*In Situ* **Transmission Electron Microscopy Studies in Gas/ Liquid Environment**

Fan Wu and Nan Yao

[116] Kim TH, Bae JH, Lee JW, Shin K, Lee JH, Kim MY, Yang CW. Temperature calibration of a specimen heating holder for transmission electron microscopy. Appl. Microsc.

[117] Tanigaki T, Ito K, Nagakubo Y, Asakawa T, Kanemura T. An in situ heating TEM analysis method for an interface reaction. J. Electron Microsc. (Tokyo). 2009; 58:281–

2015; 45:95–100. DOI: 10.9729/AM.2015.45.2.95.

287. DOI: 10.1093/jmicro/dfp020.

266 Microscopy and Analysis

Additional information is available at the end of the chapter

http://dx.doi.org/10.5772/62551

#### **Abstract**

Conventional transmission electron microscopy (TEM) typically operates under high vacuum conditions. However, *in situ* investigation under real-world conditions other than vacuum, such as gaseous or liquidus environment, is essential to obtain practical information for materials including catalysts, fuel cells, biological molecules, lithium ion batteries, etc. Therefore, the ability to study gas/liquid–solid interactions with atomic resolution under ambient conditions in TEM promises new insights into the growth, properties, and functionality of nanomaterials. Different platforms have been devel‐ oped for *in situ* TEM observations in ambient environment and can be classified into two categories: open-cell configuration and sealed gas/liquid cell configuration. The sealed cell technique has various advantages over the open-cell approach. This chapter serves as a review of windowed gas/liquid cells for *in situ* TEM observations.

**Keywords:** *In situ* TEM, sealed gas cell, sealed liquid cell, lithium ion battery, opencell configuration

## **1. Introduction**

Transmission electron microscopy (TEM) is one of the most powerful techniques to character‐ ize structure and chemistry of solids at the atomic scale. The simultaneous acquisition of nanoscale chemical analysis, atomic resolution images, and diffraction patterns provides comprehensive information that other characterization tools cannot compete with. Conven‐ tional TEM typically operates under high vacuum conditions ~1.5 × 10−7 Torr [1]. However, *in situ* investigation under real-world conditions other than vacuum, such as gaseous or liq‐ uidus environment, is essential to obtain practical information [2] for materials including

© 2016 The Author(s). Licensee InTech. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

catalysts, fuel cells, biological molecules, lithium ion batteries, etc. Therefore, the ability to study gas/liquid–solid interactions with atomic resolution under ambient conditions in TEM promises new insights into the growth, properties, and functionality of nanomaterials. *In situ* controlledenvironment TEM (ETEM) [3, 4] enabling TEM study of specimens in ambient environment is necessary for various nanomaterial-based technologies, such as efficient energy conversion/use/ storage, transportation, food production, and environmental protection [5] etc.

So far, different platforms have been developed for *in situ* ETEM observations in ambient environment and can be classified into two categories: 1) platforms with an open-cell config‐ uration, and 2) platforms with a sealed gas/liquid cell configuration. The sealed-type ETEM using a sealed cell has various advantages over the open-cell approach. First of all, the reaction volume and the specimen are confined by electron-transparent top and bottom "windows," allowing gas/liquid to be introduced and sealed within a tiny space, and separated from the other parts of the TEM column. The resulting electron path length is on the order of a few microns [6, 7], much thinner than the opened-type approach and allowing much better resolution to observe lattice images. This is especially good for gas cell because the acceptable reaction pressures within the gas cell can equal or exceed a full atmosphere [6, 8–10] while maintaining the ability to record atomic resolution images [8–11]. Furthermore, much more rapid thermal response than standard heating holders and more rapid stabilization of specimen drift can be realized by integrating miniaturized, low mass heating devices [12], or laser heating [1] into the sealed cell. Therefore, a better control of the reaction process and the imaging experiments is achieved. An additional advantage is that the sealed-cell approach only modifies a small device on the tip of a TEM sample holder, thus can be used in any normal TEM without modifications to any other parts of a TEM. The cost of performing ETEM studies using the sealed-cell approach is typically a tiny fraction of the cost of a dedicated ETEM using open-type approach, because the latter requires modifications to the whole column. Thereby the sealed-cell approach allows *in situ* ETEM studies to be easily extended to many laboratories in the field. Last but not least, the sealed-cell platforms enable *in situ* ETEM characterization with the introduction of any types of volatile carbon-based electrolytes, which is impossible for open-type approach due to the high vacuum requirement inside TEM chamber.

Due to the various advantages over the open-type approach, sealed-cell approach has become the dominant way to perform ETEM studies under ambient conditions. A fast-growing number of research groups worldwide are conducting researches using this technology. This chapter discusses *in situ* ETEM studies in ambient environment by using sealed gas/liquid cells. Different designs and applications of the sealed cells for *in situ* TEM observations are summarized. Future research directions of the sealed gas/liquid cells are demonstrated for the benign development of this field.

## **2. Sealed gas cells**

TEM is one of the most powerful techniques to characterize structure and chemistry of solids at the atomic scale. The simultaneous acquisition of nanoscale chemical analysis, atomic resolution images, and diffraction patterns provides comprehensive information that other characterization tools cannot compete with. However, information about the structural and chemical changes under ambient conditions, especially under gaseous environment, is usually not available since conventional TEM operates under high vacuum conditions ~1.5 × 10−7 Torr [1]. For materials such as catalysts, fuel cells, and biological molecules, *in situ* investigation under real-world conditions other than vacuum is essential to obtain practical information [2]. Therefore, the ability to study gas-solid interactions with atomic resolution at ambient pressures in TEM promises new insights into the growth, properties, and functionality of nanomaterials. Significant improvements in scanning/TEM (S/TEM) technologies containing a gaseous environment have enabled now the atomic scale study during gas-solid interactions [5] with energy resolution in the sub-eV range, and sensitivity to detect single atoms [13]. *In situ* controlled-ETEM [3, 4] enabling TEM study of specimens in ambient environment is necessary for various nanomaterial-based technologies, such as efficient energy conversion/use/storage, transportation, food production, and environmental protection [5], etc.

catalysts, fuel cells, biological molecules, lithium ion batteries, etc. Therefore, the ability to study gas/liquid–solid interactions with atomic resolution under ambient conditions in TEM promises new insights into the growth, properties, and functionality of nanomaterials. *In situ* controlledenvironment TEM (ETEM) [3, 4] enabling TEM study of specimens in ambient environment is necessary for various nanomaterial-based technologies, such as efficient energy conversion/use/

So far, different platforms have been developed for *in situ* ETEM observations in ambient environment and can be classified into two categories: 1) platforms with an open-cell config‐ uration, and 2) platforms with a sealed gas/liquid cell configuration. The sealed-type ETEM using a sealed cell has various advantages over the open-cell approach. First of all, the reaction volume and the specimen are confined by electron-transparent top and bottom "windows," allowing gas/liquid to be introduced and sealed within a tiny space, and separated from the other parts of the TEM column. The resulting electron path length is on the order of a few microns [6, 7], much thinner than the opened-type approach and allowing much better resolution to observe lattice images. This is especially good for gas cell because the acceptable reaction pressures within the gas cell can equal or exceed a full atmosphere [6, 8–10] while maintaining the ability to record atomic resolution images [8–11]. Furthermore, much more rapid thermal response than standard heating holders and more rapid stabilization of specimen drift can be realized by integrating miniaturized, low mass heating devices [12], or laser heating [1] into the sealed cell. Therefore, a better control of the reaction process and the imaging experiments is achieved. An additional advantage is that the sealed-cell approach only modifies a small device on the tip of a TEM sample holder, thus can be used in any normal TEM without modifications to any other parts of a TEM. The cost of performing ETEM studies using the sealed-cell approach is typically a tiny fraction of the cost of a dedicated ETEM using open-type approach, because the latter requires modifications to the whole column. Thereby the sealed-cell approach allows *in situ* ETEM studies to be easily extended to many laboratories in the field. Last but not least, the sealed-cell platforms enable *in situ* ETEM characterization with the introduction of any types of volatile carbon-based electrolytes, which is impossible

storage, transportation, food production, and environmental protection [5] etc.

for open-type approach due to the high vacuum requirement inside TEM chamber.

benign development of this field.

**2. Sealed gas cells**

268 Microscopy and Analysis

Due to the various advantages over the open-type approach, sealed-cell approach has become the dominant way to perform ETEM studies under ambient conditions. A fast-growing number of research groups worldwide are conducting researches using this technology. This chapter discusses *in situ* ETEM studies in ambient environment by using sealed gas/liquid cells. Different designs and applications of the sealed cells for *in situ* TEM observations are summarized. Future research directions of the sealed gas/liquid cells are demonstrated for the

TEM is one of the most powerful techniques to characterize structure and chemistry of solids at the atomic scale. The simultaneous acquisition of nanoscale chemical analysis, atomic The original designs for ETEM observations under gaseous conditions have been around for over 70 years [14] and are made available by two main kinds of methods [15]: one is the opened type, which confines the gas near the sample by means of pressure-limiting apertures and maintain the vacuum in the remaining column by a differential pumping scheme [16–19] (e.g. in 1991, Nan Yao et. al. [20] used two pole pieces to confine the gas near the sample region, for studying supported metal catalysts during catalytic process in a TEM column); the other is the sealed type, which uses a sealed gas cell [4, 7, 8, 21–24] to enclose the sample and the highpressure gas within a tiny space. For the opened type, the pressure-limiting apertures with small holes are positioned in the objective lens in close proximity to the sample, and the differential pumping system is equipped to avoid diffusion of the gas molecules from the chamber toward other parts of TEM, especially the electron gun. Any type of specimen holder can be accepted by the opened-type ETEM. However, differential pumping ETEM has many obvious disadvantages [9], such as a long time needed to ramp up to and down from a selected temperature, difficulty of stabilizing specimen drift due to the large power consumption and heating effects of the heating unit, huge cost needed to modify the TEM column, and the long gas path (on the order of ~1 cm [2, 6, 25]) through which the electron beam must pass that limits reaction pressures to a level of about 15–20 Torr [1].

On the other hand, the sealed-type ETEM using a sealed gas cell has various advantages over the differential pumping approach. First of all, the reaction volume and the specimen are confined by electron-transparent top and bottom "windows," allowing a gas to be introduced and sealed within a tiny space, and separated from the other parts of the TEM column. The resulting gas path length is on the order of a few microns [6, 7], much thinner than the openedtype approach and allowing much better resolution to observe lattice images. Consequently, the acceptable reaction pressures within the gas cell can equal or exceed a full atmosphere [6, 8–10] while maintaining the ability to record atomic resolution images [8–11]. Furthermore, much more rapid thermal response than standard heating holders and more rapid stabilization of specimen drift are realized by integrating miniaturized, low mass heating devices [12] or laser heating [1] into the sealed gas cell. Therefore, a better control of the reaction process and the imaging experiments is achieved. An additional advantage is that the sealed-gas-cell approach only modifies a small device on the tip of a TEM sample holder, thus can be used in any normal TEM without modifications to any other parts of a TEM. The cost of performing ETEM studies using the gas-cell approach is typically a tiny fraction of the cost of a dedicated ETEM using differential-pumping approach, because the latter requires modifications to the whole column. Thereby, the sealed-gas-cell approach allows *in situ* ETEM studies to be easily extended to many laboratories in the field. Due to the various advantages, sealed-gas-cell approach has become the dominant way to perform ETEM studies under gaseous environ‐ ment. A fast-growing number of research groups worldwide are conducting researches using this technology.

Advances in 0D [26], 1D [27], and 2D [28–43] material fabrication technologies have enabled various forms of nanoscale materials, which increased the needs of *in situ* ETEM studies through the closed-type approach, i.e. sealed gas cells. Sealed gas cells enabled *in situ* TEM observations, thus allowing the evaluation of the effect of external stimuli including mechan‐ ical, electrical, and magnetic force on nanomaterials. Some advantages of *in situ* TEM obser‐ vations with sealed gas cells are listed as follows [44]:


The sealed gas cells have been applied in a variety of research fields and topics, including dynamic observation of catalytic reactions [2, 8, 21, 23–45], oxidation and reduction of metals [22], interaction between materials and ionized gas [46], de/hydrogenation processes [47], biological studies [4, 7], etc. Some groups just used their newly developed sealed cells to demonstrate their properties and improved technical limits for *in situ* TEM observations [1, 9, 10, 48, 49]. These applications are discussed in details as follows.

#### **2.1. (De)hydrogenation processes**

Hydrogen storage materials are needed for hydrogen fuel, particularly in the automotive industry. To enhance the kinetics and modify the thermodynamics of hydrogenation, nano‐ structured hydrogen storage materials are needed and study of the *in situ* hydrogenation/ dehydrogenation process on the atomic scale is essential [47]. Using the MEMS-type sealed gas cell, Tadahiro Yokosawa et al. [47] observed the hydrogenation and dehydrogenation of Pd with a very consistent precision and a nanometer resolution, allowing a distinction between hydrogenation behaviors of individual grains. The electron beam was found to have no disturbing influence on the determination of the (de)hydrogenation temperatures in the case of Pd, under normal working conditions and at pressures of 750 and 2400 Torr. The relationship between (de)hydrogenation and pressure fitted well with bulk experiments in which the pressure was varied. Fast determination of the hydrogenation and dehydrogenation temper‐ atures was allowed by realizing a very fast change in temperature.

#### **2.2. Interactions between materials and gases**

laser heating [1] into the sealed gas cell. Therefore, a better control of the reaction process and the imaging experiments is achieved. An additional advantage is that the sealed-gas-cell approach only modifies a small device on the tip of a TEM sample holder, thus can be used in any normal TEM without modifications to any other parts of a TEM. The cost of performing ETEM studies using the gas-cell approach is typically a tiny fraction of the cost of a dedicated ETEM using differential-pumping approach, because the latter requires modifications to the whole column. Thereby, the sealed-gas-cell approach allows *in situ* ETEM studies to be easily extended to many laboratories in the field. Due to the various advantages, sealed-gas-cell approach has become the dominant way to perform ETEM studies under gaseous environ‐ ment. A fast-growing number of research groups worldwide are conducting researches using

Advances in 0D [26], 1D [27], and 2D [28–43] material fabrication technologies have enabled various forms of nanoscale materials, which increased the needs of *in situ* ETEM studies through the closed-type approach, i.e. sealed gas cells. Sealed gas cells enabled *in situ* TEM observations, thus allowing the evaluation of the effect of external stimuli including mechan‐ ical, electrical, and magnetic force on nanomaterials. Some advantages of *in situ* TEM obser‐

**1.** Concurrent observations of structural, morphological, and chemical changes in ambient

**2.** The same area is observed during the whole reaction process in ambient atmosphere,

**4.** Both thermodynamic and kinetic data leading to nanomaterials synthesis or functioning

**5.** Considerable time saving as the synthesis and characterization are performed concur‐

The sealed gas cells have been applied in a variety of research fields and topics, including dynamic observation of catalytic reactions [2, 8, 21, 23–45], oxidation and reduction of metals [22], interaction between materials and ionized gas [46], de/hydrogenation processes [47], biological studies [4, 7], etc. Some groups just used their newly developed sealed cells to demonstrate their properties and improved technical limits for *in situ* TEM observations [1, 9,

Hydrogen storage materials are needed for hydrogen fuel, particularly in the automotive industry. To enhance the kinetics and modify the thermodynamics of hydrogenation, nano‐ structured hydrogen storage materials are needed and study of the *in situ* hydrogenation/ dehydrogenation process on the atomic scale is essential [47]. Using the MEMS-type sealed gas cell, Tadahiro Yokosawa et al. [47] observed the hydrogenation and dehydrogenation of Pd with a very consistent precision and a nanometer resolution, allowing a distinction between

**3.** Intermediate steps during reactions in the ambient atmosphere can be identified.

vations with sealed gas cells are listed as follows [44]:

when sample is subjected to external stimuli.

10, 48, 49]. These applications are discussed in details as follows.

in ambient atmosphere can be obtained.

rently in ambient atmosphere.

**2.1. (De)hydrogenation processes**

atmosphere are enabled.

this technology.

270 Microscopy and Analysis

In 1976, Hiroshi Fujita et. al. [50] designed and used a sealed gas cell for a 3MV-class electron microscope to observe the reaction between H2 gas and iron, as shown in **Figure 1(a)**. After electron irradiation damage in vacuum, the secondary defects were preferentially formed around the dislocations, which were linear structures in **Figure 1(a)**. Fern-leaf-like structures were formed around individual dislocation lines when the iron foil was exposed to wet H2 gas of ~1200 Torr for ~30 min during electron irradiation of 2 × 1019 e/sec.cm2 in intensity, as seen in micrograph (a′), which is an enlargement of a framed part in **Figure 1(a)**. These fern-leaflike structures were quite different from those in **Figure 1(b)**, therefore they might be some sort of Fe hydrides that were closely related to the hydrogen embrittlement of iron.

**Figure 1.** Reaction between wet H2 gas and an iron foil. Micrographs (a) and (b) were taken after a heavy electron irra‐ diation in wet H2 gas of about 1200 Torr and a vacuum of 1 × 10−6 Torr, respectively. Micrographs (a′) and (b′) are enlargement of framed parts in (a) and (b), respectively. (c)–(d): Evaporation of 18/8 type stainless steel at 650°C in a vacuum of 1 × 10−6 Torr. (e)–(f): Evaporation of 18/8 type stainless steel at higher than 700°C in Ar-10 vol % H2 gas of 760 Torr [50].

#### **2.3. Suppression of specimen evaporation**

The same sealed gas cell by Hiroshi Fujita et al. [50] was also used for suppression of the evaporation of specimen at high temperatures, thus decreasing the damage made by evapo‐ ration when metals and alloys were annealed at considerably high temperatures in vacuum. **Figure 1(c)**–**(f)** show the suppression of evaporation of 18/8 type stainless steel during annealing. The specimen (**Figure 1(c)**) was partly evaporated in a vacuum of 1 × 10−6 Torr by heating at 650°C, as seen in **Figure 1(d)**. In contrast, the evaporation of specimen was remark‐ ably suppressed in a mixed gas of 760 Torr consisting of commercially pure Ar gas and 10 volume% H2 gas even when the specimen was heated at high temperatures (more than 700°C). Microstructures in the specimen could be seen clearly even when a gas layer was as thick as ~100 μm, as shown in **Figure 1(e)** and **(f)**.

#### **2.4. Oxidation and reduction of metals**

Making use of the heating element and the enclosed gases, *in situ* observations of oxidation and reduction processes can be performed with sealed gas cells. **Figure 2** shows the oxidation process of a Cu thin film in a sealed gas cell designed by M. Komatsu et al. [22] in 2005. Initially, a small amount of Cu oxide formed during evaporation, as shown in the bright field image (**Figure 2(a)**) and the corresponding selected area electron diffraction pattern (**Figure 2(a′)**), respectively. Oxygen was then introduced into the cell to 9.75 Torr and the specimen was gradually heated to 470K. Cu oxide was found to preferentially nucleate on the film surface (**Figure 2(b)** and **(b′)**). As the temperature increased, Cu was oxidized to very fine oxide particles, as shown in **Figure 2(c)** and **(c′)**. After the specimen was heated to 670K, all particles changed to CuO (**Figure 2(d)** and **(d′)**). The CuO grains grew larger when the specimen temperature reached 770K (**Figure 2(e)** and **(e′)**). The reduction of CuO was also observed *in situ*, as shown in **Figure 2**. The film was gradually reheated in 9.75 Torr of H2. As the specimen temperature was further increased, CuO was completely reduced to Cu at 670 K (**Fig‐ ure 2(h)** and **(h′)**).

**Figure 2.** Successive stages of the oxide growth on a 100-nm-thick Cu thin film between room temperature and ~770K under 1.3 × 103 Pa of O2. (a)–(e): Bright field images. (a′)–(e′): The corresponding SAEDs. Successive stages of the re‐ duction of CuO between room temperature and ~670K under 1.3 × 10<sup>3</sup> Pa of H2. (f)–(h): Bright field images. (f′)–(h′): The corresponding SAEDs [22].

#### **2.5.** *In situ* **growth of nanostructures**

**2.3. Suppression of specimen evaporation**

272 Microscopy and Analysis

~100 μm, as shown in **Figure 1(e)** and **(f)**.

**2.4. Oxidation and reduction of metals**

**ure 2(h)** and **(h′)**).

under 1.3 × 103

The corresponding SAEDs [22].

