**2.1 Dielectric photonic crystals**

It has been demonstrated that the geometry of PhCs strongly affect whether TE or TM band gap will be generated. In particular, when the crystal is formed by isolated regions of high dielectric, the TM gap is favored; when the crystal is formed by connected regions of high dielectric, the TE gap is favored. Hence, the geometry plays a crucial role. In regards, when 2D photonic crystals are considered, only five possible geometrical arrangements exist. These are called Bravais lattices. Considering the complementary conditions for the realization of a TE/TM gaps, the best Bravais lattice to maximize the probability of obtaining a full band gap is the hexagonal one. This kind of Bravais lattice posses the highest possible rotational symmetry in two dimensions. In view of this information, we have instead chosen the square photonic crystal (Galli et al., 2002), namely a configuration not supporting a full band gap for isotropic dielectric materials (Wang et al., 2005; Proietti Zaccaria, 2008b). In this manner we expect to enhance the different results in terms of full band gap rising either from a dielectric crystal or a metallic one.

### **2.1.1 Full band gap in dielectric 2D photonic crystals**

258 Photonic Crystals – Innovative Systems, Lasers and Waveguides

the existing free electrons. This kind of mode is known as Surface Plasmon Polariton (SPP). Photonic crystals, as a translational modulation of the refractive index, have already been playing a very crucial role in plasmonics. In fact, they can provide the missing wave vector for the coupling between photons and free electrons of a metal layer (Raether, 1988). Here

a. two dimensional (2D) metallic photonic crystal for maximizing the optical band gap; b. metallic photonic crystal structures for Surface Enhanced Raman Spectroscopy (SERS); c. few molecules detection through super-hydrophobic crystals (De Angelis et al., 2011). The first topic concerns the fundamentals of SPP and PhCs. No particular application will be suggested, but mostly we will focus on the theory behind the generation of SPP inside a metallic photonic crystal. The next two topics are, on the other hand, strictly related to applications. We will concentrate our attention on Raman spectroscopy, as a very important tool for the investigation of the optical properties of many kinds of samples, such as semiconductors or proteins. In particular, metallic ordered (periodic) structures will be used either as artificial SERS substrates or as combiners of SERS and super-hydrophobic effect for few molecules detection. This description will offer a general overview of the important functions that PhCs can hold in Plasmonics and how we could start thinking more

we have chosen to face three situations relating PhCs and Plasmonics:

**2. Two dimensional metallic PhCs for maximum full band gap** 

One of the main issues when dealing with two dimensional PhCs is the maximum full band gap that the crystal can provide. Specifically, in 2D the crystal manifests two non degenerative polarizations known as TE and TM. The former refers to solutions with the electric field *in* the plane of symmetry of the crystal, the latter with the magnetic field in the same plane. In order to produce a *full* band gap, both TE and TM must provide an optical gap in the *same* spectral region. This is not an easy task and different solutions have been proposed (Joannopoulos et al., 1995), however mostly referring to dielectric PhCs. Here we shall examine how metallic PhCs can improve the chance of a common band gap between

It has been demonstrated that the geometry of PhCs strongly affect whether TE or TM band gap will be generated. In particular, when the crystal is formed by isolated regions of high dielectric, the TM gap is favored; when the crystal is formed by connected regions of high dielectric, the TE gap is favored. Hence, the geometry plays a crucial role. In regards, when 2D photonic crystals are considered, only five possible geometrical arrangements exist. These are called Bravais lattices. Considering the complementary conditions for the realization of a TE/TM gaps, the best Bravais lattice to maximize the probability of obtaining a full band gap is the hexagonal one. This kind of Bravais lattice posses the highest possible rotational symmetry in two dimensions. In view of this information, we have instead chosen the square photonic crystal (Galli et al., 2002), namely a configuration not supporting a full band gap for isotropic dielectric materials (Wang et al., 2005; Proietti Zaccaria, 2008b). In this manner we expect to enhance the different results in terms of full

intensively of PhCs realized with metallic materials.

TE and TM with respect to a dielectric structure.

band gap rising either from a dielectric crystal or a metallic one.

**2.1 Dielectric photonic crystals**

The chosen dielectric PhC is made of circular columns of high refractive index material, namely silicon (n=3.6) surrounded by air (n=1). The radius of the columns is r=300nm and lattice period P=1m. As expected, only TM polarization shows zero transmission regions, in particular three band gaps below 1m-1 are shown in Fig. 2.1. On the other hand, TE does not sustain any band gap. These results remain true even increasing the columns dielectric value or changing the columns radius. No full band gap is then found for this kind of structure.

Fig. 2.1. TM band structure for a 2D square silicon columns photonic crystal. Three band gaps are shown.

#### **2.2 Metallic photonic crystals**

The behavior of 2D metallic photonic crystals is fundamentally different from what expected by 2D dielectric photonic crystals (Zhao et al., 2009; Sakoda et al., 2001; Ito & Sakoda, 2001). In fact, the use of metallic dispersive materials strongly modifies the light behavior in periodic structures, both for TM and TE polarizations. In particular, TM polarization shows a cut-off frequency c and no modes are found below it. This implies the existence of a TM gap below c. From a physical point of view this is related to the existence of free electrons in metallic materials. Similarly, TE polarization shows a behavior which is absent in dielectric 2D photonic crystals. In fact, in specific range of frequencies, metallic photonic crystals show TE polaritonic band gap close to the plasma frequency. Physically it is associated to the creation of surface plasmon polaritons on the metallic columns of the crystal. These peculiarities of metallic PhCs have the merit of increasing the chance of a full band gap also for square crystals. Hence, maintaining the same geometrical configuration as in the previous section, namely a square structure of columns in air, we have numerically analyzed the band gap formation when metallic columns are considered instead of dielectric ones.

