the fuel and oxidiser were held together in a synthetic rubber matrix. The DB-types had homogeneous grains containing small amounts of dispersed additives. There were 12 Ctype and 6 DB-type rocket motor propellants.

Each rocket motor type was fired a number of times (see table 1) and the IR emission spectra were recorded for each test as replicate measurements. The total set of recorded IR emissionspectra thus comprised 420 measurements. The spectra were recorded by Roodt (1998) using a spectral radiometer at varying distances, i.e. 500 m, 350 m, 250 m and 200 m. The data were preprocessed in order to compensate for the varying absorbance path lengths and atmospheric conditions (Bouguer's law) as described in Roodt (1998).


Table 1. The number of middle-IR emission spectra repeat measurements taken from tests for each of the rocket motor types.

A principal components analysis was done on a standardised IR emission spectrum data set including all 420 data samples. Results showed that 86.7% of the total variance of the wavelength variables could be explained by the first two principal components. The map of squared correlation coefficients in figure 3 confirms this result.

Fig. 3. A map of squared correlation factors of the IR emission spectral absorbance values to investigate the presence of potentially redundant correlated information in the variable space.

A correlation map of the rocket motor design features in figure 4 shows that there is very little correlation between the variables representing the rocket motor parameters.

Fig. 4. A map of correlation factors of the rocket motor design parameters and chemistry to investigate the presence of potentially redundant correlated information in the underlying structure.

### **4. Construction of models**

All data in the forward mappings (prediction of emission spectra) were mean-centred and scaled to unit variance during the training of the models.

#### **4.1 Model validation**

438 Infrared Spectroscopy – Materials Science, Engineering and Technology

The data set of IR emission spectra consisted of radiometer absorbance values at 146

Two types of rocket motor propellants were used, namely a composite (C) and a doublebase (DB) type. The C-type propellants consisted of heterogeneous grains where the fuel and oxidiser were held together in a synthetic rubber matrix. The DB-types had homogeneous grains containing small amounts of dispersed additives. There were 12 C-

Each rocket motor type was fired a number of times (see table 1) and the IR emission spectra were recorded for each test as replicate measurements. The total set of recorded IR emissionspectra thus comprised 420 measurements. The spectra were recorded by Roodt (1998) using a spectral radiometer at varying distances, i.e. 500 m, 350 m, 250 m and 200 m. The data were preprocessed in order to compensate for the varying absorbance path lengths

**DB1 DB2 DB3 DB4 DB5 DB6 C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12**  39 21 31 20 20 22 15 24 24 17 18 26 44 15 25 14 23 22 Table 1. The number of middle-IR emission spectra repeat measurements taken from tests

A principal components analysis was done on a standardised IR emission spectrum data set including all 420 data samples. Results showed that 86.7% of the total variance of the wavelength variables could be explained by the first two principal components. The map of

Fig. 3. A map of squared correlation factors of the IR emission spectral absorbance values to investigate the presence of potentially redundant correlated information in the variable space.

10 20 30 40 50 60 70 80 90 100 110 120 130 140

**Wavelength Number**

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

and atmospheric conditions (Bouguer's law) as described in Roodt (1998).

squared correlation coefficients in figure 3 confirms this result.

different wavelengths in the middle-IR band (2 to 5.5 μm).

type and 6 DB-type rocket motor propellants.

for each of the rocket motor types.

**Wavelength Number**

Although the complete data set consisted of 417 measured IR spectra, it covered only 18 different rockets, i.e. it contained 399 replicates. These replicates were not used in the validation of the models. Instead, leave-one-out cross-validation (Hjorth, 1994) was used to assess the quality of the models, i.e. the set of n (= 18) independent samples was split into n-1 training samples, while the nth point was reserved for model validation. The trainingvalidation split was repeated n times until each data point had been omitted once for validation. A validation set of n predictions on the 'unseen' data was therefore derived from all the available data and a predicted residual estimate sum of squares (PRESS) was calculated on the validation set.

$$PRESS\_{l} = \sum\_{l=1}^{n} e\_{lj}^{2} \tag{14}$$

Identification of Rocket Motor Characteristics from Infrared Emission Spectra 441

Lawrence et al. (1997) have shown an example of a single-layer perceptron neural network, where the optimal model built on 200 independent data points consisted of 661 parameters. Justification for this result is given by the fact that the nonlinear optimization algorithm for a neural network does not reach a global optimum. Lawrence et al. (1997) further stated that the Vapnik-Chervonenkis (VC) dimension is somewhat conservative in estimating the lower

Partial least squares and principal components regression can be used to reduce the dimensionality of the input space, in this case attempting to reduce the degrees of freedom of the models to 14 or less without losing the most important information in the input data. The evaluation of the degrees of freedom of a nonlinear model built on a data set so close to full rank can only be possible if the degrees of freedom associated with each model can be estimated reliably. Van der Voet (1999) suggested a method of defining pseudo-degrees of

Here MSEPrs is the mean square error of resubstitution for the entire data set per output variable and MSECV is the mean square error of leave-one-out cross-validation. This method has been developed mainly to help with the estimation of the degrees of freedom of complex models and results are consistent with df = m+1, for m input variables, in the case

The regression models, cross-validation cycles and statistical analyses were programmed using the MATLAB® Release 12 software package. The neural network toolbox available in

The results in table 2 show that the NNPLS model was the most parsimonious. The NNPLS model yielded the lowest PRESS-value, SSEP-value and average pseudo-degrees of freedom.

this package was used for the training of the multilayer perceptron neural networks.

 Linear PLS MLP NNPLS Complexity 11 LD 2 H 11 LD PRESS 688.50 613.23 258.10 SSEP 76.28 52.74 45.41 X-Block %η<sup>2</sup> 99.69 NA 99.76 Y-Block %η<sup>2</sup> 90.52 93.43 94.34 Y-Block max %η<sup>2</sup> 94.98 98.31 98.98 Average RCV2 0.461 0.417 0.626 Average RCV,max2 0.541 0.548 0.746 Average R2 0.876 0.803 0.825 Average R2max 0.957 0.907 0.954 Average pdf 12.42 12.91 12.33 Parameters 220 476 307 Table 2. A summary of performance scores of each candidate model for the forward mapping problem. The Y-block variances are calculated on the overall optimised models.

��� � ��� � �������⁄������ (16)

bound for the required number of data points.

of linear regression models.

**5.1 Forward mapping** 

**5. Results** 

freedom (pdf) based on the performance of a model, as in (14)

The PRESS-values calculated for the output variables normally passed through a minimum with increased model complexity, as the model started to map random noise in the data and was used to guide the complexity of the constructed models. A residual score defined as SSP, which had the same form as equation (12), except that it was calculated on all n data points was used when training on the overall model.

The fraction of the variance (R2-value) of an output variable explained by the model is defined as the variance explained by the model over the total variance using the prediction error, SSE (SSEPj or PRESSj):

$$R^2 = \left(\sum\_{l=1}^n \{f(\mathbf{x}\_l, \mathbf{e}) - \overline{\mathbf{y}}\}^2 \Big/ \Big|\_{\sum\_{l=1}^n \{\mathbf{y}\_l - \overline{\mathbf{y}}\}^2} \right) = \text{SSM} \Big/\_{\text{SST}} = 1 - \text{SSE} \Big/\_{\text{SST}} \tag{15}$$

In the case of both the forward and reverse mappings there are a large number of output variables. In order to be able to compare the performances of the candidate models, the PRESSj- and SSEPj–values were summed over all j = 1, 2, … l output variables to yield single PRESS- and SSEP-values for each model. The model yielding the lowest PRESS-score was expected to best predict validation data and therefore best generalize the input-output relationships.

During cross-validation of the linear PLS model, model fitting was therefore repeated 18 times (once for each of the 18 rockets) for each latent dimension as the overall complexity increased. In the case of the feed-forward multilayer perceptron neural network, 18 training sessions were required each time a node was added to the hidden layer.

### **4.2 Degrees of freedom**

In the case of the forward mapping where the IR emission spectra were to be predicted by a given set of rocket motor features, there were 18 input and 146 output variables. Clearly, for a simple linear least squares model, the model requires 19 degrees of freedom (18 input variables plus the bias). However, the situation is more complicated when nonlinear models are fitted to the data.

Statistical theory requires that a regression model has to be built from an overdetermined system. For this reason it is required that there should be at least 3 to 5 lack-of-fit degrees of freedom (nlof) available as a check on the suitability of the model (Brereton, 1992; Draper & Smith, 1981). Hence in this case, for the simplest linear regression model using a total number of n sample points of which there are nr replicates, the maximum required number of model degrees of freedom, df, excluding bias, becomes: df = n –nr– nlof – 1 = 420 – 402 – 3 – 1 = 14.

For m input variables the pseudo-dimension for prediction by a multilayer perceptron neural network requires that at least m+1 independent samples are available per node for building a model (Sontag, 1998; Schmitt, 2001). It therefore appears that a larger set of data points is required to fit nonlinear models, such as neural networks that generally have a large number of parameters (weights) to fit.

Lawrence et al. (1997) have shown an example of a single-layer perceptron neural network, where the optimal model built on 200 independent data points consisted of 661 parameters. Justification for this result is given by the fact that the nonlinear optimization algorithm for a neural network does not reach a global optimum. Lawrence et al. (1997) further stated that the Vapnik-Chervonenkis (VC) dimension is somewhat conservative in estimating the lower bound for the required number of data points.

Partial least squares and principal components regression can be used to reduce the dimensionality of the input space, in this case attempting to reduce the degrees of freedom of the models to 14 or less without losing the most important information in the input data.

The evaluation of the degrees of freedom of a nonlinear model built on a data set so close to full rank can only be possible if the degrees of freedom associated with each model can be estimated reliably. Van der Voet (1999) suggested a method of defining pseudo-degrees of freedom (pdf) based on the performance of a model, as in (14)

$$pdf = n\left(1 - \sqrt{MSE\_{rs} / MSEV}\right) \tag{16}$$

Here MSEPrs is the mean square error of resubstitution for the entire data set per output variable and MSECV is the mean square error of leave-one-out cross-validation. This method has been developed mainly to help with the estimation of the degrees of freedom of complex models and results are consistent with df = m+1, for m input variables, in the case of linear regression models.

## **5. Results**

440 Infrared Spectroscopy – Materials Science, Engineering and Technology

The PRESS-values calculated for the output variables normally passed through a minimum with increased model complexity, as the model started to map random noise in the data and was used to guide the complexity of the constructed models. A residual score defined as SSP, which had the same form as equation (12), except that it was calculated on all n data

The fraction of the variance (R2-value) of an output variable explained by the model is defined as the variance explained by the model over the total variance using the prediction

> ∑ (�� � ��) � � ��� �

In the case of both the forward and reverse mappings there are a large number of output variables. In order to be able to compare the performances of the candidate models, the PRESSj- and SSEPj–values were summed over all j = 1, 2, … l output variables to yield single PRESS- and SSEP-values for each model. The model yielding the lowest PRESS-score was expected to best predict validation data and therefore best generalize the input-output

During cross-validation of the linear PLS model, model fitting was therefore repeated 18 times (once for each of the 18 rockets) for each latent dimension as the overall complexity increased. In the case of the feed-forward multilayer perceptron neural network, 18 training

In the case of the forward mapping where the IR emission spectra were to be predicted by a given set of rocket motor features, there were 18 input and 146 output variables. Clearly, for a simple linear least squares model, the model requires 19 degrees of freedom (18 input variables plus the bias). However, the situation is more complicated when nonlinear models

Statistical theory requires that a regression model has to be built from an overdetermined system. For this reason it is required that there should be at least 3 to 5 lack-of-fit degrees of freedom (nlof) available as a check on the suitability of the model (Brereton, 1992; Draper & Smith, 1981). Hence in this case, for the simplest linear regression model using a total number of n sample points of which there are nr replicates, the maximum required number of model degrees of freedom, df, excluding bias, becomes: df = n –nr– nlof – 1 = 420 – 402 – 3

For m input variables the pseudo-dimension for prediction by a multilayer perceptron neural network requires that at least m+1 independent samples are available per node for building a model (Sontag, 1998; Schmitt, 2001). It therefore appears that a larger set of data points is required to fit nonlinear models, such as neural networks that generally have a

sessions were required each time a node was added to the hidden layer.

�

� = ���

��� � =�� ���

��� � (15)

points was used when training on the overall model.

�∑ (�(��� �) � ��) � �

error, SSE (SSEPj or PRESSj):

�� =

**4.2 Degrees of freedom** 

are fitted to the data.

large number of parameters (weights) to fit.

– 1 = 14.

relationships.

�

���

The regression models, cross-validation cycles and statistical analyses were programmed using the MATLAB® Release 12 software package. The neural network toolbox available in this package was used for the training of the multilayer perceptron neural networks.

## **5.1 Forward mapping**

The results in table 2 show that the NNPLS model was the most parsimonious. The NNPLS model yielded the lowest PRESS-value, SSEP-value and average pseudo-degrees of freedom.


Table 2. A summary of performance scores of each candidate model for the forward mapping problem. The Y-block variances are calculated on the overall optimised models.

Identification of Rocket Motor Characteristics from Infrared Emission Spectra 443

The advantage of linear PLS can be seen in figure 6. The regression coefficients can be collapsed into a single coefficient per input variable, as shown in equation 7. In this way the input variables with the most significant leverages could be determined for certain ranges of wavelengths in the spectrum. The improved predictions using nonlinear PLS could be attributed to the distinctions made between DB- and C-class rocket motor designs. This suggests that with the availability of more data, it may be useful to build separate linear PLS

wavelength numbers 46-85

Fig. 6. A plot of the regression coefficients of the linear PLS model for all 146 output

The NNPLS model appears to be the best, owing to its better generalization ability. The low average RCV2–values and the relatively poor prediction on DB2 (figure 7) were not entirely unexpected, since the model had to extrapolate, as a result of the lack of data similar to DB2. The linear tendency in input-output relationships shows that some predictions on unseen data can be fairly accurate, such as that for C4 (figure 8). The overall model predictions for DB2 and C4 (trained on all data), together with their 95% confidence intervals are shown in

Tc Ec Pc D C H O N Al K F Cu Pb S Cl Si Ti Fe <sup>t</sup>

As a note of interest, Qin & McAvoy (1992) have shown that NNPLS models can be collapsed to multilayer perceptron architectures. In this case it was therefore possible to represent the best NNPLS model in the form of a single layer neural network with 29 hidden nodes using tan-sigmoidal activation functions and an output layer of 146 nodes with

Moreover, it is interesting to note that the optimal models (PLS, neural network and NNPLS) yielded similar average pseudo-degrees of freedom (MSECV/MSEPrs-ratios). The large numbers of parameters (as shown in table 2) support the conclusions by Lawrence at al. (1997) that there can be more variables than independent data points in nonlinear modelling. The pseudo-degrees of freedom appear to be a more consistent way of measuring model

complexity than simple comparison of the number of parameters of each model.

models for each of the DB- and C-class rocket motor types.

variables.


Overall Linear Regression Coefficient, B PLS

figures 9 and 10.

purely linear functions.

The Y-block variance (η2) is the percentage of the output variance explained by the model over all output variables. This is analogous to the R2–values calculated for the individual output variables. The values indicated by maxima were those where the pure error component had been subtracted. Only the linear PLS model was able to perform better than the NNPLS model on the R2-scores calculated for the overall model. The reason for this is the fact that except for C5, the C-class rocket motor irradiance spectra were most accurately predicted using the linear PLS model.

Furthermore, the NNPLS model appeared not only to retain the linear latent projections, but also introduced nonlinearity in the inner models to compensate for the shortcomings of the linear PLS algorithm. This is shown in figure 5, where the PLS inner model scores are plotted to show the shape of the curve fitted by the neural network. The output scores after the first latent dimension seem to have near linear relationships with the input scores.

Fig. 5. The target and predicted PLS output scores vs. the input scores for the first 4 latent dimensions using the overall NNPLS model.

The Y-block variance (η2) is the percentage of the output variance explained by the model over all output variables. This is analogous to the R2–values calculated for the individual output variables. The values indicated by maxima were those where the pure error component had been subtracted. Only the linear PLS model was able to perform better than the NNPLS model on the R2-scores calculated for the overall model. The reason for this is the fact that except for C5, the C-class rocket motor irradiance spectra were most accurately

Furthermore, the NNPLS model appeared not only to retain the linear latent projections, but also introduced nonlinearity in the inner models to compensate for the shortcomings of the linear PLS algorithm. This is shown in figure 5, where the PLS inner model scores are plotted to show the shape of the curve fitted by the neural network. The output scores after the first latent dimension seem to have near linear relationships with the input scores.




0

**u4**

5

**LD 4**

0

**u2**

5

10

**LD 2**


**t 2**


**t 4**

Fig. 5. The target and predicted PLS output scores vs. the input scores for the first 4 latent

**Target output scores (LV) Mean target output scores (LV) Predicted output scores (LV)**

dimensions using the overall NNPLS model.


**t**


**t 1**



0

5

10

**u3**

15

20

**LD 3**

0

10

**u1**

20

30

**LD 1**

predicted using the linear PLS model.

The advantage of linear PLS can be seen in figure 6. The regression coefficients can be collapsed into a single coefficient per input variable, as shown in equation 7. In this way the input variables with the most significant leverages could be determined for certain ranges of wavelengths in the spectrum. The improved predictions using nonlinear PLS could be attributed to the distinctions made between DB- and C-class rocket motor designs. This suggests that with the availability of more data, it may be useful to build separate linear PLS models for each of the DB- and C-class rocket motor types.

Fig. 6. A plot of the regression coefficients of the linear PLS model for all 146 output variables.

The NNPLS model appears to be the best, owing to its better generalization ability. The low average RCV2–values and the relatively poor prediction on DB2 (figure 7) were not entirely unexpected, since the model had to extrapolate, as a result of the lack of data similar to DB2. The linear tendency in input-output relationships shows that some predictions on unseen data can be fairly accurate, such as that for C4 (figure 8). The overall model predictions for DB2 and C4 (trained on all data), together with their 95% confidence intervals are shown in figures 9 and 10.

As a note of interest, Qin & McAvoy (1992) have shown that NNPLS models can be collapsed to multilayer perceptron architectures. In this case it was therefore possible to represent the best NNPLS model in the form of a single layer neural network with 29 hidden nodes using tan-sigmoidal activation functions and an output layer of 146 nodes with purely linear functions.

Moreover, it is interesting to note that the optimal models (PLS, neural network and NNPLS) yielded similar average pseudo-degrees of freedom (MSECV/MSEPrs-ratios). The large numbers of parameters (as shown in table 2) support the conclusions by Lawrence at al. (1997) that there can be more variables than independent data points in nonlinear modelling. The pseudo-degrees of freedom appear to be a more consistent way of measuring model complexity than simple comparison of the number of parameters of each model.

Identification of Rocket Motor Characteristics from Infrared Emission Spectra 445

**DB2**

Fig. 9. Examples of plume irradiance predictions for DB-class rocket motors obtained for the

**C4**

0 25 50 75 100 125 150 Wavelength Number

Fig. 10. Examples of plume irradiance predictions for C-class rocket motors obtained for the

0 25 50 75 100 125 150 Wavelength Number

overall NNPLS model using 11 latent dimensions.

**Mean target data Predicted values**

**95% prediction confidence interval**

**Mean target data Predicted values**

**95% prediction confidence interval**

0

0.1

0.2

0.3

Absorbance Units

0.4

0.5

0.6

overall NNPLS model using 11 latent dimensions.

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55

Absorbance Units

Fig. 7. Examples of plume irradiance predictions for 'unseen' DB-class rocket motors obtained during leave-one-out cross-validation of NNPLS with 11 latent dimensions.

Fig. 8. Examples of plume irradiance predictions for 'unseen' C-class rocket motors obtained during leave-one-out cross-validation of NNPLS with 11 latent dimensions.

#### **C4**

**DB2**

**2 standard deviations band for measured data**

Fig. 7. Examples of plume irradiance predictions for 'unseen' DB-class rocket motors obtained during leave-one-out cross-validation of NNPLS with 11 latent dimensions.

**C4**

**2 standard deviations band for measured data**

0 25 50 75 100 125 150 Wavelength Number

0

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45

Absorbance Units

0.1

0.2

0.3

0.4

Absorbance Units

0.5

0.6

**Mean target data Predicted values**

**Mean target data Predicted values**

0.7

Fig. 8. Examples of plume irradiance predictions for 'unseen' C-class rocket motors obtained

0 25 50 75 100 125 150 Wavelength Number

during leave-one-out cross-validation of NNPLS with 11 latent dimensions.

Fig. 9. Examples of plume irradiance predictions for DB-class rocket motors obtained for the overall NNPLS model using 11 latent dimensions.

**C4**

Fig. 10. Examples of plume irradiance predictions for C-class rocket motors obtained for the overall NNPLS model using 11 latent dimensions.

**DB2**

Identification of Rocket Motor Characteristics from Infrared Emission Spectra 447

LD X-Block %η2 Y-Block %η<sup>2</sup> Avg pdf PRESS SSEP

1 92.44 13.57 1.87 466.1 362.83

2 95.25 26.85 3.75 475.6 307.08

3 98.63 32.61 4.91 551.1 282.89

4 99.60 43.81 7.45 664.3 235.89

5 99.85 60.05 8.66 632.3 167.72

12 99.99 88.47 15.35 1971.9 48.42

Table 4. The sum-squared residuals obtained from building a linear PLS model for the

**DB5**

Fig. 11. Examples of rocket motor parameter predictions for 'unseen' rocket motors in the DB-class obtained during leave-one-out cross-validation of the NNPLS model (3 latent

**C H O N Al K F Cu Pb S Cl Si Ti Fe Tc Ec Pc Dt**

**Tc Ec Pc Dt**

reverse modelling problem.

dimensions).

**0**

**2**

**4**

**6**

**8**

**Output Value**

**10**

**12**

**14**

**Target values Predicted values**

**RMSECV band for predicted values**

### **5.2 Reverse mapping**

The optimal scores of the prediction abilities for each of the candidate models are shown in table 3. In the reverse problem it would be possible to find the optimal model complexity for each individual output variable. However, for the sake of simplicity, it is more sensible to compare the models by pooling the results for all output variables. Except for the R2-scores, the average performance scores over all output variables did not differ much from the results shown in table 3.


Table 3. A summary of performance scores of each candidate model for the reverse mapping problem. The Y-block variances are calculated on the overall optimised models.

Even though the relatively low average RCV2–value is a poor result, it does not necessarily reflect adversely on the true performance of the NNPLS model. This is owing to the data of output variables K, F, S, Si, Ti and Fe consisting of numerous zero entries and the inability of the model to handle these irregularities. The RCV2–values for C, H, O, N, Al, Cu, Pb, Cl and EC range between 0.6 and 0.8.

It was difficult to exactly determine the optimal model for linear PLS. In table 4 it is shown that a linear PLS model with 3 or 4 latent dimensions could have been chosen to increase the Y-block explained variance. This would have moved the average pdf-values closer to larger values of the other models. These results show that there is a requirement for nonlinear structures in the model building.

A few of the predictions for unseen data obtained from leave-one-out cross-validation are shown in figures 11 and 12. The square root of MSECV (RMSECV) is calculated for each individual output variable in order to obtain a measure of the standard deviation of the error of prediction. The same predictions made from the overall model built on all the available data are shown in figures 13 and 14.


The optimal scores of the prediction abilities for each of the candidate models are shown in table 3. In the reverse problem it would be possible to find the optimal model complexity for each individual output variable. However, for the sake of simplicity, it is more sensible to compare the models by pooling the results for all output variables. Except for the R2-scores, the average performance scores over all output variables did not differ much from the

Linear PLS MLP NNPLS

Complexity 2 LD 2 H 3 LD

PRESS 475.6 329.1 273.8

SSEP 307.08 245.61 171.18

X-Block %η<sup>2</sup> 95.3 NA 99.84

Y-Block %η<sup>2</sup> 26.9 41.49 59.22

Average RCV2 0.183 0.351 0.322

Average R2 0.351 0.615 0.615

Average pdf 3.75 7.07 6.19

Parameters 296 297 462

problem. The Y-block variances are calculated on the overall optimised models.

Table 3. A summary of performance scores of each candidate model for the reverse mapping

Even though the relatively low average RCV2–value is a poor result, it does not necessarily reflect adversely on the true performance of the NNPLS model. This is owing to the data of output variables K, F, S, Si, Ti and Fe consisting of numerous zero entries and the inability of the model to handle these irregularities. The RCV2–values for C, H, O, N, Al, Cu, Pb, Cl and

It was difficult to exactly determine the optimal model for linear PLS. In table 4 it is shown that a linear PLS model with 3 or 4 latent dimensions could have been chosen to increase the Y-block explained variance. This would have moved the average pdf-values closer to larger values of the other models. These results show that there is a requirement for nonlinear

A few of the predictions for unseen data obtained from leave-one-out cross-validation are shown in figures 11 and 12. The square root of MSECV (RMSECV) is calculated for each individual output variable in order to obtain a measure of the standard deviation of the error of prediction. The same predictions made from the overall model built on all the

**5.2 Reverse mapping** 

results shown in table 3.

EC range between 0.6 and 0.8.

structures in the model building.

available data are shown in figures 13 and 14.

Table 4. The sum-squared residuals obtained from building a linear PLS model for the reverse modelling problem.

Fig. 11. Examples of rocket motor parameter predictions for 'unseen' rocket motors in the DB-class obtained during leave-one-out cross-validation of the NNPLS model (3 latent dimensions).

Identification of Rocket Motor Characteristics from Infrared Emission Spectra 449

**C12**

Fig. 14. Examples of rocket motor parameter predictions for C-class rockets obtained for the

**C H O N Al K F Cu Pb S Cl Si Ti Fe Tc Ec Pc Dt**

Similarly to the forward problem the better predictions are obtained for the C-class rocket motors due to the more data available in this class of rocket designs. The physical design parameters, EC, PC and Dt show the largest confidence intervals. Even so, the predictions on

i. Mean centre or standardise the inputs and outputs, **X** and **Y**. Initialise the algorithm by setting the output scores, **u** equal to a column of **Y**. For each latent dimension, a = 1,

�� � ��� �� ⁄ � (A1)

**Tc Ec Pc Dt**

��� ‖�‖⁄ (A2)

� � �� �⁄ �� (A3)

overall optimum NNPLS model trained with 3 latent dimensions on all data points.

unseen data are better than expected given the small amount of available data.

The NIPALS algorithm for PLS according to Baffi et al, (1999a):

ii. Calculate the input weights, **w**, by regressing **X** on **u**:

**Target values Predicted values**

**95% confidence prediction interval**

v. Calculate output loadings by regressing **Y** on **t**:

2, … h follow steps ii to xiii below:

iii. Normalise **w** to unit length:

iv. Calculate the input scores:

**6. Appendix** 

**Output Value**

Fig. 12. Examples of rocket motor parameter predictions for 'unseen' rocket motors in the Cclass obtained during leave-one-out cross-validation of the NNPLS model (3 latent dimensions).

**DB5**

Fig. 13. Examples of rocket motor parameter predictions DB-class rockets obtained for the overall optimum NNPLS model trained with 3 latent dimensions on all data points.

Fig. 14. Examples of rocket motor parameter predictions for C-class rockets obtained for the overall optimum NNPLS model trained with 3 latent dimensions on all data points.

Similarly to the forward problem the better predictions are obtained for the C-class rocket motors due to the more data available in this class of rocket designs. The physical design parameters, EC, PC and Dt show the largest confidence intervals. Even so, the predictions on unseen data are better than expected given the small amount of available data.

#### **6. Appendix**

448 Infrared Spectroscopy – Materials Science, Engineering and Technology

**C12**

Fig. 12. Examples of rocket motor parameter predictions for 'unseen' rocket motors in the C-

**DB5**

**C H O N Al K F Cu Pb S Cl Si Ti Fe Tc Ec Pc Dt**

**Tc Ec Pc Dt**

**Tc Ec Pc Dt**

Fig. 13. Examples of rocket motor parameter predictions DB-class rockets obtained for the overall optimum NNPLS model trained with 3 latent dimensions on all data points.

**C H O N Al K F Cu Pb S Cl Si Ti Fe Tc Ec Pc Dt**

class obtained during leave-one-out cross-validation of the NNPLS model (3 latent

dimensions).

**0**

**2**

**4**

**6**

**8**

**Output Value**

**10**

**12**

**14**

**Target values Predicted values**

**95% confidence prediction interval**

**Target values Predicted values**

**RMSECV band for predicted values**

**Output Value**

The NIPALS algorithm for PLS according to Baffi et al, (1999a):


$$\mathbf{w}^T = \mathbf{u}^T \mathbf{X} / \mathbf{u}^T \mathbf{u} \tag{A1}$$

iii. Normalise **w** to unit length:

$$\mathbf{w} = \mathbf{w} / \|\mathbf{w}\|\tag{A2}$$

iv. Calculate the input scores:

$$\mathbf{t} = \mathbf{X}\mathbf{w}/\mathbf{w}^T\mathbf{w} \tag{A3}$$

v. Calculate output loadings by regressing **Y** on **t**:

$$\mathbf{q}^T = \mathbf{t}^T \mathbf{Y} / \mathbf{t}^T \mathbf{t} \tag{A4}$$

Identification of Rocket Motor Characteristics from Infrared Emission Spectra 451

Despite these promising results, the best candidate models for both the forward and reverse problems cannot be regarded as adequate for practical applications. However, considering the lack of available data, the results can be regarded as acceptable to motivate funding for the collection of more data and rigorous testing. The fact that the input-output relationships appear to have almost linear relationships in some latent dimensions is promising, as this

Baffi, G., Martin, E.B. & Morris, A.J. (1999a). Non-linear projection to latent structures

Baffi, G., Martin, E.B. & Morris, A.J. (1999b). Non-linear projection to latent structures

Bishop, C.M. (1995). *Neural Networks for Pattern Recognition*, Oxford University Press,

Brereton, R.G. (1992). *Multivariate Pattern Recognition in Chemometrics*, Elsevier Science

Draper, N.R. & Smith, H. (1981). *Applied Regression Analysis* (2nd ed), John Wiley and Sons,

Haykin, S. (1999). *Neural Networks - A Comprehensive Foundation*, (2nd ed)., Prentice-Hall,

Lawrence, S., Giles, C.L. & Tsoi, A.C. (1997). Lessons in neural network training: Overfitting

Lorber, A., Wangen, L.E. & Kowalski, B.R. (1987). A theoretical foundation for the PLS

Marquardt, D.W. (1963). An algorithm for least-squares estimation of non-linear parameters, *Journal of the Society for Industrial and Applied Mathematics*, Vol. 11, pp. 431-441. Qin, S.J. & McAvoy, T.J. (1992). Nonlinear PLS Modeling using neural networks. *Computers* 

Riedmiller, M. & Braun, H. (1993). A Direct adaptive method for faster backpropagation

Roodt, J.H.S. The Prediction of the Emission Spectra of Solid Rocket Propellants. (1998). *Ph.D. dissertation*, University of Stellenbosch, Stellenbosch, South Africa. Schmitt, M. (2001). *Neural Networks with local receptive fields and superlinear VC dimension*,

Sontag, E.D. (1998). *VC Dimension of Neural Networks*, State University of New Jersey, NJ,

Urban Hjorth, J.S. (1994). *Computer Intensive Statistical Methods*, Chapman and Hall, London,

Van der Voet, H. (1999). Pseudo-degrees of freedom for complex predictive models: The example of partial least squares, *Journal of Chemometrics*, Vol. 13, pp. 195-208.

learning: The RPROP algorithm, *Proceedings of the IEEE International Conference on* 

Ruhr University of Bochum, Germany, Retrieved from www.ruhr-uni-

may be harder than expected, *Fourteenth National Conference on Artificial Intelligence*,

revisited: The quadratic PLS algorithm, *Computers and Chemical Engineering*, Vol. 23,

revisited (the neural network PLS algorithm), *Computers and Chemical Engineering*,

could lead to the development of robust models.

**8. References** 

pp. 395-411.

Oxford, UK.

NY, USA.

Vol. 23, pp. 1293-1307.

Publishers, Amsterdam.

Upper Saddle River, NJ, USA.

AAAI Press, Menlo Park, CA, USA.

*Neural Networks*, Vol. 1, pp. 586-591.

bochum.de/lmi/mschmitt.

UK.

algorithm. *Journal of Chemometrics*, Vol. 1, pp. 19-31.