The same sealed gas cell by Hiroshi Fujita et al. [50] was also used for suppression of the evaporation of specimen at high temperatures, thus decreasing the damage made by evapo‐ ration when metals and alloys were annealed at considerably high temperatures in vacuum. **Figure 1(c)**–**(f)** show the suppression of evaporation of 18/8 type stainless steel during annealing. The specimen (**Figure 1(c)**) was partly evaporated in a vacuum of 1 × 10−6 Torr by heating at 650°C, as seen in **Figure 1(d)**. In contrast, the evaporation of specimen was remark‐ ably suppressed in a mixed gas of 760 Torr consisting of commercially pure Ar gas and 10 volume% H2 gas even when the specimen was heated at high temperatures (more than 700°C). Microstructures in the specimen could be seen clearly even when a gas layer was as thick as

Making use of the heating element and the enclosed gases, *in situ* observations of oxidation and reduction processes can be performed with sealed gas cells. **Figure 2** shows the oxidation process of a Cu thin film in a sealed gas cell designed by M. Komatsu et al. [22] in 2005. Initially, a small amount of Cu oxide formed during evaporation, as shown in the bright field image (**Figure 2(a)**) and the corresponding selected area electron diffraction pattern (**Figure 2(a′)**), respectively. Oxygen was then introduced into the cell to 9.75 Torr and the specimen was gradually heated to 470K. Cu oxide was found to preferentially nucleate on the film surface (**Figure 2(b)** and **(b′)**). As the temperature increased, Cu was oxidized to very fine oxide particles, as shown in **Figure 2(c)** and **(c′)**. After the specimen was heated to 670K, all particles changed to CuO (**Figure 2(d)** and **(d′)**). The CuO grains grew larger when the specimen temperature reached 770K (**Figure 2(e)** and **(e′)**). The reduction of CuO was also observed *in situ*, as shown in **Figure 2**. The film was gradually reheated in 9.75 Torr of H2. As the specimen temperature was further increased, CuO was completely reduced to Cu at 670 K (**Fig‐**

**Figure 2.** Successive stages of the oxide growth on a 100-nm-thick Cu thin film between room temperature and ~770K

duction of CuO between room temperature and ~670K under 1.3 × 10<sup>3</sup>

Pa of O2. (a)–(e): Bright field images. (a′)–(e′): The corresponding SAEDs. Successive stages of the re‐

Pa of H2. (f)–(h): Bright field images. (f′)–(h′):

*In situ* observation of the growth process of CuO whiskers was carried out in the same sealed gas cell by M. Komatsu et al. [22]. A series of electron micrographs show the successive stages of growth of Cu oxide whiskers in 30 Torr of O2 (**Figure 3**). Initially, a non-uniform oxide film formed on the Cu surface, resulting to a jagged edge (**Figure 3(a)** and **(b)**). After 40s, the oxide layer stabilized with a smoother edge (**Figure 3(c)**). Then whiskers started to grow on the oxide layer gradually (**Figure 3(d)**–**(f)**).

**Figure 3.** Successive stages of growth of Cu oxide whiskers in 30 Torr of O2 [22].

#### **2.6. Reactions with atomic/ionized gases**

The earliest report of interactions with ionized/atomic gas induced by the electron beam of TEM was made by Li Sun et al. [46] in 2011. The enhanced electron flux could increase the concentrations of both reducing electrons and oxygen ions. Below a certain threshold, oxidization dominated the system response and resulted in accelerated interaction between silver and oxygen ions. The average size of silver grains continued to decrease, as shown in **Figure 4(c)**–**(f)**. At current densities of 0.44 A cm−2, the silver grains were rod-shaped (**Figure 4(c)**, **(d)**). At current densities greater than 0.65 A cm−2, the silver grains reverted back to a more compact angular morphology (**Figure 4(f)**). Due to the increased oxygen fugacity associated with higher concentrations of ionized and atomic oxygen, all Ag2O phase was further oxidized to AgO at current densities greater than 0.65 A cm−2. Above 0.75 A cm−2, a significant portion of noncrystalline phase existed (**Figure 4(g)**). Once the electron current density increased beyond 0.77 A cm−2, new grains nucleated (**Figure 4(h)**) out of the vapor phase, exhibiting a twin structure. The reaction between the nanoparticles (NPs) and gas produced a concentration gradient around the particles that was observed as a bright ring around each silver grain. The silver oxide depletion width in the gas phase indicated a strong chemical interaction between the solid and vapor phases. The AgO vapor phase often aligned itself into 2D sheets perpendicular to the beam that subsequently became unstable precipitate clusters of new silver grains or swept through a region randomly (**Figure 4(h)**). The competi‐ tion between oxidation and electron-beam-induced reduction also provided excess heat. For significantly high fluxes of ionized oxygen, a thermal effect could induce local vaporization at the surface, and sequentially an *in situ* nanoscale reaction ion sputtering. This investigation revealed a variety of microstructural processes associated with the oxidation of Ag by atomic and ionized gas species. The electron beam was demonstrated to be an important source of both oxidation and reduction. The sealed cell approach provided an opportunity to make early observations of real-time nanoscale dynamics associated with oxidation in ionized and atomic gas, the movement of a partial pressure of a gas phase, and interactions between the condensed and vapor phases of a material. The results provided new insights into manipulating nano‐ structure and chemistry through ionized gas treatment and offered unique access to simulate reactions with atomic and ionized gas.

**Figure 4.** Microstructure of the observation area in an air-filled cell after exposure of (a) 0 s with 0.18 A cm−2 current density; (b) 20 s with 0.18 A cm−2 current density (the arrow marks a new grain); (c) 0 s with 0.44 A cm−2 current densi‐ ty; (d) 20 s with 0.44 A cm−2 current density; (e) 0 s with 0.66 A cm−2 current density; (f) 20 s with 0.66 A cm−2 current density; (g) 20 s with 0.72 A cm−2 current density; and (h) 20s with 0.8 A cm−2 current density (the arrow marks the vertical alignment of the AgO vapor phase) [46].

#### **2.7. Dynamic observation of catalysts and catalytic reactions**

One of the earliest attempts to observe a catalyst in a sealed gas cell was reported by Parkinson et al. [21] in 1989. Using a narrow-gap, sealed gas cell and a 400-kV TEM, images of the crystal lattice of ceria (0.31 nm) were recorded under flowing nitrogen gas at 20 Torr. Structural information of chemical significance became discernible at ~0.3 nm, which offered real hope of carrying out fundamental dynamic studies of the activation, reaction, and passivation of gas/solid systems at close to the atomic level.

Seventeen years later, the atomic-scale *in situ* observations of catalysts were performed by S. Giorgio et al. [23], during a chemical reaction. For the first time, Au and Pd clusters supported on Ti02 and amorphous carbon were observed with a sealed gas cell with the resolution of (111) lattice fringes. Initially, an Au cluster in vacuum was strongly contaminated, but the contam‐ ination disappeared while the faceting and the crystalline lattice were visible in the cluster after circulation of H2 at a pressure of 3 Torr at room temperature. The cluster was completely faceted after annealing until 350°C in the same reducing atmosphere, then cooling down to room temperature.

itself into 2D sheets perpendicular to the beam that subsequently became unstable precipitate clusters of new silver grains or swept through a region randomly (**Figure 4(h)**). The competi‐ tion between oxidation and electron-beam-induced reduction also provided excess heat. For significantly high fluxes of ionized oxygen, a thermal effect could induce local vaporization at the surface, and sequentially an *in situ* nanoscale reaction ion sputtering. This investigation revealed a variety of microstructural processes associated with the oxidation of Ag by atomic and ionized gas species. The electron beam was demonstrated to be an important source of both oxidation and reduction. The sealed cell approach provided an opportunity to make early observations of real-time nanoscale dynamics associated with oxidation in ionized and atomic gas, the movement of a partial pressure of a gas phase, and interactions between the condensed and vapor phases of a material. The results provided new insights into manipulating nano‐ structure and chemistry through ionized gas treatment and offered unique access to simulate

**Figure 4.** Microstructure of the observation area in an air-filled cell after exposure of (a) 0 s with 0.18 A cm−2 current density; (b) 20 s with 0.18 A cm−2 current density (the arrow marks a new grain); (c) 0 s with 0.44 A cm−2 current densi‐ ty; (d) 20 s with 0.44 A cm−2 current density; (e) 0 s with 0.66 A cm−2 current density; (f) 20 s with 0.66 A cm−2 current density; (g) 20 s with 0.72 A cm−2 current density; and (h) 20s with 0.8 A cm−2 current density (the arrow marks the

One of the earliest attempts to observe a catalyst in a sealed gas cell was reported by Parkinson et al. [21] in 1989. Using a narrow-gap, sealed gas cell and a 400-kV TEM, images of the crystal lattice of ceria (0.31 nm) were recorded under flowing nitrogen gas at 20 Torr. Structural information of chemical significance became discernible at ~0.3 nm, which offered real hope of carrying out fundamental dynamic studies of the activation, reaction, and passivation of

Seventeen years later, the atomic-scale *in situ* observations of catalysts were performed by S. Giorgio et al. [23], during a chemical reaction. For the first time, Au and Pd clusters supported on Ti02 and amorphous carbon were observed with a sealed gas cell with the resolution of (111) lattice fringes. Initially, an Au cluster in vacuum was strongly contaminated, but the contam‐ ination disappeared while the faceting and the crystalline lattice were visible in the cluster

reactions with atomic and ionized gas.

274 Microscopy and Analysis

vertical alignment of the AgO vapor phase) [46].

gas/solid systems at close to the atomic level.

**2.7. Dynamic observation of catalysts and catalytic reactions**

The resolution of *in situ* observation of catalysts was improved by a novel MEMS-type nanoreactor in 2008 [8]. More importantly, the nanoreactor facilitated the direct observation of Cu nanocrystal growth and mobility on a sub-second time scale at a higher temperature (500°C) and higher gas pressure (900 Torr of H2). The *in situ* TEM images showed atomic lattice fringes in the Cu nanocrystals with spacing of 0.18 nm, attesting the spatial resolution limit of the system. The system of Cu nanocrystals on a ZnO support is commonly used as catalyst for methanol synthesis and for conversion of hydrocarbons in fuel cells. Also, it is a prototype example of the industrially important group of 3d transition metal catalysts. The catalyst was heated in the H2 atmosphere to the maximum operation temperature of 500°C. ZnO crystallites with diameters of 20–100 nm appeared in the precursor with facetted, compact shapes (**Figure 5(a)**). The CuO appeared as smaller patches of more irregular shapes at the edges of

**Figure 5.** Image sequences of the Cu nanocrystal growth and mobility on ZnO. Nanocrystals (darker contrast) form from CuO precursors (blue arrows) during heating from room temperature to 500°C in 900 Torr H2. After growth, nanocrystals can exhibit transient mobility (white square). Crystallites on the opposite window are seen out of focus (black arrows in (a) and (c)). The frames are recorded at (a) room temperature, (b) 260°C, (c) 330°C, (d) 365°C, (e) 410°C, and (f) 500°C. All frames are averaged over four consecutive images. The exposure time for each image is 0.145 s (color online) [8].

ZnO. As temperature increased to ~260°C, the CuO patches broke up into several particles with diameters of 5–10 nm (**Figure 5(b)**–**(e)**). The state of the nanocrystals was inferred from atomic-resolution TEM images during exposure to 900 Torr H2 at 500°C. Atomic lattice fringes were clear in both the brighter ZnO support crystallites and the darker Cu nanocrystals (**Figure 6(a)**). Lattice fringes with spacings of 0.21 and 0.18 nm could be recorded in the nanocrystals [24], corresponding to (111) and (200) planes of Cu, identified by Fourier transform (**Figure 6(b)**) of the TEM image.

**Figure 6.** A representative HRTEM image of the Cu/ZnO catalyst during exposure to 900 Torr hydrogen at 500°C. (a) The image displays lattice fringes of a twinned Cu nanocrystal and of the ZnO support. (b) A Fourier transform of (a). The bright dots represent sets of lattice fringes. Their lattice spacing corresponds to the distance to the origin and re‐ veals the crystallographic identity. The large, red circle corresponds a spacing of 0.21 nm. The smallest, resolved lattice spacing is 0.18 nm. (c)–(f) [2]: *In situ* TEM images of the nanoparticulate gold catalyst supported on TiO2 recorded se‐ quentially. The time shown in the lower right-hand corners of (d)–(f) correspond to intervals measured from the time at which (c) was recorded (color online) [8].

One year later, the MEMS-type sealed gas cell developed by Tadahiro Kawasaki et. al. [2] was applied for *in situ* TEM observations of a gold nanoparticulate catalyst supported on TiO2. One percent CO in dry air was introduced to react with O2 to form CO2 on the catalyst surface. **Figure 6(c)**–**(f)** show the sequential morphologies of the gold NP. The shape of the gold particle changed markedly over a short period of time, such as the 0.4 s interval between **Figure 6(d)** and **(e)** and the 0.2 s between **(e)** and **(f)**. Various facets of the gold appeared in **Figure 6(c)**, **(d)**, and **(f)**. They sometimes disappeared and the gold particle formed a spherical shape in **Figure 6(e)**. However, the lattice fringes of the gold could not be observed due to electron scattering by the high-pressure gas.

The most recent *in situ* visualization of oscillatory behavior of Pt NPs catalyzing CO oxidation was reported by S. B. Vendelbo et al. [45] in 2014. TEM image series of the Pt NPs were acquired at windows both at the entrance and exit of the reaction zone, at a rate (1–2 frames per second) faster than the rate of the reaction oscillations, to directly visualize the NPs on this timescale. Near the reaction zone entrance, the NPs had a stationary and more spherical morphology during the oscillating reaction. In contrast, near the reaction zone exit, the Pt NPs switched between spherical and facetted morphology (**Figure 7(1)**). As the CO conversion increased rapidly, Pt NPs started a gradual transformation from the more spherical shape towards a more facetted shape. The fully facetted shape was reached within 3 s after the CO peak conversion (**Figure 7(1) III**). On decrease in the CO conversion, the NP transformed back to the more spherical shape (**Figure 7(1) IV**) and retained that shape until the CO conversion rose steeply again. Thus, the individual NPs near the exit from the reaction zone underwent oscillatory and reversible shape changes with a temporal frequency matching the oscillations in reaction power, indicating that the oscillatory CO conversion and the dynamic shape change of the Pt NPs were coupled. To address the mechanism governing the oscillatory reaction, the state of the Pt NPs was examined at the atomic scale (**Figure 7(2) (c)–(e)**). The spacing of crystalline lattice planes and the uniform contrast across the projected image of the NPs were consistent with metallic Pt. The combined high-resolution TEM and DFT analyses indicated that the Pt surfaces remained in the metallic state under the present conditions. Time-resolved series of high-resolution TEM images show that in the more spherical state, the Pt NPs were terminated by close-packed (111) planes, more open (110) planes and step sites (**Figure 7(2) (a)**, **(c)**, **(e)**), while for the more facetted state, the NPs were terminated by extended (111) planes as well as a reduced abundance of higher index terminations and steps (**Figure 7(2) (b)**, **(d)**).

ZnO. As temperature increased to ~260°C, the CuO patches broke up into several particles with diameters of 5–10 nm (**Figure 5(b)**–**(e)**). The state of the nanocrystals was inferred from atomic-resolution TEM images during exposure to 900 Torr H2 at 500°C. Atomic lattice fringes were clear in both the brighter ZnO support crystallites and the darker Cu nanocrystals (**Figure 6(a)**). Lattice fringes with spacings of 0.21 and 0.18 nm could be recorded in the nanocrystals [24], corresponding to (111) and (200) planes of Cu, identified by Fourier

**Figure 6.** A representative HRTEM image of the Cu/ZnO catalyst during exposure to 900 Torr hydrogen at 500°C. (a) The image displays lattice fringes of a twinned Cu nanocrystal and of the ZnO support. (b) A Fourier transform of (a). The bright dots represent sets of lattice fringes. Their lattice spacing corresponds to the distance to the origin and re‐ veals the crystallographic identity. The large, red circle corresponds a spacing of 0.21 nm. The smallest, resolved lattice spacing is 0.18 nm. (c)–(f) [2]: *In situ* TEM images of the nanoparticulate gold catalyst supported on TiO2 recorded se‐ quentially. The time shown in the lower right-hand corners of (d)–(f) correspond to intervals measured from the time

One year later, the MEMS-type sealed gas cell developed by Tadahiro Kawasaki et. al. [2] was applied for *in situ* TEM observations of a gold nanoparticulate catalyst supported on TiO2. One percent CO in dry air was introduced to react with O2 to form CO2 on the catalyst surface. **Figure 6(c)**–**(f)** show the sequential morphologies of the gold NP. The shape of the gold particle changed markedly over a short period of time, such as the 0.4 s interval between **Figure 6(d)** and **(e)** and the 0.2 s between **(e)** and **(f)**. Various facets of the gold appeared in **Figure 6(c)**, **(d)**, and **(f)**. They sometimes disappeared and the gold particle formed a spherical shape in **Figure 6(e)**. However, the lattice fringes of the gold could not be observed due to electron

The most recent *in situ* visualization of oscillatory behavior of Pt NPs catalyzing CO oxidation was reported by S. B. Vendelbo et al. [45] in 2014. TEM image series of the Pt NPs were acquired at windows both at the entrance and exit of the reaction zone, at a rate (1–2 frames per second) faster than the rate of the reaction oscillations, to directly visualize the NPs on this timescale. Near the reaction zone entrance, the NPs had a stationary and more spherical morphology during the oscillating reaction. In contrast, near the reaction zone exit, the Pt NPs switched between spherical and facetted morphology (**Figure 7(1)**). As the CO conversion increased rapidly, Pt NPs started a gradual transformation from the more spherical shape towards a

transform (**Figure 6(b)**) of the TEM image.

276 Microscopy and Analysis

at which (c) was recorded (color online) [8].

scattering by the high-pressure gas.

**Figure 7.** (1) Time-resolved TEM images of a Pt NP at the gas exit of the reaction zone. (2) Atomic-scale visualization of the dynamic refacetting of a Pt NP during the oscillatory CO oxidation. Time-resolved high-resolution TEM images of a Pt NP at the gas exit of the reaction zone. The gas entering the reaction zone is 1.0 bar of CO:O2:He at 4.2%:21.0%: 74.8% and nanoreactor temperature is 727K. (a)–(e): The TEM images showing the more spherical shape (a, c, e) and the more facetted shape (b, d), during the oscillatory reaction. Fast Fourier transforms included as insets in (c)–(e) re‐ veal a lattice spacing corresponding to the Pt(111) lattice planes. The orientation of the observed Pt(111) lattice fringes is consistent with the superimposed crystal lattice vectors and zone axis (color online) [45].

Apart from the homemade sealed gas cells above, commercial MEMS-type sealed gas cells have also been applied to *in situ* studies of catalysts. The membrane-type heating chip manufactured by Hummingbird Scientific (Lacey, WA, USA) provided a temperature con‐ trollable reaction platform for oxidation reactions of cobalt NPs with flowing oxygen (0.2 sccm), while ramping temperature from 150 to 250°C and 250 to 350°C at ~5°C/s [25]. **Fig‐** **ure 8(a)** shows the time-lapse images of three selected Co particles. The metallic cobalt core could shrink with a unidirectional retraction front (**Figure 8(aI)**) and a sweeping retraction front (**Figure 8(aII)**). In projection, the residual metallic puddle was faceted, which was likely shaped by the faceted hollow shell (**Figure 8(aIII)**). The quantification of the volume trajectory of the metallic core (**Figure 8(d)**) shows that the metallic core volume started to rapidly decrease when temperature reached 250°C. After the first volume-decreasing phase, volume shrinkage dwelled for a short period of time at the first plateau (① in **Figure 8(b)**). Then a second rapid decreasing phase initiated with a lower volume shrinkage rate (**Figure 8(b)**). The metallic core was finally eliminated, but at an even slower volume shrinking rate. The particle's oxide shell was in contact with other particles with upper and lower right boundaries open.

**Figure 8.** *In situ* heating of cobalt NPs in flowing oxygen. (a) Real-time reaction dynamics of the Kirkendall effect. (b) Metallic core volume trajectory of particle I in (a). ① and ② mark two diffusion stagnation plateaus. (c) Restructuring of the hollow oxide structure at 250–350°C in flowing oxygen. (d): Hollow core volume trajectory of particle I in (c). Scale bar is 10 nm [25] (color online).

#### **2.8. Biological studies**

The sealed gas cells encapsulated specimens in a thin gas layer, preventing specimens from destruction as in vacuum. Therefore, the sealed gas cells have also been widely applied into biological studies. This section will discuss the biological applications of the sealed gas cells for *in situ* TEM observation.

The first application of sealed gas cells in biology was reported by H. G. Heide in 1962 [4]. For organic specimens, it is necessary to prevent carbon removal from increasing to a rate higher than the rate of contamination, which would destruct the specimen. A rapid dehydration of the specimen can be prevented if unnecessary heating is avoided even at pressures of 100–200 Torr, which was proved by TEM pictures of small water droplets in air at 100 Torr. It was possible to prevent carbon removal in the specimen if H2, He, N2, or Ar instead of air was used at this pressure and the illuminated area was reduced to ~2 μm diameter with the double condenser.

In 2001, an *in situ* sealed gas cell was used to study the reduction of Cr (VI) by bacterium (*Shewanella oneidensis*) by T. L. Daulton et al. [7]. Bacteria from rinsed cultures were placed directly in the gas cell and examined under 97.5 Torr pressure of air saturated with water vapor, showing rod-shaped morphology typical of flagellated and non-spore-forming species (**Figure 9**). Cells remained plump/hydrated while the EPS retained moisture and appeared as a continuous capsule surrounding the cells. However, damage to the cells was observed within minutes of electron-beam exposure, arising from the primary destruction of weak Van der Waals biomolecular bonds. Direct *in situ* TEM imaging revealed two distinct populations of *S. oneidensis* in the cultures: bacteria exhibiting low image contrast (**Figure 9(a)**, **(c)**) and bacteria encrusted/impregnated with electron-dense particles (**Figure 9(b)**, **(d)**). Further examination of the encrusted bacteria showed that their gram-negative, cell envelope was electron dense (**Figure 9(d)**) and appeared darkest along the perimeter where the electron path length was the greatest. The cell envelopes of non-encrusted cells produced very low image contrast as compared to encrusted bacteria. The increase in contrast indicated that the cell envelope was saturated with absorbed elements of heavy mass, such as Cr. The binding of heavy elements in the cell envelope was associated with Cr reduction.