#### **2.2.1 Full band gap in metallic 2D photonic crystals: Drude vs. Lorentz model**

We start by considering the Drude model (Rakic´ et al., 1998) to describe the metallic parts of the crystal:

$$\varepsilon(\phi) = 1 - \frac{f\_0 o\_p^2}{\alpha^2 - i o \, \Gamma\_0} \tag{2.1}$$

Photonic Crystals for Plasmonics: From Fundamentals to Superhydrophobic Devices 261

Fig. 2.3. TE transmission spectra for a Drude square PhC. Different lines (colors) correspond

These results confirm that the use of metallic PhCs increases the chance of a full band gap even for square-like Bravais lattices. However we have to keep in mind that the modeling leading to the previous results is based on a simplified Drude description. Hence, it is now important to asset the role that possible Lorentz contributes could play in the overall gap calculation. By moving to the Drude-Lorentz description from the Drude equation (1):

0

modify the transmission spectrum compared to a simple Drude description.

 

() 1

**2.2.2 Photonic crystal with realistic metallic materials** 

case, will be enough to determine the existence of a full band gap.

 

obtained as shown in Fig. 2.4.

possible full band gaps.

*j j j*

 

> 

(2.2)

2 2

*i i*

2 2 2 0 1

and assuming k=1, *f1*=1, 1=800nm and 1=0.1p a different transmission spectrum is

We can see that considering also the Lorentz term strongly modifies the transmission spectrum to the extent that for the present case the gap around 1.1m is suppressed. It is then of fundamental importance to compare the Drude and the Drude-Lorentz models when real materials, such as Ag, Au or Al, are considered. In fact, their description implies taking into account a number of Lorentz peaks which, as we have seen, might strongly

We will consider two materials: Ag and Al. The former is usually considered for plasmonic applications in the visible range whereas the latter can play a very interesting role in the UV range. We shall calculate their transmission spectra both for TM and TE in order to identify

First we will assume a Drude description. In case of Al, the experimental parameters to build up the model are p= 14.98eV, 0=0.047eV and *f0*=0.523 (Rakic´ et al., 1998). Under these conditions the transmission/absorption spectra, both for TM and TE polarization, are shown in Fig. 2.5. We have considered two incident angles, 0o and 45o which, in the present

*f f*

*k p j p*

to different incident angles (blue: 0o; green: 45o).

where p is the plasma frequency, 0 is the damping constant and *f0* is the oscillator strength. We recall that this model is a simplified version of the realistic Drude-Lorentz model, in fact it does not describe the material resonances (absorption) of the metal. Nevertheless, the Drude model is commonly used ought to the reduced calculation complexity. Furthermore, it is often assumed that for noble metals such as Ag or Au, no difference is expected between the Drude and the Drude-Lorentz description in the visible range. We will show that this assumption is not always justified.

By choosing p=2c/P, 0=0.01p and *f0*=1 in Eq. (1), a square PhC with columns of radius r=0.472P and TM polarization, the transmission spectrum of Fig. 2.2 was obtained. The figure shows different curves each associated to a different incident angle. In fact, in order to determine a band gap, two possible ways can be chosen: i) band structure calculation; b) transmission spectrum calculation. Considering the dispersive properties of the metal, the band structure calculation presents some difficulties which make the transmission spectra method the simplest way for the calculation of band gaps. However, when this method is considered, the light must impinge on the PhC under a number of angles, namely the angles covering the irreducible Brillouin zone (Zhao et al., 2009). In our case it means to look at the range [0-45] degrees. Zero transmission regions which are common to all the simulations can be identified as band gaps. For clarity we have chosen only 0 and 45 degrees (however, for the present case, this choice will not affect the identification of optical gaps).

Fig. 2.2. TM transmission spectra for a Drude square PhC. Different lines (colors) correspond to different incident angles (blue: 0o; green: 45o). Two band gaps are found.

Fig. 2.2 clearly shows two band gaps. In particular the lowest frequency one (above 1.35m) is associated to the cut-off frequency c as confirmed by the band structure calculation in (Sakoda et al., 2001).

Interesting, when TE polarization is considered, a wide band gap is obtained as shown in Fig. 2.3. This result stresses the difference with similar geometries having only dielectric parts where TE does not provide any gap. In fact, if the TE and TM transmission spectra are overlapped, a full band gap is found.

where p is the plasma frequency, 0 is the damping constant and *f0* is the oscillator strength. We recall that this model is a simplified version of the realistic Drude-Lorentz model, in fact it does not describe the material resonances (absorption) of the metal. Nevertheless, the Drude model is commonly used ought to the reduced calculation complexity. Furthermore, it is often assumed that for noble metals such as Ag or Au, no difference is expected between the Drude and the Drude-Lorentz description in the visible