USA, Retrieved from www.math.rutgers.edu /~sontag

*and Chemical Engineering*, Vol. 16, pp. 379-391.

vi. Normalise **q** to unit length:

$$\mathbf{q} = \mathbf{q}/\|\mathbf{q}\|\tag{A5}$$

vii. Calculate new output scores **u**:

$$\mathbf{u} = \mathbf{Y}\mathbf{q}/\mathbf{q}^{\mathsf{T}}\mathbf{q} \tag{A6}$$

viii. Check for convergence on **w:** If not return to step ii ix. Calculate the input loadings, **p** by regressing **X** on **t**:

$$\mathbf{p}^T = \mathbf{t}^T \mathbf{X} / \mathbf{t}^T \mathbf{t} \tag{A7}$$

x. Calculate the inner linear regression coefficient b:

$$\mathbf{b} = \mathbf{t}^T \mathbf{u} / \mathbf{t}^T \mathbf{t} \tag{A8}$$

xi. Calculate the input residual matrix:

$$F = X - \mathbf{t}p^T\tag{A9}$$

xii. Calculate the output residual matrix:

$$E = Y - \mathbf{\hat{u}}q^T\tag{A10}$$

xiii. If additional PLS latent dimensions are required replace **X** and **Y** with **F** and **E** respectively and return to step ii for calculation of latent dimension a + 1.

#### **7. Conclusions**

The building of data-driven models in this study was constrained by the sparsity of the available data, as there were only 18 independent samples (rocket motor designs) available. In addition the input and output data were highly multivariate with 18 rocket motor design parameters and 146 spectral wavelengths in the middle IR band. One advantage is that the IR spectral measurements were repeated a number of times (4 to 44 repeats per rocket motor).

The variables (wavelengths) associated with the IR emission spectra were highly correlated. Principal components analysis (PCA), linear and nonlinear PLS showed that at least 86% of the total variance could be explained by the two primary latent dimensions. The forward and reverse modelling results showed that dimensional reduction with a linear model (PLS) produced better models than a nonlinear model (multilayer perceptron neural network trained with the back propagation algorithm) without dimensional reduction.

The NNPLS algorithm with Levenberg-Marquardt training of the inner feed forward neural network models produced the best predictions of the forward models. The average RCV2 value of 0.63 (0.75 for maximum RCV2) for all 146 output variables on unseen data was satisfactory when considering the lack of available data. The average R2–value of more than 0.80 obtained for the overall model trained on all data was also an encouraging result. The average pseudo-dimension for the NNPLS model with 11 latent dimensions was 12.33. This left about 5 lack-of-fit degrees of freedom as a check for the model complexity.

Despite these promising results, the best candidate models for both the forward and reverse problems cannot be regarded as adequate for practical applications. However, considering the lack of available data, the results can be regarded as acceptable to motivate funding for the collection of more data and rigorous testing. The fact that the input-output relationships appear to have almost linear relationships in some latent dimensions is promising, as this could lead to the development of robust models.

#### **8. References**

450 Infrared Spectroscopy – Materials Science, Engineering and Technology

xiii. If additional PLS latent dimensions are required replace **X** and **Y** with **F** and **E**

The building of data-driven models in this study was constrained by the sparsity of the available data, as there were only 18 independent samples (rocket motor designs) available. In addition the input and output data were highly multivariate with 18 rocket motor design parameters and 146 spectral wavelengths in the middle IR band. One advantage is that the IR spectral measurements were repeated a number of times (4 to 44 repeats per rocket

The variables (wavelengths) associated with the IR emission spectra were highly correlated. Principal components analysis (PCA), linear and nonlinear PLS showed that at least 86% of the total variance could be explained by the two primary latent dimensions. The forward and reverse modelling results showed that dimensional reduction with a linear model (PLS) produced better models than a nonlinear model (multilayer perceptron neural network

The NNPLS algorithm with Levenberg-Marquardt training of the inner feed forward neural network models produced the best predictions of the forward models. The average RCV2 value of 0.63 (0.75 for maximum RCV2) for all 146 output variables on unseen data was satisfactory when considering the lack of available data. The average R2–value of more than 0.80 obtained for the overall model trained on all data was also an encouraging result. The average pseudo-dimension for the NNPLS model with 11 latent dimensions was 12.33. This

trained with the back propagation algorithm) without dimensional reduction.

left about 5 lack-of-fit degrees of freedom as a check for the model complexity.

respectively and return to step ii for calculation of latent dimension a + 1.

vi. Normalise **q** to unit length:

vii. Calculate new output scores **u**:

xi. Calculate the input residual matrix:

xii. Calculate the output residual matrix:

**7. Conclusions** 

motor).

viii. Check for convergence on **w:** If not return to step ii ix. Calculate the input loadings, **p** by regressing **X** on **t**:

x. Calculate the inner linear regression coefficient b:

�� � ��� �� ⁄ � (A4)

��� ‖�‖⁄ (A5)

� � �� �� ⁄ � (A6)

�� � ��� �� ⁄ � (A7)

����� �� ⁄ � (A8)

� � � � ��� (A9)

�������� (A10)


**24** 

*China* 

**Optical Technologies for** 

*China Agricultural University* 

Yankun Peng, Yongyu Li and Jingjing Chen

**Determination of Pesticide Residue** 

Pesticides are essential for agricultural and horticultural crops production. Pesticides are commonly classified as insecticide, fungicide, herbicide, rodenticide, etc. These pesticides act against insects, rodents, weeds which are harmful in agricultural or horticultural planting. Normally, farmers use the pesticides following the instruction written in the package. In most cases, the pesticides are mixed with water and sprayed over the plants. Basically, after spraying fruits or vegetables with pesticide, a period of 10 to 14 days is required to allow the chemical to degrade. However, the full degradation of pesticide is not always achieved. In recent years, some farmers ignored to use the pesticide correctly and rationally. In order to chase a better insecticidal effect and the economic interests, the phenomenon of using pesticide excessively, or selling the fruits or vegetables just after spraying the pesticide in few days are not difficult to see. Moreover, the pesticides

Currently, several different technologies such as gas chromatography (GC), highperformance liquid chromatography (HPLC), thin-layer chromatography, supercritical fluid chromatography, chromatography-mass spectrometry, capillary electrophoresis, enzyme inhibition method, immunoassay method, and bio-sensor method are used to determine the concentration of pesticide residue. The accuracy of these technologies such as GC and HPLC is best (Gambacorta et al., 2005). However, these analysis methods have limitations of time and labor for controlling individual products. Normally, at least hours are needed to measure the pesticide concentration in a single sample because of the complication in the testing process. These instrument analysis methods as such, can be used only in laboratory for accurate analysis and statutory inspection (Luypaert et al., 2003). Biological and chemical analysis methods were developed in recent years, but there are also some flaws, such as the

Compare with the growing public requirement of food security, the traditional pesticide detection technologies are not competent because of the shortcomings such as longer detection cycle, complex testing process, testing process, lagged nature of nature of detection results, etc. Therefore, development of fast, reliable detection method or equipment of pesticides residue is imperative. And it is vital to control the pesticide concentration on agricultural products for maintaining public health conditions and

overdosing also have the potential to contaminate the soil, air, and river.

pre-treatments are needed and the demanding of experimental conditions.

protecting the entire environment.

**1. Introduction** 


## **Optical Technologies for Determination of Pesticide Residue**

Yankun Peng, Yongyu Li and Jingjing Chen *China Agricultural University China* 

## **1. Introduction**

452 Infrared Spectroscopy – Materials Science, Engineering and Technology

Wold, H. (1966). Estimation of principal components and related models by iterative least

Wold, S., Kettaneh-Wold, N. & Skagerberg, B. (1998). Nonlinear PLS modeling, *Chemometrics* 

squares, In: *Multivariate Analysis*, Academic Press, NY, USA.

*and Intelligent Laboratory Systems*, Vol. 7, pp. 53-65.

Pesticides are essential for agricultural and horticultural crops production. Pesticides are commonly classified as insecticide, fungicide, herbicide, rodenticide, etc. These pesticides act against insects, rodents, weeds which are harmful in agricultural or horticultural planting. Normally, farmers use the pesticides following the instruction written in the package. In most cases, the pesticides are mixed with water and sprayed over the plants. Basically, after spraying fruits or vegetables with pesticide, a period of 10 to 14 days is required to allow the chemical to degrade. However, the full degradation of pesticide is not always achieved. In recent years, some farmers ignored to use the pesticide correctly and rationally. In order to chase a better insecticidal effect and the economic interests, the phenomenon of using pesticide excessively, or selling the fruits or vegetables just after spraying the pesticide in few days are not difficult to see. Moreover, the pesticides overdosing also have the potential to contaminate the soil, air, and river.

Currently, several different technologies such as gas chromatography (GC), highperformance liquid chromatography (HPLC), thin-layer chromatography, supercritical fluid chromatography, chromatography-mass spectrometry, capillary electrophoresis, enzyme inhibition method, immunoassay method, and bio-sensor method are used to determine the concentration of pesticide residue. The accuracy of these technologies such as GC and HPLC is best (Gambacorta et al., 2005). However, these analysis methods have limitations of time and labor for controlling individual products. Normally, at least hours are needed to measure the pesticide concentration in a single sample because of the complication in the testing process. These instrument analysis methods as such, can be used only in laboratory for accurate analysis and statutory inspection (Luypaert et al., 2003). Biological and chemical analysis methods were developed in recent years, but there are also some flaws, such as the pre-treatments are needed and the demanding of experimental conditions.

Compare with the growing public requirement of food security, the traditional pesticide detection technologies are not competent because of the shortcomings such as longer detection cycle, complex testing process, testing process, lagged nature of nature of detection results, etc. Therefore, development of fast, reliable detection method or equipment of pesticides residue is imperative. And it is vital to control the pesticide concentration on agricultural products for maintaining public health conditions and protecting the entire environment.

Optical Technologies for Determination of Pesticide Residue 455

The following example presents the methodology for determination of chlorpyrifos based

Pesticide solution: A commercial pesticide, containing 40% chlorpyrifos (Noposion, China) was used. Chlorpyrifos is an organophosphorus pesticide, normally used in the paddy, wheat, cotton, fruit trees, and vegetables. Distilled water was prepared in order to provide the solutions with different concentrations. A total of 24 concentration levels, from 1 mg/kg to 400 mg/kg of active ingredient were diluted based on the amount of chlorpyrifos. After preparation, the solutions were kept in conical flasks and preserved in a cool place in order

Filter paper samples: It is well known that the control level of pesticide residue does not lie at the percent level but at the 10-6 level, even 10-9 level. It is hard to obtain a satisfactory result by the use of NIR spectroscopy to determine the concentration of pesticide solution. The reason being that water has several strong absorption peaks in near-infrared bands; as a result it is difficult to get the information of pesticide compared to water in the solution. In order to obtain the absorption of trace chemicals, a special method to concentrate the amount of chemicals on samples was developed. Filter paper was used as substrate, water was removed from wet substrate by drying, and then the NIR measurement was performed

Normal filter papers (Shuangquan, China), 9 cm diameter were selected. Initially, every piece of 9 cm filter-papers were sheared into four pieces each 30 mm diameter by using a special mold. Then the filter paper were kept into a special platform, which was made of polystyrene foam and pins (Figure 1). Each platform ware almost 20 cm long and 5 cm wide, and four pieces of filter paper could be placed on each platform. After putting the filter papers onto the platform, 200μL of pesticide solution was gently pipetted onto each filter paper (the amount of 200μL is the volume absorbable by filter paper without any overflow). Several pieces of filter paper samples were prepared for each concentration level. A total of

to prevent chemical degradation and contamination.

Fig. 1. Platform for filter paper.

on the dried substrate.

99 filter paper samples were prepared.

on NIR. **Samples** 

Development of safe, fast, reliable and low-cost analytical methods for the determination of pesticide residue that avoids the use of organic solvents, and reduces the contact of operator with the toxic substances is growing interest at present. In recent years, spectroscopy based procedures is regarded as a potential method which could solve the above problems. Spectroscopy analysis methods have been widely used in chemical industry, agriculture, medicine and other areas (Peng et al., 2008, 2009; ElMasry et al., 2007).

A NIR spectroscopic method and an optical imaging technology for prediction of organophosphorus pesticide are introduced as follows.

## **2. NIR spectroscopy for pesticide determination**

Among the optical analysis methods, near-infrared (NIR) spectroscopy is the most popular method because of its non-destructive nature, the low operating cost and the fast response times (Armenta et al., 2007), and it also has been successfully applied to quality control in food (Pi et al., 2009; Leroy et al., 2003; Subbiah et al., 2008), petrochemical, pharmaceutical, clinical and biomedical and environmental sectors (Ripoll et al., 2008). Near-infrared (0.7- 2.5μm; 12900-4000cm-1) spectroscopy is further classified into NIR reflectance spectroscopy and NIR transmission spectroscopy. NIR can be non-dispersive (filter-based instrumentation), dispersive and use Fourier transform-based instrumentation. Table 1 lists some NIR spectroscopic applications suitable for pesticides determination. All these researches have shown the possibility and reasonability for determination of pesticide concentration using NIR spectroscopy.


Table 1. Near-infrared spectroscopy to determinate pesticides concentration.

The following example presents the methodology for determination of chlorpyrifos based on NIR.

#### **Samples**

454 Infrared Spectroscopy – Materials Science, Engineering and Technology

Development of safe, fast, reliable and low-cost analytical methods for the determination of pesticide residue that avoids the use of organic solvents, and reduces the contact of operator with the toxic substances is growing interest at present. In recent years, spectroscopy based procedures is regarded as a potential method which could solve the above problems. Spectroscopy analysis methods have been widely used in chemical industry, agriculture,

A NIR spectroscopic method and an optical imaging technology for prediction of

Among the optical analysis methods, near-infrared (NIR) spectroscopy is the most popular method because of its non-destructive nature, the low operating cost and the fast response times (Armenta et al., 2007), and it also has been successfully applied to quality control in food (Pi et al., 2009; Leroy et al., 2003; Subbiah et al., 2008), petrochemical, pharmaceutical, clinical and biomedical and environmental sectors (Ripoll et al., 2008). Near-infrared (0.7- 2.5μm; 12900-4000cm-1) spectroscopy is further classified into NIR reflectance spectroscopy and NIR transmission spectroscopy. NIR can be non-dispersive (filter-based instrumentation), dispersive and use Fourier transform-based instrumentation. Table 1 lists some NIR spectroscopic applications suitable for pesticides determination. All these researches have shown the possibility and reasonability for determination of pesticide

**Instrumental method Determination attribute Reference**  Mid- and near-infrared Metribuzin in agrochemicals Khanmohammadi et al., 2008

Propamocarb in emulsiable pesticide

Mid- and near-infrared Describing diuron sorption in soils Forouzanhohar et al., 2009

NIR Pesticide phoxim residues Shen et al., 2009

NIR Detecting the chlorpyrifos content Liu et al., 2009

Table 1. Near-infrared spectroscopy to determinate pesticides concentration.

formulations Armenta et al.,2007

concentrate formulations Quintảs et al., 2008

Agricultural Products Makio et al., 2007

and tetrachloro-isophthalonitrile Sarawong et al., 2007

pesticide Ysacc Sato-Berrú et al., 2004

chlordecone Brunet et al., 2009

agrochemicals Xiong et al., 2010

medicine and other areas (Peng et al., 2008, 2009; ElMasry et al., 2007).

organophosphorus pesticide are introduced as follows.

**2. NIR spectroscopy for pesticide determination** 

NIR Pesticide determination in commercially

IR spectroscopy Classification of Pesticide Residues in the

NIR/ Dry extracts Determination of acephate, dichlofluanid

NIR–Raman Quantitative analysis of methyl-parathion

NIR Determination of active ingredient of

NIR Determination of soil content in

concentration using NIR spectroscopy.

Fourier transform infrared

spectroscopy

Pesticide solution: A commercial pesticide, containing 40% chlorpyrifos (Noposion, China) was used. Chlorpyrifos is an organophosphorus pesticide, normally used in the paddy, wheat, cotton, fruit trees, and vegetables. Distilled water was prepared in order to provide the solutions with different concentrations. A total of 24 concentration levels, from 1 mg/kg to 400 mg/kg of active ingredient were diluted based on the amount of chlorpyrifos. After preparation, the solutions were kept in conical flasks and preserved in a cool place in order to prevent chemical degradation and contamination.

Fig. 1. Platform for filter paper.

Filter paper samples: It is well known that the control level of pesticide residue does not lie at the percent level but at the 10-6 level, even 10-9 level. It is hard to obtain a satisfactory result by the use of NIR spectroscopy to determine the concentration of pesticide solution. The reason being that water has several strong absorption peaks in near-infrared bands; as a result it is difficult to get the information of pesticide compared to water in the solution. In order to obtain the absorption of trace chemicals, a special method to concentrate the amount of chemicals on samples was developed. Filter paper was used as substrate, water was removed from wet substrate by drying, and then the NIR measurement was performed on the dried substrate.

Normal filter papers (Shuangquan, China), 9 cm diameter were selected. Initially, every piece of 9 cm filter-papers were sheared into four pieces each 30 mm diameter by using a special mold. Then the filter paper were kept into a special platform, which was made of polystyrene foam and pins (Figure 1). Each platform ware almost 20 cm long and 5 cm wide, and four pieces of filter paper could be placed on each platform. After putting the filter papers onto the platform, 200μL of pesticide solution was gently pipetted onto each filter paper (the amount of 200μL is the volume absorbable by filter paper without any overflow). Several pieces of filter paper samples were prepared for each concentration level. A total of 99 filter paper samples were prepared.

Optical Technologies for Determination of Pesticide Residue 457

2. A random matrix was developed which has the same dimension size as spectral matrix

3. Partial least squares regression (PLSR) was used again. Leave one out cross validation was carried between the new matrix XRa and concentration matrix Y. After each step of

4. Analyzing the stability of *C* value which is the ratio of the mean value of vector *b* and

*mean b*

5. Absolute value of *Ci* was used to discriminate if each spectra variable is effective or not. All effective variables were selected to create a new independent variable matrix, and then this new matrix and Y were used to establish a new PLSR prediction model.

Original NIR spectra of total 99 filter samples are shown in figure 2, and the spectra of samples after pre-processing with MSC are shown in figure 3. It is obviously seen that the base line drift of the spectra is reduced in the figure 3 compared to figure 2 by the

For the total sets, two spectrum pre-processing methods MSC and SNV were used. Figure 4 illustrates the results of the cross validation when MSC and SNV were used as the spectrum

Fig. 2. NIR transmittance spectra of filter-paper samples with different chlorpyrifos content.

 *i*

*std b* (2)

*i*

X. Then X and Ra were joined together to be a new matrix XRa.

the standard deviation of vector *b*:

**NIR spectra** 

application of MSC.

pre-processing method.

**Results of PLSR in full bands** 

leave one out cross validation, a regression coefficient *b* was obtained.

*i*

*C*

Drying filter paper samples: Platform with filter paper samples were carefully moved into the vacuum drying oven, at room temperature for 1 hour. After drying, samples were stored into vacuum packing bags immediately and marked with different concentrations.

#### **Spectrum acquisition**

An Antaris FT-NIR spectrometer (Thermo Nicolet, Waltham, Massachusetts, USA), equipped with an InGaAs detector was used. The filter paper sample was placed in a specially modified sample cell. The spectra were acquired in the range of 4000 cm-1 to 10000 cm-1 at 8 cm-1 interval. For each sample, three points were chosen randomly for the NIR measurement, and 32 scans were co-added for each point. The sample was then removed, and the spectra were collected again in the same manner. Three spectra were obtained for each sample at the same state, and averaged spectra were calculated for further evaluation. To prevent the interference of water vapor in the air, the spectra of samples were acquired immediately after taking out from the vacuum packing bags.

#### **Pre-processing method and data analysis**

The Matlab 7.0 software (MathWorks, USA) was used for all calculations. A total of 99 filter paper samples were divided into two groups, 75 samples were selected as calibration set; the left 24 samples in each concentration level were put into validation sample set. Partial least squares regression (PLSR) was used to develop a prediction model. Multiplicative scatter correction (MSC) and standard normal variate (SNV) were used in PLSR for preprocessing of spectral data. MSC efficiently eliminates the base line drift of the spectra which in turn reflects the more detailed characteristics of the spectra, and also removes additive and/or multiplicative signal effects (Brunet et al., 2009). The main advantage of SNV is to avoid attributes in greater numeric ranges dominate those in smaller numeric ranges. The PLSR model basing on all variables of the spectra is complex, thus a special algorithm uninformative variable elimination (UVE) was used as a method for variables selection of NIR spectral data in order to develop the effective PLSR prediction model for determination of pesticide the concentrations in each sample.

UVE is an algorithm based on the regression coefficient b of PLSR (Chen et al., 2005; Wu et al., 2009). In the PLSR-NIR prediction model, there is a relationship between X (spectral matrix) and Y (concentration matrix):

$$
\Upsilon = \lambda b + \varepsilon \tag{1}
$$

where *b* is the regression coefficient vector, *e* is the error vector. The following five steps were taken to get a new spectral matrix with fewer wave bands:

1. PLSR was used to develop a prediction model in the entire wave range from 4000 cm-1 to 10000 cm-1. Cross validation was applied to the calibration set. Each time, one sample was taken out from the calibration set. A calibration model was established for the remaining samples and the model was then used to predict the sample left out. Thereafter, the sample was placed back into the calibration set and a second sample was taken out. The procedure was repeated until all samples have been left out once. The root mean square error of cross validation (RMSEcv) was calculated for each of all wavelength combinations. The best principal component (PC) number with the highest Rcv (correlation coefficient of cross validation) and lowest RMSEcv value was selected.


$$C\_i = \frac{mean\left(b\_i\right)}{std\left(b\_i\right)}\tag{2}$$

5. Absolute value of *Ci* was used to discriminate if each spectra variable is effective or not. All effective variables were selected to create a new independent variable matrix, and then this new matrix and Y were used to establish a new PLSR prediction model.

#### **NIR spectra**

456 Infrared Spectroscopy – Materials Science, Engineering and Technology

Drying filter paper samples: Platform with filter paper samples were carefully moved into the vacuum drying oven, at room temperature for 1 hour. After drying, samples were stored

An Antaris FT-NIR spectrometer (Thermo Nicolet, Waltham, Massachusetts, USA), equipped with an InGaAs detector was used. The filter paper sample was placed in a specially modified sample cell. The spectra were acquired in the range of 4000 cm-1 to 10000 cm-1 at 8 cm-1 interval. For each sample, three points were chosen randomly for the NIR measurement, and 32 scans were co-added for each point. The sample was then removed, and the spectra were collected again in the same manner. Three spectra were obtained for each sample at the same state, and averaged spectra were calculated for further evaluation. To prevent the interference of water vapor in the air, the spectra of samples were acquired

The Matlab 7.0 software (MathWorks, USA) was used for all calculations. A total of 99 filter paper samples were divided into two groups, 75 samples were selected as calibration set; the left 24 samples in each concentration level were put into validation sample set. Partial least squares regression (PLSR) was used to develop a prediction model. Multiplicative scatter correction (MSC) and standard normal variate (SNV) were used in PLSR for preprocessing of spectral data. MSC efficiently eliminates the base line drift of the spectra which in turn reflects the more detailed characteristics of the spectra, and also removes additive and/or multiplicative signal effects (Brunet et al., 2009). The main advantage of SNV is to avoid attributes in greater numeric ranges dominate those in smaller numeric ranges. The PLSR model basing on all variables of the spectra is complex, thus a special algorithm uninformative variable elimination (UVE) was used as a method for variables selection of NIR spectral data in order to develop the effective PLSR prediction model for

UVE is an algorithm based on the regression coefficient b of PLSR (Chen et al., 2005; Wu et al., 2009). In the PLSR-NIR prediction model, there is a relationship between X (spectral

where *b* is the regression coefficient vector, *e* is the error vector. The following five steps

1. PLSR was used to develop a prediction model in the entire wave range from 4000 cm-1 to 10000 cm-1. Cross validation was applied to the calibration set. Each time, one sample was taken out from the calibration set. A calibration model was established for the remaining samples and the model was then used to predict the sample left out. Thereafter, the sample was placed back into the calibration set and a second sample was taken out. The procedure was repeated until all samples have been left out once. The root mean square error of cross validation (RMSEcv) was calculated for each of all wavelength combinations. The best principal component (PC) number with the highest Rcv (correlation coefficient of cross validation) and lowest RMSEcv value was selected.

Y = X*b* + *e* (1)

into vacuum packing bags immediately and marked with different concentrations.

immediately after taking out from the vacuum packing bags.

determination of pesticide the concentrations in each sample.

were taken to get a new spectral matrix with fewer wave bands:

**Pre-processing method and data analysis** 

matrix) and Y (concentration matrix):

**Spectrum acquisition** 

Original NIR spectra of total 99 filter samples are shown in figure 2, and the spectra of samples after pre-processing with MSC are shown in figure 3. It is obviously seen that the base line drift of the spectra is reduced in the figure 3 compared to figure 2 by the application of MSC.

#### **Results of PLSR in full bands**

For the total sets, two spectrum pre-processing methods MSC and SNV were used. Figure 4 illustrates the results of the cross validation when MSC and SNV were used as the spectrum pre-processing method.

Fig. 2. NIR transmittance spectra of filter-paper samples with different chlorpyrifos content.

Optical Technologies for Determination of Pesticide Residue 459

coefficient (R) between predicted and actual data. The results obtained are shown in table 2

From table 2 we can see that PLSR method do get satisfied prediction results. However, PLSR method using full bands of the spectra for developing calibration model are timeconsuming while running the computer program. Some variables in the full bands of samples' spectra are effective while some are not. As such determining effective spectra from the full band spectra is very essential. A special algorithm, namely uninformative variable elimination (UVE) was used in this research to find out the effective variables. The

Pre-processing method LV Rcv RMSEcv (mg/kg) R RMSEP (mg/kg)

Table 2. Calibration and validation results for chlorpyrifos concentration by using PLSR

MSC and SNV were used as the pre-processing method. According the result in table1, the optimal principal component number was chosen as 10. Then UVE algorithm was used to

In figure 5, the dotted line indicates the threshold of variables selection. In the range of [1, 1557], the variables corresponding the *C* value within the threshold range are not effective,

MSC + SNV 10 0.91 0.41 0.95 0.32

LV: the optimal principal component (PC) number used in cross-validation

select the effective variables. The results are shown in figure 5.

corresponding to R = 0.95 and RMSEP= 0.32 mg/kg.

variables with useless information were eliminated.

Rcv: correlation coefficient of cross validation RMSEcv: root mean square error of cross validation

R: correlation coefficients in validation set RMSEP: root mean square error of prediction

Fig. 5. Variables selected by UVE.

method.

**Results of UVE-PLSR** 

Fig. 3. NIR transmittance spectra of filter-paper samples after MSC.

Fig. 4. Optimal PC number of prediction model for filter-paper samples.

The total sample sets were separated into calibration set and validation set. Cross validation was first used in calibration sample set to find the optimal principle component number. From figure 4 we can see the best principle component number to be 10 with corresponding highest Rcv of 0.91 and lowest RMSEcv of 0.41. Model accuracy was then evaluated on the validation set using the root mean square error of prediction (RMSEP), correlation coefficient (R) between predicted and actual data. The results obtained are shown in table 2 corresponding to R = 0.95 and RMSEP= 0.32 mg/kg.

#### **Results of UVE-PLSR**

458 Infrared Spectroscopy – Materials Science, Engineering and Technology

Fig. 3. NIR transmittance spectra of filter-paper samples after MSC.

Fig. 4. Optimal PC number of prediction model for filter-paper samples.

The total sample sets were separated into calibration set and validation set. Cross validation was first used in calibration sample set to find the optimal principle component number. From figure 4 we can see the best principle component number to be 10 with corresponding highest Rcv of 0.91 and lowest RMSEcv of 0.41. Model accuracy was then evaluated on the validation set using the root mean square error of prediction (RMSEP), correlation From table 2 we can see that PLSR method do get satisfied prediction results. However, PLSR method using full bands of the spectra for developing calibration model are timeconsuming while running the computer program. Some variables in the full bands of samples' spectra are effective while some are not. As such determining effective spectra from the full band spectra is very essential. A special algorithm, namely uninformative variable elimination (UVE) was used in this research to find out the effective variables. The variables with useless information were eliminated.


LV: the optimal principal component (PC) number used in cross-validation

Rcv: correlation coefficient of cross validation

RMSEcv: root mean square error of cross validation

R: correlation coefficients in validation set

RMSEP: root mean square error of prediction

Table 2. Calibration and validation results for chlorpyrifos concentration by using PLSR method.

MSC and SNV were used as the pre-processing method. According the result in table1, the optimal principal component number was chosen as 10. Then UVE algorithm was used to select the effective variables. The results are shown in figure 5.

Fig. 5. Variables selected by UVE.

In figure 5, the dotted line indicates the threshold of variables selection. In the range of [1, 1557], the variables corresponding the *C* value within the threshold range are not effective,

Optical Technologies for Determination of Pesticide Residue 461

(a) (b)

(c) (d)

etc. Hyperspectral imaging is a powerful tool for acquiring both spectral and spatial information from an object at contiguous wavelengths over a wide spectral range. According to determination of pesticide, hyperspectral imaging combined with fluorescence

The following example presents the methodology to determinate chlorpyrifos based on

Pesticide solution: A commercial pesticide, containing 40% chlorpyrifos (Noposion, China) was used. Methanol was prepared in order to provide the solutions with different concentrations. Five concentration levels, 0.5, 1, 2, 8 and 16 mg/kg of active ingredient were

Vegetable samples: Pollution-free rapes bought from local market were used. After washing up all the surface of rape samples by the use of distilled water, pesticide solutions were

Fig. 6. Variables selected by UVE with different random matrix.

stimulate technology could acquire a satisfactory result.

hyperspectral fluorescence imaging technology.

diluted based on the amount of chlorpyrifos.

sprayed evenly on dry rape samples' surface.

**Samples** 

and 368 variables left after ineffective variables were eliminated. A new PLSR prediction model was developed by using these 368 variables. The results showed that the correlation coefficient (Rcv) in cross validation is 0.91, the root mean square error of cross validation (RMSEcv) is 0.42 mg/kg, and the correlation coefficient (R) in validation set is 0.94, the root mean square error of prediction (RMSEP) is 0.36 mg/kg. Compared with the results of the PLSR used in full bands, the UVE-PLSR could get similar results but using fewer wave bands. In the UVE algorithm, the random matrix which was added into the original matrix was different each time, so the results would be different in every prediction model. In order to prove the stability of UVE algorithm, another 4 times of UVE-PLSR was used. The results of a total of five times UVE-PLSR are shown in table 3.


LV: the optimal principal component (PC) number used in cross-validation

Rcv: correlation coefficient of cross validation

RMSEcv: root mean square error of cross validation

R: correlation coefficients in validation set

RMSEP: root mean square error of prediction

Table 3. Prediction results of UVE-PLSR methods.

Figure 6 shows the variables selection results by the use of another 4 times UVE-PLSR based on different random matrix. As the results shown in table 3, the differences between each UVE-PLSR are small. The number of variables ranged from 281 to 395, and the prediction results were almost identical to each other. Considering the different random matrix, the Rcv (correlation coefficient of cross validation) range from 0.90 to 0.91, RMSEcv (cross validation) range from 0.42 to 0.47 mg/kg, R (validation set) is 0.94, RMSEP range from 0.36 to 0.37mg/kg which MSC and SNV were used as the pre-processing method. It could be concluded that the differences of random matrix have very weak affection in the process of developing a prediction model, and the numbers of variables used in UVE-PLSR could be declined by more than 70%. These results indicated that the prediction capability of UVE-PLSR is similar as the PLSR used in full bands. So, it can be concluded that NIR determination of pesticide is a low cost, an environment friendly and a potential method compared to the traditional methods, and the UVE-PLSR algorithm is an efficient method to select the effective variables of spectra and develop a prediction model of pesticide concentration with fewer wave bands.

#### **3. Optical imaging technology for pesticide determination**

In recent years, optical imaging technology has become popular. Hyperspectral imaging as one of optical imaging technology has been used in agriculture, biomedicine, food industry

and 368 variables left after ineffective variables were eliminated. A new PLSR prediction model was developed by using these 368 variables. The results showed that the correlation coefficient (Rcv) in cross validation is 0.91, the root mean square error of cross validation (RMSEcv) is 0.42 mg/kg, and the correlation coefficient (R) in validation set is 0.94, the root mean square error of prediction (RMSEP) is 0.36 mg/kg. Compared with the results of the PLSR used in full bands, the UVE-PLSR could get similar results but using fewer wave bands. In the UVE algorithm, the random matrix which was added into the original matrix was different each time, so the results would be different in every prediction model. In order to prove the stability of UVE algorithm, another 4 times of UVE-PLSR was used. The results

(mg/kg) R RMSEP

(mg/kg)

Variables Thresholds

of a total of five times UVE-PLSR are shown in table 3.