**Figure 9.** *S. oneidensis* imaged in the environmental cell at 100 Torr: bacteria exhibiting low contrast in bright field EC-TEM imaging (a, c), and bacteria encrusted/impregnated with electron dense particulates (b, d). The arrowhead in pan‐ el (d) points to a low contrast bacterium in the same field of view as a bacterium with electron dense particulates, illustrating the dramatic contrast difference. The low-contrast, diffuse background, best seen in panel (a), represents the extracellular polymeric substances that surround the cells [7].

#### **2.9.** *In situ* **investigations on cladding materials**

**ure 8(a)** shows the time-lapse images of three selected Co particles. The metallic cobalt core could shrink with a unidirectional retraction front (**Figure 8(aI)**) and a sweeping retraction front (**Figure 8(aII)**). In projection, the residual metallic puddle was faceted, which was likely shaped by the faceted hollow shell (**Figure 8(aIII)**). The quantification of the volume trajectory of the metallic core (**Figure 8(d)**) shows that the metallic core volume started to rapidly decrease when temperature reached 250°C. After the first volume-decreasing phase, volume shrinkage dwelled for a short period of time at the first plateau (① in **Figure 8(b)**). Then a second rapid decreasing phase initiated with a lower volume shrinkage rate (**Figure 8(b)**). The metallic core was finally eliminated, but at an even slower volume shrinking rate. The particle's oxide shell

**Figure 8.** *In situ* heating of cobalt NPs in flowing oxygen. (a) Real-time reaction dynamics of the Kirkendall effect. (b) Metallic core volume trajectory of particle I in (a). ① and ② mark two diffusion stagnation plateaus. (c) Restructuring of the hollow oxide structure at 250–350°C in flowing oxygen. (d): Hollow core volume trajectory of particle I in (c).

The sealed gas cells encapsulated specimens in a thin gas layer, preventing specimens from destruction as in vacuum. Therefore, the sealed gas cells have also been widely applied into biological studies. This section will discuss the biological applications of the sealed gas cells

The first application of sealed gas cells in biology was reported by H. G. Heide in 1962 [4]. For organic specimens, it is necessary to prevent carbon removal from increasing to a rate higher than the rate of contamination, which would destruct the specimen. A rapid dehydration of the specimen can be prevented if unnecessary heating is avoided even at pressures of 100–200 Torr, which was proved by TEM pictures of small water droplets in air at 100 Torr. It was possible to prevent carbon removal in the specimen if H2, He, N2, or Ar instead of air was used at this pressure and the illuminated area was reduced to ~2 μm diameter with the double

Scale bar is 10 nm [25] (color online).

for *in situ* TEM observation.

condenser.

**2.8. Biological studies**

278 Microscopy and Analysis

was in contact with other particles with upper and lower right boundaries open.

Cladding is the outer layer of the fuel rods, preventing radioactive fission fragments from escaping the fuel into the coolant and contaminating it. Exposures to irradiation, temperature changes, and stresses may induce microstructural changes, and ultimately result in failure of the cladding. It is thus essential to use *in situ* TEM to observe microstructural changes at the nanoscale dynamically, for predicting the performance of cladding in-service and during storage, understanding the dominant processes related to these changes and their consequen‐ ces. In 2012, a sealed gas cell developed by K. Hattar et al. [51] was used to investigate the radiation tolerance of potential Generation IV cladding materials and the degradation mechanisms in Zr-based claddings of importance for dry storage. Examination of a Zircaloy foil enclosed by top and bottom windows (**Figure 10(a)**) showed deterioration of resolution due to expected additional scattering of electrons by the 5-μm-thick air, after initial scattering by the foil. Despite loss in resolution, prominent features of the foil that were previously observed under vacuum still remained visible. After annealing at 300°C for over 15 min, negligible changes in the Zircaloy morphology occurred (**Figure 10(b)**). Following the 15-min annealing at 300°C, the temperature of the gas cell was raised to 600°C and a dramatic morphology change within the sample was observed almost instantaneously (**Figure 10(c– d)**).

**Figure 10.** Images of Zircaloy lamella at nominally atmospheric pressure. (a) Initial structure. (b) 15 min at 300°C, (c) 600°C, and (d)600°C after 10 min [51].

The same device was applied to study hydride formation in ZirloTM cladding material [12] recently. The formation of hydrides, their dissolution, and re-precipitation, is particularly important for long-term dry storage of currently used fuel assemblies, as the size, shape, and orientation of hydrides play a strong role in the mechanical properties of spent claddings. During *in situ* observation, hydrogen was introduced and maintained at a pressure of 330 Torr. The temperature was then increased at a rate of 1°C/s to approximately 400°C, and held for 90 min. **Figure 11(a)** and **(b)** show a comparison between the microstructures of the ZirloTM prior to annealing and during later stages of annealing. The disappearance of microstructural features (arrow 1) and the formation of a new grain (arrow 2) are evident. The region around the new grain (**Figure 11(c)**) and the analysis of diffraction information (**Figure 11(d)**) indicated the formation of either ε-ZrHx (x > 1.8) or γ-ZrH. These results show that *in situ* environmental heating TEM can be applied to study this mechanism at the nanoscale in order to verify predictive material models.

### **3. Sealed liquid cells**

Conventional TEM is not compatible with studies of electrochemical energy storage processes, but the development of TEM holders and sealed liquid cell (SLC) platforms encapsulating thin liquid layers promise *in situ* imaging and spectroscopy of electrochemical processes [52, 53] (e.g. electrodeposition [54] and dendrite growth [55]) on the nanoscale [56–58], by incorporat‐ ing electrodes [54, 59] in a liquid environment. One major application of SLCs for *in situ* TEM observation is for lithium ion battery(LIB) research. Unlike ex situ studies, which involve unexpected reactions due to the removal of the particles from their native and reactive environment [60], *in situ* TEM electrochemical characterization will mimic the true environ‐ ment in a commercial LIB cell. The *in situ* liquid TEM has allowed quantitative analysis of processes (e.g. NP growth from solution [61–63]), and direct observation of beam-sensitive systems (including macromolecular complexes [64, 65], soft materials [66, 67]) and of processes that span from the electrochemical deposition of metals [54, 55], to growth of different nanostructures [61, 62, 68–71]. Now it gains growing attention for LIB research.

storage, understanding the dominant processes related to these changes and their consequen‐ ces. In 2012, a sealed gas cell developed by K. Hattar et al. [51] was used to investigate the radiation tolerance of potential Generation IV cladding materials and the degradation mechanisms in Zr-based claddings of importance for dry storage. Examination of a Zircaloy foil enclosed by top and bottom windows (**Figure 10(a)**) showed deterioration of resolution due to expected additional scattering of electrons by the 5-μm-thick air, after initial scattering by the foil. Despite loss in resolution, prominent features of the foil that were previously observed under vacuum still remained visible. After annealing at 300°C for over 15 min, negligible changes in the Zircaloy morphology occurred (**Figure 10(b)**). Following the 15-min annealing at 300°C, the temperature of the gas cell was raised to 600°C and a dramatic morphology change within the sample was observed almost instantaneously (**Figure 10(c–**

**Figure 10.** Images of Zircaloy lamella at nominally atmospheric pressure. (a) Initial structure. (b) 15 min at 300°C, (c)

The same device was applied to study hydride formation in ZirloTM cladding material [12] recently. The formation of hydrides, their dissolution, and re-precipitation, is particularly important for long-term dry storage of currently used fuel assemblies, as the size, shape, and orientation of hydrides play a strong role in the mechanical properties of spent claddings. During *in situ* observation, hydrogen was introduced and maintained at a pressure of 330 Torr. The temperature was then increased at a rate of 1°C/s to approximately 400°C, and held for 90 min. **Figure 11(a)** and **(b)** show a comparison between the microstructures of the ZirloTM prior to annealing and during later stages of annealing. The disappearance of microstructural features (arrow 1) and the formation of a new grain (arrow 2) are evident. The region around the new grain (**Figure 11(c)**) and the analysis of diffraction information (**Figure 11(d)**) indicated the formation of either ε-ZrHx (x > 1.8) or γ-ZrH. These results show that *in situ* environmental heating TEM can be applied to study this mechanism at the nanoscale in order to verify

Conventional TEM is not compatible with studies of electrochemical energy storage processes, but the development of TEM holders and sealed liquid cell (SLC) platforms encapsulating thin liquid layers promise *in situ* imaging and spectroscopy of electrochemical processes [52, 53]

**d)**).

280 Microscopy and Analysis

600°C, and (d)600°C after 10 min [51].

predictive material models.

**3. Sealed liquid cells**

Apart from nanostructured anodes/cathodes, the development of platforms enabling *in situ* TEM electrochemical characterization is also required by various other aspects of LIB research. For example, one of the most well-known reactions at the electrode/electrolyte interface is the formation of the solid-electrolyte interphase (SEI), which is a reaction product of mixed composition formed on high-voltage anodes (e.g. Li metal or graphite-lithium intercalation compounds) or cathodes, by electrochemical reduction or oxidation of the electrolyte [72], respectively. The study of the SEI layer requires the use of commonly used LIB electrolytes (volatile carbonate-based solution), the ability to monitor the changes in SEI layer with cycling, time or temperature, and probes having sufficient spatial resolution to detect a reaction product layer of a few nm thick. All of these requirements make ex situ characterization inappropriate for the study of SEI, since the SEI layer is highly sensitive to moisture, air, and other kinds of contaminations [73]. The importance and critical need to develop platforms enabling *in situ* TEM electrochemical characterization of LIBs are thus obvious. Furthermore, the detailed understanding in dendritic growth of lithium metal on the electrode also requires such *in situ* TEM platforms, because the dendritic growth of lithium metal on the electrode can lead to short-circuit and thus battery failure [73].

**Figure 11.** (a) and (b) Overview of images before the start of *in situ* experiment, and after 72 min at 400°C in H2 atmos‐ phere, (c) zoomed-in image of the region of interest around arrow 2 in (b), and (d) shows the superimposition of two diffraction patterns obtained from the circled dark grain shown in (c), one of the parent α-Zr phase (diffraction spots of weaker intensity) and the other is consistent with a face-centered tetragonal hydride phase [12].

Typical commercially available Li-ion batteries usually use carbonate-based liquids as electrolytes, such as diethyl carbonate (DEC), dimethyl carbonate (DMC) mixed with ethylene carbonate (EC), etc. To enable *in situ* TEM characterization of electrochemical reactions in real lithium ion batteries, sealing those volatile liquids inside a sufficiently narrow channel for electron transmission is a wise option. The SLC platforms enable *in situ* electrochemical characterization with any types of electrolytes with a sealed-cell configuration, thus promoting the potential use of volatile carbon-based electrolytes for LIB research. One of the first TEM SCLs was created by F.M. Ross et al. [54, 74] to study Cu electro-deposition during TEM imaging. This SLC platform sealed the aqueous electrolyte by assembling two silicon chips with thin silicon nitride membranes in a face-to-face configuration. This flip-chip approach allows imaging chemical reactions in liquids with high spatial resolution [54, 57, 58, 62, 75] with different membranes of silicon nitride, silicon dioxide, or polymer, such that it has been adopted in various studies, including cell imaging [76–78] and NP synthesis [58] in solutions. For example, electrochemical deposition of polycrystalline Au [75], anisotropic electrodepo‐ sition of nickel nanograins [79], and electrochemical growth of single crystal lead dendrites through nucleation, aggregation, alignment, and attachment of randomly oriented small grains [80] were imaged by using electrochemical SLCs.

To date, large progress has been made on fabrication and testing of the design features (including sealing, assembly, alignment, etc.) of SLCs [81], which opens the opportunity to address key questions on the electrode-electrolyte interfaces in native liquid environments, e.g. Kyong Wook Noh et al. [82] captured cyclic formation and dissolution of solid-electrolyte interphase at a Sn electrode in commercial liquid. Due to the reduced length scale of the electrodes, limited electrolyte volume, low current measurements [83], high vapor pressure of commercial electrolytes, and low contrast of lithium during TEM imaging through the membrane window, the application of SLCs as *in situ* electrochemical TEM cells for LIB research is still a great challenge and very limited. The application of sealed liquid cells for *in situ* TEM electrochemical characterization of lithium ion batteries are discussed as follows.

To track the lithiation process and elucidate the lithiation mechanism, the following techniques can be used: morphological imaging, electron diffraction [84, 85], energy-dispersive X-ray (EDX), and electron energy loss spectroscopy (EELS). Morphological imaging cannot give chemical information [60], and diffraction spots are quickly obscured in thicker liquid films. Moreover, lithium scatters electrons so weakly that elastic imaging is challenging and EDX signal for lithium has a much too low energy for detection. EELS offers chemical fingerprints (core-loss EELS) and electronic structure information (valence EELS), but EELS of Li in liquid will be degraded by multiple scattering events in thick liquids [86] and the lithium K-edge (~54 eV) cannot be distinguished from many transition metal (e.g. Fe [87]) edges and the superim‐ posed bulk plasmon of the thick liquid films, which makes the majority of Li-edge spectroscopy to be ex situ [88] and core-loss EELS of the lithium practically impossible in a liquid cell. Therefore, valence EELS is the best way to interrogate electronic structure and detect the state of lithiation of battery electrodes in SLCs, because valence EELS provides strong signals due to large scattering cross-sections and low background from the liquid (the electronic structure shift usually occurs at energies below ∼6−7 eV where the electrolyte is transparent and stable [60]). The spatial resolution of valence EELS is ultimately limited by the delocalization of the low-energy excitations [89], multiple scattering in the liquid environment, and low-dose imaging conditions, to be on the nanometer scale. Megan E. Holtz et al. [60] successfully observed the lithiation state by valence energy-filtered TEM (EFTEM) in thicker liquid layers than commonly allowed by core-level spectroscopy [86], probing the low-energy regime at ∼1−10 eV. They employed ab initio theory to calculate optical gaps of the relevant solvated species, taking solution effects into account with a hybrid function [90] including a nonlinear description of the polarization response of the surrounding liquid. By combining electrochem‐ istry in the TEM with valence spectroscopic imaging and theory, they identified the lithiation state of both the electrode and electrolyte during *in situ* operation. Their work demonstrated the unique ability of an *in situ* TEM SLC to observe the Li de/insertion dynamics and degra‐ dation of LiFePO4 cathode in real time. The real-time evolution of individual grains and NPs of LiFePO4 (cathode) [60] was studied in the native environment of a battery in a liquid cell TEM (as mentioned above). Particles (lithium-rich/poor) were observed to delithiate one at a time in a mosaic fashion, with different delithiation mechanisms in neighboring particles. Core-shell structures and anisotropic growth in different particles within the same agglomer‐ ate on the electrode were directly imaged along with the phase transformations, thanks to the *in situ* SLC design. Although they used Li2SO4 aqueous electrolyte due to its high abundance, less viscosity, low weight, and nontoxicity [91], volatile electrolytes could be used in their SLCs. They imaged at 5 eV with a 5 eV wide energy window [60] to track the state of lithiation (**Figure 12**). There were clear differences between the charged (**Figure 12**, right) and the discharged state (**Figure 12**, left) in both the particles and the solution in the 5 eV spectroscopic

Typical commercially available Li-ion batteries usually use carbonate-based liquids as electrolytes, such as diethyl carbonate (DEC), dimethyl carbonate (DMC) mixed with ethylene carbonate (EC), etc. To enable *in situ* TEM characterization of electrochemical reactions in real lithium ion batteries, sealing those volatile liquids inside a sufficiently narrow channel for electron transmission is a wise option. The SLC platforms enable *in situ* electrochemical characterization with any types of electrolytes with a sealed-cell configuration, thus promoting the potential use of volatile carbon-based electrolytes for LIB research. One of the first TEM SCLs was created by F.M. Ross et al. [54, 74] to study Cu electro-deposition during TEM imaging. This SLC platform sealed the aqueous electrolyte by assembling two silicon chips with thin silicon nitride membranes in a face-to-face configuration. This flip-chip approach allows imaging chemical reactions in liquids with high spatial resolution [54, 57, 58, 62, 75] with different membranes of silicon nitride, silicon dioxide, or polymer, such that it has been adopted in various studies, including cell imaging [76–78] and NP synthesis [58] in solutions. For example, electrochemical deposition of polycrystalline Au [75], anisotropic electrodepo‐ sition of nickel nanograins [79], and electrochemical growth of single crystal lead dendrites through nucleation, aggregation, alignment, and attachment of randomly oriented small

To date, large progress has been made on fabrication and testing of the design features (including sealing, assembly, alignment, etc.) of SLCs [81], which opens the opportunity to address key questions on the electrode-electrolyte interfaces in native liquid environments, e.g. Kyong Wook Noh et al. [82] captured cyclic formation and dissolution of solid-electrolyte interphase at a Sn electrode in commercial liquid. Due to the reduced length scale of the electrodes, limited electrolyte volume, low current measurements [83], high vapor pressure of commercial electrolytes, and low contrast of lithium during TEM imaging through the membrane window, the application of SLCs as *in situ* electrochemical TEM cells for LIB research is still a great challenge and very limited. The application of sealed liquid cells for *in situ* TEM electrochemical characterization of lithium ion batteries are discussed as follows.

To track the lithiation process and elucidate the lithiation mechanism, the following techniques can be used: morphological imaging, electron diffraction [84, 85], energy-dispersive X-ray (EDX), and electron energy loss spectroscopy (EELS). Morphological imaging cannot give chemical information [60], and diffraction spots are quickly obscured in thicker liquid films. Moreover, lithium scatters electrons so weakly that elastic imaging is challenging and EDX signal for lithium has a much too low energy for detection. EELS offers chemical fingerprints (core-loss EELS) and electronic structure information (valence EELS), but EELS of Li in liquid will be degraded by multiple scattering events in thick liquids [86] and the lithium K-edge (~54 eV) cannot be distinguished from many transition metal (e.g. Fe [87]) edges and the superim‐ posed bulk plasmon of the thick liquid films, which makes the majority of Li-edge spectroscopy to be ex situ [88] and core-loss EELS of the lithium practically impossible in a liquid cell. Therefore, valence EELS is the best way to interrogate electronic structure and detect the state of lithiation of battery electrodes in SLCs, because valence EELS provides strong signals due to large scattering cross-sections and low background from the liquid (the electronic structure shift usually occurs at energies below ∼6−7 eV where the electrolyte is transparent and stable

grains [80] were imaged by using electrochemical SLCs.

282 Microscopy and Analysis

**Figure 12.** *In situ* charging and discharging of the cathode material LiFePO4 in 0.5 M Li2SO4 aqueous electrolyte: the 5 eV spectroscopic EFTEM images of charging and discharging at indicated times. Scale bar is 400 nm. Bright regions are delithiated FePO4 and dark regions are LiFePO4. There are more bright regions of FePO4 at the end of charge cycles and less during the discharges. White arrows point toward "bright" charged particles, and black arrows point toward "dark" discharged particles [60].

images. Particles showed more bright regions (corresponding to delithiated FePO4) in the charged state. The cluster of particles was brighter in the charged image as marked by black arrows, especially around the edges of the cluster. The brightest particles may correspond to completely delithiated FePO4, whereas the overall slight increase in intensity in the particles may indicate partially delithiated particles. On discharge, these bright regions of FePO4 disappeared, transitioning back to LiFePO4.

The electrochemical lithiation of Au electrode, dendritic growth of crystalline lithium, and the subsequent stripping of lithium and thinning of Li-Au layer under the applied cyclic voltam‐ metry was observed by Zeng et al. [73], using commercial LiPF6/EC/DEC electrolyte, which proved that real electrolyte of LIBS can be used in electrochemical SLCs [73]. **Figure 13 (A)**−**(J)** shows the sequential images representing the early stage of electrolyte decomposition, lithiation of gold electrode, and the subsequent growth and dissolution of lithium dendrites. **Figure 13 (k)** shows the corresponding applied electrical potential and measured electrical current from frame **(A)** to frame **(J)**. The thickness of Li-Au alloy did not change drastically at the later stage, as shown in **Figure 13(L)**. During stripping, the dissolution of plated lithium starts from the tip and the kink points as a reverse process of plating (**Figure 13(M)**). The formation of SEI layer on the other side of the electrode was also captured for better under‐ standing of correlation between cyclic stability and the passivating film formed during the charge-discharge process in real LIBs. The drawback in their design was a lack of lithium metal source inside the SLC to supply the consumed lithium ions, such that the Li ion concentration in the electrolyte changed during the reaction. Adding a lithium metal source and an additional reference electrode into the SLC is necessary for direct comparison between the electrochemical processes inside a TEM column and that in real LIBs.

**Figure 13.** (A−J) Time evolution of the growth and dissolution of Li-Au alloy and lithium dendrite; (K) the correspond‐ ing applied electric potential and measured electric current from frame A to frame J; (L) plot of Li-Au layer thickness and area as a function of time; (M) dimension and area evolution of the lithium dendrite tip as a function of time dur‐ ing cyclic voltammetry in the voltage range of 0 to −3 V at scan rate of 0.1 V/s [73].