By choosing p=2c/P, 0=0.01p and *f0*=1 in Eq. (1), a square PhC with columns of radius r=0.472P and TM polarization, the transmission spectrum of Fig. 2.2 was obtained. The figure shows different curves each associated to a different incident angle. In fact, in order to determine a band gap, two possible ways can be chosen: i) band structure calculation; b) transmission spectrum calculation. Considering the dispersive properties of the metal, the band structure calculation presents some difficulties which make the transmission spectra method the simplest way for the calculation of band gaps. However, when this method is considered, the light must impinge on the PhC under a number of angles, namely the angles covering the irreducible Brillouin zone (Zhao et al., 2009). In our case it means to look at the range [0-45] degrees. Zero transmission regions which are common to all the simulations can be identified as band gaps. For clarity we have chosen only 0 and 45 degrees (however, for the present case, this choice will not affect the identification of

Fig. 2.2. TM transmission spectra for a Drude square PhC. Different lines (colors) correspond

Fig. 2.2 clearly shows two band gaps. In particular the lowest frequency one (above 1.35m) is associated to the cut-off frequency c as confirmed by the band structure calculation in

Interesting, when TE polarization is considered, a wide band gap is obtained as shown in Fig. 2.3. This result stresses the difference with similar geometries having only dielectric parts where TE does not provide any gap. In fact, if the TE and TM transmission spectra are

to different incident angles (blue: 0o; green: 45o). Two band gaps are found.

range. We will show that this assumption is not always justified.

optical gaps).

(Sakoda et al., 2001).

overlapped, a full band gap is found.

Fig. 2.3. TE transmission spectra for a Drude square PhC. Different lines (colors) correspond to different incident angles (blue: 0o; green: 45o).

These results confirm that the use of metallic PhCs increases the chance of a full band gap even for square-like Bravais lattices. However we have to keep in mind that the modeling leading to the previous results is based on a simplified Drude description. Hence, it is now important to asset the role that possible Lorentz contributes could play in the overall gap calculation. By moving to the Drude-Lorentz description from the Drude equation (1):

$$\varepsilon(o) = 1 - \frac{f\_0 o\_p^2}{o^2 - i o \Gamma\_0} + \sum\_{j=1}^k \frac{f\_j o\_p^2}{\left(o\_j^2 - o^2\right) + i o \Gamma\_j} \tag{2.2}$$

and assuming k=1, *f1*=1, 1=800nm and 1=0.1p a different transmission spectrum is obtained as shown in Fig. 2.4.

We can see that considering also the Lorentz term strongly modifies the transmission spectrum to the extent that for the present case the gap around 1.1m is suppressed. It is then of fundamental importance to compare the Drude and the Drude-Lorentz models when real materials, such as Ag, Au or Al, are considered. In fact, their description implies taking into account a number of Lorentz peaks which, as we have seen, might strongly modify the transmission spectrum compared to a simple Drude description.

#### **2.2.2 Photonic crystal with realistic metallic materials**

We will consider two materials: Ag and Al. The former is usually considered for plasmonic applications in the visible range whereas the latter can play a very interesting role in the UV range. We shall calculate their transmission spectra both for TM and TE in order to identify possible full band gaps.

First we will assume a Drude description. In case of Al, the experimental parameters to build up the model are p= 14.98eV, 0=0.047eV and *f0*=0.523 (Rakic´ et al., 1998). Under these conditions the transmission/absorption spectra, both for TM and TE polarization, are shown in Fig. 2.5. We have considered two incident angles, 0o and 45o which, in the present case, will be enough to determine the existence of a full band gap.

Photonic Crystals for Plasmonics: From Fundamentals to Superhydrophobic Devices 263

Fig. 2.5. TM/TE absorption/transmission spectra for a Drude square PhC. Aluminium is considered. Lattice constant P=1m, cylinder radius=300nm. The source incident angle was

Moving now to PhCs based on Ag, the results for the Drude and Drude-Lorentz models are shown in Figs. 2.8 and 2.9, respectively. The optical parameters are p=9.01eV, 0=0.048eV, *f0*=0.845, 1=0.816eV, 1=3.866eV and *f1*=0.065 (Rakic´ et al., 1998). In this case the spectrum range was chosen from 50nm to 2m in order to exploit the band gap behavior from the UV

The Drude model for Ag shows a full band gap between 710nm and 750nm in a way similar to the Al structure (we recall that for both the materials we have used the same geometry). Once again, differently from square dielectric periodic structures, metallic square PhCs can provide full band gaps. Similarly to the Al case, the Drude-Lorentz model registers an increase in the absorption in the frequency region below 1.0m. In particular, the three lowest wavelengths absorption peaks shown in the Drude-Lorentz model of Fig. 2.9

Furthermore, by observing the TE spectra for the silver PhC of Fig. 2.9, it can be noticed that absorptive peaks inside zero transmission regions are shown below 400nm, namely close to the silver plasma frequency of 137nm. This is the fingerprint of polaritonic gaps created by surface plasmon polaritons on the surface of the silver columns. On the other hand, TM spectra show a zero transmission region above 1.4m which can be associated to the cut-off

correspond to three well defined resonant peaks of silver, as confirmed by Fig. 2.10.

0o and 45o.

to the IR region.

frequency c.

Fig. 2.4. TM transmission spectra of a square metallic PhC with radius r=0.472·P and material properties defined by (continue line – Drude model) p= 2c/P and damping constant =0.01·p (0o incident angle of Fig. 2.2); (dashed line – Drude-Lorentz model) p= 2c/P and damping constant 1=10· for the resonance at 800nm. Incident angle of 0o was assumed.