Variables LV Rcv RMSEcv

LV: the optimal principal component (PC) number used in cross-validation

**3. Optical imaging technology for pesticide determination** 

UVE-PLSR-1 368 7 0.91 0.42 0.94 0.36 ±29.16 UVE-PLSR-2 281 7 0.90 0.47 0.94 0.37 ±31.31 UVE-PLSR-3 395 7 0.90 0.43 0.94 0.36 ±27.61 UVE-PLSR-4 379 7 0.90 0.43 0.94 0.37 ±28.23 UVE-PLSR-5 330 7 0.90 0.43 0.94 0.36 ±30.27

Figure 6 shows the variables selection results by the use of another 4 times UVE-PLSR based on different random matrix. As the results shown in table 3, the differences between each UVE-PLSR are small. The number of variables ranged from 281 to 395, and the prediction results were almost identical to each other. Considering the different random matrix, the Rcv (correlation coefficient of cross validation) range from 0.90 to 0.91, RMSEcv (cross validation) range from 0.42 to 0.47 mg/kg, R (validation set) is 0.94, RMSEP range from 0.36 to 0.37mg/kg which MSC and SNV were used as the pre-processing method. It could be concluded that the differences of random matrix have very weak affection in the process of developing a prediction model, and the numbers of variables used in UVE-PLSR could be declined by more than 70%. These results indicated that the prediction capability of UVE-PLSR is similar as the PLSR used in full bands. So, it can be concluded that NIR determination of pesticide is a low cost, an environment friendly and a potential method compared to the traditional methods, and the UVE-PLSR algorithm is an efficient method to select the effective variables of spectra and develop a prediction model of pesticide

In recent years, optical imaging technology has become popular. Hyperspectral imaging as one of optical imaging technology has been used in agriculture, biomedicine, food industry

Model Number of

Rcv: correlation coefficient of cross validation RMSEcv: root mean square error of cross validation

Table 3. Prediction results of UVE-PLSR methods.

R: correlation coefficients in validation set RMSEP: root mean square error of prediction

concentration with fewer wave bands.

Fig. 6. Variables selected by UVE with different random matrix.

etc. Hyperspectral imaging is a powerful tool for acquiring both spectral and spatial information from an object at contiguous wavelengths over a wide spectral range. According to determination of pesticide, hyperspectral imaging combined with fluorescence stimulate technology could acquire a satisfactory result.

The following example presents the methodology to determinate chlorpyrifos based on hyperspectral fluorescence imaging technology.

#### **Samples**

Pesticide solution: A commercial pesticide, containing 40% chlorpyrifos (Noposion, China) was used. Methanol was prepared in order to provide the solutions with different concentrations. Five concentration levels, 0.5, 1, 2, 8 and 16 mg/kg of active ingredient were diluted based on the amount of chlorpyrifos.

Vegetable samples: Pollution-free rapes bought from local market were used. After washing up all the surface of rape samples by the use of distilled water, pesticide solutions were sprayed evenly on dry rape samples' surface.

Optical Technologies for Determination of Pesticide Residue 463

Fig. 8. Hyperspectral fluorescence image of 8mg/kg sample.

Fig. 9. ROI image of the sample with the chlorpyrifos concentration of 8mg/kg sample.

emission spectral intensity at the peak. It can also be seen in figure 10 that the fluorescence emission peak value reduces when the concentration of chlorpyrifos decreases. The peak emission at the wavelength 524nm might be the effect of other organic elements present in the commercial composite pesticide used in this research. The results can be used as theoretical basis for developing rapid detection instrument for vegetable pesticide residue.

#### **Data acquisition**

A hyperspectral fluorescence imaging system (Figure 7) was used which mainly consisted of a high-performance back-illuminated charge coupled device (CCD) camera (Sencicam QE Germany), an imaging spectrograph (ImSpector V10E, Spectral Imaging Ltd., Finland), and a light unit with fluorescence lamps as the light source (Foshan, China). The camera, with spatial resolution of 1376×1040, was fitted with a 25mm lens (Computar, Japanese), the spectrograph had an effective spectral region from 400 to 1100nm with a 2.8 nm spectral resolution. The light source used in this study was a pair of fluorescence lamps which had the spectral region from 340 to 600nm. The whole system was shielded with a close chamber avoiding the interferences of external lights from outside. Hyperspectral fluorescence images were acquired and ENVI 4.3 software was used for data analysis.

Fig. 7. Hyperspectral fluorescence imaging system.

#### **Data analysis**

Figure 8 shows the hyperspectral fluorescence image of 8mg/kg sample which was composed by the use of ENVI 4.3 software. In this picture, the white regions are the fluorescence of chlorpyrifos solutions. Threshold segmentation method was used to acquire the Region of Interest (Figure 9, red regions). Then average spectral curves of different samples in Region of Interest were calculated in whole wave bands (Figure 10). According to figure 10, chlorpyrifos has strong fluorescence characteristic when methanol is used as solvent. The emission spectrum of chlorpyrifos indicates that it has the peak emission at the wavelength of 437 nm, and chlorpyrifos samples with different concentration have different fluorescence

A hyperspectral fluorescence imaging system (Figure 7) was used which mainly consisted of a high-performance back-illuminated charge coupled device (CCD) camera (Sencicam QE Germany), an imaging spectrograph (ImSpector V10E, Spectral Imaging Ltd., Finland), and a light unit with fluorescence lamps as the light source (Foshan, China). The camera, with spatial resolution of 1376×1040, was fitted with a 25mm lens (Computar, Japanese), the spectrograph had an effective spectral region from 400 to 1100nm with a 2.8 nm spectral resolution. The light source used in this study was a pair of fluorescence lamps which had the spectral region from 340 to 600nm. The whole system was shielded with a close chamber avoiding the interferences of external lights from outside. Hyperspectral fluorescence

Figure 8 shows the hyperspectral fluorescence image of 8mg/kg sample which was composed by the use of ENVI 4.3 software. In this picture, the white regions are the fluorescence of chlorpyrifos solutions. Threshold segmentation method was used to acquire the Region of Interest (Figure 9, red regions). Then average spectral curves of different samples in Region of Interest were calculated in whole wave bands (Figure 10). According to figure 10, chlorpyrifos has strong fluorescence characteristic when methanol is used as solvent. The emission spectrum of chlorpyrifos indicates that it has the peak emission at the wavelength of 437 nm, and chlorpyrifos samples with different concentration have different fluorescence

images were acquired and ENVI 4.3 software was used for data analysis.

Fig. 7. Hyperspectral fluorescence imaging system.

**Data analysis** 

**Data acquisition** 

Fig. 8. Hyperspectral fluorescence image of 8mg/kg sample.

Fig. 9. ROI image of the sample with the chlorpyrifos concentration of 8mg/kg sample.

emission spectral intensity at the peak. It can also be seen in figure 10 that the fluorescence emission peak value reduces when the concentration of chlorpyrifos decreases. The peak emission at the wavelength 524nm might be the effect of other organic elements present in the commercial composite pesticide used in this research. The results can be used as theoretical basis for developing rapid detection instrument for vegetable pesticide residue.

Optical Technologies for Determination of Pesticide Residue 465

Khanmohammadi M., S. Arment, S. Garrigues and M. de la Guardia. 2008. Mid-and near-

Leroy, B., S. Lambotte, O. Dotreppr, H. Lecocq, L. Istasse, and A. Clinquart. 2003. Prediction

Peng, Y., and R. Lu. 2008. Analysis of spatially resolved hyperspectral scattering images for

Peng, Y., and J. H. Wu. 2008. Hyperspectral scattering profiles for prediction of beef

Peng, Y., J. Zhang, and J.H. Wu. 2009. Hyperspectral scattering profiles for prediction of the microbial spoilage of beef. SPIE Paper No. 7315-25, Orlando, Florida, USA. Ripoll, G., P. Alberti, B. Panea, J.L. Olleta, and C. Sanudo. 2008. Near-infrared reflectance

Saranwong, S., and S. Kawano. 2005. Rapid determination of fungicide contaminated on

Subbiah, J., C.R. Calkins, A. Samal, and G.E. Meyer. 2008. Visible/near-infrared

Wu, D., H.X. Wu, J.B. Cao, Z.H. Huang, and Y. He. 2009. Classifying the species of

Li Shuqian, Lu Lei, Chen Fusheng, et al. Rapid detection techniques of organophosphorus

Zhao Siqi, Yan Su. The detection technologies of pesticide residue [J]. Anhui Agricultural

Zhu Chunyan, Li Weikai, Li Yanmei. Detection of organophosphorus pesticide residues on

Lai Suichun, Wang Fuhua, Deng Yicai. Research situation and development of pesticide

ReneeD. JiJi, GaryA. Cooper, KarlS. Booksh. Excitation- emission matrix fluorescence based

Shengye Jin, Zhaochao Xu, Jiping Chen, Xinmiao Liang, Yongning Wu, Xuhong Qian.

Sciences, 2008, 36(10): 4176-4178. (in Chinese with English abstract)

Herald, 2008, (2): 108-108. (in Chinese with English abstract)

tomato surface. *Journal of Near infrared spectrosc*. 13:169-175.

near-infrared reflection and transmission spectra. *Meat Science*. 66: 45-54. Luypaert, J., M.H. Zhang, and D.L. Massart. 2003. Feasibility study for the use of near

Camellia sinensis (L.). *Analytica Chimica Acta.* 478: 303-312.

*Spectroscopy*. 46: 82-88.

*Technology.* 48: 52-62.

*Meat Science*. 80: 697-702.

*Waves*. 28(6): 423-427.

*Electronics in Agriculture*. 64: 225-233.

58-59. (in Chinese with English abstract)

(in Chinese with English abstract)

Analytica Chimica Acta,1999, (397): 61-72.

USA.

infrared determination of metribuzin in agrochemicals. *Journal of Vibrational* 

of technological and organoleptic properties of beef longissimus thoracis from

infrared spectroscopy in the qualitative and quantitative analysis of green tea,

assessing apple fruit firmness and soluble solids content. *Postharvest Biology and* 

tenderness. ASABE Paper No. 080004. Rhode Island convention center, Rhode,

spectroscopy for predicting chemical, instrumental and sensory quality of beef.

hyperspectral imaging for beef tenderness prediction. *Journal of Computers and* 

exopalaemo by using visible and near infrared spectra with uninformative variable elimination and successive projections algorithm. *Journal of Infrared and Millimeter* 

pesticide residue in fruits and vegetables [J]. Hubei Agricultural Science, 2004, (4):

vegetables by using FTIR/ATR method [J]. Science and Technology Innovation

residues analysis technology [J]. Guangdong Agricultural Science, 2006, (1): 76-77.

determination of carbamate pesticides and polycyclic aromatic hydrocarbons [J].

Determination of organophosphate and carbonate pesticides based on enzyme

Fig. 10. Emission spectrum of samples with different chlorpyrifos concentration.

#### **4. Conclusion**

Pesticide concentration can be readily measured with NIR spectroscopy and optical imaging technology. However the accuracy and precision could be improved. There is a need to develop rapid optical techniques for pesticide determination which could be used in the future for agro-food safety assurance. The optical technique could be one of the most useful tools along with the advancement of spectral instrument for determination of pesticide residue.

#### **5. References**


Fig. 10. Emission spectrum of samples with different chlorpyrifos concentration.

Pesticide concentration can be readily measured with NIR spectroscopy and optical imaging technology. However the accuracy and precision could be improved. There is a need to develop rapid optical techniques for pesticide determination which could be used in the future for agro-food safety assurance. The optical technique could be one of the most useful tools along with the advancement of spectral instrument for determination of pesticide

Armenta, S., S. Garrigues, and M. de la Guardia.2007. Partial least squares-near infrared

Brunet, D., T. Woignier, M. Lesueur-Jannoyer, R. Achard, L. Rangon, and B.G. Barthes. 2009.

Gambacorta, G., M. Faccia, C. Lamacchia, A. Di Luccia, and E. La Notte. 2005. Pesticide

residues in tomato grown in open field. *Food control*. 16: 629-632.

determination of pesticides in commercial formulations. *Journal of Vibrational* 

Determination of soil content in chlordecone (organochlorine pesticide) using near infrared reflectance spectroscopy (NIRS). *Environmental Pollution*. 157: 3120-3125. Chen, B., and D. Chen. 2005. The application of uninformative variables elimination in nearinfrared spectroscopy. *Spectronic Instruments and Analysis*. 04: 26-30. ElMasry, G., N. Wang, A. ElSayed, and M. Ngadi. 2007. Hyperspectral imaging for

nondestructive determination of some quality attributes for strawberry. *Journal of* 

**4. Conclusion** 

**5. References** 

*Spectroscopy*. 44: 273-278.

*Food Engineering*.81: 98-107.

residue.


**25** 

E.M. Sheregii

*Poland* 

*University of Rzeszow, Rzeszów* 

**High Resolution Far Infrared Spectra** 

**of the Semiconductor Alloys Obtained** 

**Using the Synchrotron Radiation as Source** 

It is known that far-infrared (FIR) spectra give direct information on phonon modes and impurity levels in the crystal lattices. Infrared spectroscopy enable us also to obtain information about real crystalline microstructure and interior interactions of the

A role of the semiconductor alloys in electronics and optoelectronics increases constantly. Since the electron-phonon interaction is main mechanism of the current carriers scattering in semiconductors it is important to recognize deeply the phonon spectra of the semiconductors solid solutions. The results on the semiconductor alloy FIR-spectra collected during 70, 80 and 90 decades have not been explained satisfactorily on base of simple "twomode behavior" for ternary and "three-mode behavior" for quaternary alloys. The previous reviews and books dedicated to the phonon spectra of semiconductor's compounds (Barker & Sievers, 1975;Tylor,1988;Adachi,1999;Kosevich,1999) give not a reasonable answers on the questions concerning: whether the vibrations of different dipole pairs in the solid solution are connected in the alloy lattice and form a running wave (phonons) or on the contrary: they are disseminated on great number of local modes? Another one concerning of the solid solutions microstructure – geometry of chaos: whether this geometry factor is sufficient or thermodynamic one is necessary to add. The adequate describing of this geometry factor applying to the phonon spectra interpretation should be developed both for ternary alloys

In order to give answers on these principally important question we need the credible experimental results and for this purpose in years 2001-2006 were performed in Laboratory Nationale di Frascati the six TARI (Transnational Access to the Research Infrastructure) Projects concerning the FIR-spectra of the semiconductor solid solutions obtained by measuring of optical reflectivity using synchrotron radiation as source. The brilliant properties of synchrotron radiation enable us to obtain FIR-spectra of comparably high resolution: about 1 cm-1 whereas nature sources provided 2.5 cm-1 at better case. The results on FIR-spectra obtained by this way for mercury contained semiconductor alloys – ternary as Hg1-xCdxTe as well as quaternary like ZnyCdxHg1-x-yTe – will be presented in this Chapter. There is third principally important question concerning Hg-contained solid solutions:

semiconductor solid solutions (Barker & Sievers, 1975; Robouch et al.,2001).

**1. Introduction** 

and quaternary also.

inhibition using a pH-sensitive fluorescence probe [J]. Analytica Chimica Acta, 2004, (523): 117-123.


## **High Resolution Far Infrared Spectra of the Semiconductor Alloys Obtained Using the Synchrotron Radiation as Source**

E.M. Sheregii *University of Rzeszow, Rzeszów Poland* 

### **1. Introduction**

466 Infrared Spectroscopy – Materials Science, Engineering and Technology

Salan Hassoon, Israel Schechter. A sensitive fluorescence probe for DDT-type pesticides [J].

Salan Hassoon, Israel Schechter. In situ fluorimetric determination of pesticides on

Harald Hake, Ravid Ben-Zur, Israel Schechter, Angelika Anders. Fast optical assessment of pesticide coverage on plants [J]. Analytica Chimica Acta,2007, (596): 1-8. Bengt Danielsson, Ioana Surugiu, Anatoli Dzgoev, Michael Mecklenburg, Kumaran

Xiangying Sun, Kaihao Xia, Bin Liu. Design of fluorescent self-assembled multilayers and

J.F.Garcia Reyes, E.J.LlorentMart Mnez, P.Ortega Barrales, A.Molina Diaz. Multiwavelength

Atanasse Coly, Jean-Jacques Aaron. Fluorimetric analysis of pesticides: Methods, recent

Munoz de la Pena, M.C. Mahedero, A.Bautista-Sanchez. High-performance liquid

Wang Zhongdong, Wang Yutian. Theoretical and experimental study on fluorescence

Lou Zhizai, Huang Shihua. Detecting of psticide residue in vegetable using fluorescence

Chen Jingjing, Li Yongyu, Wu Jianhu, Peng Yankun. Rapid determination of ppm-order

production, agriculture and the environment in Asia, Japan, 2009: 103-107. Chen Jingjing, Li Yongyu, Wu Jianhu, et al. Rapid determination of ppm-order

Y. Peng, R. Lu. Prediction of apple fruit Firmness and soluble solids content using

2004, (523): 117-123.

227-234.

Analytica Chimica Acta, 1998, (368): 77-82.

vegetables [J]. Analytica Chimica Acta,2000, (405): 9-15.

carbaryl and benomyl [J]. Talanta,2004, (64): 742-749.

(in Chinese with English abstract)

Chinese with English abstract)

English abstract)

2006, (82): 142-152.

26(1): 59-65. (in Chinese with English abstract)

developments and applications. Talanta, 1998, (46):815-843.

inhibition using a pH-sensitive fluorescence probe [J]. Analytica Chimica Acta,

Ramanathan. Optical detection of pesticides and drugs based on chemiluminescence-fluorescence assays [J]. Analytica Chimica Acta,2001, (426):

interfacial sensing for organophosphorus pesticides [J]. Talanta, 2008, (76): 747-751.

fluorescence based optosensor for simultaneous determination of fuberidazole,

chromatographic determination of phenylureas by photochemically-induced fluorescence detection [J]. Journal of Chromatography A, 2002, (950): 287-291. Wang Yutian, Wang Zhongdong. Study on fluorescence spectrometer for monitoring

pesticide residues on vegetables [J]. Journal of Applied Optics, 2005, 26(5): 10-13.

characteristics of common pesticides [J]. Chinese Journal of Luminescence, 2005,

technique [J]. Acta Laser Biology Sinica, 2008, 17(6): 657-660. (in Chinese with

concentration of organophosphorus pesticide based on near-infrared spectroscopy[C]. The 3rd international symposium on sustainability in food

concentration of organophosphorus pesticide based on near-infrared spectroscopy[J]. Food safety & Quality Detection Technology, 2009, 1(1): 45-50. (in

characteristics of multispectral scattering images [J]. Journal of Food Engineering,

It is known that far-infrared (FIR) spectra give direct information on phonon modes and impurity levels in the crystal lattices. Infrared spectroscopy enable us also to obtain information about real crystalline microstructure and interior interactions of the semiconductor solid solutions (Barker & Sievers, 1975; Robouch et al.,2001).

A role of the semiconductor alloys in electronics and optoelectronics increases constantly. Since the electron-phonon interaction is main mechanism of the current carriers scattering in semiconductors it is important to recognize deeply the phonon spectra of the semiconductors solid solutions. The results on the semiconductor alloy FIR-spectra collected during 70, 80 and 90 decades have not been explained satisfactorily on base of simple "twomode behavior" for ternary and "three-mode behavior" for quaternary alloys. The previous reviews and books dedicated to the phonon spectra of semiconductor's compounds (Barker & Sievers, 1975;Tylor,1988;Adachi,1999;Kosevich,1999) give not a reasonable answers on the questions concerning: whether the vibrations of different dipole pairs in the solid solution are connected in the alloy lattice and form a running wave (phonons) or on the contrary: they are disseminated on great number of local modes? Another one concerning of the solid solutions microstructure – geometry of chaos: whether this geometry factor is sufficient or thermodynamic one is necessary to add. The adequate describing of this geometry factor applying to the phonon spectra interpretation should be developed both for ternary alloys and quaternary also.

In order to give answers on these principally important question we need the credible experimental results and for this purpose in years 2001-2006 were performed in Laboratory Nationale di Frascati the six TARI (Transnational Access to the Research Infrastructure) Projects concerning the FIR-spectra of the semiconductor solid solutions obtained by measuring of optical reflectivity using synchrotron radiation as source. The brilliant properties of synchrotron radiation enable us to obtain FIR-spectra of comparably high resolution: about 1 cm-1 whereas nature sources provided 2.5 cm-1 at better case. The results on FIR-spectra obtained by this way for mercury contained semiconductor alloys – ternary as Hg1-xCdxTe as well as quaternary like ZnyCdxHg1-x-yTe – will be presented in this Chapter. There is third principally important question concerning Hg-contained solid solutions:

High Resolution Far Infrared Spectra of the

**0,0**

presented in Fig. 2 and 3.

curves R

In Fig.2 are shown the FIR-spectra as reflectivity curves R

*Hg0.8Cd0.2Te* in the spectral range 90 cm-1- 115 cm-1.

**0,2**

**0,4**

**Reflectivity** 

**0,6**

**0,8**

Semiconductor Alloys Obtained Using the Synchrotron Radiation as Source 469

 x=0.7 x=0.4 x=0.06 x=0.2, n-type x=0.2, p-type

**Te**

**Hg1-xCdx**

for the n-type *Hg0.8 Cd0.2Te* 

**100 120 140 160 180 200**

**Wawenumber, cm-1**

corresponds to the *HgTe-*like sub-band and second one – to the *CdTe*-like sub-band.. This type of reflectance spectrum again shows according to previous work (Baars & Sorgers, 1972; Amirtharaj et al.,1990; Biao,1996; Rath et al., 1995) two-mode behavior of the optical phonons in the *Hg1-xCdxTe* alloys*.* Whereas, the subtle structure of both sub-bands is clearly observed too what undoubtedly indicates on multi-mode character of phonon spectra (above two modes). Authors (Kozyrev et al.,1998) interpreted these subtle structure in frame of the V-B model but they limited consideration of the FIR-spectra in the spectral region 118 – 160 cm-1 . Whereas, in the region 90 – 116 cm-1 are observed additional lines registered earlier (Talwar, 1984; Amirtharaj et al., 1990;Biao,1996;Rath 1995). The line amplitudes of main sub-bands (HgTe-like at 118 – 130 cm-1 as well as CdTe-like at 140 – 170 cm-1) decrease when the temperature increases for both samples. Contrary, the line amplitudes of additional lines (we can call them as Additional Phonon Modes (APM) whereas the main sub-bands we can called as Canonical Phonon Modes (CPM)) increase when the temperature increases. That is clearly shown on the temperature dependence of FIR-spectra

alloy in the temperature region 30 K – 300 K. The temperature dependence of reflectivity

*like* band towards the higher frequency side with increase of the temperature and shift of the *CdTe-like* band to the lower frequency side when the temperature increases similarly to results obtained in (Roth et al.,1995). The main *TO* – phonon mode frequency of *HgTe -*like sub-band increases from 118 cm-1 at 30K to 121 cm-1 at 300 K for n-type alloy. We have inferred from the FIR-spectra that the sings of the temperature induced shifts of the *HgTe* – like and *CdTe* –like mode frequencies in the MCT alloy are opposite to each other for the composition range *x ≤* 0.3. The *CdTe* –like mode frequency decrease from 154.2 cm-1 to 152 cm-2 with the increase in temperature from 30K to 300K and the intensity of the *HgTe*-like and *CdTe*-like *TO* mode decreases and higher background is observed for the *p*-type

for p- *Hg0.8 Cd0.2Te* alloy is presented in Fig. 3. We may see the shift of the *HgTe-*

Fig. 1. Reflectivity spectra *CdxHg1-x Te* (x is changing from 0.06 to 0.7) obtained at 300K*.* 

additional lines arousing constantly in the region of frequencies lower then main HgTe-like modes. The cause of appearing of these additional lines was not explained (Baars & Sorgers, 1972;)Amirtharaj et al, 1990;Biao,1996;Rath et al., 1995). The new results as well as previous published ones (Sheregii et al, 2006; Cebulski, et al.,2008;Polit et al., 2010;Sheregii et al.,2009;Sheregii et al.,20011) but with new interpretation, will be presented here and allow us partly to respond on the formulated above questions.

## **2. Experiment**

## **2.1 Experimental technique**

The optical reflection spectra in the region from 10 to 10 000 cm-1 where the phonon frequency values (30 – 400 cm-1) of semiconductor's compounds are located, were measured in the wide temperature interval and composition region. Experiments were performed at the DAFNE-light laboratory at Frascati (Italy) using the experimental set-up described in (Cesteli Guidi et al., 2005). A BRUKER Equinox 55 FT-IR interferometer modified to collect spectra in vacuum and both the synchrotron radiation light emitted by the DAFNE storage ring as well as a mercury lamp were used as IR sources (Marcelli et al., 2005). The measurements were performed in the temperature range of 20-300 *K* at the spectral resolution of 1 cm-1 (2 cm-1 in some cases) collecting typically 200 scans within 600 s of acquisition time with a bolometer cooled down to 4.2 *K.*

The reflectivity was measured by using as a reference a gold film evaporated onto the surface of the investigated samples. This method enabled us to measure the reflectivity coefficient with an accuracy of about 0.2 %. The imaginary part curves of the dielectric function *Im* were calculated, from reflectivity spectrum, by means of the Kramers-Kronig (KK) procedure with uncertainty less than 1.5%.

## **2.2 Experimental results for ternary alloys**

In the our previous published works were presented results obtained by the same way on the ternary solid solutions Hg1-xCdxTe and Hg1-xZnxTe (Cebulski, et al.,2008;Polit et al., 2010;Sheregii et al.,2009;Sheregii et al., 2011). It was shown in these works that observed subtle structure of the two phonon sub-bands in case of ternary alloys can be successfully explained on base of the five structural cells model of H.W.Verleur and A.S. Barker (V-B model) (Verleur & Barker, 1966) thought the additional phonon lines were observed. Last one required the new hypothesis – the two wells potential model for Hg-atoms in lattice (Polit et al., 2010) – for explanation the experimental spectra. The V-B model will be presented in next sub-chapter. In this sub-chapter are exposed the FIR-spectra concerning ternary alloys in order to illustrate the fact of multi-mode behaviour – main statement of the random version of the V-B model which is necessary to interpret of the experimental FIRspectra.

The high-resolution reflection FIR-spectra obtained for the ternary Hg1-xCdxTe for compositions from x=0.06 to x=0.7 at the temperature 300K and in the spectral range 100 cm-1 to 200 cm-1 are shown in Fig.1 as reflectivity curves R Two bands which shift weakly with the composition are observed: first one around 118-128 cm-1 and second one around 145-155 cm-1. The amplitude of first band increases when the content of *HgTe* increases and amplitude of second band increases when the *CdTe* content increases. The first band

additional lines arousing constantly in the region of frequencies lower then main HgTe-like modes. The cause of appearing of these additional lines was not explained (Baars & Sorgers, 1972;)Amirtharaj et al, 1990;Biao,1996;Rath et al., 1995). The new results as well as previous published ones (Sheregii et al, 2006; Cebulski, et al.,2008;Polit et al., 2010;Sheregii et al.,2009;Sheregii et al.,20011) but with new interpretation, will be presented here and allow

The optical reflection spectra in the region from 10 to 10 000 cm-1 where the phonon frequency values (30 – 400 cm-1) of semiconductor's compounds are located, were measured in the wide temperature interval and composition region. Experiments were performed at the DAFNE-light laboratory at Frascati (Italy) using the experimental set-up described in (Cesteli Guidi et al., 2005). A BRUKER Equinox 55 FT-IR interferometer modified to collect spectra in vacuum and both the synchrotron radiation light emitted by the DAFNE storage ring as well as a mercury lamp were used as IR sources (Marcelli et al., 2005). The measurements were performed in the temperature range of 20-300 *K* at the spectral resolution of 1 cm-1 (2 cm-1 in some cases) collecting typically 200 scans within 600 s of

The reflectivity was measured by using as a reference a gold film evaporated onto the surface of the investigated samples. This method enabled us to measure the reflectivity coefficient with an accuracy of about 0.2 %. The imaginary part curves of the dielectric

In the our previous published works were presented results obtained by the same way on the ternary solid solutions Hg1-xCdxTe and Hg1-xZnxTe (Cebulski, et al.,2008;Polit et al., 2010;Sheregii et al.,2009;Sheregii et al., 2011). It was shown in these works that observed subtle structure of the two phonon sub-bands in case of ternary alloys can be successfully explained on base of the five structural cells model of H.W.Verleur and A.S. Barker (V-B model) (Verleur & Barker, 1966) thought the additional phonon lines were observed. Last one required the new hypothesis – the two wells potential model for Hg-atoms in lattice (Polit et al., 2010) – for explanation the experimental spectra. The V-B model will be presented in next sub-chapter. In this sub-chapter are exposed the FIR-spectra concerning ternary alloys in order to illustrate the fact of multi-mode behaviour – main statement of the random version of the V-B model which is necessary to interpret of the experimental FIR-

The high-resolution reflection FIR-spectra obtained for the ternary Hg1-xCdxTe for compositions from x=0.06 to x=0.7 at the temperature 300K and in the spectral range 100 cm-

with the composition are observed: first one around 118-128 cm-1 and second one around 145-155 cm-1. The amplitude of first band increases when the content of *HgTe* increases and amplitude of second band increases when the *CdTe* content increases. The first band

Two bands which shift weakly

were calculated, from reflectivity spectrum, by means of the Kramers-

us partly to respond on the formulated above questions.

acquisition time with a bolometer cooled down to 4.2 *K.*

Kronig (KK) procedure with uncertainty less than 1.5%.

1 to 200 cm-1 are shown in Fig.1 as reflectivity curves R

**2.2 Experimental results for ternary alloys** 

**2. Experiment** 

function *Im*

spectra.

**2.1 Experimental technique** 

Fig. 1. Reflectivity spectra *CdxHg1-x Te* (x is changing from 0.06 to 0.7) obtained at 300K*.* 

corresponds to the *HgTe-*like sub-band and second one – to the *CdTe*-like sub-band.. This type of reflectance spectrum again shows according to previous work (Baars & Sorgers, 1972; Amirtharaj et al.,1990; Biao,1996; Rath et al., 1995) two-mode behavior of the optical phonons in the *Hg1-xCdxTe* alloys*.* Whereas, the subtle structure of both sub-bands is clearly observed too what undoubtedly indicates on multi-mode character of phonon spectra (above two modes). Authors (Kozyrev et al.,1998) interpreted these subtle structure in frame of the V-B model but they limited consideration of the FIR-spectra in the spectral region 118 – 160 cm-1 . Whereas, in the region 90 – 116 cm-1 are observed additional lines registered earlier (Talwar, 1984; Amirtharaj et al., 1990;Biao,1996;Rath 1995). The line amplitudes of main sub-bands (HgTe-like at 118 – 130 cm-1 as well as CdTe-like at 140 – 170 cm-1) decrease when the temperature increases for both samples. Contrary, the line amplitudes of additional lines (we can call them as Additional Phonon Modes (APM) whereas the main sub-bands we can called as Canonical Phonon Modes (CPM)) increase when the temperature increases. That is clearly shown on the temperature dependence of FIR-spectra presented in Fig. 2 and 3.

In Fig.2 are shown the FIR-spectra as reflectivity curves R for the n-type *Hg0.8 Cd0.2Te*  alloy in the temperature region 30 K – 300 K. The temperature dependence of reflectivity curves R for p- *Hg0.8 Cd0.2Te* alloy is presented in Fig. 3. We may see the shift of the *HgTelike* band towards the higher frequency side with increase of the temperature and shift of the *CdTe-like* band to the lower frequency side when the temperature increases similarly to results obtained in (Roth et al.,1995). The main *TO* – phonon mode frequency of *HgTe -*like sub-band increases from 118 cm-1 at 30K to 121 cm-1 at 300 K for n-type alloy. We have inferred from the FIR-spectra that the sings of the temperature induced shifts of the *HgTe* – like and *CdTe* –like mode frequencies in the MCT alloy are opposite to each other for the composition range *x ≤* 0.3. The *CdTe* –like mode frequency decrease from 154.2 cm-1 to 152 cm-2 with the increase in temperature from 30K to 300K and the intensity of the *HgTe*-like and *CdTe*-like *TO* mode decreases and higher background is observed for the *p*-type *Hg0.8Cd0.2Te* in the spectral range 90 cm-1- 115 cm-1.

High Resolution Far Infrared Spectra of the

layer and was ideal for optical measurements.

temperatures: 30 K, 100K and 300 K.