SLCs have also been used to study the stabilities of different electrolytes commonly used for Li-ion and Li-O2 battery [92, 93]. Five different electrolytes [94], including LiAsF6 salt dissolved in three different organic solvents: (1) 1,3-dioxolane (DOL), (2) DMC, (3) a mixture of DMC and EC and LiTf in dimethyl sulfoxide (DMSO), LiPF6 in EC/DMC were studied. **Figure 14** shows six different time series of bright-field (BF) STEM images corresponding to the five electrolyte solutions and the EC/DMC solvent alone. To ensure that the observed lack of degradation products when imaging the LiTf:DMSO mixture was not a result of improper focus, the edge of the window was recorded as a reference. **Figure 14** shows that apart from LiTf in DMSO, all the other salt-containing solutions tested showed some evidence of degra‐ dation. It is worth to note that the degradations of the electrolytes were triggered by the imaging electrons (300 kV), instead of extra electrodes. This work shows that the electron-beam in the STEM can be used as an effective tool for evaluating stability and degradation in battery electrolytes by allowing direct visualization of the reductive decomposition of the electrolyte components, instead of postmortem analysis (chromatography) [95, 96]. This *in situ* approach can potentially be used for more rapid identification of next-generation electrolytes.

images. Particles showed more bright regions (corresponding to delithiated FePO4) in the charged state. The cluster of particles was brighter in the charged image as marked by black arrows, especially around the edges of the cluster. The brightest particles may correspond to completely delithiated FePO4, whereas the overall slight increase in intensity in the particles may indicate partially delithiated particles. On discharge, these bright regions of FePO4

The electrochemical lithiation of Au electrode, dendritic growth of crystalline lithium, and the subsequent stripping of lithium and thinning of Li-Au layer under the applied cyclic voltam‐ metry was observed by Zeng et al. [73], using commercial LiPF6/EC/DEC electrolyte, which proved that real electrolyte of LIBS can be used in electrochemical SLCs [73]. **Figure 13 (A)**−**(J)** shows the sequential images representing the early stage of electrolyte decomposition, lithiation of gold electrode, and the subsequent growth and dissolution of lithium dendrites. **Figure 13 (k)** shows the corresponding applied electrical potential and measured electrical current from frame **(A)** to frame **(J)**. The thickness of Li-Au alloy did not change drastically at the later stage, as shown in **Figure 13(L)**. During stripping, the dissolution of plated lithium starts from the tip and the kink points as a reverse process of plating (**Figure 13(M)**). The formation of SEI layer on the other side of the electrode was also captured for better under‐ standing of correlation between cyclic stability and the passivating film formed during the charge-discharge process in real LIBs. The drawback in their design was a lack of lithium metal source inside the SLC to supply the consumed lithium ions, such that the Li ion concentration in the electrolyte changed during the reaction. Adding a lithium metal source and an additional reference electrode into the SLC is necessary for direct comparison between the electrochemical

**Figure 13.** (A−J) Time evolution of the growth and dissolution of Li-Au alloy and lithium dendrite; (K) the correspond‐ ing applied electric potential and measured electric current from frame A to frame J; (L) plot of Li-Au layer thickness and area as a function of time; (M) dimension and area evolution of the lithium dendrite tip as a function of time dur‐

ing cyclic voltammetry in the voltage range of 0 to −3 V at scan rate of 0.1 V/s [73].

disappeared, transitioning back to LiFePO4.

284 Microscopy and Analysis

processes inside a TEM column and that in real LIBs.

**Figure 14.** e<sup>−</sup> beam-induced breakdown of different electrolytes upon irradiation. (a−e) Cropped BF STEM images showing the time evolution of five different electrolytes at serial exposure times. (f) Frames from a data set probing the stability of the solvent EC/DMC for the same dose conditions as above over 7 min of continuous irradiation. (g) TEM images of an irradiated area of the LiAsF6 in DMC mixture after separating and washing the Si chips for performing postmortem analysis. Low-magnification (left) and high-resolution TEM and consequent fast Fourier transform of the irradiated area shows the presence of LiF nanocrystals [94].

The lithiation and delithiation process of fully submerged electrodes is another important application of SLCs. For example, the de/lithiation process of Si NW electrodes during electrochemical testing was observed [88] by using *in situ* SLC platform and real electrolyte (as mentioned previously). The structural evolution of the Si NW upon lithiation is illustrated in **Figure 15(a)**−**(c)**. The pristine Cu-Si NW has an overall diameter of ∼100 nm as revealed by the dark contrast in **Figure 15(a)**. The width of the Cu coating on the Si NW was measured to be ∼80 nm. The lithiation of the Si nanowire immersed in the liquid electrolyte progressed in the core-shell fashion. The total diameter of the wire changed from 100 to 298 nm at 1658 s (**Figure 15(b)**) and to 391 nm at 2462 s (**Figure 15(c)**). The diameter as a function of lithiation time is plotted in **Figure 15(d)**. The increase of the diameter was quicker at the beginning of the lithiation and slowed down with the progression of the lithiation process. The lithiation behavior observed by the *in situ* SLC was also compared with that obtained based on the opencell configuration in their study. For the case of SLC, the Si NW was fully immersed in the liquid electrolyte so that the insertion of lithium ions into Si was from all possible directions at the same time. The lithiation of the single nanowire proceeded in a core-shell mode with a uniform shell thickness along the axial direction of the whole nanowire, providing a global view of the response of the whole single NW with lithium insertion. However, for the opencell configuration the lithium ion source was only in contact with the end of the Si nanowire, leading to the sequential lithiation process of the nanowire in only one direction.

**Figure 15.** *In situ* liquid-cell TEM observation of the lithiation of the Cu-coated Si (Cu-Si) NW. (a) TEM image showing the pristine state of the Cu-Si NW at 0 s; (b) core-shell formation of the Cu-Si NW during lithiation at 1658 s; (c) TEM image of the Cu-Si NW at 2462 s; (d) plotted width changes of the NW as a function of time. Note that, in all images from a to c, the Pt contact region is labeled by the black lines in the left of the image. The inset in panel c illustrating the cross-sectional image after anisotropic swelling of the Si nanowire upon lithium insertion with maximum volume ex‐ pansion along the <100> direction [88].

Apart from Si NW electrodes, Si NPs as electrode material were also studied by using G-SLC [56, 97], which showed that the very first lithiation at the Si-electrolyte interface had the strong orientation dependence favoring the <110> directions, but then the Li diffusion occurred isotropically after passing the initial stage regardless of the NP size. This indicated that the rate-limiting diffusion barrier is at Si-electrolyte interfaces instead of within Si or at the interfaces between lithiated and unlithiated regions. The orientation-dependent initial lithiation phenomenon was evidenced by HRTEM images as well as electron diffraction analyses, as shown in **Figure 16**. For the representative three Si NPs (whose original diameters are 34, 83, and 103 nm), their morphological and dimensional changes along <110>, <111>, and <100> directions were monitored. The selected-area electron diffraction patterns (the left ones in **Figure 16** (**a**-**c**)) indicated that all of the three Si NPs were single-crystalline with <110> zone axes. The lithiation progressed predominantly along <110> directions, leading to the aniso‐ tropic volume expansion along the same crystal orientations, as indicated by the white arrows in **Figure 16** (**a**–**c**).

The lithiation and delithiation process of fully submerged electrodes is another important application of SLCs. For example, the de/lithiation process of Si NW electrodes during electrochemical testing was observed [88] by using *in situ* SLC platform and real electrolyte (as mentioned previously). The structural evolution of the Si NW upon lithiation is illustrated in **Figure 15(a)**−**(c)**. The pristine Cu-Si NW has an overall diameter of ∼100 nm as revealed by the dark contrast in **Figure 15(a)**. The width of the Cu coating on the Si NW was measured to be ∼80 nm. The lithiation of the Si nanowire immersed in the liquid electrolyte progressed in the core-shell fashion. The total diameter of the wire changed from 100 to 298 nm at 1658 s (**Figure 15(b)**) and to 391 nm at 2462 s (**Figure 15(c)**). The diameter as a function of lithiation time is plotted in **Figure 15(d)**. The increase of the diameter was quicker at the beginning of the lithiation and slowed down with the progression of the lithiation process. The lithiation behavior observed by the *in situ* SLC was also compared with that obtained based on the opencell configuration in their study. For the case of SLC, the Si NW was fully immersed in the liquid electrolyte so that the insertion of lithium ions into Si was from all possible directions at the same time. The lithiation of the single nanowire proceeded in a core-shell mode with a uniform shell thickness along the axial direction of the whole nanowire, providing a global view of the response of the whole single NW with lithium insertion. However, for the opencell configuration the lithium ion source was only in contact with the end of the Si nanowire,

leading to the sequential lithiation process of the nanowire in only one direction.

**Figure 15.** *In situ* liquid-cell TEM observation of the lithiation of the Cu-coated Si (Cu-Si) NW. (a) TEM image showing the pristine state of the Cu-Si NW at 0 s; (b) core-shell formation of the Cu-Si NW during lithiation at 1658 s; (c) TEM image of the Cu-Si NW at 2462 s; (d) plotted width changes of the NW as a function of time. Note that, in all images from a to c, the Pt contact region is labeled by the black lines in the left of the image. The inset in panel c illustrating the cross-sectional image after anisotropic swelling of the Si nanowire upon lithium insertion with maximum volume ex‐

pansion along the <100> direction [88].

286 Microscopy and Analysis

**Figure 16.** Morphological and dimensional changes of Si NPs analyzed by GLC-TEM during the course of lithiation: (a–c) Time series bright-field TEM images of the Si NPs with initial diameters of 34, 83, and 103 nm, respectively. The white arrows indicate the Si<110> directions. The SA-EDPs in (a–c) indicate crystalline nature of the pristine Si NPs and their crystallographic orientations along <110> zone axes. The scale bars in (a–c) are 20 nm [97].

Based on the above discussions, the major research achievements to date in applications of SLCs for *in situ* TEM electrochemical characterization of LIBs are highlighted as follows:


## **4. Summary and future research directions**

In sum, sealed gas cells for *in situ* ETEM observation and sealed liquid cells for *in situ* TEM electrochemical characterization of LIBs have been reviewed in this article. Sealed cells have various advantages over the opened-type approach, thus becoming the dominant way to perform ETEM studies under gaseous/liquid environment. However, improvements are still needed in the following aspects for the benign development of this technology.

For sealed gas cells, thermal expansion/contraction and the consequent sample drift need to be minimized during *in situ* ETEM observation. The key to realize this is to improve ways of heating the sample. Localized heating sources such as laser or infrared light can be used for this purpose. Furthermore, the gas path length for electrons to go through the sealed gas cells needs to be decreased, possibly by better designs of the cell configurations. For example, if the heater can be integrated into the window instead of hanging out in the middle of the cell, the thickness of the spacer can be decreased such that a shorter gas path length is accomplished. Moreover, the pressure limit within the sealed gas cell should be increased, so that this technique can be applied to increasingly more fields of study. This can be achieved by various approaches: 1) since the major concern for limiting the pressure is the mechanical strength of the windows, either modifying the manufacturing process of the windows or replacing the current material with new ones may work well; 2) the configurations of the windows can be improved to withstand a higher pressure. For example, a single window with a large surface area is easier to break under high pressure, compared with several smaller windows distrib‐ uting evenly. Last but not least, the solid-gas reactions need to be controlled and monitored better by developing ways to measure the parameters of the sample inside the cell more precisely. The direct approach to achieve this goal is to integrate measurement devices, such as nanoscale thermometers and pressure gauges into the sealed gas cells. The indirect way is to use chemical analysis techniques to reflect temperature and pressure changes. For example, the drift of EELS peak positions can be used to tell the temperature change.

For sealed liquid cells, TEM characterization of lithium through a liquid is challenging because lithium is a weak elastic scatterer and multiple scattering from the liquid could dominate the signal, resulting to a poor spatial resolution and contrast. Consequently, the spatial resolution and contrast needs to be enhanced. The future research direction is to improve SLC designs for better spatial resolution, which can be realized by decreasing the thickness of the liquid layer, redesigning the electrode configuration, utilizing alternative viewing window materials, employing different electrolytes, controlling electron dose, optimizing spacer thickness, etc. With an improved spatial resolution, various new research frontiers, including defect structure within SEI layer during (de)lithiation process, can be explored by using SLCs in TEM, rendering it the most promising technique for *in situ* electrochemical characterization of LIBs.

## **Author details**

Based on the above discussions, the major research achievements to date in applications of SLCs for *in situ* TEM electrochemical characterization of LIBs are highlighted as follows:

**1.** The stabilities and degradation mechanisms of commercial electrolytes commonly used for Li-ion and Li-O2 battery were studied [94], providing reference for future choices on

**2.** The lithiation and delithiation process of fully submerged electrodes, such as Si NW electrodes [88, 98], Sn electrodes [82], and Si NPs [97], during electrochemical testing were

**3.** The electrochemical lithiation process, such as dendritic growth of crystalline lithium [60], stripping of lithium [60], SEI layer formation [99, 100], etc., were observed in real time with nanoscale resolution during electrochemical charge and discharge [60] using

In sum, sealed gas cells for *in situ* ETEM observation and sealed liquid cells for *in situ* TEM electrochemical characterization of LIBs have been reviewed in this article. Sealed cells have various advantages over the opened-type approach, thus becoming the dominant way to perform ETEM studies under gaseous/liquid environment. However, improvements are still

For sealed gas cells, thermal expansion/contraction and the consequent sample drift need to be minimized during *in situ* ETEM observation. The key to realize this is to improve ways of heating the sample. Localized heating sources such as laser or infrared light can be used for this purpose. Furthermore, the gas path length for electrons to go through the sealed gas cells needs to be decreased, possibly by better designs of the cell configurations. For example, if the heater can be integrated into the window instead of hanging out in the middle of the cell, the thickness of the spacer can be decreased such that a shorter gas path length is accomplished. Moreover, the pressure limit within the sealed gas cell should be increased, so that this technique can be applied to increasingly more fields of study. This can be achieved by various approaches: 1) since the major concern for limiting the pressure is the mechanical strength of the windows, either modifying the manufacturing process of the windows or replacing the current material with new ones may work well; 2) the configurations of the windows can be improved to withstand a higher pressure. For example, a single window with a large surface area is easier to break under high pressure, compared with several smaller windows distrib‐ uting evenly. Last but not least, the solid-gas reactions need to be controlled and monitored better by developing ways to measure the parameters of the sample inside the cell more precisely. The direct approach to achieve this goal is to integrate measurement devices, such as nanoscale thermometers and pressure gauges into the sealed gas cells. The indirect way is to use chemical analysis techniques to reflect temperature and pressure changes. For example,

needed in the following aspects for the benign development of this technology.

the drift of EELS peak positions can be used to tell the temperature change.

observed by using *in situ* SLC platform and real electrolyte.

electrolytes.

288 Microscopy and Analysis

commercial electrolyte [73].

**4. Summary and future research directions**

Fan Wu\* and Nan Yao

\*Address all correspondence to: fanwu@princeton.edu

Princeton Institute for the Science and Technology of Materials (PRISM), Princeton University, USA

## **References**


[21] Parkinson GM. High resolution, in-situ controlled atmosphere transmission electron microscopy (CATEM) of heterogeneous catalysts. Catal Lett. 1989;2:303–7.

[7] Daulton TL, Little BJ, Lowe K, Jones-Meehan J. In situ environmental cell–transmission electron microscopy study of microbial reduction of chromium(VI) using electron

[8] Creemer JF, Helveg S, Hoveling GH, Ullmann S, Molenbroek AM, Sarro PM, et al. Atomic-scale electron microscopy at ambient pressure. Ultramicroscopy. 2008;108:993–

[9] Allard LF, Overbury SH, Bigelow WC, Katz MB, Nackashi DP, Damiano J. Novel MEMS-based gas-cell/heating specimen holder provides advanced imaging capabili‐ ties for in situ reaction studies. Microscopy and Microanalysis. 2012;18:656–66.

[10] Yaguchi T, Suzuki M, Watabe A, Nagakubo Y, Ueda K, Kamino T. Development of a high temperature-atmospheric pressure environmental cell for high-resolution TEM.

[11] de Jonge N, Bigelow WC, Veith GM. Atmospheric pressure scanning transmission

[12] Rajasekhara S, Hattar KM, Tikare V, Dingreville RPM, Clark B. Hydride formation in cladding materials studied via in-situ environmental heating transmission electron

[13] Rose HH. Historical aspects of aberration correction. Journal of Electron Microscopy.

[14] Marton L. La microscopie electronique des objets biologiques. Bulletin de l'Academie

[15] Tanaka N, Usukura J, Kusunoki M, Saito Y, Sasaki K, Tanji T, et al. Development of an environmental high-voltage electron microscope for reaction science. Microscopy.

[16] Baker RTK, Harris PS. Controlled atmosphere electron microscopy. Journal of Physics

[17] Butler EP. In situ experiments in the transmission electron microscope. Reports on

[18] Boyes ED, Gai PL. Environmental high resolution electron microscopy and applications

[19] Sharma R. Design and applications of environmental cell transmission electron microscope for in situ observations of gas–solid reactions. Microscopy and Microanal‐

[20] N. Yao GES, R. A. Kemp, D. C. Guthrie, R. D. Cates, C. M. Bolinger. Environmental cell TEM studies of catalyst particle behavior. In: Bailey G, editor. 49th Annual Conference

Journal of Electron Microscopy. 2011;60:217–25.

electron microscopy. Nano Letters. 2010;10:1028–31.

de Belgique Classe des Sciences (5). 1937;28:672–5.

to chemical science. Ultramicroscopy. 1997;67:219–32.

of EMSA. San Francisco: San Francisco Press; 1991. p. 1028–9.

E: Scientific Instruments. 1972;5:793.

Progress in Physics. 1979;42:833.

energy loss spectroscopy. Microscopy and Microanalysis. 2001;7:470–85.

8.

290 Microscopy and Analysis

microscopy. 2012.

2009;58:77–85.

2013;62:205–15.

ysis. 2001;7:494–506.


[48] Vendelbo SB, Kooyman PJ, Creemer JF, Morana B, Mele L, Dona P, et al. Method for local temperature measurement in a nanoreactor for in situ high-resolution electron microscopy. Ultramicroscopy. 2013;133:72–9.

[34] Wu F. Planar defects in metallic thin film heterostructures: North Carolina State

[35] Bayati M, Molaei R, Wu F, Budai J, Liu Y, Narayan R, et al. Correlation between structure and semiconductor-to-metal transition characteristics of VO2TiO2/sapphire

[36] Wu F, Narayan J. Controlled epitaxial growth of body-centered cubic and face-centered cubic Cu on MgO for integration on Si. Crystal Growth & Design. 2013;13:5018–24.

[37] Molaei R, Bayati R, Wu F, Narayan J. A microstructural approach toward the effect of thickness on semiconductor-to-metal transition characteristics of VO2 epilayers.

[38] Wu F, Zhu YT, Narayan J. Grain size effect on twin density in as-deposited nanocrys‐

[39] Lee YF, Wu F, Narayan J, Schwartz J. Oxygen vacancy enhanced room-temperature ferromagnetism in Sr3SnO/c-YSZ/Si (001) heterostructures. MRS Communications.

[40] Wu F, Zhu YT, Narayan J. Macroscopic twinning strain in nanocrystalline Cu. Materials

[41] Lee YF, Wu F, Kumar R, Hunte F, Schwartz J, Narayan J. Epitaxial integration of dilute magnetic semiconductor Sr3SnO with Si (001). Applied Physics Letters. 2013;103:2101.

[42] Wu F, Wen HM, Lavernia EJ, Narayan J, Zhu YT. Twin intersection mechanisms in nanocrystalline fcc metals. Materials Science and Engineering: A. 2013;585:292–6. [43] Gupta N, Singh R, Wu F, Narayan J, McMillen C, Alapatt GF, et al. Deposition and characterization of nanostructured Cu2O thin-film for potential photovoltaic applica‐

[44] Sharma R. Experimental set up for in situ transmission electron microscopy observa‐ tions of chemical processes. Micron (Oxford, England : 1993). 2012;43:1147–55.

[45] Vendelbo SB, Elkjær CF, Falsig H, Puspitasari I, Dona P, Mele L, et al. Visualization of oscillatory behaviour of Pt nanoparticles catalysing CO oxidation. Nat Mater.

[46] Sun L, Noh KW, Wen J-G, Dillon SJ. In situ Transmission Electron Microscopy Obser‐ vation of Silver Oxidation in Ionized/Atomic Gas. Langmuir. 2011;27:14201–6.

[47] Yokosawa T, Alan T, Pandraud G, Dam B, Zandbergen H. In-situ TEM on (de)hydro‐ genation of Pd at 0.5–4.5 bar hydrogen pressure and 20–400°C. Ultramicroscopy.

thin film heterostructures. Acta Materialia. 2013;61:7805–15.

talline Cu film. Philosophical Magazine. 2013;93:4355–63.

tions. Journal of Materials Research. 2013;28:1740–6.

Journal of Applied Physics. 2014;115:4311.

University; 2014.

292 Microscopy and Analysis

2014;4:7–13.

2014;13:884–90.

2012;112:47–52.

Research Letters. 2013;2:63–9.


[75] Radisic A, Vereecken PM, Hannon JB, Searson PC, Ross FM. Quantifying electrochem‐ ical nucleation and growth of nanoscale clusters using real-time kinetic data. Nano Letters. 2006;6:238–42.

[62] Liao H-G, Cui L, Whitelam S, Zheng H. Real-Time Imaging of Pt3Fe Nanorod Growth

[63] Woehl TJ, Park C, Evans JE, Arslan I, Ristenpart WD, Browning ND. Direct observation of aggregative nanoparticle growth: kinetic modeling of the size distribution and

[64] Evans JE, Jungjohann KL, Wong PCK, Chiu P-L, Dutrow GH, Arslan I, et al. Visualizing macromolecular complexes with in situ liquid scanning transmission electron micro‐

[65] Mirsaidov UM, Zheng H, Casana Y, Matsudaira P. Imaging protein structure in water at 2.7 nm resolution by transmission electron microscopy. Biophysical Journal.