The first important result to be noticed is the existence of a full band gap in the range 700- 750nm. This achievement is consistent with the full band gaps obtained with the arbitrary Drude model introduced in Figs. 2.2 and 2.3. Furthermore, it is interesting to observe the behavior of the absorption both for TM and TE. In fact, the former shows no peaks in any band gap region whereas the latter has a smooth behavior, inside the gap regions, only for high wavelengths. It is explained recalling that only TE polarization shows a peculiar band gap which originates from the creation of surface plasmon polaritons on the metallic cylinders of the crystal, namely absorption peaks have to be observed in the gaps. The frequency region supporting the surface plasmon polariton modes starts roughly just below the plasma frequency of the metal, which in case of Al is 14.98eV=83nm, and its width depends on the geometry of the crystal.

When the Drude-Lorentz model is considered the transmission for both TM and TE is shown in Fig. 2.6. By comparison with Fig. 2.5 not substantial differences in terms of band gaps can be noticed, neither for TM nor for TE. In fact, the 700-750nm full band gap is found also in the Drude-Lorentz description. The only noticeable change is the increase of the absorption with respect to the transmission mainly below 1.0m ought to the differences in the refractive index profile between the Drude and the Drude-Lorentz model of Al. In fig. 2.7 the two models are plotted. A strong absorptive peak around 800nm is shown in the Drude-Lorentz model.

Fig. 2.4. TM transmission spectra of a square metallic PhC with radius r=0.472·P and material properties defined by (continue line – Drude model) p= 2c/P and damping constant =0.01·p (0o incident angle of Fig. 2.2); (dashed line – Drude-Lorentz model) p= 2c/P and damping constant 1=10· for the resonance at 800nm. Incident angle of

The first important result to be noticed is the existence of a full band gap in the range 700- 750nm. This achievement is consistent with the full band gaps obtained with the arbitrary Drude model introduced in Figs. 2.2 and 2.3. Furthermore, it is interesting to observe the behavior of the absorption both for TM and TE. In fact, the former shows no peaks in any band gap region whereas the latter has a smooth behavior, inside the gap regions, only for high wavelengths. It is explained recalling that only TE polarization shows a peculiar band gap which originates from the creation of surface plasmon polaritons on the metallic cylinders of the crystal, namely absorption peaks have to be observed in the gaps. The frequency region supporting the surface plasmon polariton modes starts roughly just below the plasma frequency of the metal, which in case of Al is 14.98eV=83nm, and its width

When the Drude-Lorentz model is considered the transmission for both TM and TE is shown in Fig. 2.6. By comparison with Fig. 2.5 not substantial differences in terms of band gaps can be noticed, neither for TM nor for TE. In fact, the 700-750nm full band gap is found also in the Drude-Lorentz description. The only noticeable change is the increase of the absorption with respect to the transmission mainly below 1.0m ought to the differences in the refractive index profile between the Drude and the Drude-Lorentz model of Al. In fig. 2.7 the two models are plotted. A strong absorptive peak around 800nm is shown in the

0o was assumed.

depends on the geometry of the crystal.

Drude-Lorentz model.

Fig. 2.5. TM/TE absorption/transmission spectra for a Drude square PhC. Aluminium is considered. Lattice constant P=1m, cylinder radius=300nm. The source incident angle was 0o and 45o.

Moving now to PhCs based on Ag, the results for the Drude and Drude-Lorentz models are shown in Figs. 2.8 and 2.9, respectively. The optical parameters are p=9.01eV, 0=0.048eV, *f0*=0.845, 1=0.816eV, 1=3.866eV and *f1*=0.065 (Rakic´ et al., 1998). In this case the spectrum range was chosen from 50nm to 2m in order to exploit the band gap behavior from the UV to the IR region.

The Drude model for Ag shows a full band gap between 710nm and 750nm in a way similar to the Al structure (we recall that for both the materials we have used the same geometry). Once again, differently from square dielectric periodic structures, metallic square PhCs can provide full band gaps. Similarly to the Al case, the Drude-Lorentz model registers an increase in the absorption in the frequency region below 1.0m. In particular, the three lowest wavelengths absorption peaks shown in the Drude-Lorentz model of Fig. 2.9 correspond to three well defined resonant peaks of silver, as confirmed by Fig. 2.10.

Furthermore, by observing the TE spectra for the silver PhC of Fig. 2.9, it can be noticed that absorptive peaks inside zero transmission regions are shown below 400nm, namely close to the silver plasma frequency of 137nm. This is the fingerprint of polaritonic gaps created by surface plasmon polaritons on the surface of the silver columns. On the other hand, TM spectra show a zero transmission region above 1.4m which can be associated to the cut-off frequency c.

Photonic Crystals for Plasmonics: From Fundamentals to Superhydrophobic Devices 265

value is approached, differences between the two models become appreciable. This is

Fig. 2.8. TM/TE absorption/transmission spectra for the Drude square PhC. Silver is considered. Lattice constant P=1m, cylinder radius=300nm, incident angles 0o and 45o.

models start rising that implies the failure of the Drude model at low frequencies.