1,0

0,0

0,2

0,4

0,6

Reflectivity

0,8

Semiconductor Alloys Obtained Using the Synchrotron Radiation as Source 471

solution lattices with a common anion (ZnxCdyHg1-x-yTe or ZMCT in our case), enables us to control the material parameters with one extra degree of freedom (Cebulski et al., 1998). The bulk quaternary layers of ZnxCdyHg1-x-yTe were obtained by liquid phase epitaxial technique on the CdTe substrates in A.F. Joffe Physical-Technical Institute (St.Petersbourg, Russion), the compositions of samples are shown in Table 1. The thickness of the homogeneous layer was 4 m. The surface of samples was natural (110) plane of grown

Number of sample x, mol % y, mol.%

Optical reflectivity from surface of nine ZnxCdyHg1-x-yTe samples of seven compositions in the far-infrared region was measured using the synchrotron radiation as source (high resolution FIR-spectra – the reflectivity experiment is described above)). Some of results were published earlier (Sheregii et. al., 2006). The measurements of reflectivity were performed in temperature region from 30 K to 300 K. In Fig. 4 and 5 are presented reflectivity FIR-spectra obtained for two compositions of ZnxCdyHg1-x-yTe for three

100 120 140 160 180 200

 40K 100K 300K

Wavenumber, cm-1

I 0.020 0.200 II 0.070 0.21 III 0.120 0.17 IV 0.127 0.117 V 0.180 0.120 VI 0.050 0.230 VII 0.120 0.130

Table 1. Compositions of the ZnxCdyHg1-x-yTe samples investigated.

Zn0.05Cd0.23Hg0.72Te

Fig. 4. Reflectivity Spectra of Zn0,05Cd 0,23 Hg0,72Te (sample VI).

Fig. 2. Reflectivity spectra of n- Cd0.2 Hg0.8Te in the temperature range 30 K – 300 K.

Fig. 3. Reflectivity spectra of p- Cd0.2 Hg0.8Te in the temperature range 30 K – 300 K.

In case of p-type Cd0.2 Hg0.8Te alloy the main *TO* – phonon mode frequency of *HgTe -*like sub-band increases from 118 cm-1 at 30K to approximately 122 cm-1 at 300 K.

#### **2.3 Experimental results for quaternary alloys**

The introduction of low amounts of Zn stabilizes the weak Hg–Te bonds, in crystal lattice of the MCT solid solution, while Cd destabilizes them (Sher et al.,1985). The introduction of a third metal cation (Zn for example), by substitution of matrix cations (Hg or Cd) in solid

T = 30 K T = 100 K T = 170 K T = 230 K T = 300 K

 T = 30 K T = 100 K T = 170 K T = 230 K T = 300 K

**50 75 100 125 150 175 200**

**Wavenumbers, cm-1**

**50 75 100 125 150 175 200**

**Wavenumbers, cm-1**

Fig. 3. Reflectivity spectra of p- Cd0.2 Hg0.8Te in the temperature range 30 K – 300 K.

sub-band increases from 118 cm-1 at 30K to approximately 122 cm-1 at 300 K.

In case of p-type Cd0.2 Hg0.8Te alloy the main *TO* – phonon mode frequency of *HgTe -*like

The introduction of low amounts of Zn stabilizes the weak Hg–Te bonds, in crystal lattice of the MCT solid solution, while Cd destabilizes them (Sher et al.,1985). The introduction of a third metal cation (Zn for example), by substitution of matrix cations (Hg or Cd) in solid

Fig. 2. Reflectivity spectra of n- Cd0.2 Hg0.8Te in the temperature range 30 K – 300 K.

**Te** 

**0,0**

**0,0**

**2.3 Experimental results for quaternary alloys** 

**0,2**

**0,4**

**0,6**

**Reflectivity**

**0,8**

**Hg1-xCdx**

**x=0.2, p-type**

**1,0**

**0,2**

**0,4**

**0,6**

**Reflectivity**

**0,8**

**1,0**

**Hg1-xCdx**

**x=0.2**, **n-type,** 

**Te,** 

solution lattices with a common anion (ZnxCdyHg1-x-yTe or ZMCT in our case), enables us to control the material parameters with one extra degree of freedom (Cebulski et al., 1998). The bulk quaternary layers of ZnxCdyHg1-x-yTe were obtained by liquid phase epitaxial technique on the CdTe substrates in A.F. Joffe Physical-Technical Institute (St.Petersbourg, Russion), the compositions of samples are shown in Table 1. The thickness of the homogeneous layer was 4 m. The surface of samples was natural (110) plane of grown layer and was ideal for optical measurements.


Table 1. Compositions of the ZnxCdyHg1-x-yTe samples investigated.

Optical reflectivity from surface of nine ZnxCdyHg1-x-yTe samples of seven compositions in the far-infrared region was measured using the synchrotron radiation as source (high resolution FIR-spectra – the reflectivity experiment is described above)). Some of results were published earlier (Sheregii et. al., 2006). The measurements of reflectivity were performed in temperature region from 30 K to 300 K. In Fig. 4 and 5 are presented reflectivity FIR-spectra obtained for two compositions of ZnxCdyHg1-x-yTe for three temperatures: 30 K, 100K and 300 K.

Fig. 4. Reflectivity Spectra of Zn0,05Cd 0,23 Hg0,72Te (sample VI).

High Resolution Far Infrared Spectra of the

**0**

**0**

**20**

**40**

**60**

**Im**  **80**

**100**

**Hg1-xCdx**

**T=30K**

**x=0.2, p-type**

**Te** 

**120**

**5**

**10**

**15**

**Im** 

**20**

**25**

**Hg1-xCdx**

**T=300K**

**x=0.2, p-type**

**Te** 

**30**

Semiconductor Alloys Obtained Using the Synchrotron Radiation as Source 473

 after KK Sum L1 L2 L3 L4 L5 L6 L7 L8 L9 L10

 after KK sum L1 L2 L3 L4 L5 L6 L7 L8 L9

**80 100 120 140 160 180**

**Wavenumber,cm-1**

**80 100 120 140 160 180**

**Wavenumber**, **cm-1**

Fig. 7. Imaginary part of the dielectric function of *Hg0,8Cd0,2Te p-* type in the temperature 30K.

300 *K.* The position of main *HgTe*-line for *p*-type is 118 cm-1 at 30 *K* and is the same as for *n*type sample while the oscillator strengths of these lines for n- and p-type samples are drastically different: 62500 cm-2 for *n*-type and 39000 cm-2 for *p*-type, respectively. The damping factor is nearly two times larger for *p*-type *Hg0,8 Cd0,2 Te* because the line shape is

It is interesting to consider in details the temperature behavior of the observed phonon modes for both *n*- and *p*-type *Hg0.8Cd0.2Te.* In Figures 8 and 9 are shown the temperature dependences of frequencies for observed phonon lines of the *HgTe*-like and *CdTe*-like subbands (CPMs) of *n*- and *p*-type samples. It is seen that only one *HgTe*-like mode is observed

much asymmetric and wider in comparison with the *n*-type *Hg0,8 Cd0,2 Te.* 

at 30K and two *CdTe*-like mode for *n- Hg0.8Cd0.2Te* (see Fig.7).

Fig. 6. Imaginary part of the dielectric function od *Hg0,8Cd0,2Te p-* type in the temperature 300K.

Fig. 5. Reflectivity Spectra of Zn0,12Cd 0,13 Hg0,75Te (sample VII).

These curves are similar to typical reflection spectra but these curves have a much richer structure of spectra, as were observed for ternary alloys. It is seen three main bands at 130 cm-1, 160 cm-1 and 180 cm-1 can be point out in the reflective spectra. However, each of these sub-bands has additional subtle structures, which point to the superposition of a greater number of lines. With increasing of temperature from 30 K to 300 K the subtle structure of observed sub-bands became more smooth.

#### **3. Discussion**

#### **3.1 Spectral analyses of the ternary alloy FIR-spectra**

The lines corresponding to phonon modes are clearly observed on the *Im* curves calculated by Kramers-Kroning analyses from the experimental FIR reflectivity curves *R().*  In Fig.6 and 7 are shown *Im*curves for *p*-Hg0.8Cd0.2Te obtained for temperature 300 K and 30 K, respectively. In Fig. 6 we can see considerable asymmetry of HgTe-bands caused by additional lines in the range of 90 cm-1 –115 cm-1. That are the additional lines origin of which is discussed during last two decades. The dispersion analysis of the CPMs and APMs was performed by approximating the *Im*curves by the Lorentzian sum

$$\operatorname{Im} \mathcal{E}(oo) = \sum\_{i=1}^{k} \frac{S\_i \mathcal{V}\_i oo}{(o\_{\text{TOi}}^2 - o^2)^2 + o^2 \mathcal{V}\_i^2} \tag{1}$$

where *Si, TOi* and *i* are the oscillator strength, frequency and damping parameters of the *i*phonon mode, respectively. The results of spectral analysis for *p*-*Hg0.8Cd0.2Te* are presented in Fig.6 and for *n*-*Hg0.8Cd0.2Te in Fig.* 7. Parameters of Lorentzian's oscillators used for fitting the *Im*-curves are shown in Table 2.

In Table 3 the oscillator strengths sum for APM are shown separately as (SHgTe)add. There are nine well-resolved oscillators for *p*-type *Hg0.8Cd0.2Te* at 30 *K* and eleven for this sample at

 40K 100K 300K

curves for *p*-Hg0.8Cd0.2Te obtained for temperature 300 K

curves by the Lorentzian sum

(1)

curves

*).* 

100 120 140 160 180 200

Wavenumber, cm-1

These curves are similar to typical reflection spectra but these curves have a much richer structure of spectra, as were observed for ternary alloys. It is seen three main bands at 130 cm-1, 160 cm-1 and 180 cm-1 can be point out in the reflective spectra. However, each of these sub-bands has additional subtle structures, which point to the superposition of a greater number of lines. With increasing of temperature from 30 K to 300 K the subtle structure of

0,0

observed sub-bands became more smooth.

was performed by approximating the *Im*

**3.1 Spectral analyses of the ternary alloy FIR-spectra** 

 


The lines corresponding to phonon modes are clearly observed on the *Im*

calculated by Kramers-Kroning analyses from the experimental FIR reflectivity curves *R(*

phonon mode, respectively. The results of spectral analysis for *p*-*Hg0.8Cd0.2Te* are presented in Fig.6 and for *n*-*Hg0.8Cd0.2Te in Fig.* 7. Parameters of Lorentzian's oscillators used for fitting

In Table 3 the oscillator strengths sum for APM are shown separately as (SHgTe)add. There are nine well-resolved oscillators for *p*-type *Hg0.8Cd0.2Te* at 30 *K* and eleven for this sample at

1 Im ( ) ( ) *k*

 

2 22 2 2

 

*i* are the oscillator strength, frequency and damping parameters of the *i*-

*i i i TOi i S* 

and 30 K, respectively. In Fig. 6 we can see considerable asymmetry of HgTe-bands caused by additional lines in the range of 90 cm-1 –115 cm-1. That are the additional lines origin of which is discussed during last two decades. The dispersion analysis of the CPMs and APMs

**3. Discussion** 

where *Si,*

the *Im*

*TOi* and 

In Fig.6 and 7 are shown *Im*

0,2

0,4

0,6

Reflectivity

0,8

1,0

Zn0.12Cd0.13Hg0.75Te

Fig. 5. Reflectivity Spectra of Zn0,12Cd 0,13 Hg0,75Te (sample VII).

Fig. 6. Imaginary part of the dielectric function od *Hg0,8Cd0,2Te p-* type in the temperature 300K.

Fig. 7. Imaginary part of the dielectric function of *Hg0,8Cd0,2Te p-* type in the temperature 30K.

300 *K.* The position of main *HgTe*-line for *p*-type is 118 cm-1 at 30 *K* and is the same as for *n*type sample while the oscillator strengths of these lines for n- and p-type samples are drastically different: 62500 cm-2 for *n*-type and 39000 cm-2 for *p*-type, respectively. The damping factor is nearly two times larger for *p*-type *Hg0,8 Cd0,2 Te* because the line shape is much asymmetric and wider in comparison with the *n*-type *Hg0,8 Cd0,2 Te.* 

It is interesting to consider in details the temperature behavior of the observed phonon modes for both *n*- and *p*-type *Hg0.8Cd0.2Te.* In Figures 8 and 9 are shown the temperature dependences of frequencies for observed phonon lines of the *HgTe*-like and *CdTe*-like subbands (CPMs) of *n*- and *p*-type samples. It is seen that only one *HgTe*-like mode is observed at 30K and two *CdTe*-like mode for *n- Hg0.8Cd0.2Te* (see Fig.7).

High Resolution Far Infrared Spectra of the

**80**

**90**

**100**

**110**

**Wavenumber, cm-1**

corresponding APM modes.

**90**

phonon spectra of the *n-* and *p*- *Hg0.8Cd0.2Te*.

**100**

**Hg1-xCdx**

**x=0.2, n-type**

**Te** 

**110**

**120**

**Wavenumber,cm-1**

*Hg0.8Cd0.2Te*.

**130**

**140**

**150**

**T1V**

**T0**

**120**

**130**

**140**

**150**

Semiconductor Alloys Obtained Using the Synchrotron Radiation as Source 475

**T2 T3**

**T2**

**0 50 100 150 200 250 300 350**

**Temperature, K**

**0 50 100 150 200 250 300 350**

Temperature,K

samples. Moreover, in previous works was not shown such drastic difference between the

The composition frequency dependencies for all observed phonon modes in p-type MCTsystem at the temperature of 300K is presented on Figure 10. It is seen that these dependencies are similar to that one obtained in (Kozyrev et al., 1998) but the APMs observed in the spectral region 104 – 116 cm-1 are presented here also (are absented in work

Fig. 9. The temperature dependencies of the phonon mode frequencies for the *n-* type

Fig. 8. The temperature dependencies of the phonon mode frequencies for the *p-* type The *Hg0.8Cd0.2Te*, shown in Fig. 2 and 3 as well as in Table II and III. T0, T1 and T2 are tetrahedra

generated the corresponding CPM modes. The Tnv are tetrahedra generated by the

**el.tr**

**Hg1-xCdx**

**x=0.2, p-type**

**Te** 

**T0V**

**T0V T1V**

**T0 T1 T2 T1V**

**T2V**

**T1**

**T2**

**T2V**

**T1**

**T3 T1V**

**T1**


Table 2. Parameters of Lorentzian's oscillators used for fitting the *Im*curves of the *p*- *Hg0.8Cd0.2Te* for the temperatures 30 K and 300K.


Table 3. The oscillator's sums of the CPM and AVM for *p*- *Hg0.8Cd0.2Te* at 30 K and 300K.

When the temperature is higher than 100K the splitting on two *HgTe*-like modes takes place and at last at 300 *K* the three *HgTe*-like CPMs are displayed in case of *n*-type sample. Whereas in the region 90 – 115 cm-1 , one weak line is observed at 108 cm-1 which amplitude increases with increasing of temperature and after 230 *K* this line is splitted on three ones in the range 106 – 118 cm-1.

We can see a considerably larger number of lines for p-type sample in comparison with *n*type sample but the temperature shift of the phonon mode frequencies is similar. These results obtained for the *n-* and *p*-*Hg0.8Cd0.2Te* at 30 *K* agree generally with data presented in (Rath et al., 1995) but in this work was not performed a comparison for *n*- and *p*-type

5 500

1 4200

8 1200 9 1300

*<sup>j</sup>*

Table 3. The oscillator's sums of the CPM and AVM for *p*- *Hg0.8Cd0.2Te* at 30 K and 300K.

When the temperature is higher than 100K the splitting on two *HgTe*-like modes takes place and at last at 300 *K* the three *HgTe*-like CPMs are displayed in case of *n*-type sample. Whereas in the region 90 – 115 cm-1 , one weak line is observed at 108 cm-1 which amplitude increases with increasing of temperature and after 230 *K* this line is splitted on three ones in

We can see a considerably larger number of lines for p-type sample in comparison with *n*type sample but the temperature shift of the phonon mode frequencies is similar. These results obtained for the *n-* and *p*-*Hg0.8Cd0.2Te* at 30 *K* agree generally with data presented in (Rath et al., 1995) but in this work was not performed a comparison for *n*- and *p*-type

∑SHgTe ∑SCdTe CPM AVM CPM AVM

*HgTe <sup>j</sup> add s* 

curves of the

S

4 48000

2 1500 3 4700

3 6100 4 21100

6 1000

Table 2. Parameters of Lorentzian's oscillators used for fitting the *Im*

*p*- *Hg0.8Cd0.2Te* for the temperatures 30 K and 300K.

*s*

30 2.8 0.5 300 1.54 1.19

[K] 2

10 3500 11 4890

> *i i HgTe HgTe*

*S*

 7 5800 8 8700

1 600

6 1300

2 4200

5 830

Temperatur e [K]

30

300

Temperature

the range 106 – 118 cm-1.

Numbe r of Line

Fig. 8. The temperature dependencies of the phonon mode frequencies for the *p-* type The *Hg0.8Cd0.2Te*, shown in Fig. 2 and 3 as well as in Table II and III. T0, T1 and T2 are tetrahedra generated the corresponding CPM modes. The Tnv are tetrahedra generated by the corresponding APM modes.

Fig. 9. The temperature dependencies of the phonon mode frequencies for the *n-* type *Hg0.8Cd0.2Te*.

samples. Moreover, in previous works was not shown such drastic difference between the phonon spectra of the *n-* and *p*- *Hg0.8Cd0.2Te*.

The composition frequency dependencies for all observed phonon modes in p-type MCTsystem at the temperature of 300K is presented on Figure 10. It is seen that these dependencies are similar to that one obtained in (Kozyrev et al., 1998) but the APMs observed in the spectral region 104 – 116 cm-1 are presented here also (are absented in work

High Resolution Far Infrared Spectra of the

in centre.

where 4 4! *n n n* !4 !

1

, 4 4

respectively in the alloys *AY1-yZy* :

It is easy to determine that

1 3 

4

, <sup>4</sup>

2 

6

find all configurations in lattice of alloy with composition *x* must be equal to 1:

*n*

*n*

*n*

*n*

4

*n*

The probabilities to find a some of cation *A* or *B in tetrahedron Tn in lattice* respectively are:

*<sup>n</sup> P x x x n* 

> ( ) (1 ) <sup>4</sup> *B n n*

*<sup>n</sup> Px x x n* 

The same equations there are for the probabilities to find the certain of anions *Z* or *Y*

( ) (1 ) <sup>4</sup> *Z n n*

*<sup>n</sup> Py y y <sup>n</sup>* 

*<sup>n</sup> P y y y <sup>n</sup>* 

4 4

0 4 

lattice.

Semiconductor Alloys Obtained Using the Synchrotron Radiation as Source 477

where anions *Z* are substituted by anions *Y* – the tetrahedra will be looked at similarly because in zinc-blend lattice we can represent a basic unit as tetrahedron in two versions: centred by anion and surrounded by four cations or oppositely: four anions surround cation

The probability to find the *Tn* configuration in ideal lattice of the *AxB1-xZ* or *AY1-yZy* ternary

<sup>4</sup> 4 ( ) (1 ) *n n Px x x <sup>n</sup> <sup>n</sup>* 

is the number of combinations with *n* elements in the fourth set:

solution what is equal to the ratio of the *B-Z* ion pairs per whole number of ion pairs in

It is obvious that probability ( ) *P x <sup>n</sup>* must be function of composition *x* because increasing of *x* means increasing of the *B*-atoms number in lattice what leads to increasing of the tetrahedron's number with high value of *n* (not higher then 4). The sum of probabilities to

() 1 *<sup>n</sup>*

<sup>4</sup> 4 4 ( ) (1 ) <sup>4</sup> *A n n*

<sup>4</sup> 4

<sup>4</sup> 4

<sup>4</sup> 4 4 ( ) (1 ) <sup>4</sup> *Y n n*

(2)

, *x* is a mol composition of *BZ* compound in the solid

*P x* (3)

(4)

(5)

(4')

(5')

solid solution can be calculated using the Bernoulli polynomial (Ziman, 1979):

Fig. 10. The composition dependencies of the phonon mode frequencies for *p-Hg1- xCdx Te,* at temperature 300K.

of these authors). The amplitudes of these lines decrease with increasing of the CdTecontain. That are the same lines which temperature behavior were described above for the *n-* and *p-*type *Hg0.8Cd0.2Te*. It is undoubtedly that these lines are related to the *HgTe*-pairs oscillations. It is possible to state that the APMs reproduce the CPM of HgTe-like band but are shifted to lower frequencies.

#### **3.2 The random V-B model for ternary alloys**

To understand and interpret the experimental data on the phonon spectra of the solid solutions it is necessary to describe mathematically the non-regular distribution of atoms in its lattices. It occurs that such description is possible in case of the strongly chaotically (stochastically) homogenous distribution what require a very great number of atoms and a very carefully mixed alloys. These conditions are fulfilled generally in case of the high quality homogenous semiconductor solid solutions of the III-V and II-VI semiconductor compounds. In this case we can use the Bernoulli equation (Ziman, 1979) describing a probability to occur a one from *n* equivalent events what can be apply to the probability to find one from *n* configurations in the solid solution lattice.

The crystalline structure of the most III-V and II-VI compounds (possessed zinc-blend or wurzit structure as was mentioned above) is characterized by basic cell – tetrahedron – each with a central ion surrounded in the first coordination shell by four nearest neighbours (NN) at the vertices. In a *AxB1-xZ* ternary solid solution with substitution of the cation B by cation A, different tetrahedron configurations *Tn* (*n* is the number of *B*-atoms in the tetrahedron) coexist simultaneously: 2 strictly-binary ones corresponding to the *AZ* and the *BZ* compounds, whose lattices are characterized by the tedrahedron units *T0* and *T4* (configurations), respectively and 3 strictly-ternary ones actually characterized by the configurations *T1, T2* and *T3*. The similar configurations exist in a *AYxZ1-x* solid solution

**0,0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8**

**HgTe x CdTe** 

Fig. 10. The composition dependencies of the phonon mode frequencies for *p-Hg1- xCdx Te,* at

of these authors). The amplitudes of these lines decrease with increasing of the CdTecontain. That are the same lines which temperature behavior were described above for the *n-* and *p-*type *Hg0.8Cd0.2Te*. It is undoubtedly that these lines are related to the *HgTe*-pairs oscillations. It is possible to state that the APMs reproduce the CPM of HgTe-like band but

To understand and interpret the experimental data on the phonon spectra of the solid solutions it is necessary to describe mathematically the non-regular distribution of atoms in its lattices. It occurs that such description is possible in case of the strongly chaotically (stochastically) homogenous distribution what require a very great number of atoms and a very carefully mixed alloys. These conditions are fulfilled generally in case of the high quality homogenous semiconductor solid solutions of the III-V and II-VI semiconductor compounds. In this case we can use the Bernoulli equation (Ziman, 1979) describing a probability to occur a one from *n* equivalent events what can be apply to the probability to

The crystalline structure of the most III-V and II-VI compounds (possessed zinc-blend or wurzit structure as was mentioned above) is characterized by basic cell – tetrahedron – each with a central ion surrounded in the first coordination shell by four nearest neighbours (NN) at the vertices. In a *AxB1-xZ* ternary solid solution with substitution of the cation B by cation A, different tetrahedron configurations *Tn* (*n* is the number of *B*-atoms in the tetrahedron) coexist simultaneously: 2 strictly-binary ones corresponding to the *AZ* and the *BZ* compounds, whose lattices are characterized by the tedrahedron units *T0* and *T4* (configurations), respectively and 3 strictly-ternary ones actually characterized by the configurations *T1, T2* and *T3*. The similar configurations exist in a *AYxZ1-x* solid solution

**T0**

**T0V T1V**

**T1**

**<sup>T</sup> <sup>T</sup> <sup>2</sup> <sup>3</sup>**

**T2V**

**T1V T1 T2 T3 T4**

**100**

are shifted to lower frequencies.

**3.2 The random V-B model for ternary alloys** 

find one from *n* configurations in the solid solution lattice.

**110**

**120**

**130**

**Wavenumber, cm-1**

temperature 300K.

**140**

**150**

**160**

where anions *Z* are substituted by anions *Y* – the tetrahedra will be looked at similarly because in zinc-blend lattice we can represent a basic unit as tetrahedron in two versions: centred by anion and surrounded by four cations or oppositely: four anions surround cation in centre.

The probability to find the *Tn* configuration in ideal lattice of the *AxB1-xZ* or *AY1-yZy* ternary solid solution can be calculated using the Bernoulli polynomial (Ziman, 1979):

$$P\_n(\mathbf{x}) = \binom{4}{n} (1-\mathbf{x})^{4-n} \mathbf{x}^n \tag{2}$$

where 4 4! *n n n* !4 ! is the number of combinations with *n* elements in the fourth set:

4 4 1 0 4 , 4 4 4 1 3 , <sup>4</sup> 6 2 , *x* is a mol composition of *BZ* compound in the solid solution what is equal to the ratio of the *B-Z* ion pairs per whole number of ion pairs in lattice.

It is obvious that probability ( ) *P x <sup>n</sup>* must be function of composition *x* because increasing of *x* means increasing of the *B*-atoms number in lattice what leads to increasing of the tetrahedron's number with high value of *n* (not higher then 4). The sum of probabilities to find all configurations in lattice of alloy with composition *x* must be equal to 1:

$$\sum\_{n}^{4} P\_n(\mathbf{x}) = 1 \tag{3}$$

The probabilities to find a some of cation *A* or *B in tetrahedron Tn in lattice* respectively are:

$$P\_n^A(\mathbf{x}) = \frac{4-n}{4} \binom{4}{n} (1-\mathbf{x})^{4-n} \mathbf{x}^n \tag{4}$$

$$P\_n^B(\mathbf{x}) = \frac{n}{4} \binom{4}{n} (1-\mathbf{x})^{4-n} \mathbf{x}^n \tag{5}$$

The same equations there are for the probabilities to find the certain of anions *Z* or *Y* respectively in the alloys *AY1-yZy* :

$$P\_n^Z(y) = \frac{n}{4} \binom{4}{n} (1-y)^{4-n} y^n \tag{4'}$$

$$P\_n^Y(y) = \frac{4-n}{4} \binom{4}{n} (1-y)^{4-n} y^n \tag{5'}$$

It is easy to determine that

$$\sum\_{n}^{4} P\_{n}^{A}(\mathbf{x}) = \mathbf{1} - \mathbf{x} \tag{6}$$

High Resolution Far Infrared Spectra of the

molar percent of the component *AZ*).

well as for APM (*Tnv)*.

The dissipation of the *Im*

Semiconductor Alloys Obtained Using the Synchrotron Radiation as Source 479

<sup>0</sup> ( ) *A Z n AZ*

<sup>0</sup> ( ) (1 ) *A Y n AY*

Therefore, within this approximation the experimental Im -curves (obtained by

(11) what means the proportionality of the oscillator sum to the contain of the each component in alloy (*N0x* is equal to molar percent of the *BZ* component and *N0(1-x) –* to the

Some deviations of experimental data from the dependences (10) or (11) indicate evidently on considerable role of defects or others structural factors (non-random distribution).

The probability to find the atoms Hg and Cd in the particular tetrahedra *Tn* (*n* is number of the Cd-atoms in tetrahedral) in the *Hg0.8Cd0.2Te* lattice should be taken into account using formulas (4) and (5). If *x*=0.2 the values of ( ) *Hg P x <sup>n</sup>* for different *n* are equal to: 0.410 (*n*=4), 0.307(*n*=2), 0.077(*n*=3) and 0.006(*n*=4) while the ( ) *Cd P x <sup>n</sup>* values are: 0.102(*n*=1), 0.077(*n*=2), 0.192(*n*=3) and 0.002(*n*=4). At T=30 K all HgTe-like CPMs oscillate at the same frequency because tetrahedra with different number *n* are not deformed and we observed a degeneration of vibrational modes (Hg-Te and Cd-Te bonds have the same length). If T=300 K the splitting of the mode frequency takes place (see Fig.9): the most strong line at 122.6 cm-1 should be generate by Hg-Te dipoles in the *T0* tetrahedron while the line at 125.0 cm-1 – by this dipoles in the *T1* one and very small line at 128.6 – in the *T2*. So, the frequency consequence takes place for HgTe-like modes: HgTe0 < HgTe1 < HgTe2 < HgTe3 according with work (Kozyrev et al., 1998). Analogical analyses for CdTe-like modes shown that the line at 151.5 cm-1 is generated by Cd-Te dipoles in *T1* tetrahedron and the line at 147.3 cm-1 – by the same dipoles in *T2* one. The frequency consequence for CdTe-like modes is: CdTe1 > CdTe2 > CdTe3 > CdTe4 what agree with the data of work (Kozyrev et al., 1998) also. It allow to find to what basic cells (tetrahedra) belongs each observed vibrational mode generating by Hg-Te and Cd-Te dipoles: corresponding tetrahedra are shown in Fig. 9 for CPM (*Tn*) as

*S y f Ny* (10')

*<sup>S</sup> y f <sup>N</sup> <sup>y</sup>* (11')


or *B Z Sn*

*)-*curve) enable us to

and to verify the sums (10) or

In case of the *AY1-yZy* alloys the similar oscillator strengths sums must be fulfilled:

*n*

*n*

Kramers-Kroning transformation from experimentally measured *R(*

**3.3 Identification of observed lines in case of the n-Hg0.2Cd0.8Te alloys** 

**3.4 Identification of observed lines in case of the p-Hg0.2Cd0.8Te alloys** 

*Hg0.8Cd0.2Te* sample (see Fig. 6). The parameters of these oscillators are presented in Tables 2 and 3. There are eight well-resolved oscillators for *p*-type *Hg0.8Cd0.2Te* at 30 *K* and eleven for this sample at 300 *K.* The temperature dependencies of the phonon mode frequencies for *p*-*Hg0.8Cd0.2Te* are presented in Fig. 8. We can see a considerably larger number of lines here in comparison with *n*-type sample but the temperature shift of the phonon mode frequencies is similar. Analogically was fined for what basic cells (tetrahedra) belongs each observed

find the *Si* values, to identify that with certain *<sup>A</sup> <sup>Z</sup> Sn*

$$\sum\_{n}^{4} P\_{n}^{B}(\mathbf{x}) = \mathbf{x} \tag{7}$$

The same one takes place for the *AY1-yZy* alloys:

$$\sum\_{n}^{4} P\_n^{Z}(y) = 1 - y \tag{6'}$$

$$\sum\_{n}^{4} P\_{n}^{Y}(y) = y$$

It is necessary to note that (4) and (5) are simultaneously the probabilities to find in the solid solution lattice the ion pairs *A-Z* and *B-Z*, respectively (in case of the *AY1-yZy* alloys, the probabilities to find the ion pairs *A-Z* and *A-Y* according Eqns. (4') and (5'), respectively).

The oscillator strength of the vibrational mode generated by a *A-Z*-dipole in the *Tn* configuration is (Robouch et al., 2001):

$$S\_n^{A-Z}(\mathbf{x}) = f\_{AZ} N\_0 P\_n^A(\mathbf{x}) \tag{8}$$

where *fAZ* is the oscillator strength of the single dipole *A-Z*-pair, *N0* is total number of dipole pairs in the solid solution crystal, probability ( ) *<sup>A</sup> P x <sup>n</sup>* is determined by (4).

It is important to remember that three assumption are introduced in this consideration:


If these conditions are fulfilled, the oscillator sum rule

$$\sum\_{n} S\_{n}^{A-Z}(\mathbf{x}) = \sum\_{n} f\_{AZ} N\_{0} P\_{n}^{A}(\mathbf{x}) = f\_{AZ} N\_{0} \sum\_{n} P\_{n}^{A}(\mathbf{x}) = f\_{AZ} N\_{0} (1 - \mathbf{x}) \tag{9}$$

has to be satisfied.

Similarly for *B-Z* dipole pairs:

$$\mathbf{N}\_{n}^{B-Z}(\mathbf{x}) = f\_{BZ} \mathbf{N}\_{0} P\_{n}^{B}(\mathbf{x}) \tag{10}$$

and the oscillator sum rule

$$\sum\_{n} S\_{n}^{B-Z}(\mathbf{x}) = \sum\_{n} f\_{BZ} N\_{0} P\_{n}^{B}(\mathbf{x}) = f\_{AZ} N\_{0} \sum\_{n} P\_{n}^{B}(\mathbf{x}) = f\_{BZ} N\_{0} \mathbf{x} \tag{11}$$

() 1 *<sup>A</sup> n n*

> ( ) *<sup>B</sup> n n*

() 1 *<sup>Z</sup> n n*

> ( ) *<sup>Y</sup> n n*

It is necessary to note that (4) and (5) are simultaneously the probabilities to find in the solid solution lattice the ion pairs *A-Z* and *B-Z*, respectively (in case of the *AY1-yZy* alloys, the probabilities to find the ion pairs *A-Z* and *A-Y* according Eqns. (4') and (5'), respectively). The oscillator strength of the vibrational mode generated by a *A-Z*-dipole in the *Tn*

where *fAZ* is the oscillator strength of the single dipole *A-Z*-pair, *N0* is total number of dipole

2. the alloy lattice is ideally homogenous and a random distribution of atoms in lattice

3. the oscillator strengths of the single dipole pairs for different configurations Tn are the

00 0 ( ) ( ) ( ) (1 ) *A Z <sup>A</sup> <sup>A</sup> n AZ n AZ n AZ*

00 0 ( ) ( ) ( ) *B Z <sup>B</sup> <sup>B</sup> n BZ n AZ n BZ*

*S x <sup>f</sup> NP x <sup>f</sup> N Px <sup>f</sup> N x* (9)

*S x <sup>f</sup> NP x <sup>f</sup> N Px <sup>f</sup> N x* (11)

It is important to remember that three assumption are introduced in this consideration:

pairs in the solid solution crystal, probability ( ) *<sup>A</sup> P x <sup>n</sup>* is determined by (4).