[66] Huang T-W, Liu S-Y, Chuang Y-J, Hsieh H-Y, Tsai C-Y, Wu W-J, et al. Dynamics of hydrogen nanobubbles in KLH protein solution studied with in situ wet-TEM. Soft

[67] Proetto MT, Rush AM, Chien M-P, Abellan Baeza P, Patterson JP, Thompson MP, et al. Dynamics of soft nanomaterials captured by transmission electron microscopy in liquid

[68] Evans JE, Jungjohann KL, Browning ND, Arslan I. Controlled growth of nanoparticles from solution with in situ liquid transmission electron microscopy. Nano Letters.

[69] Zheng H, Claridge SA, Minor AM, Alivisatos AP, Dahmen U. Nanocrystal diffusion in a liquid thin film observed by in situ transmission electron microscopy. Nano Lett.

[70] Jungjohann KL, Bliznakov S, Sutter PW, Stach EA, Sutter EA. In situ liquid cell electron microscopy of the solution growth of Au–Pd core-shell nanostructures. Nano Letters.

[71] Parent LR, Robinson DB, Cappillino PJ, Hartnett RJ, Abellan P, Evans JE, et al. In situ observation of directed nanoparticle aggregation during the synthesis of ordered

[72] Aurbach D, Markovsky B, Levi MD, Levi E, Schechter A, Moshkovich M, et al. New insights into the interactions between electrode materials and electrolyte solutions for

advanced nonaqueous batteries. Journal of Power Sources. 1999;81–82:95–111.

[73] Zeng Z, Liang W-I, Liao H-G, Xin HL, Chu Y-H, Zheng H. Visualization of electrode– electrolyte interfaces in LiPF6/EC/DEC electrolyte for lithium ion batteries via in situ

[74] Ross FM. Growth processes and phase transformations studied by in situ transmission

nanoporous metal in soft templates. chemistry of materials. 2014;26:1426–33.

water. Journal of the American Chemical Society. 2014;136:1162–5.

in Solution. Science. 2012;336:1011–4.

growth rate. Nano Lett. 2014;14:373–8.

2012;102:L15-L7.

294 Microscopy and Analysis

2011;11:2809–13.

2009;9:2460–5.

2013;13:2964–70.

TEM. Nano Letters. 2014;14:1745–50.

electron microscopy. IBM J Res Dev. 2000;44:489–501.

Matter. 2013;9:8856–61.

scopy. Micron (Oxford, England : 1993). 2012;43:1085–90.


Society of America, Microbeam Analysis Society, Microscopical Society of Canada. 2013;19:1027–35.


electrochemical transmission electron microscopy. microscopy and microanalysis. 2014;20:1029–37.

[100] Sacci RL, Dudney NJ, More KL, Parent LR, Arslan I, Browning ND, et al. Direct visualization of initial SEI morphology and growth kinetics during lithium deposition by in situ electrochemical transmission electron microscopy. Chemical Communica‐ tions. 2014;50:2104–7.

Society of America, Microbeam Analysis Society, Microscopical Society of Canada.

[87] Moreau P, Boucher F. Revisiting lithium K and iron M(2),(3) edge superimposition: the case of lithium battery material LiFePO(4). Micron (Oxford, England : 1993). 2012;43:16–

[88] Gu M, Parent LR, Mehdi BL, Unocic RR, McDowell MT, Sacci RL, et al. Demonstration of an electrochemical liquid cell for operando transmission electron microscopy observation of the lithiation/delithiation behavior of Si nanowire battery anodes. Nano

[89] Muller DA, Silcox J. Delocalization in inelastic scattering. Ultramicroscopy.

[90] Ernzerhof M, Scuseria GE. Assessment of the Perdew–Burke–Ernzerhof exchangecorrelation functional. The Journal of Chemical Physics. 1999;110:5029–36.

[91] Wang Y, Yi J, Xia Y. Recent progress in aqueous lithium-ion batteries. Advanced Energy

[92] Xu K. Nonaqueous liquid electrolytes for lithium-based rechargeable batteries.

[93] Nasybulin E, Xu W, Engelhard MH, Nie Z, Burton SD, Cosimbescu L, et al. Effects of electrolyte salts on the performance of Li–O2 batteries. The Journal of Physical Chem‐

[94] Abellan P, Mehdi BL, Parent LR, Gu M, Park C, Xu W, et al. Probing the degradation mechanisms in electrolyte solutions for Li-ion batteries by in situ transmission electron

[95] Gachot G, Grugeon S, Armand M, Pilard S, Guenot P, Tarascon J-M, et al. Deciphering the multi-step degradation mechanisms of carbonate-based electrolyte in Li batteries.

[96] Gachot Gg, Ribière P, Mathiron D, Grugeon S, Armand M, Leriche J-B, et al. Gas chromatography/mass spectrometry as a suitable tool for the Li-ion battery electrolyte

[97] Yuk JM, Seo HK, Choi JW, Lee JY. Anisotropic lithiation onset in silicon nanoparticle anode revealed by in situ graphene liquid cell electron microscopy. ACS Nano.

[98] Layla Mehdi B, Gu M, Parent LR, Xu W, Nasybulin EN, Chen X, et al. In-situ electro‐ chemical transmission electron microscopy for battery research. Microscopy and

[99] Unocic RR, Sun X-G, Sacci RL, Adamczyk LA, Alsem DH, Dai S, et al. Direct visuali‐ zation of solid electrolyte interphase formation in lithium-ion batteries with in situ

degradation mechanisms study. Analytical Chemistry. 2010;83:478–85.

2013;19:1027–35.

Letters. 2013;13:6106–12.

Materials. 2012;2:830–40.

istry C. 2013;117:2635–45.

2014;8:7478–85.

Microanalysis. 2014;20:484–92.

Chemical Reviews. 2004;104:4303–418.

microscopy. Nano Letters. 2014;14:1293–9.

Journal of Power Sources. 2008;178:409–21.

1995;59:195–213.

21.

296 Microscopy and Analysis

## **Chapter 12**

## **Advanced Scanning Tunneling Microscopy for Nanoscale Analysis of Semiconductor Devices**

Leonid Bolotov and Toshihiko Kanayama

Additional information is available at the end of the chapter

http://dx.doi.org/10.5772/62552

#### **Abstract**

Significant attention has been addressed to high-spatial resolution analysis of modern sub-100-nm electronic devices to achieve new functions and energy-efficient opera‐ tions. The chapter presents a review of ongoing research on charge carrier distribution analysis in nanoscale Si devices by using scanning tunneling microscopy (STM) employing advanced operation modes: a gap-modulation method, a molecule-assisted probing method, and a dual-imaging method. The described methods rely on detection and analysis of tunneling current, which is strongly localized within an atomic dimension. Representative examples of applications to nanoscale analysis of Si device crosssections and nanowires are given. Advantages, difficulties, and limitations of the advanced STM methods are discussed in comparison with other techniques used in a field of device metrology.

**Keywords:** scanning tunneling microscopy, semiconductor devices, charge carrier distribution, resonant electron tunneling, silicon-on-insulator, photocarrier profiling, fullerene molecule

### **1. Introduction**

Since invention of solid-state electric junctions, charge carrier distribution has become the primary requirement of electronic device design to achieve desirable device performance. Typically, a spatial distribution of charge carriers in semiconductor devices is created by introduction of electronic impurity atoms with particular electron configuration allowing to donate a free electron to the host semiconductor (donor impurity) or to trap a valence electron (acceptor impurity) from the host material. Thus, the host semiconductor with donor impuri‐

© 2016 The Author(s). Licensee InTech. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

ty atoms has become a negative-charge (electrons) conductor and is called *n-type*. The host semiconductor with acceptor impurity atoms has become a positive-charge (holes) conductor and is called *p-type*. Typical semiconductor devices have concentration of impurity atoms in a rangeof 1015–1021/cm3 ,whichis less than1%oftotalnumberof atoms.Defects andatomvacancy often behave like impurity atoms.

Early days, charge carrier distribution was derived from spatial distributions of impurity atoms in semiconductor materials. Secondary ion mass spectrometry (SIMS) has been used to obtain a depth distribution profile of impurity atoms in semiconductor materials by sputtering with high-energy ions. As modern high-performance Si devices such as complementary metaloxide-semiconductor (CMOS) transistors are less than 100 nm in size, and have complex material structures, the 1D SIMS profiling becomes inadequate. **Figure 1** shows a typical structure of a metal-oxide-semiconductor field effect transistor (MOSFET) consisting of gate, channel, and source/drain regions with high impurity concentrations.

**Figure 1.** (a) A sketch of a MOSFET device in a cross-section. (b) 3D view of a charge distribution in a MOSFET meas‐ ured by the STM gap modulation method. Charge concentration is emphasized by color: blue color for low charge con‐ centration in p-Si channel and red color for high charge concentration in source (S), drain (D), and gate (G).

Recently, a new technique of tree-dimensional (3D) atom mapping, which is called atom probe tomography, was introduced based on counting of atom ions ejected from a needle-like device specimen [1–6]. Aside from complexity of the sample preparation and 3D data reconstruction of the atom probe technique, the charge carrier distribution is assumed to be equal to that of impurity atoms. However, the carrier distribution deviates significantly from the impurity atom distribution as a result of internal electric field at material interfaces, trapped charges in oxide, and fractional activation of impurity atoms in areas of high impurity concentration. Therefore, techniques allowing to measure local distribution of charge carriers within the electronic device interior have been a focus of attention from scientific and practical points of view.

Significant attention has been addressed to high-spatial resolution analysis of modern sub-100 nm electronic devices, nanowire devices which meet miniaturization to less than 10 nm in order to achieve new functions and energy-efficient operation. Last decade, various techniques have been developed for charge carrier mapping. A common high-resolution imaging technique, scanning electron microscopy (SEM), has been upgraded with an energy-filtering option, allowing us to obtain the image contrast as a function of the surface electrostatic potential [7–10].

ty atoms has become a negative-charge (electrons) conductor and is called *n-type*. The host semiconductor with acceptor impurity atoms has become a positive-charge (holes) conductor and is called *p-type*. Typical semiconductor devices have concentration of impurity atoms in a

Early days, charge carrier distribution was derived from spatial distributions of impurity atoms in semiconductor materials. Secondary ion mass spectrometry (SIMS) has been used to obtain a depth distribution profile of impurity atoms in semiconductor materials by sputtering with high-energy ions. As modern high-performance Si devices such as complementary metaloxide-semiconductor (CMOS) transistors are less than 100 nm in size, and have complex material structures, the 1D SIMS profiling becomes inadequate. **Figure 1** shows a typical structure of a metal-oxide-semiconductor field effect transistor (MOSFET) consisting of gate,

**Figure 1.** (a) A sketch of a MOSFET device in a cross-section. (b) 3D view of a charge distribution in a MOSFET meas‐ ured by the STM gap modulation method. Charge concentration is emphasized by color: blue color for low charge con‐

Recently, a new technique of tree-dimensional (3D) atom mapping, which is called atom probe tomography, was introduced based on counting of atom ions ejected from a needle-like device specimen [1–6]. Aside from complexity of the sample preparation and 3D data reconstruction of the atom probe technique, the charge carrier distribution is assumed to be equal to that of impurity atoms. However, the carrier distribution deviates significantly from the impurity atom distribution as a result of internal electric field at material interfaces, trapped charges in oxide, and fractional activation of impurity atoms in areas of high impurity concentration. Therefore, techniques allowing to measure local distribution of charge carriers within the electronic device interior have been a focus of attention from scientific and practical points of

Significant attention has been addressed to high-spatial resolution analysis of modern sub-100 nm electronic devices, nanowire devices which meet miniaturization to less than 10 nm in order to achieve new functions and energy-efficient operation. Last decade, various techniques have been developed for charge carrier mapping. A common high-resolution imaging

centration in p-Si channel and red color for high charge concentration in source (S), drain (D), and gate (G).

channel, and source/drain regions with high impurity concentrations.

,whichis less than1%oftotalnumberof atoms.Defects andatomvacancy

rangeof 1015–1021/cm3

300 Microscopy and Analysis

view.

often behave like impurity atoms.

Scanning probe techniques are an important tool for local probing of electric properties and have played important roles in scientific research on electronic materials and in evaluations of device structures in fabrication processes. Scanning probe microscopy (SPM) techniques are based on the ability to position a sharp probe electrode in very close proximity with high precision to the sample surface under investigation [11]. Different physical quantities can be measured by the probe including electric tunneling current, atomic and electrostatic forces, or other types of probe-sample interactions. By moving the probe laterally over the sample surface and performing measurements at different locations, two-dimensional distributions of surface atomic structure, electric current, electrostatic potential, or other properties can be obtained.

SPM techniques employed in local electrical measurements are atomic force microscopy with a conductive probe (c-AFM) [12], scanning spreading resistance microscopy (SSRM) [13], scanning Kelvin probe microscopy (SKPM) [14], and scanning tunneling microscopy (STM) [15]. These scanning probe techniques create two-dimensional (2D) maps of variations in the surface electric potential or electric current density along a cross-section of a semiconductor device, when the surface states, defects, adsorbates, and foreign particles on the cross-sectional surface do not affect the initial charge carrier distribution. In majority of cases, certain surface treatments of the cross-sectional surface are applied prior to measurements to eliminate undesirable surface effects. Quantitative impurity profiles by SSRM and SKPM have been demonstrated for high impurity concentrations, where a spatial resolution on the order of the probe tip radius (~5 nm) was obtained under optimum conditions [16–19].

STM has been used for impurity distribution measurements in Si devices by analyzing currentvoltage spectra [20–23]. To derive quantitatively variation in the charge carrier distribution from STM measurements, one must analyze complex dependence of the tunneling current on the bias voltage, the tunneling gap, and the band-bending potential beneath the STM probe tip on a semiconductor surface. Thus, simulations of STM operation are an essential part of the data analysis.

In this chapter, we focus on advanced STM-based spectroscopy techniques as nanoscale methods for two-dimensional (2D) charge carrier analysis. It represents original development of scanning probe microscopy methods for Si device metrology with ultimate spatial resolu‐ tion. We describe the principles of the advanced STM methods and give representative examples of applications to nanoscale analysis of Si CMOS devices and nanowires. Advan‐ tages, difficulties, and limitations of the advanced STM modes will be discussed in comparison with other techniques used in a field of device metrology.

The chapter begins with description of device cross-section preparation methods and essential features of STM measurements on a semiconductor surface. Measurement principles of original STM-based techniques and application examples will be given. Current development in STM simulations will be outlined. Prospects toward research in new 2D materials will be elaborated.

## **2. Preparation of Si device cross-sections**

**Figure 2** shows a common way for making solid crystal cross-sections. The process includes a number of steps. (1) Cleavage and/or *dicing* of a thin crystal wafer are used to define a desired location of the cross-sectional plane. (2) *Chemical-mechanical planarization-polishing* (CMP) and focus ion beam (FIB) techniques are applied to tune location of the cross-sectional plane with a sub-micrometer accuracy. (3) Chemical and electric *passivation* of the cross-sectional Si surface by hydrogenation or thin oxide is carried out to prevent distortion of original charge carrier distributions by surface states and contamination.

**Figure 2.** (a) Fabrication of a device cross-section for STM measurements. (b)–(c) STM images (a set-point: 300 pA, 2.0 V) of an oxide-passivated Si surface before (b) and after (c) 1 ML C60 film formation. Color scale is 0.8 nm (b) and 2.5 nm (c). Insert shows an image of 10 C60 molecules.

Chemical and electric passivation of solid surfaces is the subject of extended research in catalysis to control on charge transfer process and chemical reactions in solid-liquid and solidgas interfaces [24]. Moreover, chemical and electric passivation of semiconductor surfaces are a basic process in fabrication of modern Si devices, enabling to reduce off-state leakage current and photocarrier losses in solar cells [25]. Without passivation treatment, silicon surfaces have pronounced bands of surface states, which dominate the contrast of the STM images, so that it becomes difficult to characterize the underlying electrical interfaces. Therefore, passivation of Si surfaces by hydrogenation or oxidation has been employed in order to reproducibly prepare uniform surfaces of device cross-sections and to obtain very low density of surface states.

### **2.1. Passivation by hydrogenation**

in STM simulations will be outlined. Prospects toward research in new 2D materials will be

**Figure 2** shows a common way for making solid crystal cross-sections. The process includes a number of steps. (1) Cleavage and/or *dicing* of a thin crystal wafer are used to define a desired location of the cross-sectional plane. (2) *Chemical-mechanical planarization-polishing* (CMP) and focus ion beam (FIB) techniques are applied to tune location of the cross-sectional plane with a sub-micrometer accuracy. (3) Chemical and electric *passivation* of the cross-sectional Si surface by hydrogenation or thin oxide is carried out to prevent distortion of original charge carrier

**Figure 2.** (a) Fabrication of a device cross-section for STM measurements. (b)–(c) STM images (a set-point: 300 pA, 2.0 V) of an oxide-passivated Si surface before (b) and after (c) 1 ML C60 film formation. Color scale is 0.8 nm (b) and 2.5

Chemical and electric passivation of solid surfaces is the subject of extended research in catalysis to control on charge transfer process and chemical reactions in solid-liquid and solid-

elaborated.

302 Microscopy and Analysis

**2. Preparation of Si device cross-sections**

distributions by surface states and contamination.

nm (c). Insert shows an image of 10 C60 molecules.

Hydrogenation of Si surfaces is achieved by etching in fluoric acid solutions. Etching removes the native Si oxide and terminates the Si dangling bonds with hydrogen atoms making a stable, passivated surface with very low density of surface states in the Si band gap [26]. A number of investigations have confirmed that tunneling spectra of such stabilized Si surfaces show variation with dopant type and concentration due to passivation of dangling bond states and the suppression of surface states [27–32].

Si(111) surfaces can be atomically flattened by wet treatment in NH4F aqueous solutions [33]. In the procedure, the samples were dipped in a 5% HF solution to remove the residual oxide layer, then immersed in a 40% NH4F solution at room temperature, and rinsed in ultrapure water for 1 min. This treatment renders the Si surface mono-hydride, well suited for STM analysis. In this treatment, hydrogen also reacts with near-surface impurity atoms forming electrically inactive complexes, thus, changing the initial charge distribution. To reactivate the impurity atoms, heating of the samples around 200–250°C is necessary [33, 34].

To prepare atomically flat Si(001) surfaces, a combined process is adopted, which consists of wet treatment using a fluoric acid solution and subsequent annealing in H2 atmosphere at ~600°C and a pressure of ~2 × 103 Pa [35]. The authors showed the formation of an atomically flat Si(001) surface that have well-ordered step-terrace structures in the active device area. The flattening was attributed to the enhanced migration of Si atoms when anisotropic etching was suppressed.

#### **2.2. Passivation by an ultrathin oxide**

Hydrogenation of Si surfaces may not always be compatible with processing steps in a particular application, as Si surface etching usually introduces topographic contrast due to etching rate dependence on doping concentration, crystal orientation, and material composi‐ tion. An alternative way to passivate Si surface is oxidation. The passivation of Si surfaces by controlled growth of ultra-thin oxide layer relies on the layer-by-layer oxidation kinetics at low oxygen pressure [36–38]. We adopted the preparation of cross-sectional surfaces of Si devices as follows [39, 40]. First, dicing and ultra-fine polishing are used to expose either (100) or (110) surfaces of the device. The polished surfaces are cleaned by few cycles of etching in dilute fluoric acid solution and wet-oxidation in H2SO4:H2O2 (3:1) solution to remove a damage layer. Finally, ultra-thin (~0.3 nm) oxide layer is grown at ~600°C under an O2 pressure of 3 × 10−3 Pa following etch-cleaning in HF:HCl (1:19). This procedure left a flat surface without any ordered structure as seen in **Figure 2(b)**, where the atomically flat terraces are separated by atomic steps of 0.24–0.27 nm in height. The oxide thickness was 0.32–0.35 nm as determined by x-ray photoelectron spectroscopy, and by scanning reflection electron microscopy (SREM). The low-pressure oxidation process results in a residual density of surface state traps of ~1012 cm−2 for Si(100) surfaces [24, 41, 42], which is suitable for STM spectroscopy analysis.

## **2.3. Formation of C60 monolayer films**

When a well-defined mono-molecular layer is prepared on a passivated surface, its molecular level can be utilized to quantitatively analyze the electrical properties of the underlying substrate. We call this method as a molecule-assisted spectroscopy. For this purpose, mono‐ molecular thick films of C60 (fullerene) were formed by vapor sublimation of C60 to the oxidized Si surfaces to a thickness of 3–5 molecular layers. The excess of C<sup>60</sup> layers was removed by sample heating at 170–190°C for 10 min. Because electrostatic interaction between the molecule and the underlying Si is stronger than the Van der Waals interaction between molecules within the film, a C60 molecules adjacent to the Si surface remain at high coverage (~80%) as seen in **Figure 2(c)** [41].

## **3. Tunneling microscopy: basics**

The STM operation principle is based on quantum mechanical phenomenon—electron tunneling through a potential barrier formed by a gap between the outermost atoms on the metal tip and the sample. When the gap is about 1 nm or less, electrons from the STM tip can penetrate into the sample with certain probability owing to the wave nature of the quantum particle.