**3. Metallic PhCs for surface enhanced Raman spectroscopy** 

In conclusion, we have shown that metallic PhCs can provide a full band gap not possible for analogues dielectric PhCs. This characteristic is related to the dispersive properties of the metallic parts of the crystal. Furthermore, we have shown that the Drude and Drude-Lorentz model provide the same results as long as the frequency range is far away from the metal plasma frequency. In fact, when approaching p, discrepancies between the two

In recent years, plasmonics based sensor device such as surface enhanced Raman spectroscopy (SERS) has attracted lots of attention to the scientific community worldwide. SERS is a technique using which an increase in optical signal of the molecule situated in the vicinity of nano-metallic surface can be observed when electromagnetic light is being impinged on it (Haynes et al., 2005). There are many techniques such as electron beam lithography (EBL) (Das et al., 2009; Kahl et al., 1998), colloidal technique (Kneipp et al., 1997; Nie & Emory, 1997; Coluccio et al., 2009), metal island film (Constantino et al., 2001), etc. by which SERS substrate can be fabricated. There is always a competition regarding the quality of the nanostructure, fabrication area, time and cost. EBL is the most efficient technique to

related to the typical absorption peaks of Ag located before 400nm.

Fig. 2.6. TM/TE absorption/transmission spectra for a Drude-Lorentz square PhC. Aluminium is considered. Lattice constant P=1m, cylinder radius=300nm, incident angles 0o and 45o.

Fig. 2.7. Real (continuous/blue line) and imaginary (dashed/green line) parts of Al refractive index for the Drude and the Drude-Lorentz models.

Finally a consideration about the Drude and the Drude-Lorentz models for the Ag PhC. The simulations have shown that when moving to the Drude-Lorentz model the transmission spectrum remains similar to the Drude counterpart for energy far away from p. When this

Fig. 2.6. TM/TE absorption/transmission spectra for a Drude-Lorentz square PhC.

Fig. 2.7. Real (continuous/blue line) and imaginary (dashed/green line) parts of Al

Finally a consideration about the Drude and the Drude-Lorentz models for the Ag PhC. The simulations have shown that when moving to the Drude-Lorentz model the transmission spectrum remains similar to the Drude counterpart for energy far away from p. When this

refractive index for the Drude and the Drude-Lorentz models.

0o and 45o.

Aluminium is considered. Lattice constant P=1m, cylinder radius=300nm, incident angles

value is approached, differences between the two models become appreciable. This is related to the typical absorption peaks of Ag located before 400nm.

Fig. 2.8. TM/TE absorption/transmission spectra for the Drude square PhC. Silver is considered. Lattice constant P=1m, cylinder radius=300nm, incident angles 0o and 45o.

In conclusion, we have shown that metallic PhCs can provide a full band gap not possible for analogues dielectric PhCs. This characteristic is related to the dispersive properties of the metallic parts of the crystal. Furthermore, we have shown that the Drude and Drude-Lorentz model provide the same results as long as the frequency range is far away from the metal plasma frequency. In fact, when approaching p, discrepancies between the two models start rising that implies the failure of the Drude model at low frequencies.

### **3. Metallic PhCs for surface enhanced Raman spectroscopy**

In recent years, plasmonics based sensor device such as surface enhanced Raman spectroscopy (SERS) has attracted lots of attention to the scientific community worldwide. SERS is a technique using which an increase in optical signal of the molecule situated in the vicinity of nano-metallic surface can be observed when electromagnetic light is being impinged on it (Haynes et al., 2005). There are many techniques such as electron beam lithography (EBL) (Das et al., 2009; Kahl et al., 1998), colloidal technique (Kneipp et al., 1997; Nie & Emory, 1997; Coluccio et al., 2009), metal island film (Constantino et al., 2001), etc. by which SERS substrate can be fabricated. There is always a competition regarding the quality of the nanostructure, fabrication area, time and cost. EBL is the most efficient technique to

Photonic Crystals for Plasmonics: From Fundamentals to Superhydrophobic Devices 267

porous alumina (APA) which could be a trade-off between the two above limits such as

In the present work, APA substrate, having hexagonal periodicity, has been fabricated. The fabricated APA substrates were used as templates for the preparation of nanopatterned gold surfaces, obtained by gold film deposition of ~25 nm thickness covering the APA features. Using this technique, we are able to achieve reproducible SERS substrates with the wall thickness and pore diameter down to 40 and 60 nm, respectively. It is noticeable that the substrate shows very efficient SERS signal even in presence of intrinsically fluorescent molecules. The substrate surface morphology was characterized by both atomic force microscopy (AFM) and scanning electron microscopy (SEM), and the performance as a SERS substrate was tested with cresyl violet (CV). These substances fluoresce at distinct wavelengths in the visible spectrum from red to violet region. Even if it is known that nanometallic surface acts as a fluorescence quenching substrate (Dulkeith et al., 2002), it is the first time, in our knowledge, that the large area SERS substrate on APA template was

In order to provide the surfaces with plasmon functionality, gold was thermally evaporated from a tungsten boat onto the APA substrates (APA), starting the deposition at a base chamber pressure of 2.0x10-6 mbar and proceeding at a 0.5 Å/s until a final total thickness of ~25 nm was reached. During deposition the sample holder was rotated at 1 RPM in order to improve the uniformity and the morphological quality of the gold layer. The substance of interest was deposited over resulting gold-coated APA substrate (termed 'AuAPA') using chemisorption technique. In this process, the substrate was dipped in a solution containing the molecule of interest. After incubation, the substrate was removed from the solution, gently rinsed to remove excess molecules those are not attached directly to the metal surface. Thereafter, the samples were dried in N2 flow, and finally stored in a desiccator before SERS measurements. Cresyl violet (CV) dyes was tested for SERS probes with the