*n n n*

*n n n*

*Px x* (6)

*Px x* (7)

*Py y* (6')

*Py y* (7')

<sup>0</sup> () () *A Z <sup>A</sup> S x <sup>n</sup> AZ n <sup>f</sup> NP x* (8)

<sup>0</sup> () () *B Z <sup>B</sup> S x <sup>n</sup> BZ n <sup>f</sup> NP x* (10)

4

4

4

4

The same one takes place for the *AY1-yZy* alloys:

configuration is (Robouch et al., 2001):

1. the role of defects is negligible;

has to be satisfied.

Similarly for *B-Z* dipole pairs:

and the oscillator sum rule

takes place (stochastic homogeneity);

same e.g. *fAZ* or *fBZ* depends not on index *n*. If these conditions are fulfilled, the oscillator sum rule In case of the *AY1-yZy* alloys the similar oscillator strengths sums must be fulfilled:

$$\sum\_{n} S\_{n}^{A-Z}(y) = f\_{AZ} N\_0 y \tag{10'}$$

$$\sum\_{n} S\_{n}^{A-Y}(y) = f\_{AY} N\_0 (1 - y) \tag{11'}$$

Therefore, within this approximation the experimental Im -curves (obtained by Kramers-Kroning transformation from experimentally measured *R()-*curve) enable us to find the *Si* values, to identify that with certain *<sup>A</sup> <sup>Z</sup> Sn* or *B Z Sn* and to verify the sums (10) or (11) what means the proportionality of the oscillator sum to the contain of the each component in alloy (*N0x* is equal to molar percent of the *BZ* component and *N0(1-x) –* to the molar percent of the component *AZ*).

Some deviations of experimental data from the dependences (10) or (11) indicate evidently on considerable role of defects or others structural factors (non-random distribution).

#### **3.3 Identification of observed lines in case of the n-Hg0.2Cd0.8Te alloys**

The probability to find the atoms Hg and Cd in the particular tetrahedra *Tn* (*n* is number of the Cd-atoms in tetrahedral) in the *Hg0.8Cd0.2Te* lattice should be taken into account using formulas (4) and (5). If *x*=0.2 the values of ( ) *Hg P x <sup>n</sup>* for different *n* are equal to: 0.410 (*n*=4), 0.307(*n*=2), 0.077(*n*=3) and 0.006(*n*=4) while the ( ) *Cd P x <sup>n</sup>* values are: 0.102(*n*=1), 0.077(*n*=2), 0.192(*n*=3) and 0.002(*n*=4). At T=30 K all HgTe-like CPMs oscillate at the same frequency because tetrahedra with different number *n* are not deformed and we observed a degeneration of vibrational modes (Hg-Te and Cd-Te bonds have the same length). If T=300 K the splitting of the mode frequency takes place (see Fig.9): the most strong line at 122.6 cm-1 should be generate by Hg-Te dipoles in the *T0* tetrahedron while the line at 125.0 cm-1 – by this dipoles in the *T1* one and very small line at 128.6 – in the *T2*. So, the frequency consequence takes place for HgTe-like modes: HgTe0 < HgTe1 < HgTe2 < HgTe3 according with work (Kozyrev et al., 1998). Analogical analyses for CdTe-like modes shown that the line at 151.5 cm-1 is generated by Cd-Te dipoles in *T1* tetrahedron and the line at 147.3 cm-1 – by the same dipoles in *T2* one. The frequency consequence for CdTe-like modes is: CdTe1 > CdTe2 > CdTe3 > CdTe4 what agree with the data of work (Kozyrev et al., 1998) also. It allow to find to what basic cells (tetrahedra) belongs each observed vibrational mode generating by Hg-Te and Cd-Te dipoles: corresponding tetrahedra are shown in Fig. 9 for CPM (*Tn*) as well as for APM (*Tnv)*.

#### **3.4 Identification of observed lines in case of the p-Hg0.2Cd0.8Te alloys**

The dissipation of the *Im* -curves on the Lorentzians was carried out for the *p*-type *Hg0.8Cd0.2Te* sample (see Fig. 6). The parameters of these oscillators are presented in Tables 2 and 3. There are eight well-resolved oscillators for *p*-type *Hg0.8Cd0.2Te* at 30 *K* and eleven for this sample at 300 *K.* The temperature dependencies of the phonon mode frequencies for *p*-*Hg0.8Cd0.2Te* are presented in Fig. 8. We can see a considerably larger number of lines here in comparison with *n*-type sample but the temperature shift of the phonon mode frequencies is similar. Analogically was fined for what basic cells (tetrahedra) belongs each observed

High Resolution Far Infrared Spectra of the

**0,1**

calculated according the Equation

s=0,04+12exp(-0,09/kT).

K also.

1967).

**0,2**

**Oscillator strengh,s**

**0,3**

**0,4**

Semiconductor Alloys Obtained Using the Synchrotron Radiation as Source 481

 B- experimental point C- theoretical curve

**0 50 100 150 200 250 300**

It is clear that the temperature dependencies of the oscillator strength sum of the lines observed in the region of the 104 – 116 cm-1, presented here, do not confirm the hypothesis that these lines are related to the *Hg*-vacancies. There are others doubtful circumvents, namely: in case of *n*-type *Hg0.8Cd0.2Te* the single very weak line at 107 cm-1 is observed at 30

If we assume that this line is caused by Hg- vacancies it is necessary to agree that the vacancy density must be not less than 1018 cm-3. While the data of the positron annihilation for *n*-type *HgCdTe* shown values of the *Hg*-vacancy concentration closer to 1015 cm-3 (Krause et al., 1990). It is necessary to note that method of the positron annihilation, seems to be direct method of the vacancy concentration measurement, in case of HgCdTe use the data of Hall-effect (in determination of the specific positron trapping rate) identifying the hole concentration to the concentration of Hg-vacancies. It is not completely correct because in *HgCdTe* there is always background of the electrically native compensated *Hg*-vacancies and real level of the Hg-vacancies is naturally higher than the hole concentration. Nevertheless, the Hg-vacancy density over 1018 cm-3 in the *n*- *Hg0.8Cd0.2Te Te* of high quality (very high

The temperature dependences of the SOSS for discussed lines lead to the activation energy of process to be equal to 75 – 90 meV what could be as substantial argument for the model of two potential wells (Hg-atoms in lattice of HgCdTe) applying by *J.A. Sussman* (Sussman,

J.A. Sussman (Sussman, 1967) proposed this model for the binary compounds. From this theory arise that a cation in the crystal lattice could have the two positions: first stable

electron mobility of 2.5x105 V/ms) is absolutely impossible.

**3.6 Two valley potential model and quasi quaternary alloys** 

Fig. 12. The temperature dependence of the sum of the additional modes oscillator strengths for the *n* –type *Hg0.8Cd0.2Te,* B- experimental points, C is approximated curve

**Temperature, K** 

vibrational mode generating by Hg-Te and Cd-Te dipoles: corresponding tetrahedra are shown in Fig. 8 for CPM (*Tn*) as well as for APM (*Tnv)*.

#### **3.5 Additional phonon modes**

There are important guiding principles that the lines in the region 104 *cm-1*- 116 *cm-1* are related to the *Hg*-vacancies (Cebulski et al., 2008). This hypothesis can be verified by temperature dependences of the specific oscillator strength sum (SOSS) of the lines observed in this region. These temperature dependences are presented in Fig. 11 for *n-Hg0.8Cd0.2Te Te* and Fig. 12 for *p- Hg0.8Cd0.2Te*. It is shown in Fig.11 that temperature dependencies of the SOSS of APM for *p*–type *Hg0.8Cd0.2Te,* have the exponential character described by function

$$\sum s\_{Hg\,Te}^{add} = 0.5 + 12 \exp(-0.075 \,/\, kT) \tag{12}$$

with activation energy equal to 75 meV. It is too small energy in comparison with the Hgvacancy activation energy to be equal to about 1eV (Chandra et al., 2003).

Fig. 11. The temperature dependence of the sum of the additional modes oscillator strengths for the *p*–type *Hg0.8Cd0.2Te,* B- experimental points, C is approximated curve calculated according the Equation s=0,5+12exp(-0,075/kT).

The Fig.12 presents the temperature dependence of the SOSS of the same lines for *n*–type *Hg0.8Cd0.2Te*. This dependence is described by exponential function similar to (12):

$$\sum s\_{HgTe}^{add} = 0.04 + 12 \exp(-0.09 \,/\, kT) \tag{13}$$

with activation energy equal to 90 meV, which is larger than for p-type but is too small to be an activation energy for Hg-vacancies.

vibrational mode generating by Hg-Te and Cd-Te dipoles: corresponding tetrahedra are

There are important guiding principles that the lines in the region 104 *cm-1*- 116 *cm-1* are related to the *Hg*-vacancies (Cebulski et al., 2008). This hypothesis can be verified by temperature dependences of the specific oscillator strength sum (SOSS) of the lines observed in this region. These temperature dependences are presented in Fig. 11 for *n-Hg0.8Cd0.2Te Te* and Fig. 12 for *p- Hg0.8Cd0.2Te*. It is shown in Fig.11 that temperature dependencies of the SOSS of APM for *p*–type *Hg0.8Cd0.2Te,* have the exponential character

0.5 12exp( 0.075 / ) *add*

with activation energy equal to 75 meV. It is too small energy in comparison with the Hg-

 B- experimental point C- theoretical curve

**0 50 100 150 200 250 300**

The Fig.12 presents the temperature dependence of the SOSS of the same lines for *n*–type

0.04 12exp( 0.09 / ) *add*

with activation energy equal to 90 meV, which is larger than for p-type but is too small to be

*HgTe <sup>s</sup> kT* (13)

Fig. 11. The temperature dependence of the sum of the additional modes oscillator strengths for the *p*–type *Hg0.8Cd0.2Te,* B- experimental points, C is approximated curve

*Hg0.8Cd0.2Te*. This dependence is described by exponential function similar to (12):

**Temperature, K** 

vacancy activation energy to be equal to about 1eV (Chandra et al., 2003).

*HgTe <sup>s</sup> kT* (12)

shown in Fig. 8 for CPM (*Tn*) as well as for APM (*Tnv)*.

**3.5 Additional phonon modes** 

**0,4**

calculated according the Equation

an activation energy for Hg-vacancies.

s=0,5+12exp(-0,075/kT).

**0,6**

**0,8**

**Oscillator strengh,s**

**1,0**

**1,2**

described by function

Fig. 12. The temperature dependence of the sum of the additional modes oscillator strengths for the *n* –type *Hg0.8Cd0.2Te,* B- experimental points, C is approximated curve calculated according the Equation s=0,04+12exp(-0,09/kT).

It is clear that the temperature dependencies of the oscillator strength sum of the lines observed in the region of the 104 – 116 cm-1, presented here, do not confirm the hypothesis that these lines are related to the *Hg*-vacancies. There are others doubtful circumvents, namely: in case of *n*-type *Hg0.8Cd0.2Te* the single very weak line at 107 cm-1 is observed at 30 K also.

If we assume that this line is caused by Hg- vacancies it is necessary to agree that the vacancy density must be not less than 1018 cm-3. While the data of the positron annihilation for *n*-type *HgCdTe* shown values of the *Hg*-vacancy concentration closer to 1015 cm-3 (Krause et al., 1990). It is necessary to note that method of the positron annihilation, seems to be direct method of the vacancy concentration measurement, in case of HgCdTe use the data of Hall-effect (in determination of the specific positron trapping rate) identifying the hole concentration to the concentration of Hg-vacancies. It is not completely correct because in *HgCdTe* there is always background of the electrically native compensated *Hg*-vacancies and real level of the Hg-vacancies is naturally higher than the hole concentration. Nevertheless, the Hg-vacancy density over 1018 cm-3 in the *n*- *Hg0.8Cd0.2Te Te* of high quality (very high electron mobility of 2.5x105 V/ms) is absolutely impossible.

The temperature dependences of the SOSS for discussed lines lead to the activation energy of process to be equal to 75 – 90 meV what could be as substantial argument for the model of two potential wells (Hg-atoms in lattice of HgCdTe) applying by *J.A. Sussman* (Sussman, 1967).

#### **3.6 Two valley potential model and quasi quaternary alloys**

J.A. Sussman (Sussman, 1967) proposed this model for the binary compounds. From this theory arise that a cation in the crystal lattice could have the two positions: first stable

High Resolution Far Infrared Spectra of the

the range 104 – 116 cm-1 is equal to

and 300 *K*.

disappeared.

HgCdTe solid solutions.

Material

Semiconductor Alloys Obtained Using the Synchrotron Radiation as Source 483

two versions: first one is realized in the lattice consisted only from the *HgI*-atoms ( that are *CPM*) and second one occurs in the lattice included the *HgII*-atoms too (that are *APM*). The theory of the quasi-quaternary alloys contained the two kind of *Hg* –atom position developed in (Cebulski et al., 2008) enable us to determine the *HgII* concentration on base of phonon spectra. The details of such consideration will be presented in next sub-chapter. As arise from this consideration, the sum of the specific oscillator strengths of *AVM* observed in

> ( 1), ( ) *II II Hg Te*

This simple expression enable us to determine experimentally (using the phonon spectra according to the the Eqn. (10)) the *y* – molar part of the *HgII*-atoms, assuming that: i) the lines corresponding to the *AVM* generating by tetrahedra bearing by *HgII* are identified correctly, ii) the specific oscillator strength of the HgII -Te oscillations in tetrahedra with *HgII*-atoms are the same as in tetrahedra without *HgII* atoms as was mentioned above. We assume that these conditions are fulfill in case of measured materials of *n-* and *p*-type *Hg0.8Cd0.2Te.* The calculated values of the molar fraction *y* of the *HgII*-Te obtained from phonon spectra (sum of the specific oscillator strength) are presented in Table IV for 30 *K*

Sum of the specific oscillator

*n*-type *Hg0.8Cd0.2Te* < 0.043 0.37 < 0.7 6.3

*p*-type *Hg0.8Cd0.2Te* 0.50 1.1987 9.7 20.67

Reassuming we can affirm that in case HgCdTe of the *p*-type it consists of two sublattices: one sublattice contains the atoms of mercury in the stable state with the shorter length bonds of HgTe, second sublattice contains atoms Hg in the metastable state with the longer bond of HgTe. The phonon spectra of HgTe-like modes are reproduced for each above mentioned sublattices. Increase of the temperature leads to the increase of the number of HgII atoms (metastable state) and the same to the enlargement of tensions in the lattice what leads to splitting HgTe-like mod CPM in the n-type the material. Because of that differences between phonon spectra of n and p type in the room temperature are

One can also affirm, that in the temperature 30K in the material *n-Cd0.2 Hg0.8Te* is observed one HgTe-like mode and two CdTe - like of the mode (CPM) would confirm the percolation model of authors (Pages et al., 2009). It indicates (non directly) that the bond percolation thresholds *xc* for the HgCdTe alloys is larger then 0.19, namely *xc*≥0.2. Nevertheless, generally the V-B model developed for random case is confirmed completely for the

Table 4. Molar fraction of the *HgII* -atoms determine from phonon spectra

*HgTe n Hg S yf y* (15)

strength for AVM *y*, mol. %

30 *K* 300 *K* 30 *K* 300 *K*

position energetically more deep, second one is metastable state with higher energy and suitably with more long bond. This model related to HgTe and HgCdTe means that the Hg atoms can be shifted from the vertex position in tetrahedra (stable position) to a non centered position (metastable position). According to Sussman's theory such transition from stable to metastable state, means that the Hg-Te bonds become longer. The probability for such transition is described by

$$\mathbf{W} \triangleq \mathbf{w} \exp(-\boldsymbol{\Delta E} / \boldsymbol{\kappa} \mathbf{T}) \tag{14}$$

where *E* is the energy difference between the two states – stable and metastable one and *w* is the assumed probability at the absolute zero.

The temperature dependences of *SOSS* for APM shown in Figures 11 and 12 and described by relations (12) and (13) enable us to determine the *E*. Therefore, in the case of *p*-type *Hg0.8Cd0.2Te* the energy transition from the stable position of the *Hg* atoms to the metastable position is 75 meV and 90 meV for the n-type one. This difference can be explained by the fact that for *p*-type material where the considerable path of the crystal lattice is non relaxed, the density of metastable states is large than in *n-*type what could change the deep of the energy minimum (value *E2*) for stable position. The ratio of the *SOSSs* of additional lines (104 -116 cm-1) for *p-* and *n*-type materials is about one order. Therefore, the density of the metastable states and stable ones should differ with the same value. Simultaneously, the length of the *Hg-Te* bonds is longer for the metastable states in comparison with stable one. This difference have been appeared in *X*-ray analyses (Polit et al., 2010): 6.4604 Å for *n*-*Hg0.8Cd0.2Te* and 6.4648 Å for *p*- *Hg0.8Cd0.2Te -* the density of metastable states is larger of one order in *p*-type material and that causes statistically more long bonds Hg-Te.

#### **3.7 General description of the HgCdTe phonon spectra**

The general description of the phonon spectra is based on three Figures: 4, 5 and 6. These Figures present the temperature dependences of the HgTe-like and CdTe-like mode frequencies for *n*- and *p*-type *Hg0.8Cd0.2Te* (Fig. 4 and 5) as well as the composition dependences of the same modes at the room temperature (Fig. 6). If temperature increases, the number of Hg-atoms occupied the meta-stable positions (*HgII* ) increases also and the deformation of crystal lattice rises, respectively. The last factor can cause the removing of degeneracy of the HgTe-like CPMs in *n-Hg0,8Cd0,2 Te* when the temperature increases over the 100 K (see Fig. 5): the AVMs appear simultaneously, too. Indeed, the AVM at 112 cm-1 (beside very weak from 30 K at 108 cm-1) take place after 100 K in n-type *Hg0.8Cd0.2Te* and after 200 K appear additionally one AVM at 115-116 cm-1. The presence of *HgII* in a tetrahedron leads to the stretching of bonds which in its turn causes the shift of the *Hg-Te*  oscillation frequency towards smaller frequencies. This effect can occur in three kinds of tetrahedra: 1) containing *3 Hg-*atoms in stable position (*HgI*) and one *HgII ;* 2) containing two *HgI,* one Cd-atom and one *HgI*I; 3) containing one *HgI*, two *Cd-*atoms and one *HgII.* The frequencies of *Hg-Te* oscillations in these tetrahedra should be arranged in the next sequence: the lowest frequency corresponds to the *Hg-Te* oscillations in the tetrahedron of first type and most higher corresponds to the oscillations in the tetrahedron of third type.

The lines in the range of 135 -137cm-1 are generated as could be assumed, by the oscillation of *Cd -Te* pair in the tetrahedra containing two *HgI ,* one *Cd-*atom and one *HgII.* Therefore, the Figures 9,10 and 11 enable us to assume that the phonon spectra in *MCT* are reproduced in

position energetically more deep, second one is metastable state with higher energy and suitably with more long bond. This model related to HgTe and HgCdTe means that the Hg atoms can be shifted from the vertex position in tetrahedra (stable position) to a non centered position (metastable position). According to Sussman's theory such transition from stable to metastable state, means that the Hg-Te bonds become longer. The probability for

*E* is the energy difference between the two states – stable and metastable one and *w*

The temperature dependences of *SOSS* for APM shown in Figures 11 and 12 and described

*Hg0.8Cd0.2Te* the energy transition from the stable position of the *Hg* atoms to the metastable position is 75 meV and 90 meV for the n-type one. This difference can be explained by the fact that for *p*-type material where the considerable path of the crystal lattice is non relaxed, the density of metastable states is large than in *n-*type what could change the deep of the energy minimum (value *E2*) for stable position. The ratio of the *SOSSs* of additional lines (104 -116 cm-1) for *p-* and *n*-type materials is about one order. Therefore, the density of the metastable states and stable ones should differ with the same value. Simultaneously, the length of the *Hg-Te* bonds is longer for the metastable states in comparison with stable one. This difference have been appeared in *X*-ray analyses (Polit et al., 2010): 6.4604 Å for *n*-*Hg0.8Cd0.2Te* and 6.4648 Å for *p*- *Hg0.8Cd0.2Te -* the density of metastable states is larger of one

The general description of the phonon spectra is based on three Figures: 4, 5 and 6. These Figures present the temperature dependences of the HgTe-like and CdTe-like mode frequencies for *n*- and *p*-type *Hg0.8Cd0.2Te* (Fig. 4 and 5) as well as the composition dependences of the same modes at the room temperature (Fig. 6). If temperature increases, the number of Hg-atoms occupied the meta-stable positions (*HgII* ) increases also and the deformation of crystal lattice rises, respectively. The last factor can cause the removing of degeneracy of the HgTe-like CPMs in *n-Hg0,8Cd0,2 Te* when the temperature increases over the 100 K (see Fig. 5): the AVMs appear simultaneously, too. Indeed, the AVM at 112 cm-1 (beside very weak from 30 K at 108 cm-1) take place after 100 K in n-type *Hg0.8Cd0.2Te* and after 200 K appear additionally one AVM at 115-116 cm-1. The presence of *HgII* in a tetrahedron leads to the stretching of bonds which in its turn causes the shift of the *Hg-Te*  oscillation frequency towards smaller frequencies. This effect can occur in three kinds of tetrahedra: 1) containing *3 Hg-*atoms in stable position (*HgI*) and one *HgII ;* 2) containing two *HgI,* one Cd-atom and one *HgI*I; 3) containing one *HgI*, two *Cd-*atoms and one *HgII.* The frequencies of *Hg-Te* oscillations in these tetrahedra should be arranged in the next sequence: the lowest frequency corresponds to the *Hg-Te* oscillations in the tetrahedron of first type and most higher corresponds to the oscillations in the tetrahedron of third type. The lines in the range of 135 -137cm-1 are generated as could be assumed, by the oscillation of *Cd -Te* pair in the tetrahedra containing two *HgI,* one *Cd-*atom and one *HgII.* Therefore, the Figures 9,10 and 11 enable us to assume that the phonon spectra in *MCT* are reproduced in

order in *p*-type material and that causes statistically more long bonds Hg-Te.

W = w exp (14)

*E*. Therefore, in the case of *p*-type

such transition is described by

is the assumed probability at the absolute zero.

by relations (12) and (13) enable us to determine the

**3.7 General description of the HgCdTe phonon spectra** 

where  two versions: first one is realized in the lattice consisted only from the *HgI*-atoms ( that are *CPM*) and second one occurs in the lattice included the *HgII*-atoms too (that are *APM*). The theory of the quasi-quaternary alloys contained the two kind of *Hg* –atom position developed in (Cebulski et al., 2008) enable us to determine the *HgII* concentration on base of phonon spectra. The details of such consideration will be presented in next sub-chapter. As arise from this consideration, the sum of the specific oscillator strengths of *AVM* observed in the range 104 – 116 cm-1 is equal to

$$S\_{\text{(n-1)},Hg^{\text{(l)}}}^{Hg^{\text{(l)}}Te}(y) = f\_{Hg^{\text{Te}}e}y \tag{15}$$

This simple expression enable us to determine experimentally (using the phonon spectra according to the the Eqn. (10)) the *y* – molar part of the *HgII*-atoms, assuming that: i) the lines corresponding to the *AVM* generating by tetrahedra bearing by *HgII* are identified correctly, ii) the specific oscillator strength of the HgII -Te oscillations in tetrahedra with *HgII*-atoms are the same as in tetrahedra without *HgII* atoms as was mentioned above. We assume that these conditions are fulfill in case of measured materials of *n-* and *p*-type *Hg0.8Cd0.2Te.* The calculated values of the molar fraction *y* of the *HgII*-Te obtained from phonon spectra (sum of the specific oscillator strength) are presented in Table IV for 30 *K* and 300 *K*.


Table 4. Molar fraction of the *HgII* -atoms determine from phonon spectra

Reassuming we can affirm that in case HgCdTe of the *p*-type it consists of two sublattices: one sublattice contains the atoms of mercury in the stable state with the shorter length bonds of HgTe, second sublattice contains atoms Hg in the metastable state with the longer bond of HgTe. The phonon spectra of HgTe-like modes are reproduced for each above mentioned sublattices. Increase of the temperature leads to the increase of the number of HgII atoms (metastable state) and the same to the enlargement of tensions in the lattice what leads to splitting HgTe-like mod CPM in the n-type the material. Because of that differences between phonon spectra of n and p type in the room temperature are disappeared.

One can also affirm, that in the temperature 30K in the material *n-Cd0.2 Hg0.8Te* is observed one HgTe-like mode and two CdTe - like of the mode (CPM) would confirm the percolation model of authors (Pages et al., 2009). It indicates (non directly) that the bond percolation thresholds *xc* for the HgCdTe alloys is larger then 0.19, namely *xc*≥0.2. Nevertheless, generally the V-B model developed for random case is confirmed completely for the HgCdTe solid solutions.

High Resolution Far Infrared Spectra of the

Zn0.05Cd0.23Hg0.72Te

0

0

phonon modes – from 108 to 190 cm-1.

10

20

30

40

Im

50

60

70

20

40

60

Im

80

100

Semiconductor Alloys Obtained Using the Synchrotron Radiation as Source 485

 Im (KK) Im (Sum) L1 L2 L3 L4 L5 L6 L7 L8 L9 L10

> Im (KK) Im (Sum) L1 L2 L3 L4 L5 L6 L7 L8 L9 L10 L11 L12 L13 L14 L15 L16 L17 L18

80 100 120 140 160 180 200

Wavenumber, cm-1

80 100 120 140 160 180 200

Wavenumber, cm-1

The present analysis shows that the investigated quaternary ZMCT has lager number of the resonance frequencies in the observed spectra compared to the ternary solid solutions of the binary HgTe, CdTe and ZnTe (118 to 180 cm-1) and extends the frequency region of the

Fig. 14. Spectral analysis for Sample Zn0,05Cd 0,23 Hg0,72Te (sample VI).

Fig. 15. Spectral analysis for sample Zn0,12Cd 0,13 Hg0,75Te (sample VII).

Zn0.12Cd0.13Hg0.75Te

In Fig. 13 are shown the values of the oscillator strengths sums (OSS) for the Hg-Te dipoles and for Cd-Te dipoles for each samples investigated. The data are presented in two way: i) only OSS for CPMs are included (open circles and squares), for x=0.2 there are two open circles because first one (upper open circle) is regarded to *n-Cd0.2 Hg0.8Te* and lower open circle – to *p-Cd0.2 Hg0.8Te* ; ii) in the OSS are included the APM OSS also (filled circles and squares). That enable us to obtain the dependencies of the OSS on composition. As follow from Fig. 13 if the oscillator strengths of APM are included in the sum of the oscillator strengths for the modes generated by Hg-Te dipoles as well as Cd-Te dipoles the OSS are proportional to the contain of correspond compound: to the *x* in case of Cd-Te dipoles and to the *1-x* in case of Hg-Te ones. As was mentioned above (see Eqns. (9) and (11)) these dependences are considered as a criterion of applying the random V-B model to the phonon spectra interpretation of the ternary solid solutions. Therefore, the random version of the V-B model satisfactorily explains the high resolution FIR-spectra of ternary HgCdTe solid solutions if APM are included into consideration.

#### **3.8 Spectral analyses of the FIR-spectra for quaternary alloys**

The Kramers – Kronig analysis was applied to determine the position of observed lines. In Fig. 14 and 15 are shown the curves of imaginary part of dielectric function Im *x,y* for compositions VI and VII (see Table 1) respectively, obtained by Kramers-Kroning transformation from the reflectivity curves presented in Fig. 4 and 5 at 30 K. The imaginary part of dielectric function for ZnxCdyHg(1-x-y)Te solid solution can be presented as the superposition of Lorentzians as it is follow from Eqn. (1). The fittings by the Lorentzian sums are presented in Fig.14 and 15 too. The parameters of Lorentzians are presented in Tab. 5 and 6 respectively.

In Fig. 13 are shown the values of the oscillator strengths sums (OSS) for the Hg-Te dipoles and for Cd-Te dipoles for each samples investigated. The data are presented in two way: i) only OSS for CPMs are included (open circles and squares), for x=0.2 there are two open circles because first one (upper open circle) is regarded to *n-Cd0.2 Hg0.8Te* and lower open circle – to *p-Cd0.2 Hg0.8Te* ; ii) in the OSS are included the APM OSS also (filled circles and squares). That enable us to obtain the dependencies of the OSS on composition. As follow from Fig. 13 if the oscillator strengths of APM are included in the sum of the oscillator strengths for the modes generated by Hg-Te dipoles as well as Cd-Te dipoles the OSS are proportional to the contain of correspond compound: to the *x* in case of Cd-Te dipoles and to the *1-x* in case of Hg-Te ones. As was mentioned above (see Eqns. (9) and (11)) these dependences are considered as a criterion of applying the random V-B model to the phonon spectra interpretation of the ternary solid solutions. Therefore, the random version of the V-B model satisfactorily explains the high resolution FIR-spectra of ternary HgCdTe solid

Fig. 13. The oscillator strength sum dependence on composition for the HgCdTe alloys

Fig. 14 and 15 are shown the curves of imaginary part of dielectric function Im

The Kramers – Kronig analysis was applied to determine the position of observed lines. In

0,0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1,0

CPM - Hg-Te - Hg-Te CPM - Cd-Te - Cd-Te)

Composition x, mol.

compositions VI and VII (see Table 1) respectively, obtained by Kramers-Kroning transformation from the reflectivity curves presented in Fig. 4 and 5 at 30 K. The imaginary part of dielectric function for ZnxCdyHg(1-x-y)Te solid solution can be presented as the superposition of Lorentzians as it is follow from Eqn. (1). The fittings by the Lorentzian sums are presented in Fig.14 and 15 too. The parameters of Lorentzians are presented in

*x,y*for

**3.8 Spectral analyses of the FIR-spectra for quaternary alloys** 

Tab. 5 and 6 respectively.

0,0 0,5 1,0 1,5 2,0 2,5 3,0 3,5 4,0

Oscillator's strenght sum

solutions if APM are included into consideration.

Fig. 14. Spectral analysis for Sample Zn0,05Cd 0,23 Hg0,72Te (sample VI).

Fig. 15. Spectral analysis for sample Zn0,12Cd 0,13 Hg0,75Te (sample VII).

The present analysis shows that the investigated quaternary ZMCT has lager number of the resonance frequencies in the observed spectra compared to the ternary solid solutions of the binary HgTe, CdTe and ZnTe (118 to 180 cm-1) and extends the frequency region of the phonon modes – from 108 to 190 cm-1.

High Resolution Far Infrared Spectra of the

and 18 ones to fit the Im

where matrixes

therefore

Tn,m in lattice are:

for cation *A*;

for cation *B*, and

It is necessary to use 10 oscillators to fit the Im

**3.9 The random V-B model for quaternary alloys** 

and *m* B-cations. This probability is equal to:

4 *n m* 

It is seen that if y=0 and m=0, the Eqn.(16) is:

and the Eqn. (2) takes place for an alloy *AxC1-xZ*.

,

and

*n m*

4 *n* 

,

*n*

*n*

*n m*

Semiconductor Alloys Obtained Using the Synchrotron Radiation as Source 487

observed lines in obtained spectra will be performed in the frame of the V-B model developed for random distribution of atoms in the lattice for quaternary solid solutions.

We consider here the four-component solid solution *AxBy C 1-x-yZ* with three kinds of cations *A, B* and *C* and with the same anion *Z*. The lattice of quaternary alloy contents 15 basic units (tetrahedra): three binary *AZ, BZ, CZ,* nine strictly ternary *ABZ, ACZ, BCZ* and three strictly quaternary *ABCZ*. If quaternary alloy have *x* mol part of *AZ* compound and *y* mol part of *BZ* we can determine the probability to find in lattice the tetrahedron *Tnm* with *n* A-cations

4 4

*n P xy x y xy m n* 

<sup>4</sup>

are the same number of combinations as in Eqn. (2).