Under external electric field, electron tunneling creates a measurable electric current, the tunneling current. In the single particle approximation, the tunneling current density is given by a difference in the particle flow across the gap from the STM tip and that from the semi‐ conductor and is expressed as an integral over particle's energy

$$\begin{aligned} J\_{\text{new}} &= J\_{\rightarrow} - J\_{\leftarrow} = \bigcap\_{0} \rho\_{ap} \left( E \right) \cdot f \left( E \right) \cdot \rho\_{sample} \left( qV\_{gap} - E \right) \cdot \left[ 1 - f \left( qV\_{gap} - E \right) \right] \\ &\cdot T \left( Z, E, V\_{gap} \right) dE, \end{aligned} \tag{1}$$

where *T*(*Z*, *E*, *V*), the transmission factor, is a function of gap width (*Z*), electron energy (*E*), and external gap voltage (*Vgap*). *ρtip*(*E*) and *ρsample*(*E*) are the density of electron states at the surface of the STM tip and the sample, respectively. *f*(*E*) is the Fermi function describing which energy states are occupied with electrons.

Here, we outline the important features of the STM technique essential for analysis of charge carrier distribution in semiconductors. They are

**•** tunneling barrier shape,

10−3 Pa following etch-cleaning in HF:HCl (1:19). This procedure left a flat surface without any ordered structure as seen in **Figure 2(b)**, where the atomically flat terraces are separated by atomic steps of 0.24–0.27 nm in height. The oxide thickness was 0.32–0.35 nm as determined by x-ray photoelectron spectroscopy, and by scanning reflection electron microscopy (SREM). The low-pressure oxidation process results in a residual density of surface state traps of ~1012 cm−2 for Si(100) surfaces [24, 41, 42], which is suitable for STM spectroscopy analysis.

When a well-defined mono-molecular layer is prepared on a passivated surface, its molecular level can be utilized to quantitatively analyze the electrical properties of the underlying substrate. We call this method as a molecule-assisted spectroscopy. For this purpose, mono‐ molecular thick films of C60 (fullerene) were formed by vapor sublimation of C60 to the oxidized Si surfaces to a thickness of 3–5 molecular layers. The excess of C<sup>60</sup> layers was removed by sample heating at 170–190°C for 10 min. Because electrostatic interaction between the molecule and the underlying Si is stronger than the Van der Waals interaction between molecules within the film, a C60 molecules adjacent to the Si surface remain at high coverage (~80%) as seen in

The STM operation principle is based on quantum mechanical phenomenon—electron tunneling through a potential barrier formed by a gap between the outermost atoms on the metal tip and the sample. When the gap is about 1 nm or less, electrons from the STM tip can penetrate into the sample with certain probability owing to the wave nature of the quantum

Under external electric field, electron tunneling creates a measurable electric current, the tunneling current. In the single particle approximation, the tunneling current density is given by a difference in the particle flow across the gap from the STM tip and that from the semi‐

*tun tip sample gap gap*

*J J J E f E qV E f qV E*

® ¬ = - = × × - ×- - é ù

where *T*(*Z*, *E*, *V*), the transmission factor, is a function of gap width (*Z*), electron energy (*E*), and external gap voltage (*Vgap*). *ρtip*(*E*) and *ρsample*(*E*) are the density of electron states at the surface of the STM tip and the sample, respectively. *f*(*E*) is the Fermi function describing which

Here, we outline the important features of the STM technique essential for analysis of charge

 r

( ) ( ) ( ) ( )

1

<sup>ò</sup> (1)

ë û

conductor and is expressed as an integral over particle's energy

( ) 0

×

energy states are occupied with electrons.

carrier distribution in semiconductors. They are

*T Z E V dE*

r

¥

,, ,

*gap*

**2.3. Formation of C60 monolayer films**

**3. Tunneling microscopy: basics**

**Figure 2(c)** [41].

304 Microscopy and Analysis

particle.


**Figure 3.** The principle of scanning tunneling microscopy of a semiconductor. (a) An STM setup, (b) an energy band diagram of a tunnel junction, and (c) a charge balance diagram.

The tunneling barrier shape determines the electron transmission factor and the value of the tunneling current. **Figure 3** shows an STM measurements setup and an energy band diagram of an ideal STM junction for n-type Si. Rectangular shape of the tunneling barrier is used in simple STM models. The actual potential barrier profile is different because of image potential lowering (*Δϕ*) owing to strong Coulomb interaction between charge and image charge in conductive materials [43–45]. Also, the tunneling gap may include an insulating layer such as ultrathin oxide and a molecular film with different dielectric properties. Therefore, the tunneling electrons experience an effective potential barrier of a barrier height (BH) given by

$$BH = \left(\Phi\_{\rm M} + E\_{\rm F} - qV\_{gap}\right) / \mathcal{D} - \Delta\phi,\tag{2}$$

where *ΦM* is the work function of the metal tip and *EF* is the Fermi energy of the semiconductor, *q* is the elementary charge. For electron energy smaller than *BH*, the transmission factor is approximated by [46].

$$T\left(E, V\right) = \exp\left(-a \cdot \sqrt{BH \cdot Z}\right),\tag{3}$$

The tunneling constant *α* = 10.2 when the gap width is in units of nanometer and *BH* – in eV.

Because electric charge density in semiconductors is lower than that in metals, applied electric field penetrates deep beneath the semiconductor surface. To maintain the charge neutrality, a band-bending region is created beneath the STM probe. The applied voltage *VS* is shared between the gap and the band-bending region and is given by

$$V\_S = V\_{gap} + V\_{bb} + \phi\_{\text{MS}},\tag{4}$$

where last term *ϕMS* = (*Φ<sup>M</sup>* − *EF*) is an electrostatic potential difference between the work function of the STM tip and the semiconductor Fermi energy, and *χ* is the electron affinity of the semiconductor. In thermal equilibrium and *VS* = 0 , the charge neutrality is conserved, and the electric charge in the STM tip (*QM*) is equal to the local electric charge at the semiconductor surface beneath the STM tip. At *VS* = 0, the band-bending region is created owing to the electrostatic potential difference *ϕMS*. **Figure 3** illustrates the case when an electron depletion region is formed for n-type Si under an external positive bias voltage *VS* > 0 to the sample. For n-Si, the surface charge (*QSS*) includes positive charge of impurity atoms (*QN*) and mobile carriers (holes) (*Q*+):

$$
\mathcal{Q}\_M = \mathcal{Q}\_{\mathfrak{S}^3} = \mathcal{Q}\_\* + \mathcal{Q}\_N,\tag{5}
$$

According to the Gauss law [43, 47, 48], the voltage across the gap is given by

$$\left|V\_{gap} = \frac{\left|\underline{Q}\_{\rm SS}\right|}{\underline{\epsilon}\_0} \cdot Z\_0\right. \tag{6}$$

where *∈*0, and *∈Si* are the permittivity of the vacuum gap and Si, respectively. The depth of the band-bending region (*w*) depends on the electric field screening by the electric charge in the semiconductor and is given by

$$\mathbf{w} \equiv \sqrt{\frac{2 \in\_{\text{Si}} V\_{bb}}{Q\_{\text{cr}}}},\tag{7}$$

It is straightforward that the tunneling current strongly depends on the local electric charge at the semiconductor surface. When there were surface states and interface traps, these trapped charges would alter the initial charge carrier distribution, and great care must be taken to prepare clean, well-defined cross-sectional surfaces. In fact, conventional furnace oxidation produces a gap-state density of about 1010 cm−2 for Si(100) and less than 10<sup>12</sup> cm−2 for Si(111) surfaces [47]. Low-pressure oxidation below 600°C results in a density of ~10<sup>12</sup> cm−2 for Si(100) surfaces [26, 41, 42]. The surface oxidation effectively reduces density of surface states on Si surfaces, making that the current behavior becomes dependent on charge carrier concentration in the Si bulk beneath the STM probe.

Topographic STM images of a sample surface are formed when the STM probe is moved along the surface while keeping pre-determined tunneling current value (*Itun*) at an applied voltage (*VS*) by adjusting the gap width with a piezoelectric scanning unit. The STM technique offers ultimate spatial resolution down to a sub-nanometer range because tunneling current is strongly localized around the outermost atom of the STM tip owing to exponential current decay with the tip-sample distance. Three advanced STM-based modes discussed below rely on measurements and analysis of the tunneling current and, thus, offer high spatial resolution. Details of the SPM system construction and operation have been reviewed in original papers and textbooks [11].

## **4. Advanced STM modes**

*T EV* ( , exp ) = -× × ( a

between the gap and the band-bending region and is given by

carriers (holes) (*Q*+):

306 Microscopy and Analysis

the semiconductor and is given by

The tunneling constant *α* = 10.2 when the gap width is in units of nanometer and *BH* – in eV.

Because electric charge density in semiconductors is lower than that in metals, applied electric field penetrates deep beneath the semiconductor surface. To maintain the charge neutrality, a band-bending region is created beneath the STM probe. The applied voltage *VS* is shared

f

0

0 , *SS*

where *∈*0, and *∈Si* are the permittivity of the vacuum gap and Si, respectively. The depth of the band-bending region (*w*) depends on the electric field screening by the electric charge in

> <sup>2</sup> , *Si bb SS*

It is straightforward that the tunneling current strongly depends on the local electric charge at the semiconductor surface. When there were surface states and interface traps, these trapped charges would alter the initial charge carrier distribution, and great care must be taken to prepare clean, well-defined cross-sectional surfaces. In fact, conventional furnace oxidation

*<sup>V</sup> <sup>w</sup> <sup>Q</sup>*

where last term *ϕMS* = (*Φ<sup>M</sup>* − *EF*) is an electrostatic potential difference between the work function of the STM tip and the semiconductor Fermi energy, and *χ* is the electron affinity of the semiconductor. In thermal equilibrium and *VS* = 0 , the charge neutrality is conserved, and the electric charge in the STM tip (*QM*) is equal to the local electric charge at the semiconductor surface beneath the STM tip. At *VS* = 0, the band-bending region is created owing to the electrostatic potential difference *ϕMS*. **Figure 3** illustrates the case when an electron depletion region is formed for n-type Si under an external positive bias voltage *VS* > 0 to the sample. For n-Si, the surface charge (*QSS*) includes positive charge of impurity atoms (*QN*) and mobile

, *VV V S gap bb MS* = ++

According to the Gauss law [43, 47, 48], the voltage across the gap is given by

*gap*

*BH Z* ), (3)

(4)

, *Q Q QQ M SS* = =++ *<sup>N</sup>* (5)

*<sup>Q</sup> V Z* = × <sup>Î</sup> (6)

<sup>Î</sup> @ (7)

To study charge carrier distribution in semiconductor devices, we describe three STM-based techniques: a vacuum gap modulation method, a molecule-assisted probing method, and a dual-imaging method.

#### **4.1. Vacuum gap modulation method**

A vibrating electrode technique was used to measure the surface potential on solid surfaces by using the Kelvin method [49]. Present-day noncontact atomic force microscopy (nc-AFM) uses vibrating probes for detecting atomic, electrostatic and magnetic forces [50]. In metals, mechanical modulation of the tunnel barrier has been applied as a method to evaluate local work function of the sample [46, 51–54]. In semiconductors, a model of STM junction consid‐ ering both transparency of the tunnel barrier and the band-bending potential was elaborated [22, 23].

When the STM probe vibrates normal to the sample surface, the gap width changes as

$$Z = Z\_0 - dz \cdot \sin\left(\alpha t\right),\tag{8}$$

where *ω* = 2*π*∙ *f* is the angular frequency, *dz* is an amplitude of the vibration. For small vibration amplitude, *dz* ≪ *Z*0, the transmission factor periodically changes with the time-dependent change of both the gap width and the gap voltage. When the STM probe approaches toward the surface, *Vgap* is reduced while increasing the surface potential (*Vbb*). A change of the gap voltage *Vgap* is related to the mean charge *QSS* at the surface by the Gauss law [43, 47, 48] and is expressed as

$$dV\_{\text{gap}} = -d\nu = -\frac{\left|\underline{Q}\_{\text{SS}}\right|}{\underline{\varepsilon}\_0} \cdot d\underline{z},\tag{9}$$

where *dψ* is a change of the band-bending potential.

To determine the tunneling current response (*dI*) to a time-dependent variation of the gap width, the tunneling current is expressed as

$$I\_{\rm new} \left( \alpha \text{ot} \right) = I\_0 + dI\_1 \cdot \sin \left( \alpha \text{ot} \right), \tag{10}$$

where *I*<sup>0</sup> is the mean tunneling current. In the linear approximation [46], the current response is dominated by variation of the mean transparency of the vacuum gap. Thus, in-phase amplitude of the tunneling current response is given as

$$\left| \mathrm{d}I\_1 \propto I\_0 \cdot \frac{\left| \underline{Q}\_{\mathrm{SS}} \right|}{\underline{\mathbf{e}}\_0} \cdot d\underline{z}. \tag{11}$$

In our experiments, the mean tunneling current *I*<sup>0</sup> is held constant; thus, the quantity (*dI*1/*dz*) is proportional to the local charge density at the surface beneath the STM tip under the bias voltage. There is a 90°-phase-shifted current component representing a displacement current owing to change in the STM junction capacitance as discussed in details in Reference [55]. We used the capacitive signal for fine-tuning of the signal phase in the measurements of in-phase current by a lock-in technique.

In the model above, terms due to the shape of the tunnel barrier and capacitance effects associated with modulation of the band-bending region beneath the STM probe are neglected, albeit the effects are essential at high frequency and low impurity concentration [55].

When the modulation of band-bending region is taken into account, the tunneling current response is given by two terms (Appendix A)

$$
\left(\frac{dI}{dz}\right) \propto I\_0 \cdot K\_\circ; \ K\_\circ = \alpha \sqrt{BH} - \left(\frac{\alpha Z\_0}{4\sqrt{BH}} + \beta\right) \cdot \frac{\left|\underline{Q}\_{\text{SS}}\right|}{\underline{\varepsilon}\_0}, \tag{12}
$$

The first term represents the contribution of the gap width modulation, and the second term accounts for variations of *Vgap* and *Vbb* .

It is constructive to take a look at origin of charge *QSS* for n-type and p-type Si under positive bias voltage. In *n*-Si in **Figure 3**, the electric field from the STM probe repels mobile elec‐ trons deep into the bulk creating a surface depletion region, and *QSS = QN + Q+* ≈ *QN* > 0. The larger the bias voltage, the larger the amount of positive charge accumulated beneath the STM probe. As a consequence, the amplitude of the current response (*dI*) depends predominant‐

ly on density of accumulated positive charge. On the contrary, in p-type Si under the same polarity bias, the electric field attracts mobile majority carriers (holes) to the surface reduc‐ ing amount of negative charge of acceptor impurities (*QP*) beneath the STM probe. As a consequence, the amplitude of the current response (*dI*) depends predominantly on small amount of accumulated positive charge, and *QSS = QP* + *Q+* ≈ *Q+*. At the position of electrical p-n junction, the balance of positive and negative charges exists, and *QSS* ≈ 0. Thus, we are able to derive position of electrical p-n junction through analysis of the (*dI*/*dz*) profiles. In addition, detection of charge centres near the Si surface at a depth of ~1 nm has been reported for epitaxial Si layers [56].

Experimentally, differential tunneling current (*dI*/*dZ*) maps were obtained by vibrating the STM probe normal to the sample surface. The STM probe-sample gap was vibrated at a frequency of 12–50 kHz and an amplitude of 20–50 pm while keeping the vacuum gap at constant mean tunneling current *I*0 (the constant current mode). In-phase current response *dI* was measured with a lock-in amplifier at each point in the topographical image. The vibration frequency was selected sufficiently larger than the feedback circuit bandwidth (~10 kHz) and away from the electromechanical resonances of the STM measurement system.

#### **4.2. Molecule-assisted probing method**

0 , *SS*

> w

To determine the tunneling current response (*dI*) to a time-dependent variation of the gap

where *I*<sup>0</sup> is the mean tunneling current. In the linear approximation [46], the current response is dominated by variation of the mean transparency of the vacuum gap. Thus, in-phase

0

In our experiments, the mean tunneling current *I*<sup>0</sup> is held constant; thus, the quantity (*dI*1/*dz*) is proportional to the local charge density at the surface beneath the STM tip under the bias voltage. There is a 90°-phase-shifted current component representing a displacement current owing to change in the STM junction capacitance as discussed in details in Reference [55]. We used the capacitive signal for fine-tuning of the signal phase in the measurements of in-phase

In the model above, terms due to the shape of the tunnel barrier and capacitance effects associated with modulation of the band-bending region beneath the STM probe are neglected,

When the modulation of band-bending region is taken into account, the tunneling current

The first term represents the contribution of the gap width modulation, and the second term

It is constructive to take a look at origin of charge *QSS* for n-type and p-type Si under positive bias voltage. In *n*-Si in **Figure 3**, the electric field from the STM probe repels mobile elec‐ trons deep into the bulk creating a surface depletion region, and *QSS = QN + Q+* ≈ *QN* > 0. The larger the bias voltage, the larger the amount of positive charge accumulated beneath the STM probe. As a consequence, the amplitude of the current response (*dI*) depends predominant‐

0

 bæ ö æ ö µ = - +× ç ÷ ç ÷ è ø <sup>è</sup> <sup>ø</sup> <sup>Î</sup> <sup>×</sup> (12)

; ,. <sup>4</sup>

a

albeit the effects are essential at high frequency and low impurity concentration [55].

*dI <sup>Z</sup> QSS I K K BH*

*dz BH*

a

<sup>Î</sup> (9)

) =+ × 0 1 sin , ( ) (10)

. *QSS dI I dz* µ× × <sup>Î</sup> (11)

0

*<sup>Q</sup> dV d dz* =- =- × y

*I t I dI t tun* (

1 0

*gap*

w

where *dψ* is a change of the band-bending potential.

amplitude of the tunneling current response is given as

width, the tunneling current is expressed as

308 Microscopy and Analysis

current by a lock-in technique.

response is given by two terms (Appendix A)

accounts for variations of *Vgap* and *Vbb* .

033

The ability of specific molecules to selective reactions on the surface is well known in catalysis. Recently, functionalization of SPM probes by attaching functional groups to achieve the chemical selectivity in recognition of DNA sequences and biological molecules has been performed, for example, see [57–59].

The method described here is different. A molecule-assisted probing method makes use of a discrete energy level of an adsorbed molecule as a *marker* of the local Fermi energy. It takes advantage of resonant electron tunneling (RET) to monitor the energy level of the marker molecule, such as fullerene C60, introduced into a tunneling barrier between the STM probe and the oxidized Si surface. The fact that the C60-derived conductance peaks shift in energy depending on dopant concentration in the underlying substrate makes this technique usable as a probing method of the charge carrier profiling on semiconductors [39, 41, 60]. The C60 molecule was selected as it satisfies the selection criteria: small size, chemical stability, and an energy position of molecular orbital outside of the Si energy band gap.

A model of a double-barrier junction (DBJ) was elaborated based on the theory of planar resonant tunnel diodes [61] and alignment of molecular states [62]. **Figure 4(a)** and **Fig‐ ure 4(b)** show the experimental setup and an energy band diagram of an ideal DBJ consisting of the vacuum gap (B1), the C60 layer and the thin oxide (B2) under a resonant injection bias *VRET*. *EA* is the electron affinity of the C60 layer, and *Ei* is the Fermi energy for intrinsic Si. At the resonance condition, the Fermi energy of the STM tip aligns with the lowest unoccupied molecular orbital (LUMO), and thus, the strength of electric field in the vacuum gap is given by *F* = (*ΦM* − *EA*)/*Z*0. For an ideal oxide and neutrality of C60, continuity of the electric displace‐ ment is preserved across the DBJ, and the RET voltage is given by

$$V\_{RET} = \varepsilon\_{Sl} F \cdot \left( Z\_0 + \frac{d\_{c\_{i0}}}{\varepsilon\_{c\_{i0}}} + \frac{d\_{a\alpha}}{\varepsilon\_{a\alpha}} \right) + V\_{hb}, \tag{13}$$

where *d*60 and *dox* are the thickness of *C*<sup>60</sup> molecule and the oxide, respectively. *∈C*60 and *∈ox* are the permittivity of C60 and oxide, respectively. *Vbb* voltage is obtained as a function of the electric field *F* at the Si surface by solving the 3D Poisson equation at quasi-equilibrium.

To measure the RET voltage, mono-molecular fullerene films were prepared by vapor sublimation of C60 to the oxidized Si surfaces at room temperature followed by re-evaporation of excess molecules as described in Section 2.3. Differential conductance (*dI*/*dV*) − *V* spectra in **Figure 4(c)** were obtained at a constant probe-sample gap by using a lock-in technique where a small *ac* voltage (20 mVpp, 50 kHz) was superimposed on the sample bias voltage. The initial tunneling conditions were set with a tunneling current of 200 pA at a set-point voltage of 2.5 V. Each (*dI*/*dV*) − *V* spectrum was fitted to Lorentzian function to determine a voltage of the C60-derived conductance peak, the RET voltage [41, 64]. For high conductance of the tunnel gap, the STM tip is close to the molecule layer, and another transport mechanism, the single electron tunneling [66], becomes apparent and hinders the RET voltage detection. Thus, optimization of the gap width is required.

**Figure 4.** Molecule-assisted probing method. (a) A setup. (b) An energy band diagram of a double-barrier junction un‐ der the resonance conditions. B1 is the tunneling gap, and B2 is thin oxide. (c) (dI/dV) spectra of C60 on p-type Si sub‐ strates with a boron concentration of 8 × 1014 cm− 3 (curve 1), 4 × 1015 cm− 3(curve 2), 3 × 1018 cm− 3(curve 3), and without C60. (d) RET voltage as a function of the Si Fermi energy (*EF* − *Ei* ) from measurements (symbols) and 3D numerical cal‐ culations for oxide thickness of 0.3 nm (broken line) and 0.7 nm (solid line) according to Eq. (13) and Reference [41].