Fig. 3.1 shows the schematic picture of honey-comb structures, the SEM and AFM image of AuAPA SERS substrate. AFM and SEM images show the clear formation of SERS honey-

CV (Fig. 3.2a for optical absorption spectrum and molecular structure) is an organic dye molecule intensively used in biology and medicine for histological stain. It is an effective stain applied for highlighting acidic components of tissues and is commonly used for nerve tissue sections. SERS measurements were carried out for CV deposited on AuAPA SERS substrate at different positions in the range of 300-1400 cm-1. Raman measurements were also performed after deposition of a 2 l drop of CV on a flat non-patterned silicon wafer substrate (see Fig. 3.2a) which can be considered as a positive control sample. The Raman spectrum of CV on Si surface is relatively featureless, with an exponentially increasing fluorescence background and a single characteristic peak of CV at ~591 cm-1 with low intensity. In the inset of Fig. 3.2a, the zoomed Raman spectrum of CV in the range of 400-650

In Fig. 3.2a, SERS substrate background measurement (no dye on AuAPA pattern) was carried out and shown in the inset of Fig. 3.2a, which shows a flat background response

comb structures. The gold-coating over APA template is fixed to the 25 nm.

large fabrication area and SERS enhancement factor.

employed for SERS analysis in such detail.

concentration in the order of 10-6 M.

cm-1 is also shown.

Fig. 2.9. TM/TE absorption/transmission spectra for the Drude-Lorentz square PhC. Silver is considered. Lattice constant P=1m, cylinder radius=300nm, incident angles 0o and 45o.

Fig. 2.10. Real (continuous/blue line) and imaginary (dashed/green line) parts of Ag refractive index for the Drude-Lorentz model. Three resonant peaks are shown in the UV region. Silver plasma frequency p=9.04eV~137nm.

reproduce the nano-patterned SERS substrates but faces a critical limitations when it is about large area fabrication whereas colloidal technique is well known for large area metal deposition but showing minimum reproducibility. In the past, Zhang et al. (2006) reported a new technique called nanosphere lithography in which shadow evaporation through selfassembled arrays of polystyrene nanosphere was used to fabricate a large area SERS substrate. Herein, we propose a new way to fabricate a SERS substrate based on anodic

Fig. 2.9. TM/TE absorption/transmission spectra for the Drude-Lorentz square PhC. Silver is considered. Lattice constant P=1m, cylinder radius=300nm, incident angles 0o and 45o.

Fig. 2.10. Real (continuous/blue line) and imaginary (dashed/green line) parts of Ag refractive index for the Drude-Lorentz model. Three resonant peaks are shown in the UV

reproduce the nano-patterned SERS substrates but faces a critical limitations when it is about large area fabrication whereas colloidal technique is well known for large area metal deposition but showing minimum reproducibility. In the past, Zhang et al. (2006) reported a new technique called nanosphere lithography in which shadow evaporation through selfassembled arrays of polystyrene nanosphere was used to fabricate a large area SERS substrate. Herein, we propose a new way to fabricate a SERS substrate based on anodic

region. Silver plasma frequency p=9.04eV~137nm.

porous alumina (APA) which could be a trade-off between the two above limits such as large fabrication area and SERS enhancement factor.

In the present work, APA substrate, having hexagonal periodicity, has been fabricated. The fabricated APA substrates were used as templates for the preparation of nanopatterned gold surfaces, obtained by gold film deposition of ~25 nm thickness covering the APA features. Using this technique, we are able to achieve reproducible SERS substrates with the wall thickness and pore diameter down to 40 and 60 nm, respectively. It is noticeable that the substrate shows very efficient SERS signal even in presence of intrinsically fluorescent molecules. The substrate surface morphology was characterized by both atomic force microscopy (AFM) and scanning electron microscopy (SEM), and the performance as a SERS substrate was tested with cresyl violet (CV). These substances fluoresce at distinct wavelengths in the visible spectrum from red to violet region. Even if it is known that nanometallic surface acts as a fluorescence quenching substrate (Dulkeith et al., 2002), it is the first time, in our knowledge, that the large area SERS substrate on APA template was employed for SERS analysis in such detail.

In order to provide the surfaces with plasmon functionality, gold was thermally evaporated from a tungsten boat onto the APA substrates (APA), starting the deposition at a base chamber pressure of 2.0x10-6 mbar and proceeding at a 0.5 Å/s until a final total thickness of ~25 nm was reached. During deposition the sample holder was rotated at 1 RPM in order to improve the uniformity and the morphological quality of the gold layer. The substance of interest was deposited over resulting gold-coated APA substrate (termed 'AuAPA') using chemisorption technique. In this process, the substrate was dipped in a solution containing the molecule of interest. After incubation, the substrate was removed from the solution, gently rinsed to remove excess molecules those are not attached directly to the metal surface. Thereafter, the samples were dried in N2 flow, and finally stored in a desiccator before SERS measurements. Cresyl violet (CV) dyes was tested for SERS probes with the concentration in the order of 10-6 M.

Fig. 3.1 shows the schematic picture of honey-comb structures, the SEM and AFM image of AuAPA SERS substrate. AFM and SEM images show the clear formation of SERS honeycomb structures. The gold-coating over APA template is fixed to the 25 nm.