(,) 1 *n m n m*

 <sup>4</sup> <sup>4</sup> () 1 1 1 *<sup>m</sup> <sup>n</sup> P x <sup>n</sup> x x n*

At least, it is possible to consider when *1-x-y =* 0*, 4-n-m* = 0. In this case *y = 1-x* and *m = 4-n*,

4 4 (,) 1 *n m*

*n P xy x y m n* 

and we obtain the probability to find a tetrahedron *Tn* for an alloy *AxByZ.* It means that Eqn. (16) really correctly described the random distribution of atoms in the quaternary solid solutions *AxBy C 1-x-yZ*. The correspond probabilities to find particular cations in tetrahedron

<sup>4</sup> 4 4

<sup>4</sup> 4 4

( ) (1 ) <sup>4</sup> *A n m nm*

( ) (1 ) <sup>4</sup> *B n m nm*

*n n P x xy x m n*

*m n P x xy x m n*

*x,y)* curve (x=0.12, y=0.13) – sample VII. The identification of

*x,y)* curve (x=0.05, y=0.23) – sample VI,

(16)

(17)

(18)

, (19)

, (20)


Table 5. Parameters of Lorentzians presented in Fig.10 for sample VII


Table 6. Parameters of Lorentzians presented in Fig.9 for sample VI.

It is necessary to use 10 oscillators to fit the Im *x,y)* curve (x=0.05, y=0.23) – sample VI, and 18 ones to fit the Im *x,y)* curve (x=0.12, y=0.13) – sample VII. The identification of observed lines in obtained spectra will be performed in the frame of the V-B model developed for random distribution of atoms in the lattice for quaternary solid solutions.

#### **3.9 The random V-B model for quaternary alloys**

486 Infrared Spectroscopy – Materials Science, Engineering and Technology

**Frequency (cm-1) Oscilator strength (cm-2) Damping factor (cm-1)** 

107,20 4847 9,02 112,50 7341 7,02 119,00 4130 4,00 122,00 7000 3,50 124,50 8090 3,00 127,47 19700 3,50 131,00 27000 5,57 134,80 5700 6,20 139,10 4700 4,40 141,10 1000 4,40 149,30 900 2,40 152,40 890 4,10 160,20 4410 2,40 163,30 4000 2,40 173,00 8100 4,00 175,30 9800 4,00 178,00 483 6,40 181,70 5000 4,10

Table 5. Parameters of Lorentzians presented in Fig.10 for sample VII

Table 6. Parameters of Lorentzians presented in Fig.9 for sample VI.

**Frequency (cm-1) Oscilator strength (cm-2) Damping factor (cm-1)**  117,52 1600 4,02 120,92 6000 2,52 122,80 17041 1,41 124,00 860 1,30 125,00 1000 4,50 150,00 200 3,00 160,00 900 1,57 161,00 3000 1,20 162,00 4300 1,00 172,00 1000 3,80

We consider here the four-component solid solution *AxBy C 1-x-yZ* with three kinds of cations *A, B* and *C* and with the same anion *Z*. The lattice of quaternary alloy contents 15 basic units (tetrahedra): three binary *AZ, BZ, CZ,* nine strictly ternary *ABZ, ACZ, BCZ* and three strictly quaternary *ABCZ*. If quaternary alloy have *x* mol part of *AZ* compound and *y* mol part of *BZ* we can determine the probability to find in lattice the tetrahedron *Tnm* with *n* A-cations and *m* B-cations. This probability is equal to:

$$P\_{n,m}(\mathbf{x}, y) = \binom{4 - n}{m} \binom{4}{n} (1 - \mathbf{x} - y)^{4 - n - m} \mathbf{x}^n y^m \tag{16}$$

where matrixes 4 *n m* and 4 *n* are the same number of combinations as in Eqn. (2).

It is seen that if y=0 and m=0, the Eqn.(16) is:

$$P\_n(\mathbf{x}) = \mathbf{1} \cdot \binom{4}{n} (1-\mathbf{x})^{4-m} \cdot \mathbf{x}^n \cdot \mathbf{1} \tag{17}$$

and the Eqn. (2) takes place for an alloy *AxC1-xZ*.

At least, it is possible to consider when *1-x-y =* 0*, 4-n-m* = 0. In this case *y = 1-x* and *m = 4-n*, therefore

$$P\_{n,m}(\mathbf{x}, \mathbf{y}) = \binom{4-n}{m} \binom{4}{n} \cdot \mathbf{1} \cdot \mathbf{x}^n \cdot \mathbf{y}^m \tag{18}$$

and we obtain the probability to find a tetrahedron *Tn* for an alloy *AxByZ.* It means that Eqn. (16) really correctly described the random distribution of atoms in the quaternary solid solutions *AxBy C 1-x-yZ*. The correspond probabilities to find particular cations in tetrahedron Tn,m in lattice are:

$$P\_n^A(\mathbf{x}) = \frac{n}{4} \binom{4-n}{m} \binom{4}{n} \mathbf{x}^n y^m (1-\mathbf{x})^{4-n-m} \tag{19}$$

for cation *A*;

$$P\_n^{B}(\mathbf{x}) = \frac{m}{4} \binom{4-n}{m} \binom{4}{n} \mathbf{x}^n y^m (1-\mathbf{x})^{4-n-m} \,, \tag{20}$$

for cation *B*, and

High Resolution Far Infrared Spectra of the

1Hg 3Cd Hg - Te 6·10-3 - 1,2·10-3

2Zn 1Cd 1Hg Hg - Te 4·10-3 - 6·10-6

3Hg 1Cd Cd - Te 0,09 - 0,04

1Zn 3Cd Cd - Te 1,9·10-3 - 4,8·10-4

2Hg 2Zn Zn - Te 0,05 - 7·10-4

2Zn 2Cd Cd - Te 1,4·10-3 - 5·10-5

1Hg 3Zn Zn - Te 1,2·10-2 - 1,9·10-5

3Zn 1Cd Cd - Te 7·10-4 - 2·10-6

1Zn 1Cd 2Hg Cd -Te 0,032 - 0,007

1Zn 3Cd Zn - Te 1,6·10-4 - 6,5·10-4

2Zn 1Cd 1Hg Cd - Te 8·10-3 - 2·10-4

**Tethraedra Dipole-**

2HgII2Cd

1HgII3Cd

1Hg 3Zn

Semiconductor Alloys Obtained Using the Synchrotron Radiation as Source 489

 **n-type** 

**Composition VII, p-type** 

**Nr of line** 

**pairs Probablities Composition VI,**

1Zn 1Cd 2HgII Hg - Te 0,002 –0,007 **107,2 cm-1**<sup>1</sup>

1Zn 2Cd 1HgII Hg - Te 8·10-3- 2·10-3 **112,5 cm-1**<sup>2</sup>

4 Hg Hg - Te 0,37 - 0,24 **117,5 cm-1 119.0 cm-1** 3 1Zn 1Cd 2Hg Hg - Te 0,06 – 0,01 **120,92 cm-1 122,3 cm-1** 4 2Hg 2Cd Hg - Te 0,05 - 7·10-4 **124,5 cm-1** 5 3Hg 1Zn Hg - Te 0,18 - 2·10-2 **122,8 cm-1 127,5 cm-1** 6 3Hg 1Cd Hg - Te 0,28 - 0,13 **124,0 cm-1 131,0 cm-1** 7

2Hg 2Zn Hg - Te 0,05 - 7·10-4 **125. 0 cm-1 134,8 cm-1** 8 1Zn 1Cd 1Hg Hg - Te 0,07 - 0,02 **139,1 cm-1** 9

4Cd Cd - Te 0.02 - 2·10-4 **141,1 cm-1** 10 1Hg 3Cd Cd - Te 0,02 - 0,004 **150,0 cm-1 149,3 cm-1** 11 2 Hg 2Cd Cd - Te 0,07 - 0,02 **160 cm-1 152,4 cm-1** 12 1Zn 2Cd 1Hg Cd - Te 0,015 – 0,004 **161,5 cm-1 160,2 cm-1** 13

3Hg 1Zn Zn - Te 0,056 - 9·10-3 **162,0 cm-1 163,0 cm-1** 14

$$P\_n^{\mathbb{C}}(\mathbf{x}) = \frac{4-n-m}{4} \binom{4-n}{m} \binom{4}{n} \mathbf{x}^n y^m (1-\mathbf{x})^{4-n-m} \tag{21}$$

for cation *C*.

By this way the Eqns. (2, 4, 5, 16, 19 – 21) represent the complete description of the random atom distribution in the ternary *AxB1-xZ* and quaternary *AxBy C 1-x-yZ* solid solutions with substitution of cations. The four-component solid solution *AxByC 1-x-yZ* in ideally random case described by relations (16) and (19-21) consists from fifteen structural units – tetrahedra – which can generate 66 optically active phonon (vibrational) modes. These number of modes arose by next way: three strictly binary tetrahedra generate three vibrational modes AZ-like, BZ-like and CZ-like, nine (3x3) strictly ternary tetrahedra generate 9x6=54 vibrational modes and three strictly quaternary tetrahedra generate 3x3=9 vibrational modes: in sum 66 vibrational modes. In practice the most of these modes are degenerated (have the same frequencies): for example, the AZ-like modes generated in tetrahedra ABZ could have the same frequencies as AZ-like modes in tetrahedra ACZ. The same concerns the BZ-like and CZ-like modes. By this way the number of distinguished modes should be 30.

The expression for the oscillator strengths are similar as for ternary alloys:

$$\mathbf{N}\_{n,m}^{A-Z}(\mathbf{x},\mathbf{y}) = f\_{AZ}\mathbf{N}\_0 P\_{n,m}^A(\mathbf{x},\mathbf{y})\tag{22}$$

where probability , , *<sup>A</sup> P xy n m* is determined by (19). Similarly are given the expression for , *B Z Sn m* and , *C Z Sn m* . The corresponding oscillator sum rule are:

$$\sum S\_{n,m}^{A-Z} = N\_0 f\_{AZ} (1 - x - y) \tag{23}$$

$$\sum \mathbf{S}\_{n,m}^{B-Z} = \mathbf{N}\_0 f\_{BZ} \mathbf{x} \tag{24}$$

$$\sum \mathbf{S}\_{n,m}^{\mathcal{C}-Z} = \mathbf{N}\_0 f\_{\mathcal{C}Z} \mathbf{y} \tag{25}$$

The role of these oscillator sum rules would be the same as in ternary alloys but in the practice it is more difficult to relies the verification of the random distribution of atoms because the number of theoretically possible modes is very large and this factor prevent calculation of the oscillator sum rule.

#### **3.10 Identification of observed lines in case of the ZnxCdyHg1-x-yTe alloys**

An oscillator strength of particular mode enables us to determine the fraction of basic tetrahedral cells and interpret observed lines in phonon spectra. The attempt to interpret the spectra for sample VI (Fig. 9) and sample VII (Fig.10) is presented in Table 7. Here are presented results of the probabilities to find one of three dipole pairs (Hg-Te, Cd-Te, Zn-Te) in corresponding tetrahedron in lattice calculated according Eqn. (19-21) practically for all possibilities configurations in lattice of the ZnxCdyHg1-x-yTe alloys. There are observed modes generated by dipoles in cells probabilities to find of which in lattice is not less than 0.02. Summarizing the above mentioned results on the quaternary A1-x-yBxCyZ alloys, it is

By this way the Eqns. (2, 4, 5, 16, 19 – 21) represent the complete description of the random atom distribution in the ternary *AxB1-xZ* and quaternary *AxBy C 1-x-yZ* solid solutions with substitution of cations. The four-component solid solution *AxByC 1-x-yZ* in ideally random case described by relations (16) and (19-21) consists from fifteen structural units – tetrahedra – which can generate 66 optically active phonon (vibrational) modes. These number of modes arose by next way: three strictly binary tetrahedra generate three vibrational modes AZ-like, BZ-like and CZ-like, nine (3x3) strictly ternary tetrahedra generate 9x6=54 vibrational modes and three strictly quaternary tetrahedra generate 3x3=9 vibrational modes: in sum 66 vibrational modes. In practice the most of these modes are degenerated (have the same frequencies): for example, the AZ-like modes generated in tetrahedra ABZ could have the same frequencies as AZ-like modes in tetrahedra ACZ. The same concerns the BZ-like and CZ-like modes. By this way the number of distinguished modes should be

, 0 , (,) (,) *A Z <sup>A</sup> S xy f NP xy n m AZ n m*

where probability , , *<sup>A</sup> P xy n m* is determined by (19). Similarly are given the expression for

, 0 (1 ) *A Z S Nf x y n m AZ*

, 0 *B Z S Nf x n m BZ*

, 0 *C Z S Nf y n m CZ*

The role of these oscillator sum rules would be the same as in ternary alloys but in the practice it is more difficult to relies the verification of the random distribution of atoms because the number of theoretically possible modes is very large and this factor prevent

An oscillator strength of particular mode enables us to determine the fraction of basic tetrahedral cells and interpret observed lines in phonon spectra. The attempt to interpret the spectra for sample VI (Fig. 9) and sample VII (Fig.10) is presented in Table 7. Here are presented results of the probabilities to find one of three dipole pairs (Hg-Te, Cd-Te, Zn-Te) in corresponding tetrahedron in lattice calculated according Eqn. (19-21) practically for all possibilities configurations in lattice of the ZnxCdyHg1-x-yTe alloys. There are observed modes generated by dipoles in cells probabilities to find of which in lattice is not less than 0.02. Summarizing the above mentioned results on the quaternary A1-x-yBxCyZ alloys, it is

**3.10 Identification of observed lines in case of the ZnxCdyHg1-x-yTe alloys** 

(22)

(23)

(24)

(25)

*n m n P x xy x m n*

The expression for the oscillator strengths are similar as for ternary alloys:

. The corresponding oscillator sum rule are:

*n*

for cation *C*.

30.

, *B Z Sn m*

and ,

*C Z Sn m*

calculation of the oscillator sum rule.

<sup>4</sup> 4 4 4 ( ) (1 ) <sup>4</sup> *C n m nm*

 

(21)


High Resolution Far Infrared Spectra of the

assumption (Sheregii & Ugrin, 1992)).

8587 (1998)

512 (1990)

*Sol.C*,6, 1999 (2009)

*B*, 82, 014306 (2010)

Sussmann J.A. ,*J.Phys.Solids* ,28,1643,(1967 ) Talwar D.N., *J.Appl.Phys.* 56, 1601 (1984)

Baars J. and Sorgers F., *Solid State Commun*.10, 875 (1972) Barker A.S. and Sievers J., *Rev. Modern Phys*. 47, S1 (1975)

Triboulet R., *Appl. Phys. Lett.* 92, 121904 (2008)

Chandra D., Schaake H.F., Kinch M.A., *J. Electronics Materials*, 32, 810, (2003). Kosevich A.M., *The crystal lattice,* WILEY-VCH, Berlin-NewYork, (1999) Kozyrev S.P., L.K. Vodopyanov, R.Triboulet, *Phys. Rev. B,* 58, 3, 1374 (1998).

Pagès O., Souhabi J., Postnikov A. V., and Chafi A., *Phys. Rev. B* 80, 035204 (2009)

Sher A., Chen A.B., Spicer W.E. and Shih C.K., *J. Vac. Sci. Technol*. A3, 105 (1985)

Robouch B., Kisiel A. and Sheregii E. M., *Phys. Rev B.*, 64, 073204 (2001)

Sheregii E. M. and Ugrin Yu. O., *Sol. State Comm.*, 83, 1043 (1992)

Science Publishers, Amsterdam, pp. 35-131, (1988)

Rath S., Jain K.P., Abbi S.C., Julien C., Balkanski M., *Phys . Rev. B* ,52, 24, 17172 (1995)

Sheregii E. M., J. Cebulski, A. Marcelli, M. Piccinini, *China J. Phys.*, 102, 045504 (2011) Sheregii E.M., Cebulski J., Marcelli A. and Piccinini M., *Phys. Rev. Lett.* 102, 045504 (2009) Sheregii E.M., J. Polit , J. Cebulski, A. Marcelli, M. Castelli Guidi, B. Robouch, P. Calvani, M. Piccinini,, A. Kisiel, I. V. Ivanov-Omskii, *Infrared Physics & Technology* 49, 13 (2006)

*Optical Society of America*, A22, 2810 (2005)

oscillator strengths.

**5. References** 

Semiconductor Alloys Obtained Using the Synchrotron Radiation as Source 491

The results described above can give affirmative answer on the question: whether geometry of chaos e.g. the Bernoulli equation is enough to describe the oscillator strengths of observed lines in FIR-spectra if the Additional Phonon Modes will be involved in sums of the

Presented here cycle of researches dedicated to the ternary Hg1-xCdxTe cannot confirm but allow us to assume that the HgTe-like CPMs for x≤0.2 are extending and dispersion relation should be exist for them (as was shown the data on Magnetophonon Resonance confirm this

Adachi Sadao, *Optical Properties of Crystalline Solids and Amorphous Semiconductors. Materials and Fundamental Principles*, Kluver Academic Publishers, Boston, 1999 Amirtharaj P.M., Dhart N.K, Baars J. and Seelewind H., *Semicond. Sci. Technol.* 5, S68 (1990)

Cebulski J., Sheregii E. M., Polit J., Marcelli A., Piccinini M., Kisiel A., Kucherenko I.,

Cestelli Guidi M., Piccinini M., Marcelli A., Nucara A., Calvani P., Burattini E., *Journal of the* 

Krause R., Klimakow A., Kiessling F.M., Polity A., Gille P., Schenk M., *J. Cryst.Growth*., 101,

Marcelli A., Cesteli Guidi M., Piccinini M., Innocenzi P., Malfatti L., Xu W., *Phys. Stat.* 

Polit J., E.M. Sheregii, J. Cebulski, A. Marcelli, B. Robouch, A. Kisiel, A. Mycielski, *Phys. Rev.* 

Taylor D.W. in: Elliot R.J., Ipatova I.P. (Eds), *Optical properties of mixed crystals*, Elsevier

Biao Li.,.Chu J.H, Ye H.J., Guo S.P., Jiang W., Tang D.Y., *Appl.Phys.Lett.* 68, 3272 (1996) Cebulski J., Gebicki W., Ivanov-Omskii V.I., Polit J., Sheregii E.M., *J. Phys.: Cond. Matter* 10,


Table 7. Interpretation of observed lines in FIR spectra of ZnxCdyHg1-x-y Te

possible to conclude that 21 different modes are distinguished among the high resolution FIR-spectra of the seven composition of the ZnxCdyHg1-x-yTe alloys measured. By this way, the V-B random model developed for the quaternary alloys enable us to identify the observed structure of the sub-bands in the high resolution FIR-spectra ffor the ZnxCdyHg1 xyTe alloys.

The next step is calculation of the OSS for certain dipole pairs. It is reason to consider the dependence on composition of the OSS for Zn-Te dipoles (the ZnTe contain is changed from 0.05 to 0.18). In Table 8 are shown calculated OSS for this dipoles.


Table 8. The oscillator strength sum for Zn-Te dipoles in measured samples of ZnxCdyHg1-x-yTe

From Table 8 follow that OSS for Zn-Te dipoles is really approximately proportional to contain of ZnTe (values of *x*) in the ZnxCdyHg1-x-yTe alloys. Therefore, this important consequence of the V-B random model (Eqn.25) is fulfilled for the semiconductor quaternary alloys also.

## **4. Conclusion**

Those, the high resolution FIR-spectra of the ternary HgCdTe and quaternary HgZnCdTe alloys obtained by using of the synchrotron radiation as source enable us to decipher the tangled phonon spectra in these kinds of the semiconductor solid solutions applying the random version of the V-B model for its interpretation.

The results described above can give affirmative answer on the question: whether geometry of chaos e.g. the Bernoulli equation is enough to describe the oscillator strengths of observed lines in FIR-spectra if the Additional Phonon Modes will be involved in sums of the oscillator strengths.

Presented here cycle of researches dedicated to the ternary Hg1-xCdxTe cannot confirm but allow us to assume that the HgTe-like CPMs for x≤0.2 are extending and dispersion relation should be exist for them (as was shown the data on Magnetophonon Resonance confirm this assumption (Sheregii & Ugrin, 1992)).

## **5. References**

490 Infrared Spectroscopy – Materials Science, Engineering and Technology

 **n-type** 

**Composition VII, p-type** 

**Nr of line** 

**pairs Probablities Composition VI,**

1Zn 1Cd 2Hg Zn - Te 0,032 - 0,007 **172,0 cm-1 173,0 cm-1** 15 2Zn 1Cd 1Hg Zn - Te 0,02 - 4·10-4 **175,3 cm-1** 16 1Zn 2Cd 1Hg Zn- Te 7·10-3 - 2·10-3 **178,0 cm-1** 17 4Zn Zn - Te 1·10-3 - 2·10-6 **181,7 cm-1** 18

possible to conclude that 21 different modes are distinguished among the high resolution FIR-spectra of the seven composition of the ZnxCdyHg1-x-yTe alloys measured. By this way, the V-B random model developed for the quaternary alloys enable us to identify the observed structure of the sub-bands in the high resolution FIR-spectra ffor the ZnxCdyHg1-

The next step is calculation of the OSS for certain dipole pairs. It is reason to consider the dependence on composition of the OSS for Zn-Te dipoles (the ZnTe contain is changed from

From Table 8 follow that OSS for Zn-Te dipoles is really approximately proportional to contain of ZnTe (values of *x*) in the ZnxCdyHg1-x-yTe alloys. Therefore, this important consequence of the V-B random model (Eqn.25) is fulfilled for the semiconductor quaternary

Those, the high resolution FIR-spectra of the ternary HgCdTe and quaternary HgZnCdTe alloys obtained by using of the synchrotron radiation as source enable us to decipher the tangled phonon spectra in these kinds of the semiconductor solid solutions applying the

Number of sample x, mol OSS for Zn-Te

I 0.02 0.065 VI 0.05 0.197 II 0.07 0.211 VII 0.12 0.769 V 0.18 0.907 Table 8. The oscillator strength sum for Zn-Te dipoles in measured samples of

Table 7. Interpretation of observed lines in FIR spectra of ZnxCdyHg1-x-y Te

0.05 to 0.18). In Table 8 are shown calculated OSS for this dipoles.

random version of the V-B model for its interpretation.

**Tethraedra Dipole-**

xyTe alloys.

ZnxCdyHg1-x-yTe

alloys also.

**4. Conclusion** 

2Zn 2Cd Zn - Te 5·10-5 - 1,4·10-3

3Zn 1Cd Zn - Te 2·10-3 - 5·10-6



Piccinini,, A. Kisiel, I. V. Ivanov-Omskii, *Infrared Physics & Technology* 49, 13 (2006) Sussmann J.A. ,*J.Phys.Solids* ,28,1643,(1967 )

**Effective Reaction** 

*Vienna University of Technology* 

*Brenntag CEE GmbH* 

*Austria* 

Daniel Lumpi and Christian Braunshier

**Monitoring of Intermediates by ATR-IR** 

**Spectroscopy Utilizing Fibre Optic Probes** 

The use of reaction monitoring in order to determine operation parameters in organic synthesis and pharmaceutical chemistry still commonly relies on off-line approaches. However, application of on-line or in particular in-line methodologies provides highly valuable data with respect to process optimization and scale-up (Bakeev, 2005; Rubin et al., 2006). This statement is especially true for time-resolved spectroscopic *in-situ* techniques, which allow to gain insights into key intermediate formation or structures, and therefore also provide valuable information for mechanistic considerations (Minnich et al., 2007; Wiss

One of the major advantages of in-line techniques (both *in-situ* and real-time) over off-line approaches is that the investigation occurs inside the reaction system, thus eliminating sample alterations prior to analysis. These alterations during probing, including the loss of inertness or changes of reaction conditions, may result in erroneous readings; especially when directly compared to the (batch) process. This dramatically affects investigations at low temperatures. Both sampling and standard bypass approaches, which do not ensure constant thermal conditions in the course of analysis, potentially lead to incorrect results. For obtaining real-time information on chemical composition of samples in gas, liquid or solid phase, mid-infrared (IR) spectroscopy proved to be highly versatile, especially when performed with the attenuated total reflectance (ATR) technique (Grunwaldt & Baiker, 2005;

In this chapter, we focus on in-line monitoring of both highly sensitive and reactive organic key intermediates (reagents) by mid-IR fibre probes based on the ATR technique (Fig. 1). At first, we give a short introduction to mid-IR spectroscopy (also briefly commenting on alternative spectroscopic methods), outline advantages of IR fibre optics as well as ATR technologies and provide a brief overview of fibres suitable for mid-IR fibre applications. The last part is intended to introduce the interested reader to fibreoptic probes available and typical characteristics, referring to literature about chemical and physical properties of modern IR fibre materials, showcasing potential areas of

Marziano et al., 2000; Minnich et al., 2007; Zogg et al., 2004).

**1. Introduction** 

et al., 2006).

application.

Verleur H.W. and Barker A.S., *Phys. Rev.* 149,715 (1966). Ziman J.M., *Models of disorder*, Cambridge University Press, Cambridge, England, 1979 **26** 

## **Effective Reaction Monitoring of Intermediates by ATR-IR Spectroscopy Utilizing Fibre Optic Probes**

Daniel Lumpi and Christian Braunshier *Vienna University of Technology Brenntag CEE GmbH Austria* 

## **1. Introduction**

492 Infrared Spectroscopy – Materials Science, Engineering and Technology

Verleur H.W. and Barker A.S., *Phys. Rev.* 149,715 (1966).

The use of reaction monitoring in order to determine operation parameters in organic synthesis and pharmaceutical chemistry still commonly relies on off-line approaches. However, application of on-line or in particular in-line methodologies provides highly valuable data with respect to process optimization and scale-up (Bakeev, 2005; Rubin et al., 2006). This statement is especially true for time-resolved spectroscopic *in-situ* techniques, which allow to gain insights into key intermediate formation or structures, and therefore also provide valuable information for mechanistic considerations (Minnich et al., 2007; Wiss et al., 2006).

One of the major advantages of in-line techniques (both *in-situ* and real-time) over off-line approaches is that the investigation occurs inside the reaction system, thus eliminating sample alterations prior to analysis. These alterations during probing, including the loss of inertness or changes of reaction conditions, may result in erroneous readings; especially when directly compared to the (batch) process. This dramatically affects investigations at low temperatures. Both sampling and standard bypass approaches, which do not ensure constant thermal conditions in the course of analysis, potentially lead to incorrect results. For obtaining real-time information on chemical composition of samples in gas, liquid or solid phase, mid-infrared (IR) spectroscopy proved to be highly versatile, especially when performed with the attenuated total reflectance (ATR) technique (Grunwaldt & Baiker, 2005; Marziano et al., 2000; Minnich et al., 2007; Zogg et al., 2004).

In this chapter, we focus on in-line monitoring of both highly sensitive and reactive organic key intermediates (reagents) by mid-IR fibre probes based on the ATR technique (Fig. 1). At first, we give a short introduction to mid-IR spectroscopy (also briefly commenting on alternative spectroscopic methods), outline advantages of IR fibre optics as well as ATR technologies and provide a brief overview of fibres suitable for mid-IR fibre applications. The last part is intended to introduce the interested reader to fibreoptic probes available and typical characteristics, referring to literature about chemical and physical properties of modern IR fibre materials, showcasing potential areas of application.

Effective Reaction Monitoring of Intermediates

be conveniently acquired (Melling & Thomson, 2002).

information *in vivo* (Brancaleon et al., 2000).

**2.2 Attenuated total reflection** 

(Fig. 2).

by ATR-IR Spectroscopy Utilizing Fibre Optic Probes 495

within the typically stationary IR spectrometer by enabling the direct placement of the fibre-

The technologies of mid-IR spectroscopy providing highly relevant physico-chemical information, and the flexibility of fibre-optic probes offering new possibilities of application to measure samples in gas, liquid and solid phase, result in a breakthrough in molecular spectroscopy. Nowadays, spectroscopy utilizing IR fibre probes is routinely used in research laboratories, process development facilities and industrial quality control. This routine application can be explained by the fact that data, often not available by other methods, can

Additionally, ATR-based mid-IR fibre optic probes represent an entirely non-invasive technique, which has recently been shown to be a promising tool e.g. for biotechnological applications (Mazarevica, 2004) and, even more impressively, for obtaining spectroscopic

In principle, the sensor constructions rely on five basic sensing schemes: transmission, reflection, grazing angle reflection, attenuated total reflection (ATR), and a variant of the ATR effect known as the fibre-evanescent wave sensor (FEWS) (Melling & Thomson, 2002). The investigations described in chapters 3 and 4 are based on the use of the ATR technique, focusing on its advantageous properties, especially in terms of mechanical robustness.

The majority of reported mid-IR fibre probes rely on the well-established attenuated total reflection (ATR) technique, revealing many advantages in the general applicability over e.g. absorption measurements in short pathway flow cells. In the ATR method only a thin film (a few micrometers) at the proximity of the ATR element is subject to the measurement. The thickness of the analyzed film is defined by the penetration depth of the evanescent field

Fig. 2. Principle of total reflection of an infrared beam at the boundary of the ATR element to a medium with lower refractive index *n*2 < *n*1 (*n*1, *n*2, refractive indices; θ, angle of incidence; *E*, exponentially decaying evanescent field; *d*p, penetration depth; λ, wavelength of incident

The interactions of the incident and the internally reflected electromagnetic waves generate an exponentially decaying evanescent field, which penetrates the adjacent medium to a

radiation). Figure reproduced from Mizaikoff & Lendl, 2002.

optic probe inside the reaction system of interest (Melling & Thomson, 2002).

Fig. 1. Mid-IR fibre and ATR probe (left) and IR probe focused on the ATR element (right).

In the second part, we report on recent results of in-line investigations successfully utilizing mid-IR fibres on organometallic species. Hence, explorations on the formation of sodium alkoxy intermediates (performed in our group, Lumpi et al.) and their impact on reaction optimizations towards monodisperse oligo (ethylene glycols) are outlined. Subsequently, we switch to monitoring of organolithium compounds. In these projects Weymeels et al. as well as Gupta et al. were able to demonstrate ATR-IR fibre probe applications on metallation reactions and Lumpi et al. on metal halogen exchange reactions under cryogenic conditions.

## **2. Monitoring by mid-infrared spectroscopy**

Spectroscopy in the mid-infrared of a spectral range from approximately 4000 cm-1 to 400 cm-1 (2.5 μm to 25 μm) emerged as an effective tool for both qualitative and quantitative analysis. Within this range most of the fundamental molecular vibrations, the first overtones and combination frequencies occur. These typically relatively sharp absorption bands generally possess high absorption coefficients. Not only do these desirable spectroscopic properties facilitate an identification of molecules by its specific spectral "fingerprint", but also comprise valuable structural information (e.g. functional groups, substitution patterns, etc.) (Melling & Thomson, 2002).

The distinctive absorption bands associated with individual molecules enable the analysis of individual components in even complex mixtures by either evaluating isolated bands or by applying modern chemometric methods (e.g. Principal compound analysis), which process the entire spectral information. As a consequence, mid-IR spectroscopy represents a widely applicable tool for investigations of dynamic processes (e.g. chemical reactions, phase transitions, sedimentations, etc.). Moreover, information about interactions of the analyst with the surrounding media can be acquired because vibrational modes tend to be affected by the molecule's environment (Raichlin & Katzir, 2008).

Therefore, from many perspectives, mid-IR spectroscopy provides clearly more information than spectroscopy in other regions of the spectrum, such as the visible or the near-infrared range (Raichlin & Katzir, 2008).

#### **2.1 Mid-infrared optical fibre probes**

Modern-technology fibre optics offers important and versatile tools in spectroscopy. In the field of vibrational spectrometry fibre optics had a great influence on near-IR and Raman spectroscopy. The development of mid-IR transparent fibres (discussed in chapter 2.3) in the last decades had a significant impact on IR methods (Lendl & Mizaikoff, 2002). The fibre application makes it possible to overturn the established method of analyzing samples

Fig. 1. Mid-IR fibre and ATR probe (left) and IR probe focused on the ATR element (right).

**2. Monitoring by mid-infrared spectroscopy** 

substitution patterns, etc.) (Melling & Thomson, 2002).

range (Raichlin & Katzir, 2008).

**2.1 Mid-infrared optical fibre probes** 

affected by the molecule's environment (Raichlin & Katzir, 2008).

In the second part, we report on recent results of in-line investigations successfully utilizing mid-IR fibres on organometallic species. Hence, explorations on the formation of sodium alkoxy intermediates (performed in our group, Lumpi et al.) and their impact on reaction optimizations towards monodisperse oligo (ethylene glycols) are outlined. Subsequently, we switch to monitoring of organolithium compounds. In these projects Weymeels et al. as well as Gupta et al. were able to demonstrate ATR-IR fibre probe applications on metallation reactions and Lumpi et al. on metal halogen exchange reactions under cryogenic conditions.