The measured RET voltage obtained for uniformly doped Si wafers with different dopant concentrations is shown in **Figure 4(d)**. The data are well reproduced by the numerical calculations according to Eq. (13) where STM probe emitter was modeled as a cone with a hemispherical end and a radius of curvature of 10 nm, and *Z*0 = 1 nm, *dC*60 = 1 nm , and *dox* = 0.3 nm, *ΦM* = 4.5 eV for W(111) probes and *EA* = 2.6 eV. The good agreement between the calculated RET voltage and the experimental data for uniform-doped wafers verifies the calibration relationship for Si [41, 63].

The spatial resolution of the method is restricted to the size of the *marker* molecule and to the electric field penetration length. It has been demonstrated by the (dI/dV) mapping that the RET peaks are localized within the C60 core (~1 nm) due to their origin in resonant tunneling mediated by one lowest unoccupied molecular orbital (LUMO+1) of C60 [41]. Since the LUMO +1 was localized at the pentagonal rings [65] and C60 molecule rotates at room temperature, the observed peak intensity represents the orientation-averaged orbital conductance of C60. The estimate of the penetration depth is a Debye length of ~1.5 nm for *p*-Si under large positive bias, though the length depends on the dopant concentration for n-Si [41, 63].

#### **4.3. A dual-imaging method**

60 60

where *d*60 and *dox* are the thickness of *C*<sup>60</sup> molecule and the oxide, respectively. *∈C*60 and *∈ox* are the permittivity of C60 and oxide, respectively. *Vbb* voltage is obtained as a function of the electric

To measure the RET voltage, mono-molecular fullerene films were prepared by vapor sublimation of C60 to the oxidized Si surfaces at room temperature followed by re-evaporation of excess molecules as described in Section 2.3. Differential conductance (*dI*/*dV*) − *V* spectra in **Figure 4(c)** were obtained at a constant probe-sample gap by using a lock-in technique where a small *ac* voltage (20 mVpp, 50 kHz) was superimposed on the sample bias voltage. The initial tunneling conditions were set with a tunneling current of 200 pA at a set-point voltage of 2.5 V. Each (*dI*/*dV*) − *V* spectrum was fitted to Lorentzian function to determine a voltage of the C60-derived conductance peak, the RET voltage [41, 64]. For high conductance of the tunnel gap, the STM tip is close to the molecule layer, and another transport mechanism, the single electron tunneling [66], becomes apparent and hinders the RET voltage detection. Thus,

**Figure 4.** Molecule-assisted probing method. (a) A setup. (b) An energy band diagram of a double-barrier junction un‐ der the resonance conditions. B1 is the tunneling gap, and B2 is thin oxide. (c) (dI/dV) spectra of C60 on p-type Si sub‐ strates with a boron concentration of 8 × 1014 cm− 3 (curve 1), 4 × 1015 cm− 3(curve 2), 3 × 1018 cm− 3(curve 3), and without

culations for oxide thickness of 0.3 nm (broken line) and 0.7 nm (solid line) according to Eq. (13) and Reference [41].

The measured RET voltage obtained for uniformly doped Si wafers with different dopant concentrations is shown in **Figure 4(d)**. The data are well reproduced by the numerical calculations according to Eq. (13) where STM probe emitter was modeled as a cone with a

*RET Si bb*

ò

*<sup>d</sup> <sup>d</sup> V FZ <sup>V</sup>* æ ö = ×+ + + ç ÷ è ø

field *F* at the Si surface by solving the 3D Poisson equation at quasi-equilibrium.

ò

optimization of the gap width is required.

310 Microscopy and Analysis

C60. (d) RET voltage as a function of the Si Fermi energy (*EF* − *Ei*

<sup>0</sup> , *<sup>C</sup> ox*

(13)

) from measurements (symbols) and 3D numerical cal‐

 ò

*C ox*

STM technique is limited to conductive surfaces and is inapplicable to the imaging of novel device structures, including insulator surfaces such as silicon-on-insulator (SOI) devices. Strong interest to such measurements is stimulated by the fact that discrete dopant distribution enables attractive applications such as quantum computing [67] and single-electron devices [68]. Therefore, a dual-imaging method was developed to enable simultaneous measurements of electric current and interaction force acting on the scanning probe. It was achieved by attaching an STM metal tip to a special force sensor [67–76].

**Figure 5** shows the experimental setup for the simultaneous measurement of tunneling current (*Itun*) and force between the metal probe tip and the Si surface. In our technique, the interaction force gradient between the metal probe tip and the surface was detected as a shift in the resonance frequency (*Δf*) of a quartz length extension resonator (qLER) which vibrated at ~1 MHz (Q factor ~50,000) with an amplitude of 0.05–0.3 nm [67–70]. The probe tips were made of a tungsten wire with a diameter of 10 μm. The wire was attached to the quartz resonator and sharpened by the focused ion beam technique (FIB). Typically, the probe tips had a diameter of Ø30 nm and the aspect ratio of more than 10, resulting in small stray capacitance. Detection of the frequency shift by electric means makes such sensors suitable for measure‐ ments in ultra-high vacuum environment and at different temperature, which are often required in nanomaterial and nanoscale device research.

The advantages of our multimode scanning probe microscopy (MSPM) system are


**•** the force detection is performed in a *noncontact* manner, which is suitable for measurements of solid crystals and thin films.

In the CC mode, a force gradient map is measured while the mean gap (*Z0*) maintains a setpoint tunneling current. Typically, the measurement condition corresponds to a gap of approximately 1 nm, as estimated from the distance dependence of the tunneling current [72]. The spatial variation of the frequency shift (Δ*f*) reflects variations in the interaction force caused by charge carriers, impurity charges, and surface imperfections as illustrated in **Figure 5(b)**. When a donor is present in proximity to the STM tip, the attractive force acting on the tip increases owing to Coulomb interaction between the donor charge and the image charge induced in the STM tip, leading to measurable change in the *Δf* value [75, 76]. The interaction strength depends on the depth of the donor location and the electrostatic screening by mobile carriers. Experimentally, lateral extent of 5–10 nm and a detection depth of ~1 nm have been reported for phosphorus and boron atoms in Si [32, 33, 76]. Change in the interaction force on grains with different work function was employed for recognizing crystal orientation of sub-10-nm-size grains in nano-crystalline TiN films [77].

**Figure 5.** Dual-imaging method. (a) A measurement setup. (b) A sketch of interaction force acting on a vibrating STM probe. (c) (*Itun-Z*) and (*Δf-Z*) spectra showing ranges of repulsive interaction (1–2) and attractive Coulomb interaction (2–3) for an oxide-passivated Si(111) surface (a set-point: 30 pA, 2.0 V). (d) A measured (*Δf-VS*) spectrum at position 3(blue curve), and a result of fitting to Eq. (14) (red curve).

In the CF mode, a tunneling current (*Itun*) map is measured while the mean gap (*Z0*) is main‐ tained at a constant frequency shift. There are two ranges in distance dependences of *Itun and Δf* as indicated in **Figure 5(c)** for an oxide-passivated Si(111) surface. At short distances (range 1–2), repulsive interaction dominates, and current exponentially grows when the STM tip approaches the surface. At longer distances (range 2–3), the electrostatic Coulomb interaction dominates. There is an optimal distance indicated as position 2 in **Figure 5(c)** where the sensitivity to electrostatic force is maximum [72]. At this distance, the (*Δf* − *VS*) spectrum has the largest curvature.

Under the applied voltage *VS*, the electrostatic force gradient between the probe tip and the sample is expressed according to the theory in References [73, 78] for small vibration amplitude

$$
\Delta f \propto \frac{\partial F}{\partial z} = -\frac{1}{2} (V\_s - CPD)^2 \cdot \frac{\partial^2 C}{\partial z^2},
\tag{14}
$$

where *C* is the effective tip-sample capacitance. *CPD*, the contact potential difference, refers to the difference between the work function of the metal probe (*ΦM*) and the Fermi energy of the underlying Si (*EF*), and is given by

$$CPD = \frac{1}{q} \left( E\_F - \Phi\_M \right),\tag{15}$$

where *q* is the elementary charge. A local value of the CPD voltage, which is determined by local charge concentration in the underlying Si, can be obtained by fitting of the spectrum to Eq. (14). In the example in **Figure 5(d)**, a CPD voltage of +0.8 V was obtained for an oxidized p-Si(111) surface. The CPD voltage mapping was employed in 2D analysis of the built-in potential in small Si MOSFET devices [79] and p-n junctions [72] showing the attainable spatial resolution better than 3 nm. Particular applications of the CF mode also include analysis of impurity distribution profiles from *Itun* maps measured at different bias voltage [80], nonuniform distribution of photocarrier in Si stripes [81], and nanoscale conductance switching in phase-change GeSbTe thin films [82].

## **5. Application examples**

**•** the force detection is performed in a *noncontact* manner, which is suitable for measurements

In the CC mode, a force gradient map is measured while the mean gap (*Z0*) maintains a setpoint tunneling current. Typically, the measurement condition corresponds to a gap of approximately 1 nm, as estimated from the distance dependence of the tunneling current [72]. The spatial variation of the frequency shift (Δ*f*) reflects variations in the interaction force caused by charge carriers, impurity charges, and surface imperfections as illustrated in **Figure 5(b)**. When a donor is present in proximity to the STM tip, the attractive force acting on the tip increases owing to Coulomb interaction between the donor charge and the image charge induced in the STM tip, leading to measurable change in the *Δf* value [75, 76]. The interaction strength depends on the depth of the donor location and the electrostatic screening by mobile carriers. Experimentally, lateral extent of 5–10 nm and a detection depth of ~1 nm have been reported for phosphorus and boron atoms in Si [32, 33, 76]. Change in the interaction force on grains with different work function was employed for recognizing crystal orientation of

**Figure 5.** Dual-imaging method. (a) A measurement setup. (b) A sketch of interaction force acting on a vibrating STM probe. (c) (*Itun-Z*) and (*Δf-Z*) spectra showing ranges of repulsive interaction (1–2) and attractive Coulomb interaction (2–3) for an oxide-passivated Si(111) surface (a set-point: 30 pA, 2.0 V). (d) A measured (*Δf-VS*) spectrum at position

of solid crystals and thin films.

312 Microscopy and Analysis

sub-10-nm-size grains in nano-crystalline TiN films [77].

3(blue curve), and a result of fitting to Eq. (14) (red curve).

#### **5.1. Channel length in small MOSFET**

For STM measurements, cross-sections of Si MOSFETs were prepared by ultra-fine polishing to expose (110) surfaces and were passivated by ultra-thin oxide layer as described in Sec‐ tion 2.2. Si *n*-type MOSFET with nominal gate lengths (*LG*) in the range of 20–150 nm were fabricated according to a process described in Reference [83]. The measurements were done with W(111) crystal probes in an ultrahigh vacuum (~4 × 10−9 Pa) at room temperature.

**Figure 6.** (a) Topographic image of a cross-section containing two small Si MOSFET devices. (b) A (dI/dZ) map of a device with a gate length of 31 nm (a set-point: 230 pA, 3.4 V, dz = 20 pm). (c)–(d) Line profiles measured at 12 nm depth beneath the gate electrodes showing the electric channel length (*LS-D*). (e) Profiles calculated by Eq. (12) for ex‐ pected impurity distribution. (f) Measured electric channel length (symbols) as a function of gate length. Line is the calculation result.

Topographic image of two small MOSFET is shown in **Figure 6(a)**, where the gate electrodes are surrounded by two black cavities produced by sidewall oxide etching during the surface preparation. The source/drain (S/D) extensions on the left- and right-hand sides of the gate electrode are seen as bright stripes in the (dI/dZ) map in **Figure 6(b)**. Depletion regions separate the S/D extensions from the p-type channel beneath the gate electrode and the Si bulk. The extension depth is ~18 nm as measured from the gate oxide. The electric channel length (*LS* − *<sup>D</sup>*) was determined as the distance between 2 minima in (dI/dZ) line profiles measured at a depth of 12 nm beneath the gate oxide as indicated in **Figure 6(c, d)**. Calculated profiles of the K3 factor in **Figure 6(e)** reproduce the measured (dI/dZ) profiles, confirming that each minimum in (dI/dZ) signal represents the position of the electric p-n junction. *LG* was determined from STM topographs. Results summarized in **Figure 6(f)** give an overlap value of 6 ± 1 nm, which is in excellent agreement with a transverse straggle of 7 nm for an implanted ion energy of 25 keV. An accuracy of the channel measurements was about 1 nm at 3.4 V, while the measure‐ ments were affected by random positions of individual ionized dopant atoms in the extension regions.

#### **5.2. Super-junction devices fabricated by the channeling ion implantation**

The C60-assisted probing technique has been actually applied to quantitative analysis of charge carrier profiles on cross-sections of power MOSFET, where the precise control over the doping profile is essential to obtain low ON-state resistance and high breakdown voltage [39, 40]. **Figure 7(a)** depicts a schematic structure of a super-junction power MOSFET. Two *p*-type islands were formed by multiple boron ion implantations into the low-doped *n*-type epitaxial layer with a carrier density of ~1 × 1016 cm3 . In **Figure 7(b)**, we clearly see that two p-type islands are separately formed with the same peak concentrations, confirming the anticipated dopant concentration. Moreover, the experimental data revealed an extension of island 1 beyond the expected depth, which is attributed to a scatter-less travel of boron ions through Si crystal at high implantation energy, the ion channeling effect[84].

**Figure 7.** (a) Schematic structure of a super-junction device showing two p-Si islands made by boron ion implantation. (b) Depth profiles of the RET voltage along center of the device: measured data (symbols) taken with 20-nm steps. Pro‐ files (lines) were calculated for the two boron density profiles shown in (c). Reproduced with the permission from Ref‐ erence [39].

#### **5.3. Length-dependent resistivity of Si nanowires**

**Figure 6.** (a) Topographic image of a cross-section containing two small Si MOSFET devices. (b) A (dI/dZ) map of a device with a gate length of 31 nm (a set-point: 230 pA, 3.4 V, dz = 20 pm). (c)–(d) Line profiles measured at 12 nm depth beneath the gate electrodes showing the electric channel length (*LS-D*). (e) Profiles calculated by Eq. (12) for ex‐ pected impurity distribution. (f) Measured electric channel length (symbols) as a function of gate length. Line is the

Topographic image of two small MOSFET is shown in **Figure 6(a)**, where the gate electrodes are surrounded by two black cavities produced by sidewall oxide etching during the surface preparation. The source/drain (S/D) extensions on the left- and right-hand sides of the gate electrode are seen as bright stripes in the (dI/dZ) map in **Figure 6(b)**. Depletion regions separate the S/D extensions from the p-type channel beneath the gate electrode and the Si bulk. The extension depth is ~18 nm as measured from the gate oxide. The electric channel length (*LS* − *<sup>D</sup>*) was determined as the distance between 2 minima in (dI/dZ) line profiles measured at a depth of 12 nm beneath the gate oxide as indicated in **Figure 6(c, d)**. Calculated profiles of the K3 factor in **Figure 6(e)** reproduce the measured (dI/dZ) profiles, confirming that each minimum in (dI/dZ) signal represents the position of the electric p-n junction. *LG* was determined from STM topographs. Results summarized in **Figure 6(f)** give an overlap value of 6 ± 1 nm, which is in excellent agreement with a transverse straggle of 7 nm for an implanted ion energy of 25 keV. An accuracy of the channel measurements was about 1 nm at 3.4 V, while the measure‐ ments were affected by random positions of individual ionized dopant atoms in the extension

**5.2. Super-junction devices fabricated by the channeling ion implantation**

layer with a carrier density of ~1 × 1016 cm3

The C60-assisted probing technique has been actually applied to quantitative analysis of charge carrier profiles on cross-sections of power MOSFET, where the precise control over the doping profile is essential to obtain low ON-state resistance and high breakdown voltage [39, 40]. **Figure 7(a)** depicts a schematic structure of a super-junction power MOSFET. Two *p*-type islands were formed by multiple boron ion implantations into the low-doped *n*-type epitaxial

are separately formed with the same peak concentrations, confirming the anticipated dopant concentration. Moreover, the experimental data revealed an extension of island 1 beyond the

. In **Figure 7(b)**, we clearly see that two p-type islands

calculation result.

314 Microscopy and Analysis

regions.

The ability of the dual-imaging method for characterization of modern silicon-on-insulator (SOI) devices is illustrated by analysis of the structure and electric conductance of SOI nanowires (NW) with different surface passivation. Note that the NW is the promising structure for sub-10-nm MOSFETs and for such functional devices as chemical sensors. **Figure 8** shows high-resolution measurements of a Si NW with a cross-section area of 20 × 20 nm2 acquired at a set point of *Δf* = 0.6 Hz, *dz* = 95 pm, *VS* = − 1.5 V. We see in **Figure 8(c)** the current gradually decreases in the NW interior with the distance from the Si pad owing to the dependence of the NW resistivity on its length. We note that an apparent NW width in the current map is about 2-fold of that in the topograph. As the NW is protruded above the buried oxide (BOX) by 20 nm, a side surface of the sharp tip touches the NW as illustrated in the insert of **Figure 8(c)**, and this results in a so-called "sidewall" current outside the Si NW body. The current value and fluctuations were reduced for the NW passivated with an ultrathin oxide layer compared to the hydrogen passivation. The tunneling current decreased within a distance of ~300 nm from the Si pad electrode for both types of surface termination. At the negative voltage, the tunneling current is defined by electrons traveling from large Si pad through the SOI nanowire, and the current value is determined by resistivity of the NW volume and the surface conduction. The macroscopic conduction model including the conductance contributions of the nanowire volume and the surface states confirmed the length-dependent conductance of thin Si nanowires [85].

**Figure 8.** (a) An experiment setup. (b) Topographic image of silicon-on-insulator nanowire with a cross-section of 20 × 20 nm2 , and (c) corresponding current map acquired at -1.5 V and Δ*f* = 0.6 Hz, *dz* = 95 pm. (d) Current profiles along A-A line for Si nanowires after hydrogen-passivation (curve 1), oxide passivation (curve 2), and along B-B line (curve 3). Adopted from Reference [85] (Copyright 2013 Trans. Mat. Res. Soc. Japan).

#### **5.4. Wavelength-dependent photocarrier distribution across strained Si stripes**

Photo-carrier generation in semiconductors is a fundamental process utilized in solar cells and photo-detectors. For reduced size of modern detectors, the role of structural elements in carrier accumulation and transport has been increasing [86]. In particular, photocarrier distribution on textured surfaces of Si can be a factor to improve the efficiency of solar cells. Analysis of spatial distribution of photocurrent (PC) in strained Si stripes under tilted illumination gives an insight into photocarrier behavior near the stripe edges with an effective spatial resolution of ~10 nm [81].

**Figure 9** shows the sample structure and the measurement setup, where inhomogeneous light intensity profile was created under tilted (50° off-normal) illumination and different light wavelength (*λ*). Strained Si stripes of 50–1000 nm in width and 300 nm in height were fabricated on Si(001) wafer, and separated by SiO2. The stripe surface was passivated by an ultrathin oxide as described in Section 2.2. The light intensity was mechanically modulated at frequency of ~3 kHz, and the PC signal was measured by a lock-in unit. Topographs and PC maps were measured by the dual-imaging method where the tip-sample gap was set by a set-point of *Δf* = 1.2 Hz, *dz* = 130 pm, and *VS* = − 0.8 V, using the CF mode.

Topographic image in **Figure 9(b)** shows uniform surface of the Si stripe. The PC signal was not uniform, and large at a distance of ~50 nm from the stripe edge on the light illumination side, when stripes were illuminated with laser light and an intensity of 12 mW/cm2 as seen in **Figure 9(c)**. Large PC signal at stripe edges was observed irrespective of the scanning direc‐ tions, when light with *λ* = 405 and 364 nm was used as seen in line profiles in **Figure 9**(**d**, **e**). In contrast, illumination with red light (*λ* = 675 nm) produced uniform PC distribution. As the absorption depth in Si is ~11 nm for *λ* = 364 nm, ~130 nm for *λ* = 405 nm, and ~4000 nm for *λ* = 675 nm [87], the respective illumination produces different light intensity profiles. Calcu‐

lated PC profiles in **Figure 9(f)** reproduced the observed PC distributions when a rectangular bar geometry, non-coherent light, and a photocarrier diffusion length of 100 nm were used [81].

**Figure 9.** (a) Photocurrent (PC) measurement setup. (b) A topograph and (c) corresponding PC map of a Si stripe un‐ der illumination with *λ* = 405 nm. (d)–(e) Measured line profiles of height (black lines) and PC (dotted lines) across the stripe edge under tilted illumination for two wavelengths (*λ*). (a set-point: Δ*f* = 1.2 Hz, dZ = 130 pm, *VS* = − 0.8 V). (f) PC line profiles calculated for a rectangular bar exposed to light at top and side surfaces. Adopted from Reference [81] (Copyright 2012 The Japan Society of Applied Physics).

We note that the relative intensity of a PC peak at a position of ~30 nm for *λ* = 364 nm is ~3.2 fold the signal in the stripe interior. Enhancement of light intensity by ~3.5-fold at strained Si stripe edges has been reported for *λ* = 364 nm [88, 89]. The enhancement mechanism may be related to increased photocarrier generation owing to interference of coherent laser light [81], narrowing of the Si energy gap under stress [90] or increase in the tunneling probability through electromagnetic field coupling to the sharp STM tip [91].