CV (Fig. 3.2a for optical absorption spectrum and molecular structure) is an organic dye molecule intensively used in biology and medicine for histological stain. It is an effective stain applied for highlighting acidic components of tissues and is commonly used for nerve tissue sections. SERS measurements were carried out for CV deposited on AuAPA SERS substrate at different positions in the range of 300-1400 cm-1. Raman measurements were also performed after deposition of a 2 l drop of CV on a flat non-patterned silicon wafer substrate (see Fig. 3.2a) which can be considered as a positive control sample. The Raman spectrum of CV on Si surface is relatively featureless, with an exponentially increasing fluorescence background and a single characteristic peak of CV at ~591 cm-1 with low intensity. In the inset of Fig. 3.2a, the zoomed Raman spectrum of CV in the range of 400-650 cm-1 is also shown.

In Fig. 3.2a, SERS substrate background measurement (no dye on AuAPA pattern) was carried out and shown in the inset of Fig. 3.2a, which shows a flat background response

Photonic Crystals for Plasmonics: From Fundamentals to Superhydrophobic Devices 269

observed in the Fig. 3.2b it is most probable that the CV is oriented in such a way that the – N-H2 group is closer to the gold film, leading to the strong Raman scattering and, consequently, higher Raman signal. Additionally, in Fig. 3.2b, the high exponential background appeared in Fig. 3.2a for the CV on silicon substrate has disappeared. Hence, significant fluorescence quenching is also illustrated, as already observed in the past for gold nanoparticles (Dulkeith et al., 2002). SERS substrate background measurement (no dye on AuAPA pattern) was carried out and found a flat background response without any

Fig. 3.3. a) The overall honey-comb structure. b) The overall electric field distribution in x-y plane, in this case, the polarization of incident light is along x-direction; c) absolute value of the electric field profile along the highlighted horizontal line. The excitation light is = 633

In order to demonstrate the enhanced electric field due to honey-comb nanostructure, we have also performed a series of simulations with the intent of calculating the near field electromagnetic distribution on AuAPA substrates. This kind of analysis can, in fact, provide information on the level of excitation of any sample deposited on a SERS substrate. In Fig. 3.3, simulations of the electric field due to the LSP generation on SERS substrates at z=0 nm (i.e. on the front surface of substrate) using CST software are presented. In particular, a 25 nm thick golden slab showing honeycomb pattern with periodicity of 100 nm, and sub-wavelength air holes (Lezec et al., 2004) with diameter of 60 nm, respectively, was simulated. The source light was fixed to 633 nm with linear polarization along the xaxis. The overall simulated structure is shown in Fig. 3.3a. For all the periodicity patterns, the results show an electric field strongly affected by the material discontinuity. In fact, the absolute value of the total electric field shows some nodes at the gold-air interface. The xpolarization creates anti-phase hot spots along the x-axis which determines the abrupt change of the electric field as in Fig. 3.3b. The profile, shown in Fig. 3.3c, gives even a better image of this behavior. The sharp doubled horns existing inside the holes are due to the

nm on gold surface with dielectric function real= 9.83 and imaginary= 1.97.

Raman band at all.

without any Raman band at all. Various measurements were carried out on this SERS AuAPA substrate at different points, giving identical Raman vibrational frequency of CV.

Fig. 3.1. a) Illustration of Honey-comb structure; SEM and AFM images after gold-coating of 25 nm are shown in Fig. b) and c).

Fig. 3.2. a) Cresyl violet deposited on Si substrate, showing high fluorescence background. Molecular structure, optical absorption spectrum and zoomed Raman spectrum of CV in the range of 400-650 cm-1 are shown in the inset of the same Fig.; b) SERS spectrum of CV is also illustrated, showing the CV spectrum due to an efficient fluorescence quenching by honey-comb structure.

SERS spectra of the CV molecule, performed on the AuAPA substrates with pore diameter (~60 nm) and wall thickness (~40 nm) (Fig. 3.2b), the characteristic vibrational bands of CV are observed (Vogel et al., 2000). Intense Raman bands centred at around 348, 591, 675 and 1186 cm-1 can be attributed to the out of plane sceleton deformation, combination of in-plane N-H2 and ring bending, ring deformation and combination of N-H2 rocking and C-Hx rocking, respectively (Vogel et al., 2000; Kudelski, 2005). From the vibrational bands

without any Raman band at all. Various measurements were carried out on this SERS AuAPA substrate at different points, giving identical Raman vibrational frequency of CV.

Fig. 3.1. a) Illustration of Honey-comb structure; SEM and AFM images after gold-coating of

Fig. 3.2. a) Cresyl violet deposited on Si substrate, showing high fluorescence background. Molecular structure, optical absorption spectrum and zoomed Raman spectrum of CV in the range of 400-650 cm-1 are shown in the inset of the same Fig.; b) SERS spectrum of CV is also illustrated, showing the CV spectrum due to an efficient fluorescence quenching by

SERS spectra of the CV molecule, performed on the AuAPA substrates with pore diameter (~60 nm) and wall thickness (~40 nm) (Fig. 3.2b), the characteristic vibrational bands of CV are observed (Vogel et al., 2000). Intense Raman bands centred at around 348, 591, 675 and 1186 cm-1 can be attributed to the out of plane sceleton deformation, combination of in-plane N-H2 and ring bending, ring deformation and combination of N-H2 rocking and C-Hx rocking, respectively (Vogel et al., 2000; Kudelski, 2005). From the vibrational bands

25 nm are shown in Fig. b) and c).

honey-comb structure.

observed in the Fig. 3.2b it is most probable that the CV is oriented in such a way that the – N-H2 group is closer to the gold film, leading to the strong Raman scattering and, consequently, higher Raman signal. Additionally, in Fig. 3.2b, the high exponential background appeared in Fig. 3.2a for the CV on silicon substrate has disappeared. Hence, significant fluorescence quenching is also illustrated, as already observed in the past for gold nanoparticles (Dulkeith et al., 2002). SERS substrate background measurement (no dye on AuAPA pattern) was carried out and found a flat background response without any Raman band at all.