Spectroscopy in the mid-infrared of a spectral range from approximately 4000 cm-1 to 400 cm-1 (2.5 μm to 25 μm) emerged as an effective tool for both qualitative and quantitative analysis. Within this range most of the fundamental molecular vibrations, the first overtones and combination frequencies occur. These typically relatively sharp absorption bands generally possess high absorption coefficients. Not only do these desirable spectroscopic properties facilitate an identification of molecules by its specific spectral "fingerprint", but also comprise valuable structural information (e.g. functional groups,

The distinctive absorption bands associated with individual molecules enable the analysis of individual components in even complex mixtures by either evaluating isolated bands or by applying modern chemometric methods (e.g. Principal compound analysis), which process the entire spectral information. As a consequence, mid-IR spectroscopy represents a widely applicable tool for investigations of dynamic processes (e.g. chemical reactions, phase transitions, sedimentations, etc.). Moreover, information about interactions of the analyst with the surrounding media can be acquired because vibrational modes tend to be

Therefore, from many perspectives, mid-IR spectroscopy provides clearly more information than spectroscopy in other regions of the spectrum, such as the visible or the near-infrared

Modern-technology fibre optics offers important and versatile tools in spectroscopy. In the field of vibrational spectrometry fibre optics had a great influence on near-IR and Raman spectroscopy. The development of mid-IR transparent fibres (discussed in chapter 2.3) in the last decades had a significant impact on IR methods (Lendl & Mizaikoff, 2002). The fibre application makes it possible to overturn the established method of analyzing samples within the typically stationary IR spectrometer by enabling the direct placement of the fibreoptic probe inside the reaction system of interest (Melling & Thomson, 2002).

The technologies of mid-IR spectroscopy providing highly relevant physico-chemical information, and the flexibility of fibre-optic probes offering new possibilities of application to measure samples in gas, liquid and solid phase, result in a breakthrough in molecular spectroscopy. Nowadays, spectroscopy utilizing IR fibre probes is routinely used in research laboratories, process development facilities and industrial quality control. This routine application can be explained by the fact that data, often not available by other methods, can be conveniently acquired (Melling & Thomson, 2002).

Additionally, ATR-based mid-IR fibre optic probes represent an entirely non-invasive technique, which has recently been shown to be a promising tool e.g. for biotechnological applications (Mazarevica, 2004) and, even more impressively, for obtaining spectroscopic information *in vivo* (Brancaleon et al., 2000).

In principle, the sensor constructions rely on five basic sensing schemes: transmission, reflection, grazing angle reflection, attenuated total reflection (ATR), and a variant of the ATR effect known as the fibre-evanescent wave sensor (FEWS) (Melling & Thomson, 2002). The investigations described in chapters 3 and 4 are based on the use of the ATR technique, focusing on its advantageous properties, especially in terms of mechanical robustness.

## **2.2 Attenuated total reflection**

The majority of reported mid-IR fibre probes rely on the well-established attenuated total reflection (ATR) technique, revealing many advantages in the general applicability over e.g. absorption measurements in short pathway flow cells. In the ATR method only a thin film (a few micrometers) at the proximity of the ATR element is subject to the measurement. The thickness of the analyzed film is defined by the penetration depth of the evanescent field (Fig. 2).

Fig. 2. Principle of total reflection of an infrared beam at the boundary of the ATR element to a medium with lower refractive index *n*2 < *n*1 (*n*1, *n*2, refractive indices; θ, angle of incidence; *E*, exponentially decaying evanescent field; *d*p, penetration depth; λ, wavelength of incident radiation). Figure reproduced from Mizaikoff & Lendl, 2002.

The interactions of the incident and the internally reflected electromagnetic waves generate an exponentially decaying evanescent field, which penetrates the adjacent medium to a

Effective Reaction Monitoring of Intermediates

from Harrington, 2010.

this chapter.

by ATR-IR Spectroscopy Utilizing Fibre Optic Probes 497

Fig. 3. Composite loss spectra for some common IR fibre optics: ZBLAN fluoride glass , SC sapphire, chalcogenide glass, PC AgBrCl, and hollow glass waveguide; plot reproduced

(e.g. reaction mixtures), relying on the evanescent wave principle, are AgBrCl, sapphire, chalcogenide and HFMG (Harrington, 2004). Polycrystalline silver halide fibres have been shown to be a promising candidate for mid-IR (ATR) fibre probes, especially for

The combination of the flexible structure and IR transmission of AgBrCl fibres ensures a convenient analytical approach; thus, also being applied in investigations presented later in

Besides IR spectroscopy, RINMR (rapid injection nuclear magnetic resonance) experiments also received considerable attention in the field of spectroscopic investigations of highly reactive species, especially under cryogenic conditions. The RINMR methodology, often based on the developments of J. F. McGarrity (McGarrity, 1981), C. A. Ogle and H. R. Loosli was successfully applied by several research groups to investigate reactive intermediates also at low temperatures and short time scales. In contrast to conventional NMR studies the rapid injection design relies on a piston-driven syringe injection assembly above the vessel inside the bore of the spectrometer magnet. This setup simultaneously provides turbulent mixing in the sample. In their first studies McGarrity et al. could establish that butyllithium in THF exists in equilibrium of the tetramer and the dimer complex with the proportion of dimer increasing as the temperature is decreased (McGarrity, 1985a). Moreover, kinetic examination proved that the dimeric butyllithium is more reactive toward the applied electrophiles than the tetramer by a factor of 10 (McGarrity, 1985b). Improved designs of RINMR systems implementing features such as multiple reactant and faster rapid injection were developed in

In conclusion, RINMR is a powerful tool for monitoring reactive intermediates, directly providing highly relevant structural data. However, the experimental complexity of this technique, in contrast to ATR-IR fibre probe applications, certainly limits its versatility.

measurements at wavelength > 10 μm (Brandstetter, 2009).

**2.4 Alternative spectroscopic methods for reaction monitoring** 

the last decade by P. J. Hore et al., H. J. Reich et al. and S. E. Denmark et al.

certain depth. The depth of penetration depends on the irradiation wavelength, the incident angle and the refractive indices of both the ATR element and the contact medium. The equations describing this behavior are given in Fig. 2 (Mizaikoff & Lendl, 2002).

## **2.3 Infrared transparent fibres**

This chapter gives a brief insight into the most important, for the main part commercially available, IR fibre optics. Detailed reviews on IR fibres are given in the literature by J.A. Harrington (Harrington, 2010) and, with a special focus on mid-IR applications by B. Lendl and B. Mizaikoff (Lendl & Mizaikoff, 2002).

The basic requirements for mid-IR fibres include physical properties such as transparency over the spectral range requested for the intended investigations, robustness (mechanically), stability (thermally and chemically) as well as adequate flexibility (Lendl & Mizaikoff, 2002). The characteristic of the optical transparency is typically evaluated by focusing on relevant loss mechanisms. The most important losses include intrinsic and extrinsic losses, Fresnel losses and bending losses (Sanghera & Aggarwal, 1998, as cited in Lendl & Mizaikoff, 2002). Available mid-IR fibre optics meet these challenges to different extents.

First developments on non-silica based IR transparent fibres from chalcogenide glasses, mainly arsenic sulphide, were published in 1965, exhibiting losses higher than 10 dB/m (Kapany & Simms, 1965, as cited in Harrington, 2010). Due to an elevated demand for IR fibres in short-haul applications increased research efforts were reported from the mid-1970s onwards (Harrington, 2010). Up to date, both optical and mechanical characteristics of IR fibres cannot compete with silica fibres (which are not applicable in the mid-IR region due to a transmission only up to approximately 2.5 μm). Losses in the range of a few decibels per meter still limit these to short-haul applications. Nevertheless, modern mid-IR fibres for short-haul have already enabled a broad variety of developments in spectroscopy and important usage in practical (e.g. medical) applications (Minnich et al., 2007, and references therein).

A logical categorization of the most important IR fibres can be illustrated as follows: glass, crystalline and hollow waveguides. Table 1 outlines this categorization also providing further subdivision based on materials and structures (Harrington, 2010).


Table 1. Categories of the most important of IR fibers; data reproduced from Harrington, 2010.

A graphical comparison of attenuation losses of the most relevant mid-IR fibres is given in Fig. 3. Among these, materials suitable for optic chemical sensor applications in liquid phase

certain depth. The depth of penetration depends on the irradiation wavelength, the incident angle and the refractive indices of both the ATR element and the contact medium. The

This chapter gives a brief insight into the most important, for the main part commercially available, IR fibre optics. Detailed reviews on IR fibres are given in the literature by J.A. Harrington (Harrington, 2010) and, with a special focus on mid-IR applications by B. Lendl

The basic requirements for mid-IR fibres include physical properties such as transparency over the spectral range requested for the intended investigations, robustness (mechanically), stability (thermally and chemically) as well as adequate flexibility (Lendl & Mizaikoff, 2002). The characteristic of the optical transparency is typically evaluated by focusing on relevant loss mechanisms. The most important losses include intrinsic and extrinsic losses, Fresnel losses and bending losses (Sanghera & Aggarwal, 1998, as cited in Lendl & Mizaikoff, 2002).

First developments on non-silica based IR transparent fibres from chalcogenide glasses, mainly arsenic sulphide, were published in 1965, exhibiting losses higher than 10 dB/m (Kapany & Simms, 1965, as cited in Harrington, 2010). Due to an elevated demand for IR fibres in short-haul applications increased research efforts were reported from the mid-1970s onwards (Harrington, 2010). Up to date, both optical and mechanical characteristics of IR fibres cannot compete with silica fibres (which are not applicable in the mid-IR region due to a transmission only up to approximately 2.5 μm). Losses in the range of a few decibels per meter still limit these to short-haul applications. Nevertheless, modern mid-IR fibres for short-haul have already enabled a broad variety of developments in spectroscopy and important usage in practical (e.g. medical) applications (Minnich et al., 2007, and references

A logical categorization of the most important IR fibres can be illustrated as follows: glass, crystalline and hollow waveguides. Table 1 outlines this categorization also providing

ZrF4-BaF2-LaF3-AlF3-NAF (ZBLAN)

GeO2-PbO

AgBrCl Sapphire

AsS3 and AsGeTeSe

Hollow glass waveguide Hollow sapphire at 10.6 μm

Main Subcategory Examples

Table 1. Categories of the most important of IR fibers; data reproduced from Harrington, 2010.

A graphical comparison of attenuation losses of the most relevant mid-IR fibres is given in Fig. 3. Among these, materials suitable for optic chemical sensor applications in liquid phase

equations describing this behavior are given in Fig. 2 (Mizaikoff & Lendl, 2002).

Available mid-IR fibre optics meet these challenges to different extents.

further subdivision based on materials and structures (Harrington, 2010).

Heavy metal fluoride (HMFG)

Germante Chalcogenide

Single crystal (SC)

Refractive index <1

Crystal Polycrystalline (PC)

Hollow waveguide Metal/dielectric film

**2.3 Infrared transparent fibres** 

therein).

Glass

and B. Mizaikoff (Lendl & Mizaikoff, 2002).

Fig. 3. Composite loss spectra for some common IR fibre optics: ZBLAN fluoride glass , SC sapphire, chalcogenide glass, PC AgBrCl, and hollow glass waveguide; plot reproduced from Harrington, 2010.

(e.g. reaction mixtures), relying on the evanescent wave principle, are AgBrCl, sapphire, chalcogenide and HFMG (Harrington, 2004). Polycrystalline silver halide fibres have been shown to be a promising candidate for mid-IR (ATR) fibre probes, especially for measurements at wavelength > 10 μm (Brandstetter, 2009).

The combination of the flexible structure and IR transmission of AgBrCl fibres ensures a convenient analytical approach; thus, also being applied in investigations presented later in this chapter.

## **2.4 Alternative spectroscopic methods for reaction monitoring**

Besides IR spectroscopy, RINMR (rapid injection nuclear magnetic resonance) experiments also received considerable attention in the field of spectroscopic investigations of highly reactive species, especially under cryogenic conditions. The RINMR methodology, often based on the developments of J. F. McGarrity (McGarrity, 1981), C. A. Ogle and H. R. Loosli was successfully applied by several research groups to investigate reactive intermediates also at low temperatures and short time scales. In contrast to conventional NMR studies the rapid injection design relies on a piston-driven syringe injection assembly above the vessel inside the bore of the spectrometer magnet. This setup simultaneously provides turbulent mixing in the sample. In their first studies McGarrity et al. could establish that butyllithium in THF exists in equilibrium of the tetramer and the dimer complex with the proportion of dimer increasing as the temperature is decreased (McGarrity, 1985a). Moreover, kinetic examination proved that the dimeric butyllithium is more reactive toward the applied electrophiles than the tetramer by a factor of 10 (McGarrity, 1985b). Improved designs of RINMR systems implementing features such as multiple reactant and faster rapid injection were developed in the last decade by P. J. Hore et al., H. J. Reich et al. and S. E. Denmark et al.

In conclusion, RINMR is a powerful tool for monitoring reactive intermediates, directly providing highly relevant structural data. However, the experimental complexity of this technique, in contrast to ATR-IR fibre probe applications, certainly limits its versatility.

Effective Reaction Monitoring of Intermediates

spectral range from 600 to 2000 wavenumbers.

**3.2 Technical details** 

**3.3 Inline monitoring** 

0.01 0.02 0.03 0.04 0.05 0.06 0.07

**absorbance [a.u.]**

**3.3.1 Single band analysis** 

by ATR-IR Spectroscopy Utilizing Fibre Optic Probes 499

reaction step shown in Scheme 1) is reported to take at least 18 h to reach completion. To prove this information, it was necessary to monitor the conversion in an inert, anhydrous reaction medium. Due to the lack of other proper methods of analysis, a mid-IR fibre optic probe was chosen for fast in-line monitoring of the chemical reaction under investigation.

The ATR fibre system was built up by a FT-IR spectrometer Bruker Matrix F® in connection with an ATR fibre probe (A.R.T. Photonics, Berlin; Ø 12 mm) and a MCT (mercury cadmium telluride) detector (Belov Technology, Co., Inc.). The probe was directly inserted through the ground neck of the reaction vessel and comprised two 1 m silver halide fibres (Ø 1 mm) connected to a conical two bounce diamond ATR element housed in a rod of hastelloy. Using this set-up it was possible to follow the reactions to be studied in real-time covering a

Two possibilities for the analysis of the obtained data have been applied. Single band

Determination of reaction progress by tracking changes in absorbance values at selected wavenumbers is described. Absorbance values at characteristic wavenumbers for the

> 1134 cm-1 1491 cm-1

analysis on the one hand, and multivariante curve resolution on the other hand.

substrate and the product respectively are plotted against reaction times in Figure 4.

Fig. 4. (a) Left: ATR-IR in-line monitoring for the deprotonation of 13,13,13-triphenyl-3,6,9,12-tetraoxatridecanol (substance **1b**). Distortions for t ≤ 10 min are attributed to

equilibration effects (temperature and concentration). (b) Right: Representative 3-D example

0 15 30 45 60 75 90 **time [min]**

of measurement. Figure reproduced from Lumpi et al., 2009.

Scheme 1. Synthesis of glycols **3a-d**. Scheme reproduced from Lumpi et al., 2009.

### **2.5 Reaction monitoring of organometallic compounds**

As mentioned before, the scientific results discussed in this contribution focus on monitoring of reactive organometallic species by ATR-IR spectroscopy utilizing mid-IR fibre probes. While alkoxide species, the conjugated base of an alcohol, are rather generally applied in synthetic chemistry (e.g. bases, nucleophiles and ligands) organolithium compounds require further introduction.

Since W. Schlenk and J. Holtz reported on the first syntheses of organolithium species in 1917 these powerful reagents or intermediates have gained enormous importance in the field of synthetic and pharmaceutical chemistry (Rappoport & Marek, 2004; Wu & Huang, 2006). Synthetic operations are generally carried out at low temperatures as a consequence of the high reactivity of lithium reagents; in some specific examples even below -100 °C (Rappoport & Marek, 2004). Nowadays, modern lithiation chemistry represents a wellestablished technique also receiving considerable attention in industrial processes (Rathman & Bailey, 2009; Wu & Huang, 2006). Despite the broad application of cryogenic temperature reactions dynamic analysis in order to monitor reactive lithium species remains a challenging task.

## **3.** *In-situ* **IR monitoring of alkoxides**

In this chapter, we focus on monitoring the formation of alkoxides, as this is an important intermediate for the synthesis of ether and ester containing substances. The conventional approach is to transform deprotonated moieties, followed by classical analytical methods (e.g. thin layer chromatography, HPLC or NMR spectroscopy). There are only limited descriptions of in-line methods in the literature, therefore an investigation is of high interest.

#### **3.1 Introduction**

A study (Lumpi et al., 2009) is presented in detail, where the usage of in-line ATR-IR was demonstrated in order to optimize the synthesis of monodisperse oligo(ethylene glycols) (OEG). This substance class has a wide range of applications in many fields of science and industry. They can be used as synthons for crown ether-type derivatives, as non-ionic surfactants, as templates for the synthesis of porous inorganic materials, and, more recently, functional mono-layers were applied to develop biocompatible material. The physical and chemical properties of these modified materials often depend to a large extent on the number of repetition units of the OEG moiety. For a systematic investigation of the influence of chain length, novel polystyrene-oligo(oxyethylene) graft copolymers containing monodisperse OEG units have been synthesized (Braunshier et al., 2008). For the preparation of these resins, access to well-defined oligo(ethylene glycols) of up to 12 units was required. Despite the widespread utility of OEGs, their synthesis remains a challenging task. The published synthetic methods for commercially unavailable or expensive representatives (n>4) are usually time-consuming and/or include extensive purification procedures. The most efficient strategy for synthesis of OEGs is based on bidirectional chain elongation (Keegstra et al., 1992).

Based on this approach, the key steps should be optimized in order to shorten reaction times from periods as long as several days to more acceptable values. The formation of corresponding alkoxides by deprotonation of mono-trityl protected glycols **1a-b** (first

$$\begin{bmatrix} \mathbf{0} \\ \mathbf{T}\_{\star}^{\top} \mathbf{0} \end{bmatrix} \mathbf{O} \mathbf{T} \begin{bmatrix} \mathbf{0} \\ \mathbf{0} \end{bmatrix} \xrightarrow{\mathbf{0} \cdot \mathbf{T} \begin{bmatrix} \mathbf{0} \\ \mathbf{0} \end{bmatrix}} \mathbf{T} \begin{bmatrix} \mathbf{0} \\ \mathbf{0} \end{bmatrix} \mathbf{S} \begin{bmatrix} \mathbf{0} \\ \mathbf{0} \end{bmatrix} \mathbf{S} \mathbf{T} + \begin{bmatrix} \mathbf{0} \\ \mathbf{0} \end{bmatrix} \mathbf{O} \mathbf{T} + \begin{bmatrix} \mathbf{0} \\ \mathbf{0} \end{bmatrix} \mathbf{O} \mathbf{T} \begin{bmatrix} \mathbf{0} \\ \mathbf{0} \end{bmatrix}$$

Scheme 1. Synthesis of glycols **3a-d**. Scheme reproduced from Lumpi et al., 2009.

reaction step shown in Scheme 1) is reported to take at least 18 h to reach completion. To prove this information, it was necessary to monitor the conversion in an inert, anhydrous reaction medium. Due to the lack of other proper methods of analysis, a mid-IR fibre optic probe was chosen for fast in-line monitoring of the chemical reaction under investigation.

## **3.2 Technical details**

498 Infrared Spectroscopy – Materials Science, Engineering and Technology

As mentioned before, the scientific results discussed in this contribution focus on monitoring of reactive organometallic species by ATR-IR spectroscopy utilizing mid-IR fibre probes. While alkoxide species, the conjugated base of an alcohol, are rather generally applied in synthetic chemistry (e.g. bases, nucleophiles and ligands) organolithium

Since W. Schlenk and J. Holtz reported on the first syntheses of organolithium species in 1917 these powerful reagents or intermediates have gained enormous importance in the field of synthetic and pharmaceutical chemistry (Rappoport & Marek, 2004; Wu & Huang, 2006). Synthetic operations are generally carried out at low temperatures as a consequence of the high reactivity of lithium reagents; in some specific examples even below -100 °C (Rappoport & Marek, 2004). Nowadays, modern lithiation chemistry represents a wellestablished technique also receiving considerable attention in industrial processes (Rathman & Bailey, 2009; Wu & Huang, 2006). Despite the broad application of cryogenic temperature reactions dynamic analysis in order to monitor reactive lithium species remains a

In this chapter, we focus on monitoring the formation of alkoxides, as this is an important intermediate for the synthesis of ether and ester containing substances. The conventional approach is to transform deprotonated moieties, followed by classical analytical methods (e.g. thin layer chromatography, HPLC or NMR spectroscopy). There are only limited descriptions of in-line methods in the literature, therefore an investigation is of high interest.

A study (Lumpi et al., 2009) is presented in detail, where the usage of in-line ATR-IR was demonstrated in order to optimize the synthesis of monodisperse oligo(ethylene glycols) (OEG). This substance class has a wide range of applications in many fields of science and industry. They can be used as synthons for crown ether-type derivatives, as non-ionic surfactants, as templates for the synthesis of porous inorganic materials, and, more recently, functional mono-layers were applied to develop biocompatible material. The physical and chemical properties of these modified materials often depend to a large extent on the number of repetition units of the OEG moiety. For a systematic investigation of the influence of chain length, novel polystyrene-oligo(oxyethylene) graft copolymers containing monodisperse OEG units have been synthesized (Braunshier et al., 2008). For the preparation of these resins, access to well-defined oligo(ethylene glycols) of up to 12 units was required. Despite the widespread utility of OEGs, their synthesis remains a challenging task. The published synthetic methods for commercially unavailable or expensive representatives (n>4) are usually time-consuming and/or include extensive purification procedures. The most efficient strategy for synthesis of OEGs is based on bidirectional chain

Based on this approach, the key steps should be optimized in order to shorten reaction times from periods as long as several days to more acceptable values. The formation of corresponding alkoxides by deprotonation of mono-trityl protected glycols **1a-b** (first

**2.5 Reaction monitoring of organometallic compounds** 

compounds require further introduction.

**3.** *In-situ* **IR monitoring of alkoxides** 

challenging task.

**3.1 Introduction** 

elongation (Keegstra et al., 1992).

The ATR fibre system was built up by a FT-IR spectrometer Bruker Matrix F® in connection with an ATR fibre probe (A.R.T. Photonics, Berlin; Ø 12 mm) and a MCT (mercury cadmium telluride) detector (Belov Technology, Co., Inc.). The probe was directly inserted through the ground neck of the reaction vessel and comprised two 1 m silver halide fibres (Ø 1 mm) connected to a conical two bounce diamond ATR element housed in a rod of hastelloy. Using this set-up it was possible to follow the reactions to be studied in real-time covering a spectral range from 600 to 2000 wavenumbers.

#### **3.3 Inline monitoring**

Two possibilities for the analysis of the obtained data have been applied. Single band analysis on the one hand, and multivariante curve resolution on the other hand.

#### **3.3.1 Single band analysis**

Determination of reaction progress by tracking changes in absorbance values at selected wavenumbers is described. Absorbance values at characteristic wavenumbers for the substrate and the product respectively are plotted against reaction times in Figure 4.

Fig. 4. (a) Left: ATR-IR in-line monitoring for the deprotonation of 13,13,13-triphenyl-3,6,9,12-tetraoxatridecanol (substance **1b**). Distortions for t ≤ 10 min are attributed to equilibration effects (temperature and concentration). (b) Right: Representative 3-D example of measurement. Figure reproduced from Lumpi et al., 2009.

Effective Reaction Monitoring of Intermediates

0,00



0,00

0,05

0,10

absorption [a.u.]

0,15

0,05

0,45

0,85

absorption [a.u.]

1,25

0,03

0,06

0,09

0,12

absorption [a.u.]

0,15

by ATR-IR Spectroscopy Utilizing Fibre Optic Probes 501

substance 1b

deprotonated substance 1b

0 15 30 45 60 75 90

1800 1600 1400 1200 1000 800 600

1800 1600 1400 1200 1000 800 600

(b)

(c)

Fig. 5. MCR-ALS calculations of the deprotonation of trityl-protected 1b; (a) calculated absorption profiles, (b) spectra calculations, (c) pure substance 1b spectrum. MCR-ALS-Parameters: [efa\_matrix]=efa(dep10,90); min value of log efa plot=1; number of factors=2; als2004; Data matrix=dep10; Init estimate=efa\_matrix; Non negative Concentration nnls; number of spec with non neg=2; spec equal height; Plots are optimum in the iteration Nr.

935 Std.dev of residuals vs. exp. data=0.0011238 Fitting error (lack of fit, lof) in %(PCA)=2.8319e-014 Fitting error (lack of fit, lof) in %(exp)=4.881 Percent of variance explained at the optimum is=99.7618. Figure reproduced from Lumpi et al., 2009.

(a)

IR-spectrum of 1b (calculated)

IR-spectrum of 1b (recorded in THF)

IR-spectrum of deprotonated 1b (calculated)

time [min]

wavenumber [1/cm]

wavenumber [1/cm]

The blue curve derives from the starting alcohol; the green graph originates from the deprotonated moiety. The graph clearly shows that after 90 min no significant changes of absorption values can be observed, thus being an indication for the end of the reaction.

### **3.3.2 Multivariate curve resolution**

Multivariate curve resolution – alternating least squares (MCR-ALS) – was additionally applied for the analysis of the IR data set. MCR-ALS is a modern chemometric method for the resolution of multiple component responses in unknown unresolved reaction mixtures (Tauler, 1995). This technique decomposes the recorded data set into smaller matrices containing information on the spectra and the concentration profiles of each component involved in the reaction. MCR-ALS can be applied to the analysis of a global instrumental response such as a set of mid-IR spectra recorded from a chemical reaction over time. In this case a matrix D (n x w) is obtained with n being the number of spectra and w representing the number of wavelengths. The MCR technique has the target to decompose D into the pure contributions of the components of the reaction according to the equation:

$$\mathbf{D} = \mathbf{C}\mathbf{S}^{\mathsf{T}} + \mathbf{E} \tag{1}$$

C contains the concentration profiles for all involved compounds and ST represents the corresponding spectra. MCR-ALS solves this equation in an iterative, alternating least squares manner by minimizing the residual matrix E. For the MCR analysis the data matrix D may be augmented with pure component spectra. Furthermore, several data sets may be analyzed simultaneously. During calculation meaningful constraints may be applied with the aim to guide the iteration process toward a mathematically as well as chemically meaningful solution.

Despite all advantages, MCR-ALS algorithms written in MATLAB have the disadvantage of how the selected constraints are implemented in the execution of the MATLAB routine. This process can be troublesome and sophisticated, particularly in complex cases where several data matrices are simultaneously analyzed and/or different constraints are applied. In order to overcome these difficulties and taking advantage of the better MATLAB tools to create graphical user interfaces, an improved MCR-ALS toolbox with a user-friendly graphical interface was presented by Jaumont et al., 2005.

Examples for meaningful constraints in the chemical system under investigation in the paper by Lumpi et al., 2009 are non-negativity of concentrations and spectral intensities as well as unimodality. As a result, quantitative information on the amount of spectral contribution of each component in every spectrum of the data set is obtained. 99.72 % of the spectral variance of the recorded data could be explained with two components. Therefore it was concluded that the two components needed to be considered in modeling the data set, as shown in Figure 5. The reaction under study came to completion after 90 minutes.

## **3.4 Results and discussion**

Both techniques for IR data analysis presented, showed clearly that time for deprotonation is much shorter than described in the literature. The subsequent nucleophilic reaction of the alkoxides with the tosylated glycols **2a-b** lead to substances **3a-d**.

The blue curve derives from the starting alcohol; the green graph originates from the deprotonated moiety. The graph clearly shows that after 90 min no significant changes of absorption values can be observed, thus being an indication for the end of the reaction.

Multivariate curve resolution – alternating least squares (MCR-ALS) – was additionally applied for the analysis of the IR data set. MCR-ALS is a modern chemometric method for the resolution of multiple component responses in unknown unresolved reaction mixtures (Tauler, 1995). This technique decomposes the recorded data set into smaller matrices containing information on the spectra and the concentration profiles of each component involved in the reaction. MCR-ALS can be applied to the analysis of a global instrumental response such as a set of mid-IR spectra recorded from a chemical reaction over time. In this case a matrix D (n x w) is obtained with n being the number of spectra and w representing the number of wavelengths. The MCR technique has the target to decompose D into the

C contains the concentration profiles for all involved compounds and ST represents the corresponding spectra. MCR-ALS solves this equation in an iterative, alternating least squares manner by minimizing the residual matrix E. For the MCR analysis the data matrix D may be augmented with pure component spectra. Furthermore, several data sets may be analyzed simultaneously. During calculation meaningful constraints may be applied with the aim to guide the iteration process toward a mathematically as well as chemically

Despite all advantages, MCR-ALS algorithms written in MATLAB have the disadvantage of how the selected constraints are implemented in the execution of the MATLAB routine. This process can be troublesome and sophisticated, particularly in complex cases where several data matrices are simultaneously analyzed and/or different constraints are applied. In order to overcome these difficulties and taking advantage of the better MATLAB tools to create graphical user interfaces, an improved MCR-ALS toolbox with a user-friendly graphical

Examples for meaningful constraints in the chemical system under investigation in the paper by Lumpi et al., 2009 are non-negativity of concentrations and spectral intensities as well as unimodality. As a result, quantitative information on the amount of spectral contribution of each component in every spectrum of the data set is obtained. 99.72 % of the spectral variance of the recorded data could be explained with two components. Therefore it was concluded that the two components needed to be considered in modeling the data set, as shown in Figure 5. The reaction under study came to completion after 90

Both techniques for IR data analysis presented, showed clearly that time for deprotonation is much shorter than described in the literature. The subsequent nucleophilic reaction of the

alkoxides with the tosylated glycols **2a-b** lead to substances **3a-d**.

D = CST + E (1)

pure contributions of the components of the reaction according to the equation:

**3.3.2 Multivariate curve resolution** 

meaningful solution.

minutes.

**3.4 Results and discussion** 

interface was presented by Jaumont et al., 2005.

Fig. 5. MCR-ALS calculations of the deprotonation of trityl-protected 1b; (a) calculated absorption profiles, (b) spectra calculations, (c) pure substance 1b spectrum. MCR-ALS-Parameters: [efa\_matrix]=efa(dep10,90); min value of log efa plot=1; number of factors=2; als2004; Data matrix=dep10; Init estimate=efa\_matrix; Non negative Concentration nnls; number of spec with non neg=2; spec equal height; Plots are optimum in the iteration Nr. 935 Std.dev of residuals vs. exp. data=0.0011238 Fitting error (lack of fit, lof) in %(PCA)=2.8319e-014 Fitting error (lack of fit, lof) in %(exp)=4.881 Percent of variance explained at the optimum is=99.7618. Figure reproduced from Lumpi et al., 2009.


Table 2. Reaction times and yields for the preparation of substances **3a-d**. aTimes given refer to deprotonation and overall reaction time respectively. Modified from Lumpi et al., 2009.

To obtain the target compounds (OEGs **4a-d**), the protecting groups have to be cleaved off. Virtually all published procedures use hydrogenolysis under high-pressure conditions in the presence of palladium for several days to achieve this final transformation. Apart from long reaction time, this procedure suffers from some more disadvantages. The most serious one is the need for equipment allowing to perform gas reactions under high pressure, which might be a limiting factor. Moreover, the use of halogenated organic solvents, e.g. dichloromethane and transition metal catalysts, might become troublesome, if the final product is intended to be used in the field of pharmaceutics or biology, especially when the procedure is performed on industrial scale.

Effective Reaction Monitoring of Intermediates

selected in the following examples.

butyllithium (BuLi) were investigated.