Fig. 3.3. a) The overall honey-comb structure. b) The overall electric field distribution in x-y plane, in this case, the polarization of incident light is along x-direction; c) absolute value of the electric field profile along the highlighted horizontal line. The excitation light is = 633 nm on gold surface with dielectric function real= 9.83 and imaginary= 1.97.

In order to demonstrate the enhanced electric field due to honey-comb nanostructure, we have also performed a series of simulations with the intent of calculating the near field electromagnetic distribution on AuAPA substrates. This kind of analysis can, in fact, provide information on the level of excitation of any sample deposited on a SERS substrate. In Fig. 3.3, simulations of the electric field due to the LSP generation on SERS substrates at z=0 nm (i.e. on the front surface of substrate) using CST software are presented. In particular, a 25 nm thick golden slab showing honeycomb pattern with periodicity of 100 nm, and sub-wavelength air holes (Lezec et al., 2004) with diameter of 60 nm, respectively, was simulated. The source light was fixed to 633 nm with linear polarization along the xaxis. The overall simulated structure is shown in Fig. 3.3a. For all the periodicity patterns, the results show an electric field strongly affected by the material discontinuity. In fact, the absolute value of the total electric field shows some nodes at the gold-air interface. The xpolarization creates anti-phase hot spots along the x-axis which determines the abrupt change of the electric field as in Fig. 3.3b. The profile, shown in Fig. 3.3c, gives even a better image of this behavior. The sharp doubled horns existing inside the holes are due to the

Photonic Crystals for Plasmonics: From Fundamentals to Superhydrophobic Devices 271

incorporate biophotonic devices. Well assessed mechanisms such as (i) superhydrophobicity and (ii) SERS (Surface Enhanced Raman Scattering) are recapitulated and integrated into a single micro- nano- system. The combination of these two would represent a boost towards the detection and the analysis of few (or single) molecules. In the following, SHSs and nano optics based photonic devices are treated separately. After an introduction of the most important features of these, their combination and the effects thereof are discussed. Notice that, differently from plasmonic devices, SHSs are exposed to a good detail in that they are

In short, the major novelty, here, is the simultaneous use of wetting mechanisms (that arise due to the superhydrophobic surfaces, SHSs) and sensitive materials (that include, randomly distributed silver nanograin aggregates, regular arrays of metallic nano dots, adiabatic nanofocusing cones) to increase the response of nano optics based spectroscopy devices (see the cartoon representation of Fig. 4.1). By doing so, otherwise inaccessible

It is well known that a drop post upon a solid surface develops a contact with the solid described by the sole parameter e (Fig. 4.2A) that is the equilibrium contact angle at the interface between the liquid and the solid. e obeys the celebrated Young equation (Young,

cos *SV SL <sup>e</sup>*

where ij is the surface tension between the phase i and j, and the letters S, L, V stand for the solid, liquid and vapour, and thus equation (1) may be regarded as a simple balance of

*LV*

(4.1)

introduced here for the first time in the text.

information about the biological moieties at study is disclosed.

Fig. 4.1. Raman spectroscopy through super-hydrophobicity.

**4.1 Superhydrophobic surfaces** 

1805):

hexagonal geometry of the holes. Important features emerging from the simulations are both the mean (Emean) and the effective (Eeff) value of the electric field outside the holes (shaded regions in Fig. 3.3c). The former is simply defined as the average field whereas the latter is given by the difference between the maximum and minimum values in the hole.

In conclusion, home-built AuAPA substrates with pore size of 60 nm and wall thickness of 40 nm was utilized for large area SERS substrates. CV was deposited using chemisorption technique, using which a monolayer of molecules can be achieved. CST simulations of the electric field distribution on these ordered and reproducible SERS substrates were also performed, keeping the structural parameters as close as possible to the experimental honey-comb nanostructures. Theoretical results follow the same trend of experimental findings. The major advantage of using nanoporous alumina substrates, as compared to the traditional colloidal coating or lithographic processing, is a good trade-off between the high enhancement factor '*G*' obtained and the large surface area produced. The respective SERS enhancement factor '*G*' is estimated to be ~4106.

Further research should be made in order to optimize '*G*' on the basis of the substrate parameters of pore size, wall thickness, and thickness of the gold coating. The next step towards reproducible micro-fabricated SERS devices would be the transfer of APA to technological substrates such as optical glass or silicon wafer, and combining APA patterning with optical lithography. In this way, large-area SERS bio- and chemical sensors assays for e.g. parallel biomedical screening of different body liquids or even tissues could be carried out, conveniently. The easy and inexpensive processing required for APA SERS fabrication would also make these substrates disposable, opening the way to their large scale applications. Such a high SERS enhancement as demonstrated here can provide single molecule sensitivity in a sensor based on labeling using fluorescence dyes.