**4.2.1 Technical details** 

**4.2.2 Results and discussion** 

**4.2** *In-situ* **infrared studies applying BuLi and LTMP** 

and (5) the reaction was quenched by deuterium oxide.

which can be a source of degradation and/or competitive reactions.

by ATR-IR Spectroscopy Utilizing Fibre Optic Probes 503

heteroaromatic substances. They are applied in many fields of chemistry, either for pharmaceuticals or as building blocks for various applications within materials science, e.g. in Organic Light Emitting Transistors (OLET). As a matter of fact, extensive efforts have been applied to a variety of synthetic methodologies. The most prominent strategy is based on lithium moieties, formed through deprotonation, which allows multiple functionalizations. Organolithium chemistry is known to be highly sophisticated due to the fact that analytical methods for the monitoring of reactive intermediates, often only stable at low temperatures, are rare. Furthermore, side reactions related to the aggregation state and structures of the reactive species are common. Over the last years, only a few studies on monitoring of metallation by in-situ infrared spectroscopy have been carried out (Kondo et al., 1999; Sun & Collum, 2000; Pippel et al., 2001; Zhao & Collum, 2003). One reason might be that IR spectroscopy is preferred mostly for strong absorbing groups, like carbonyl moieties, which are often missing in heteroaromatics. Some of the most important reagents for lithiation are n-butyllithium (BuLi), lithium tetramethylpiperidine (LTMP) and lithium diisopropylamide (LDA), which have been

Here, the work of Weymeels et al., 2005, in which the kinetics and mechanism of the deprotonation of 3,5-dichloropyridine by lithium tetramethylpiperidine (LTMP) and n-

The IR spectra were recorded with a ReactIRe4000 equipped with a DiComp ATR probe (ASI Applied Systems, Mettler Toledo). The following steps of the experiments have been performed: (1) THF was cooled to -75°C; (2) the spectral baseline was reset to zero and the spectra recording was started; (3) the substrate was added; (4) the base was added dropwise;

The absorption bands of 3,5-dichloropyridine instantly decreased upon the addition of the base. As a result, the absorbance values associated with the aryllithium species appeared. By comparing the absorbance bands obtained using LTMP (Fig. 6(a)) and BuLi (Fig. 6(b)), it is clearly visible that two values (753 and 1007 cm-1) out of the three attributed to 3,5-dichloro-4-pyridyllithium are identical, while the third (1139 cm-1 using LTMP and 1143 cm-1 using BuLi) is different. A possible explanation was that lithium bears different ligands, which is even more sophisticated due to the presence of the ring nitrogen atom. Studies comparing the consumption of LTMP and BuLi respectively for achieving quantitative protonation have been applied. It was assumed that a competitive formation of a complex between the generated aryllithium and LTMP could be responsible for different consumptions (1,25 equiv. LTMP *vs* 1,0 equiv. BuLi). To sum up, transition structures between the substrate and the lithio derivative could be detected by using IR in-situ monitoring. Moreover, IR monitoring assured that reaction conditions (time, temperature, number of equivalents) had been selected in a right way. This is a crucial requirement to circumvent overestimation,

Scheme 2. Synthesis of glycols **4a-d**; hydrogenolysis *vs*. acidic cleavage. Scheme reproduced from Lumpi et al., 2009.

In the work presented, this deprotection step was substituted for a safe, fast and inexpensive procedure. Acidic cleavage by acetic acid in water for only 2 h was performed to obtain the pure OEGs. Comparing this new protocol to hydrogenolysis, the advantages are the following: dramatically shortened reaction times (2 h *vs*. 4 days), easier work-up and higher product quality. In summary, an optimized protocol for the synthesis of monodisperse OEGs up to 12 units has been reported. In contrast to other approaches described in the literature, neither special equipment for high pressure hydrogenolysis nor any chromatographic purification is needed for the key steps of the sequence. In-line ATR-IR spectroscopy was shown to be a powerful analytical tool for the effective monitoring of such "problematic" processes.

#### **4.** *In-situ* **monitoring in organolithium chemistry**

#### **4.1 Introduction**

In this chapter, we describe monitoring in organolithium chemistry, one of the most important fields within organic synthesis, especially for the functionalization of

Entry Glycol Tosylate Time Product Yield 1 2 ha 3 % 1 **1a 2a** 4/84 **3a** 98 2 **1a 2b** 4/84 **3b** 97 3 **1b 2a** 2/60 **3c** 98 4 **1b 2b** 2/60 **3d** 95 Table 2. Reaction times and yields for the preparation of substances **3a-d**. aTimes given refer to deprotonation and overall reaction time respectively. Modified from Lumpi et al., 2009.

To obtain the target compounds (OEGs **4a-d**), the protecting groups have to be cleaved off. Virtually all published procedures use hydrogenolysis under high-pressure conditions in the presence of palladium for several days to achieve this final transformation. Apart from long reaction time, this procedure suffers from some more disadvantages. The most serious one is the need for equipment allowing to perform gas reactions under high pressure, which might be a limiting factor. Moreover, the use of halogenated organic solvents, e.g. dichloromethane and transition metal catalysts, might become troublesome, if the final product is intended to be used in the field of pharmaceutics or biology, especially when the

Scheme 2. Synthesis of glycols **4a-d**; hydrogenolysis *vs*. acidic cleavage. Scheme reproduced

In the work presented, this deprotection step was substituted for a safe, fast and inexpensive procedure. Acidic cleavage by acetic acid in water for only 2 h was performed to obtain the pure OEGs. Comparing this new protocol to hydrogenolysis, the advantages are the following: dramatically shortened reaction times (2 h *vs*. 4 days), easier work-up and higher product quality. In summary, an optimized protocol for the synthesis of monodisperse OEGs up to 12 units has been reported. In contrast to other approaches described in the literature, neither special equipment for high pressure hydrogenolysis nor any chromatographic purification is needed for the key steps of the sequence. In-line ATR-IR spectroscopy was shown to be a

In this chapter, we describe monitoring in organolithium chemistry, one of the most important fields within organic synthesis, especially for the functionalization of

powerful analytical tool for the effective monitoring of such "problematic" processes.

**4.** *In-situ* **monitoring in organolithium chemistry** 

procedure is performed on industrial scale.

from Lumpi et al., 2009.

**4.1 Introduction** 

heteroaromatic substances. They are applied in many fields of chemistry, either for

pharmaceuticals or as building blocks for various applications within materials science, e.g. in Organic Light Emitting Transistors (OLET). As a matter of fact, extensive efforts have been applied to a variety of synthetic methodologies. The most prominent strategy is based on lithium moieties, formed through deprotonation, which allows multiple functionalizations. Organolithium chemistry is known to be highly sophisticated due to the fact that analytical methods for the monitoring of reactive intermediates, often only stable at low temperatures, are rare. Furthermore, side reactions related to the aggregation state and structures of the reactive species are common. Over the last years, only a few studies on monitoring of metallation by in-situ infrared spectroscopy have been carried out (Kondo et al., 1999; Sun & Collum, 2000; Pippel et al., 2001; Zhao & Collum, 2003). One reason might be that IR spectroscopy is preferred mostly for strong absorbing groups, like carbonyl moieties, which are often missing in heteroaromatics. Some of the most important reagents for lithiation are n-butyllithium (BuLi), lithium tetramethylpiperidine (LTMP) and lithium diisopropylamide (LDA), which have been selected in the following examples.

#### **4.2** *In-situ* **infrared studies applying BuLi and LTMP**

Here, the work of Weymeels et al., 2005, in which the kinetics and mechanism of the deprotonation of 3,5-dichloropyridine by lithium tetramethylpiperidine (LTMP) and nbutyllithium (BuLi) were investigated.

### **4.2.1 Technical details**

The IR spectra were recorded with a ReactIRe4000 equipped with a DiComp ATR probe (ASI Applied Systems, Mettler Toledo). The following steps of the experiments have been performed: (1) THF was cooled to -75°C; (2) the spectral baseline was reset to zero and the spectra recording was started; (3) the substrate was added; (4) the base was added dropwise; and (5) the reaction was quenched by deuterium oxide.

### **4.2.2 Results and discussion**

The absorption bands of 3,5-dichloropyridine instantly decreased upon the addition of the base. As a result, the absorbance values associated with the aryllithium species appeared. By comparing the absorbance bands obtained using LTMP (Fig. 6(a)) and BuLi (Fig. 6(b)), it is clearly visible that two values (753 and 1007 cm-1) out of the three attributed to 3,5-dichloro-4-pyridyllithium are identical, while the third (1139 cm-1 using LTMP and 1143 cm-1 using BuLi) is different. A possible explanation was that lithium bears different ligands, which is even more sophisticated due to the presence of the ring nitrogen atom. Studies comparing the consumption of LTMP and BuLi respectively for achieving quantitative protonation have been applied. It was assumed that a competitive formation of a complex between the generated aryllithium and LTMP could be responsible for different consumptions (1,25 equiv. LTMP *vs* 1,0 equiv. BuLi). To sum up, transition structures between the substrate and the lithio derivative could be detected by using IR in-situ monitoring. Moreover, IR monitoring assured that reaction conditions (time, temperature, number of equivalents) had been selected in a right way. This is a crucial requirement to circumvent overestimation, which can be a source of degradation and/or competitive reactions.

Effective Reaction Monitoring of Intermediates

injections and a nitrogen line for inertness.

**4.3.2 Results and discussion** 

**4.3.1 Technical details** 

by ATR-IR Spectroscopy Utilizing Fibre Optic Probes 505

The topic of the presented contribution (Gupta et al., 2008) is ortholithiation on a range of arenes mediated by LDA in combination with catalytic amounts of LiCl (0,5% relative to LDA). The lithiation reactions were monitored using in-situ IR spectroscopy following both the disappearance of the arene and the formation of the resulting aryllithium moiety. In addition, 19F NMR spectroscopic analysis was performed, providing comparable results. The challenge was to monitor the Li-species as sensitive key intermediate, only stable at

The spectra were recorded by applying an in-situ IR spectrometer fitted with a 30-bounce, silicon-tipped probe. The spectra were acquired in 16 scans, and a representative reaction was carried out as follows: The IR probe was inserted into an oven-dried, cylindrical flask fitted with a magnetic stir bar and a T-joint. The T-joint was equipped with a septum for

Reaction rate studies investigating the effect of the LiCl and other lithium salts as catalyst were performed. Furthermore, the effect of the influence of the substrate was part of the study. In conclusion, despite some detected irregularities, the obtained results might be important for industrial application. Rate variations that go undetected in the lab could give way to unexpected and potentially costly variations on production scales. A representative example of in-situ monitoring is shown in Figure 7. A significant difference in the half-lives

(t1/2) can be identified, indicating an acceleration of the lithiation by addition of LiCl.

Fig. 7. Plot of IR absorbances (black – 1507 cm-1, red – 1418 cm-1) versus time for the ortholithiation of 1,4-difluorobenzene (0.10 M) with LDA (0.12 M) in THF at -78 °C: (A) no

added LiCl; (B) 0.5 mol% LiCl. Figure reproduced from Gupta et al., 2008.

cryogenic temperatures. Therefore classical analytical methods could not be applied.

Fig. 6. Progression of the reaction between 3,5-dichloropyridine and (a) LTMP or (b) BuLi, and subsequent deuteriolysis. Figure reproduced from Weymeels et al., 2005.

#### **4.3** *In-situ* **studies applying LDA**

Lithium diisopropylamide (LDA) is the most prominent and preferred reagent for reactions requiring a strong non-nucleophilic base, therefore being one of the most important reagents in organic chemistry (Collum et al., 2007). The central importance of LDA motivated several research teams to examine mechanism and transition state as well. Based on ATR-IR spectroscopy Collum et al. presented some studies on lithiation reaction.

The topic of the presented contribution (Gupta et al., 2008) is ortholithiation on a range of arenes mediated by LDA in combination with catalytic amounts of LiCl (0,5% relative to LDA). The lithiation reactions were monitored using in-situ IR spectroscopy following both the disappearance of the arene and the formation of the resulting aryllithium moiety. In addition, 19F NMR spectroscopic analysis was performed, providing comparable results. The challenge was to monitor the Li-species as sensitive key intermediate, only stable at cryogenic temperatures. Therefore classical analytical methods could not be applied.

## **4.3.1 Technical details**

504 Infrared Spectroscopy – Materials Science, Engineering and Technology

Fig. 6. Progression of the reaction between 3,5-dichloropyridine and (a) LTMP or (b) BuLi,

Lithium diisopropylamide (LDA) is the most prominent and preferred reagent for reactions requiring a strong non-nucleophilic base, therefore being one of the most important reagents in organic chemistry (Collum et al., 2007). The central importance of LDA motivated several research teams to examine mechanism and transition state as well. Based on ATR-IR

and subsequent deuteriolysis. Figure reproduced from Weymeels et al., 2005.

spectroscopy Collum et al. presented some studies on lithiation reaction.

**4.3** *In-situ* **studies applying LDA** 

The spectra were recorded by applying an in-situ IR spectrometer fitted with a 30-bounce, silicon-tipped probe. The spectra were acquired in 16 scans, and a representative reaction was carried out as follows: The IR probe was inserted into an oven-dried, cylindrical flask fitted with a magnetic stir bar and a T-joint. The T-joint was equipped with a septum for injections and a nitrogen line for inertness.

### **4.3.2 Results and discussion**

Reaction rate studies investigating the effect of the LiCl and other lithium salts as catalyst were performed. Furthermore, the effect of the influence of the substrate was part of the study. In conclusion, despite some detected irregularities, the obtained results might be important for industrial application. Rate variations that go undetected in the lab could give way to unexpected and potentially costly variations on production scales. A representative example of in-situ monitoring is shown in Figure 7. A significant difference in the half-lives (t1/2) can be identified, indicating an acceleration of the lithiation by addition of LiCl.

Fig. 7. Plot of IR absorbances (black – 1507 cm-1, red – 1418 cm-1) versus time for the ortholithiation of 1,4-difluorobenzene (0.10 M) with LDA (0.12 M) in THF at -78 °C: (A) no added LiCl; (B) 0.5 mol% LiCl. Figure reproduced from Gupta et al., 2008.

Effective Reaction Monitoring of Intermediates

by ATR-IR Spectroscopy Utilizing Fibre Optic Probes 507

Fig. 8. Double-sided Halogen Dance reaction; A: reaction scheme; LDA (2.5 equiv.), -40 °C. i: metallation reaction, ii: Halogen Dance reaction, iii: excess of methanol; B: 3D plot of spectra (750 - 1050 cm-1) recorded during reaction progress (0 - 160 s); C: monitoring of Li-species LDA and intermediate **2**; D: intermediate **3** formation as extracted from the spectral data set

good agreement of kinetic parameters of LDA consumption and intermediate formation

Due to overlapping absorption bands the MCR-ALS algorithm (chapter 3.3.2.) was applied to the spectral data set prior to analysis in order to compare the consumption of intermediate **2** and the formation of product **3** (Fig 8, D). The results again disclose a good agreement of kinetic data supporting the aforementioned conclusion, being highly valuable

In summary, a methodology utilizing an FFT based correction procedure for a convenient monitoring of metal halogen exchange (but also metallation) reactions is presented. By applying this technique a mechanistic investigation of a complex double-sided Halogen

In conclusion, it has been demonstrated that mid-infrared (ATR) spectroscopy utilizing modern optical fibre probes is an effective methodology for in-line monitoring of highly reactive species. Therefore, this non-invasive technique has emerged as a versatile tool for direct reaction monitoring, which was outlined using the example of organometallic

*via* MCR-ALS algorithm. Figure reproduced from Lumpi et al., 2012.

for potential application of this reaction (e.g. sequential Halogen Dance).

(Fig 8, C).

Dance reaction could be realized.

**5. Conclusion** 

intermediates.

### **4.4 Investigations of metal halogen exchange reactions**

In contrast to the previous chapters dealing with the deprotonation (metallation) of the substrate this chapter focuses on metal halogen exchange reactions towards the desired organolithium intermediates using BuLi. This type of reaction is of particular importance for the selective synthesis of certain substitution patterns (Rappoport & Marek, 2004).

The outlined contribution (Lumpi et al., 2012) presents investigations on metal halogen exchange reactions by inline monitoring of organolithium species under both inert and cryogenic conditions. Starting from relatively simple substrates the exploration of a complex Halogen Dance reaction sequence was realized, which allows the convenient synthesis of precursors for e.g. thiophene ring-opening reactions (Bobrovsky et al., 2008). In order to acquire reliable spectroscopic data *via* an optical ATR-IR fibre probe a procedure to correct the effects of (co-)sine type fringes, which are observed during the fast and exothermic metal halogen exchange reactions, has been developed.

## **4.4.1 Technical details**

The instrumental setup was identical to the setup described in chapter 3.2. The fibre probe was mounted through a ground neck into a 4-neck round bottom flask, which was subsequently charged with dry solvent and cooled *via* a cooling bath. After a steady temperature in the vessel was reached a background spectrum was recorded, the measurement started and the lithium species added. Finally, the respective reactant was rapidly added *via* a syringe.

### **4.4.2 FFT correction procedure**

Temperature deviations between the sample spectra and the background acquired during dynamic reaction processes represent one of the major challenges in IR spectroscopy at cryogenic temperatures. During these investigations the application of an ATR-IR fibre probe to relatively fast and exothermic metal halogen exchange reactions resulted in spectra heavily overlaid with (co-)sine-type artifacts. The introduced fringes lead to spectral data not utilizable for further interpretation.

To overcome these limitations a correction procedure based on fast Fourier transformation (FFT) was implemented to the data evaluation process. This high pass filter significantly reduced the fringes (attributed to differing thermal expansion coefficients of the probe materials) and rendered it possible to achieve a reliable monitoring of metal halogen exchange reactions at temperatures even below –80 °C.

#### **4.4.3 Results and discussion**

An application of the developed methodology to simple substrates for metal halogen exchange reactions but also to metallation procedures afforded reliable results. Thus, the technique was utilized for a detailed kinetic investigation of a double-sided Halogen Dance (Fig. 8, A) reaction towards 4,4'-dibromo-2,2'-bithiophene (Bobrovsky et al., 2008).

In this multi-step sequence (Fig. 8, A) the first metallation could be shown to proceed faster than the second lithiation towards **3** leading to an accumulation of intermediate **2** (Fig. 8, B). This assumption of the first reaction step exclusively consuming LDA to form **2** and the second Halogen Dance being realized in a next step leading to **3** could be verified by a

In contrast to the previous chapters dealing with the deprotonation (metallation) of the substrate this chapter focuses on metal halogen exchange reactions towards the desired organolithium intermediates using BuLi. This type of reaction is of particular importance for

The outlined contribution (Lumpi et al., 2012) presents investigations on metal halogen exchange reactions by inline monitoring of organolithium species under both inert and cryogenic conditions. Starting from relatively simple substrates the exploration of a complex Halogen Dance reaction sequence was realized, which allows the convenient synthesis of precursors for e.g. thiophene ring-opening reactions (Bobrovsky et al., 2008). In order to acquire reliable spectroscopic data *via* an optical ATR-IR fibre probe a procedure to correct the effects of (co-)sine type fringes, which are observed during the fast and exothermic metal

The instrumental setup was identical to the setup described in chapter 3.2. The fibre probe was mounted through a ground neck into a 4-neck round bottom flask, which was subsequently charged with dry solvent and cooled *via* a cooling bath. After a steady temperature in the vessel was reached a background spectrum was recorded, the measurement started and the

Temperature deviations between the sample spectra and the background acquired during dynamic reaction processes represent one of the major challenges in IR spectroscopy at cryogenic temperatures. During these investigations the application of an ATR-IR fibre probe to relatively fast and exothermic metal halogen exchange reactions resulted in spectra heavily overlaid with (co-)sine-type artifacts. The introduced fringes lead to spectral data

To overcome these limitations a correction procedure based on fast Fourier transformation (FFT) was implemented to the data evaluation process. This high pass filter significantly reduced the fringes (attributed to differing thermal expansion coefficients of the probe materials) and rendered it possible to achieve a reliable monitoring of metal halogen

An application of the developed methodology to simple substrates for metal halogen exchange reactions but also to metallation procedures afforded reliable results. Thus, the technique was utilized for a detailed kinetic investigation of a double-sided Halogen Dance

In this multi-step sequence (Fig. 8, A) the first metallation could be shown to proceed faster than the second lithiation towards **3** leading to an accumulation of intermediate **2** (Fig. 8, B). This assumption of the first reaction step exclusively consuming LDA to form **2** and the second Halogen Dance being realized in a next step leading to **3** could be verified by a

(Fig. 8, A) reaction towards 4,4'-dibromo-2,2'-bithiophene (Bobrovsky et al., 2008).

lithium species added. Finally, the respective reactant was rapidly added *via* a syringe.

the selective synthesis of certain substitution patterns (Rappoport & Marek, 2004).

**4.4 Investigations of metal halogen exchange reactions** 

halogen exchange reactions, has been developed.

**4.4.1 Technical details** 

**4.4.2 FFT correction procedure** 

not utilizable for further interpretation.

**4.4.3 Results and discussion** 

exchange reactions at temperatures even below –80 °C.

Fig. 8. Double-sided Halogen Dance reaction; A: reaction scheme; LDA (2.5 equiv.), -40 °C. i: metallation reaction, ii: Halogen Dance reaction, iii: excess of methanol; B: 3D plot of spectra (750 - 1050 cm-1) recorded during reaction progress (0 - 160 s); C: monitoring of Li-species LDA and intermediate **2**; D: intermediate **3** formation as extracted from the spectral data set *via* MCR-ALS algorithm. Figure reproduced from Lumpi et al., 2012.

good agreement of kinetic parameters of LDA consumption and intermediate formation (Fig 8, C).

Due to overlapping absorption bands the MCR-ALS algorithm (chapter 3.3.2.) was applied to the spectral data set prior to analysis in order to compare the consumption of intermediate **2** and the formation of product **3** (Fig 8, D). The results again disclose a good agreement of kinetic data supporting the aforementioned conclusion, being highly valuable for potential application of this reaction (e.g. sequential Halogen Dance).

In summary, a methodology utilizing an FFT based correction procedure for a convenient monitoring of metal halogen exchange (but also metallation) reactions is presented. By applying this technique a mechanistic investigation of a complex double-sided Halogen Dance reaction could be realized.

## **5. Conclusion**

In conclusion, it has been demonstrated that mid-infrared (ATR) spectroscopy utilizing modern optical fibre probes is an effective methodology for in-line monitoring of highly reactive species. Therefore, this non-invasive technique has emerged as a versatile tool for direct reaction monitoring, which was outlined using the example of organometallic intermediates.

Effective Reaction Monitoring of Intermediates

2451–2453, ISSN: 1359-7345

Vol.58, No.7, pp. 804-810, ISSN 1943-3530

No.1, pp. 59-65, ISSN 0030-4921

by ATR-IR Spectroscopy Utilizing Fibre Optic Probes 509

Marziano, I.; Sharp, D. C. A.; Dunn, P. J. & Hailey, P. A. (2000). On-Line Mid-IR

McGarrity, J. F.; Ogle, C. A.; Brich, Z.; Loosli, H. R. (1985b). A rapid-injection (RI) NMR

McGarrity, J. F.; Prodolliet, J. & Smyth, T. (1981). Rapid injection NMR: a simple technique

McGarrity, John F.; Ogle, Craig A. (1985a). High-field proton NMR study of the aggregation

Melling P. J. & Thomson M. (2002). Fiber-optic Probes for Mid-infrared Spectrometry, In:

Minnich, C. B.; Buskens, P.; Steffens, H. C.; Bäuerlein, P. S.; Butvina, L. N.; Küpper, L.;

Mizaikoff, B. & Lendl, B. (2002). Sensor Systems Based on Mid-infrared Transparent Fibers,

Pippel, D. J.; Weisenburger, G. A.; Faibish, N. C. & Beak, P. (2001). Kinetics and Mechanism

Raichlin Y. & Katzir A. (2008). Fiber-Optic Evanescent Wave Spectroscopy in the Middle Infrared. *Applied Spectroscopy*, Vol.62, No.2, pp. 55A-72A, ISSN 1943-3530 Rappoport, Z. & Marek, I. (2004). *The chemistry of organolithium compounds*, Wiley, ISBN 0-

Rathman, T. L. & Bailey, W. F. (2009). Optimization of Organolithium Reactions. *Organic Process Research & Development*, Vol.13, No.2, pp. 144-151, ISSN 1520-586X Rubin, A. E.; Tummala, S.; Both, D. A.; Wang, C. & Delany, E. J. (2006). Emerging

Sun, X. & Collum, D. B. (2000). Lithium Diisopropylamide-Mediated Enolizations: Solvent-

Tauler, R. (1995). Multivariate curve resolution applied to second order data. *Chemometrics and Intelligent Laboratory Systems* Vol.30, pp. 133-146, ISSN 0169-7439

*American Chemical Society*, Vol.107, No.7, 1810-1815, ISSN 0002-7863

*Chemical Society*, Vol.107, No.7, 1805-1810, ISSN 0002-7863

Wiley & Sons Ltd., ISBN 0471988472, Chichester

Wiley & Sons Ltd., ISBN 0471988472, Chichester

*S.)*, Vol.106, No.7, pp. 2794-2810, ISSN 1520-6890

Vol.122, pp. 2452–2458, ISSN 0002-7863

Vol.11, No.1, pp. 94-97, ISSN 1520-586X

4927, ISSN 0002-7863

470-84339-X, Chichester

monitoring on organolithium compounds. *Chemical Communications* Vol.48, pp.

Spectroscopy as a Real-Time Approach in Monitoring Hydrogenation Reactions, *Organic Process Research & Development*, Vol.4, No.5, pp. 357-361, ISSN 1520-586X Mazarevica, G.; Diewok, J.; Baena, J. R.; Rosenberg, E. & Lendl, B. (2004). On-Line

Fermentation Monitoring by Mid-infrared Spectroscopy. *Applied Spectroscopy*,

study of the reactivity of butyllithium aggregates in tetrahydrofuran. *Journal of the* 

for the observation of reactive intermediates. *Organic Magnetic Resonance*, Vol.17,

and complexation of n-butyllithium in tetrahydrofuran. *Journal of the American* 

*Handbook of Vibrational Spectroscopy,* J. M. Chalmers & P. R. Griffiths (Eds.), John

Leitner, W.; Liauw, M. A. & Greiner, L. (2007). Highly Flexible Fibre-Optic ATR-IR Probe for Inline Reaction Monitoring. *Organic Process Research & Development*,

In: *Handbook of Vibrational Spectroscopy,* J. M. Chalmers & P. R. Griffiths (Eds.), John

of the (−)-Sparteine-Mediated Deprotonation of (E)-N-Boc-N-(p-methoxyphenyl)-3 cyclohexylallylamine. *Journal of the American Chemical Society* Vol.123, pp. 4919–

Technologies Supporting Chemical Process R&D and Their Increasing Impact on Productivity in the Pharmaceutical Industry. *Chemical Reviews (Washington, DC, U.* 

Dependent Mixed Aggregation Effects. *Journal of the American Chemical Society*

#### **6. References**


Bakeev, K. A. (Ed.). (2005). *Process Analytical Technology - Spectroscopic tools and* 

Bobrovsky, R.; Hametner, C.; Kalt, W. & Froehlich J. (2008). Selective Halogen Dance

Brancaleon, L.; Bamberg, M. P. & Kollias, N. (2000). Spectral differences between stratum

Brandstetter, M. (2009). Ultrasonically enhanced in-line Attenuated Total Reflection (ATR)

Collum, D. B.; McNeil, A. J. & Ramirez, A. (2007). Lithium Diisopropylamide: Solution

Grunwaldt, J. -D. & Baiker, A. (2005). In situ spectroscopic investigation of heterogeneous

Gupta, L.; Hoepker, A. C.; Singh, K. J. & Collum, D. B. (2008). Lithium Diisopropylamide-

Harrington, J. A. (2004). Introduction to Infrared Fiber Optics, In: *Infrared Fibers and Their Applications*, J. A. Harrington, SPIE Press, Bellingham, ISBN 978-0819452184 Harrington**,** J. A. (2010). Infrared Fibres, In: *Handbook of Optics* (3rd Ed.), M. Bass (Ed.), 12.1-

Jaumot, J.; Gargallo, R.; de Juana, A. & Tauler, R. (2004). A graphical user-friendly interface

Lendl, B. & Mizaikoff, B. (2002). Optical Fibers for Mid-infrared Spectrometry, In: *Handbook* 

Lumpi, D.; Braunshier, C.; Hametner C.; Horkel, E.; Zachhuber, B.; Lendl, B. & Froehlich J.

Lumpi, D.; Wagner, C.; Schoepf, M.; Horkel, E.; Ramer, G.; Lendl, B. & Froehlich J. (2012).

12.13, The McGraw-Hill Companies, ISBN 978-0-07-163314-7

*Chemistry* Vol.57, pp. 6678–6680, ISSN 0022-3263

Sons Ltd., ISBN 0471988472, Chinchester

Publishing Ltd., ISBN 9781405121033, Oxford/UK

Vol.54, No.8, pp. 1175-1182, ISSN 0003-7028

*Edition* Vol.46, pp. 3002 – 3017, ISSN 1521-3773

Vol.7, No.20, 3526-3539, ISSN 1463-9084

Vol.74, pp. 2231–2233, ISSN 0022-3263

*implementation strategies for the Chemical and Pharmaceutical Industries*, Blackwell

Reactions at 5,5'-Dibromo-2,2'-bithiophene. *Heterocycles* Vol.76, No.2, pp. 1249–

corneum and sebaceous molecular components in the mid-IR. *Applied Spectroscopy*,

infrared absorption spectroscopy of suspensions, Diploma Thesis, Institute of Chemical Technologies and Analytics, Vienna University of Technology, Vienna Braunshier, C. ; Hametner, C. ; Froehlich, J. ; Schnoeller, J. & Hutter, H. (2008). Novel

monodisperse PEG-grafted polystyrene resins: synthesis and application in gelphase 13C NMR spectroscopy. *Tetrahedron Letters* Vol.49, pp. 7103–7105, ISSN: 0040-

Kinetics and Implications for Organic Synthesis. *Angewandte Chemie International* 

catalysts and reaction media at high pressure. *Physical Chemistry Chemical Physics*,

Mediated Ortholithiations: Lithium Chloride Catalysis. *Journal of Organic Chemistry*

for MCR-ALS: a new tool for multivariate curve resolution in MATLAB. *Chemometrics and Intelligent Laboratory Systems* Vol.76, pp. 101–110, ISSN 0169-7439 Keegstra, E. M. D.; Zwikker, J. W.; Roest, M. R. & Jenneskens, L. W. (1992). A highly

selective synthesis of monodisperse oligo(ethylene glycols). *Journal of Organic* 

*of Vibrational Spectroscopy,* J. M. Chalmers & P. R. Griffiths (Eds.), John Wiley &

(2009). Convenient multigram synthesis of monodisperse oligo(ethylene glycols): effective reaction monitoring by infrared spectroscopy using an attenuated total reflection fibre optic probe. *Tetrahedron Letters* Vol.50, pp. 6469–6471, ISSN: 0040-

Fibre optic ATR-IR spectroscopy at cryogenic temperatures: in-line reaction

**6. References** 

4039

4039

1259, ISSN: 0385-5414

monitoring on organolithium compounds. *Chemical Communications* Vol.48, pp. 2451–2453, ISSN: 1359-7345


Weymeels, E.; Awad, H.; Bischoff, L.; Mongin, F.; Trécourt, F.; Quéguiner, G. & Marsais, F.

Wu, G. & Huang, M. (2006). Organolithium Reagents in Pharmaceutical Asymmetric

Zhao, P. & Collum, D. B. (2003). Lithium Hexamethyldisilazide/Triethylamine-Mediated

Zogg, A.; Fischer, U. & Hungerbuehler, K. (2004). Identification of kinetic and

*Research & Development*, Vol.10, No.2, pp. 349-353, ISSN 1520-586X

ISSN 1520-6890

No.24, 5795-5806, ISSN 0009-2509

0002-7863

(2005). On the deprotonation of 3,5-dichloropyridine using lithium bases: in situ infrared spectroscopic studies. *Tetrahedron* Vol.61, pp. 3245-3249, ISSN 0040-4020 Wiss, J.; Fleury, C. & Onken, U. (2006). Safety Improvement of Chemical Processes Involving

Azides by Online Monitoring of the Hydrazoic Acid Concentration. *Organic Process* 

Processes. *Chemical Reviews (Washington, DC, U. S.)*, Vol.106, No.7, pp. 2596–2616,

Ketone Enolization: Remarkable Rate Accelerations Stemming from a Dimer-Based Mechanism. *Journal of the American Chemical Society* Vol.125, pp. 4008–4009, ISSN

thermodynamic reaction parameters from online calorimetric and IR-ATR data using a new combined evaluation algorithm. *Chemical Engineering Science*, Vol.59,

## *Edited by Theophile Theophanides*

The present book is a definitive review in the field of Infrared (IR) and Near Infrared (NIR) Spectroscopies, which are powerful, non invasive imaging techniques. This book brings together multidisciplinary chapters written by leading authorities in the area. The book provides a thorough overview of progress in the field of applications of IR and NIR spectroscopy in Materials Science, Engineering and Technology. Through a presentation of diverse applications, this book aims at bridging various disciplines and provides a platform for collaborations among scientists.

Infrared Spectroscopy - Materials Science, Engineering and Technology

Infrared Spectroscopy

Materials Science, Engineering

and Technology

*Edited by Theophile Theophanides*

Photo by LuFeeTheBear / iStock