## **6. Simulations of tunneling current spectra**

**Figure 8.** (a) An experiment setup. (b) Topographic image of silicon-on-insulator nanowire with a cross-section of

along A-A line for Si nanowires after hydrogen-passivation (curve 1), oxide passivation (curve 2), and along B-B line

Photo-carrier generation in semiconductors is a fundamental process utilized in solar cells and photo-detectors. For reduced size of modern detectors, the role of structural elements in carrier accumulation and transport has been increasing [86]. In particular, photocarrier distribution on textured surfaces of Si can be a factor to improve the efficiency of solar cells. Analysis of spatial distribution of photocurrent (PC) in strained Si stripes under tilted illumination gives an insight into photocarrier behavior near the stripe edges with an effective spatial resolution

**Figure 9** shows the sample structure and the measurement setup, where inhomogeneous light intensity profile was created under tilted (50° off-normal) illumination and different light wavelength (*λ*). Strained Si stripes of 50–1000 nm in width and 300 nm in height were fabricated on Si(001) wafer, and separated by SiO2. The stripe surface was passivated by an ultrathin oxide as described in Section 2.2. The light intensity was mechanically modulated at frequency of ~3 kHz, and the PC signal was measured by a lock-in unit. Topographs and PC maps were measured by the dual-imaging method where the tip-sample gap was set by a set-point of

Topographic image in **Figure 9(b)** shows uniform surface of the Si stripe. The PC signal was not uniform, and large at a distance of ~50 nm from the stripe edge on the light illumination

**Figure 9(c)**. Large PC signal at stripe edges was observed irrespective of the scanning direc‐ tions, when light with *λ* = 405 and 364 nm was used as seen in line profiles in **Figure 9**(**d**, **e**). In contrast, illumination with red light (*λ* = 675 nm) produced uniform PC distribution. As the absorption depth in Si is ~11 nm for *λ* = 364 nm, ~130 nm for *λ* = 405 nm, and ~4000 nm for *λ* = 675 nm [87], the respective illumination produces different light intensity profiles. Calcu‐

as seen in

side, when stripes were illuminated with laser light and an intensity of 12 mW/cm2

**5.4. Wavelength-dependent photocarrier distribution across strained Si stripes**

(curve 3). Adopted from Reference [85] (Copyright 2013 Trans. Mat. Res. Soc. Japan).

*Δf* = 1.2 Hz, *dz* = 130 pm, and *VS* = − 0.8 V, using the CF mode.

, and (c) corresponding current map acquired at -1.5 V and Δ*f* = 0.6 Hz, *dz* = 95 pm. (d) Current profiles

20 × 20 nm2

316 Microscopy and Analysis

of ~10 nm [81].

STM has the capability to 2D impurity profiling by employing advanced STM methods as shown above. Although, accurate analysis of charge carrier distributions in actual 2D and 3D device structures has been a substantial challenge. STM tunneling current is a complex function of structural, material, and electronic parameters of the system consisting of a 3D probe tip and a semiconductor. On the basis of fundamental theory, there have been theoretical discussions of 1D and 2D treatments for the STM junction geometry. A 3D numerical simulator has been reported that solves the 3D potential distribution of the sample STM probe system and calculates the tunneling current, so-called *the potential-based model* [23, 92, 93]. However, to describe the precise physics of STM measurements, the charge carrier flow in the sample must be included, as evidenced by the NW measurements in **Figure 8**. Recently, new model evolves solving the charge carrier transport between a probe tip and a sample consistently with the current continuity equation, so-called *the current-continuity model*. The currentcontinuity model accounts for charge carrier transport between states in an STM probe and the conduction and the valence band of Si and was implemented on the basis of a technology computer-aided design (TCAD) semiconductor device simulator code [94]. It is a significant advancement in the field.

An analysis based on the current-continuity model has been applied successfully to extracting impurity distribution profiles in a MOSFET from experimental current maps measured by the dual-imaging method [80], and for evaluating photocarrier dynamics in Si nanowires with a cross-section of 10 × 10 nm2 [95].

The remaining challenge is to include the effect of single impurity scattering on charge carrier transport in nanoscale devices. The impurity scattering for a thin semiconductor wire has been solved using the 3D Green function approach and the numerical Monte-Carlo method [96]. An atomistic view into an impurity atom appearance in STM images has been elaborated within the framework of a self-consistent-charge density functional tight-binding method (SCCDFTB), for example, see [97, 98].

## **7. Conclusion**

Advanced STM-based methods for 2D analysis of charge carrier distributions in semiconduc‐ tor devices with high spatial resolution represent the substantial development of scanning probe microscopy. The described methods rely on detection and analysis of tunneling current which is strongly localized within an atomic dimension. This leads to significant improvement in the sensitivity and spatial resolution for measuring local electric characteristics of Si devices and nanowires, when effects of surface states are suppressed by adequate surface treatment.

The gap modulation method can attain an ultimate spatial resolution comparable to that of STM topographic images in p-n junction regions, and can detect individual charged impurity atoms along the surface at a depth of few nanometers. Quantitative evaluation of charge distributions can be derived by comparing experimental data and simulations of the under‐ lying charge concentration. The accuracy relies on the ability of the simulation to account for quantum phenomena, and further development of simulations based on the current-continuity model will be essential.

The capability of the molecule-assisted probing method has been demonstrated with the use of C60 molecules. A spatial resolution of ~1 nm is determined by the size of the molecule. However, the C60 film on oxidized Si surfaces leaves ~20% uncovered areas. The coverage can be increased by the use of chemically modified C60 or other small molecules those formed a monomolecular-thick film on SiO2 surface. For high conductance of the tunnel gap, another transport mechanism, the single electron tunneling [66], becomes dominant and obscures the RET voltage measurements. Thus, optimization of the gap width is required.

The presented methods can be used for measuring on rough surfaces, but careful data analysis should be performed to discard "artifacts." In the gap modulation method, the tip vibration amplitude (*dz*) varies with tilt angle of the underlying surface, causing changes in the (dI/dZ) signal. In the dual-imaging method, large "sidewall" current such as shown in **Figure 8** must be considered in data analysis. Also, atomically ordered surfaces can be obtained by cleavage, yet, to attain ultimate spatial resolution, STM measurements in well-controlled environment such as in an ultrahigh vacuum are necessary, where we can avoid undesirable effects caused by absorption of charged particles and molecules from air.

To summarize, specific features of the presented 2D STM-based methods are (a) noncontact, stress-free measurements allowing analysis of delicate sample structures; (b) high spatial sensitivity to electrostatic field, which is substantial advancement in comparison with scanning Kelvin probe microscopy; (c) the ability to study nanoscale structures with a lateral size of 20 nm and below, which are inaccessible by other techniques.

Further applications of the advanced STM methods will contribute to high-spatial resolution analysis of modern sub-100-nm electronic devices, functional nanowire devices, and novel devices incorporating two-dimensional materials such as graphene and topological superlat‐ tices. It will advance our understanding of charge carrier transport at nanoscale and encourage inventing novel energy-efficient devices.

## **8. Appendix A**

with the current continuity equation, so-called *the current-continuity model*. The currentcontinuity model accounts for charge carrier transport between states in an STM probe and the conduction and the valence band of Si and was implemented on the basis of a technology computer-aided design (TCAD) semiconductor device simulator code [94]. It is a significant

An analysis based on the current-continuity model has been applied successfully to extracting impurity distribution profiles in a MOSFET from experimental current maps measured by the dual-imaging method [80], and for evaluating photocarrier dynamics in Si nanowires with a

The remaining challenge is to include the effect of single impurity scattering on charge carrier transport in nanoscale devices. The impurity scattering for a thin semiconductor wire has been solved using the 3D Green function approach and the numerical Monte-Carlo method [96]. An atomistic view into an impurity atom appearance in STM images has been elaborated within the framework of a self-consistent-charge density functional tight-binding method

Advanced STM-based methods for 2D analysis of charge carrier distributions in semiconduc‐ tor devices with high spatial resolution represent the substantial development of scanning probe microscopy. The described methods rely on detection and analysis of tunneling current which is strongly localized within an atomic dimension. This leads to significant improvement in the sensitivity and spatial resolution for measuring local electric characteristics of Si devices and nanowires, when effects of surface states are suppressed by adequate surface treatment. The gap modulation method can attain an ultimate spatial resolution comparable to that of STM topographic images in p-n junction regions, and can detect individual charged impurity atoms along the surface at a depth of few nanometers. Quantitative evaluation of charge distributions can be derived by comparing experimental data and simulations of the under‐ lying charge concentration. The accuracy relies on the ability of the simulation to account for quantum phenomena, and further development of simulations based on the current-continuity

The capability of the molecule-assisted probing method has been demonstrated with the use of C60 molecules. A spatial resolution of ~1 nm is determined by the size of the molecule. However, the C60 film on oxidized Si surfaces leaves ~20% uncovered areas. The coverage can be increased by the use of chemically modified C60 or other small molecules those formed a monomolecular-thick film on SiO2 surface. For high conductance of the tunnel gap, another transport mechanism, the single electron tunneling [66], becomes dominant and obscures the

The presented methods can be used for measuring on rough surfaces, but careful data analysis should be performed to discard "artifacts." In the gap modulation method, the tip vibration

RET voltage measurements. Thus, optimization of the gap width is required.

advancement in the field.

318 Microscopy and Analysis

cross-section of 10 × 10 nm2

**7. Conclusion**

model will be essential.

(SCCDFTB), for example, see [97, 98].

[95].

The tunneling current is described as a periodic function as

$$I\_{\rm nu}\left(t\right) = I\_0 + dI \cdot \sin\alpha t = I\_0 \cdot \left(1 + K\_3 \cdot d\boldsymbol{\varpi} \cdot \sin\alpha t\right),\tag{A1}$$

The mean tunneling current is given in terms of the thermionic emission approximation including the vacuum tunneling term according to Reference [99] as

$$I\_0 = A\_\circ^\ast T^2 \cdot \exp\left(-\alpha \sqrt{BH \cdot Z\_0}\right) \cdot \exp\left(-\beta V\_{bb}\right) \cdot \left[\exp\left(\beta V\_{gap}\right) - 1\right],\tag{A2}$$

For *dz* ≪ *Z*0, the factor *K3* is derived considering only linear terms of *dz* and *dψ*, and is given by

$$K\_3 = \alpha \sqrt{BH} - \left(\frac{\alpha Z\_0}{4\sqrt{BH}} + \beta\right) \cdot \frac{\left|\underline{Q}\_{\text{IS}}\right|}{\underline{\epsilon}\_0},\tag{A3}$$

The area charge concentration at the Si surface (*QSS*) is obtained by solving the Poisson equation. An analytic solution for a 1D abrupt junction is given by [47]

$$\mathcal{Q}\_{\rm ss} = -\frac{\Xi\_{\rm Si}}{\mathcal{J} \cdot \Lambda} \cdot G\left(V\_{\rm sb}\right),\tag{A4}$$

$$G\left(\mathbf{y}\right) = \left[\left(e^{-\beta\mathbf{y}} + \beta\mathbf{y} - 1\right) + \frac{n}{p\_0}\left(e^{\beta\mathbf{y}} - \beta\mathbf{y} - 1\right)\right]^{1/2} \ge 0,\tag{A5}$$

$$
\Lambda = \sqrt{\frac{\epsilon\_N}{\beta \mathcal{Q}\_N}}\,,\tag{A6}
$$

Λ is the extrinsic Debye length, and volume densities of positive (*p*0) and negative (*n*) charge are in the Si bulk. The factor *β* = 1/*kBT*, and *kB* is the Boltzman constant, *T* is temperature.

For 3D structures, a charge concentration at the semiconductor surface (*QSS*) is obtained by numerically solving the Poisson equation.

#### **Acknowledgements**

The authors would like to thank colleagues of Nanoelectronics Research Institute (AIST, Japan) for valuable discussions and constructive comments motivating the research works.

#### **Author details**

Leonid Bolotov\* and Toshihiko Kanayama

\*Address all correspondence to: bolotov.leonid@aist.go.jp

National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba, Ibaraki, Japan

#### **References**


[3] Miller M. K., Cerezo A., Hetherington M. G., and Smith G. D. W. Atom Probe Field Ion Microscopy. New York: Oxford University Press; 1996. 509 p. ISBN-13: 978–0198513872.

( ), <sup>Λ</sup>

 b

 b

0 <sup>1</sup> 1 0, *y y <sup>n</sup> Gy e y e y p*

Λ is the extrinsic Debye length, and volume densities of positive (*p*0) and negative (*n*) charge are in the Si bulk. The factor *β* = 1/*kBT*, and *kB* is the Boltzman constant, *T* is temperature.

For 3D structures, a charge concentration at the semiconductor surface (*QSS*) is obtained by

The authors would like to thank colleagues of Nanoelectronics Research Institute (AIST, Japan)

National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba, Ibaraki,

[1] Müller E. W, Panitz J. A., and McLane S. B.: The atom probe field ion microscope.

[2] Kelly T. F. and Miller M. K.: Atom probe tomography. Review of Scientific Instruments.

Review of Scientific Instruments. 1968; 39: 83–86. DOI:10.1063/1.1683116

for valuable discussions and constructive comments motivating the research works.

and Toshihiko Kanayama

\*Address all correspondence to: bolotov.leonid@aist.go.jp

2007; 78: 031101. DOI:10.1063/1.2709758

 - é ù = + -+ - - ³ ê ú ë û

> Λ , *Si* b*QN*

<sup>Î</sup> = - (A4)

(A5)

(A6)

1/2

*Si Q GV SS bb* b<sup>×</sup> <sup>×</sup>

( ) ( ) ( )

= ò

b

b

numerically solving the Poisson equation.

**Acknowledgements**

320 Microscopy and Analysis

**Author details**

Leonid Bolotov\*

**References**

Japan


[27] Liu L., Yu J. and Lyding J. W.: Atom-resolved three-dimensional mapping of boron dopants in Si(100) by scanning tunneling microscopy. Applied Physics Letters. 2001; 78: 386–388. DOI:10.1063/1.1339260

[15] Binnig G., Rohrer H., Gerber C., and Weibel E.: Surface studies by scanning tunneling microscopy. Physical Review Letters. 1982; 49: 57–82. DOI:10.1103/PhysRevLett.49.57

[16] Alvarez D., Hartwich J., Fouchier M., Eyben P., and Vandervorst W.: Sub-5-nm-spatial resolution in scanning spreading resistance microscopy using full-diamond tips.

[17] Zhang L., Ohuchi K., Adachi K., Ishimaru K., Takayanagi M., and Nishiyama A.: Highresolution characterization of ultrashallow junctions by measuring in vacuum with scanning spreading resistance microscopy. Applied Physics Letters. 2007; 90: 192103.

[18] Moraru D., Ligowski M., Yokoi K., Mizuno T., and Tabe M.: Single-electron transfer by inter-dopant coupling tuning in doped nanowire silicon-on-insulator field-effect transistors. Applied Physics Express. 2009; 2: 071201. DOI:10.1143/APEX.2.071201

[19] DeWolf P., Stephenson R., Trenkler T., Clarysse T., Hantschel T., and Vandervorst W.: Status and review of two-dimensional carrier and dopant profiling using scanning probe microscopy. Journal of Vacuum Science and Technology B. 2000; 18: 361–368.

[20] Jager N. D., Marso M., Salmeron M., Weber E. R., Urban K., and Ebert P.: Physics of imaging p−n junctions by scanning tunneling microscopy and spectroscopy. Physical

[21] Fukutome H., Arimoto H., Hasegawa S., and Nakashima H.: Two-dimensional characterization of carrier concentration in metal-oxide-semiconductor field-effect transistors with the use of scanning tunneling microscopy. Journal of Vacuum Science

[22] Weimer M., Kramar J., and Baldeschwieler J. D.: Band bending and the apparent barrier height in scanning tunneling microscopy. Physical Reviews B. 1989; 39: 5572–5577. DOI:

[23] Feenstra R. M.: Electrostatic potential for a hyperbolic probe tip near a semiconductor. Journal of Vacuum Science and Technology B. 2003; 21: 2080–2088. DOI:

[24] Morrison S. R. The chemical physics of surfaces. New York: Plenum Press; 1977. 415p.

[25] Mariani G., Scofield A. C., Hung C.-H., and Huffaker D. L.: Gas nanopillar-array solar cells employing in situ surface passivation. Nature Communications. 2013; 4: 1497.

[26] Yablonovitch E., Allara D. L., Chang C. C., Gmitter T., and Bright T. B.: Unusually low surface-recombination velocity on silicon and germanium surfaces. Physical Review

Reviews B. 2003; 67: 165307. DOI:10.1103/PhysRevB.67.165307

and Technology B. 2004; 22: 358–363. DOI:10.1116/1.1627792

Letters. 1986; 57: 249–252. DOI:10.1103/PhysRevLett.57.249

Applied Physics Letters. 2003; 82: 1724–1726. DOI:10.1063/1.1559931

DOI:10.1063/1.2736206

322 Microscopy and Analysis

DOI:10.1116/1.591198

10.1103/PhysRevB.39.5572

DOI:10.1007/978-1-4615-8007-2

DOI:10.1038/ncomms2509

10.1116/1.1606466


oscillation technique. Thin Solid Films. 2004; 455–456: 759–763. DOI:10.1016/j.tsf. 2003.11.262


[51] Binnig G. and Roher H.: Scanning tunneling microscopy. Surface Science. 1983; 126: 236–244. DOI:10.1016/0039-6028(83)90716-1

oscillation technique. Thin Solid Films. 2004; 455–456: 759–763. DOI:10.1016/j.tsf.

[39] Bolotov L., Nishizawa M., Miura Y., and Kanayama T.: Carrier concentration profiling on oxidized surfaces of Si device cross sections by resonant electron tunneling scanning probe spectroscopy. Journal of Vacuum Science and Technology B. 2008; 26: 415–419.

[40] Kanayama T., Nishizawa M., and Bolotov L.: Dopant and carrier concentration profiling with atomic resolution by scanning tunneling microscopy. ECS Transactions.

[41] Bolotov L., Uchida N., and Kanayama T.: Scanning tunneling spectroscopy of atomic clusters deposited on oxidized silicon surfaces: induced surface dipole and resonant electron injection. Journal of Physics: Condensed Matter. 2003; 15: S3065–S3081. DOI:

[42] Bitzer T., Rada T., Richardson N. V., Dittrich T., and Koch F.: Gap state formation during the initial oxidation of Si(100)-2×1. Applied Physics Letters. 2000; 77: 3779–3783. DOI:

[43] Simmons J. G., Hsueh F. L., and Faraone L.: Two-carrier conduction in MOS tunneloxides II-theory. Solid State Electronics. 1984; 27: 1131–1139. DOI:

[44] Simmons J. G.: Generalized formula for the electric tunnel effect between similar electrodes separated by a thin insulating film. Journal of Applied Physics. 1963; 34:

[45] Simmons J. G.: Electric tunnel effect between dissimilar electrodes separated by a thin insulating film. Journal of Applied Physics. 1963; 34: 2581–2590. DOI:10.1063/1.1729774

[46] Tersoff J. and Hamann D. R.: Theory and application for the scanning tunneling microscope. Physical Review Letters. 1983; 50: 1998–2001. DOI:10.1103/PhysRevLett.

[47] Sze S. M. Physics of Semiconductor Devices. New York: Wiley; 1981. 868 p. ISBN:

[48] Filip V., Wong H., and Nicolaescu D.: Quantum charge transportation in metal-oxide-Si structures with ultrathin oxide. Journal of Vacuum Science and Technology B. 2006;

[49] The Kelvin Probe Principles – KP Technology Ltd. [Internet]. Available from: http://

[50] Morita S., Wiesendanger R., and Meyer E., editors. Noncontact Atomic Force Micro‐

www.kelvinprobe.info/technique-theory.htm [Accessed 2016-02-10].

scopy. Berlin: Springer; 2002. 440 p. DOI:10.1007/978-3-642-56019-4

2003.11.262

324 Microscopy and Analysis

DOI:10.1116/1.2802103.

10.1088/0953-8984/15/42/006.

10.1016/0038-1101(84)90055-8

1793–1803. DOI:10.1063/1.1702682

24: 38–45. DOI:10.1116/1.2138720

10.1063/1.1330222

50.1998

0-471-05661-8.

2009; 19(1): 117–124. DOI:10.1149/1.3118937


tunneling microscopy. Japanese Journal of Applied Physics. 2015; 54: 04DA03. DOI: 10.7567/JJAP.54.04DA03

[78] Hasegawa Y. and Eguchi T.: Potential profile around step edges of Si surface measured by nc-AFM. Applied Surface Science. 2002; 188: 386–390. DOI:10.1016/ S0169-4332(01)00955-2

[64] Mazur U. and Hipps K. W.: Resonant tunneling in metal phthalocyanines. Journal of

[65] Grobis M., Wachowiak A., Yamachika R., and Crommie M. F.: Tuning negative differential resistance in a molecular film. Applied Physics Letters. 2005; 86: 204102.

[66] Grabert H. and Devoret M., editors. Single Charge Tunneling. New York: Springer;

[67] Kane B. E.: A silicon-based nuclear spin quantum computer. Nature. 1998; 393: 133–

[68] Likharev K. K.: Dynamics of some single flux quantum devices: I. Parametric quantron. IEEE Transactions on Magnetics. 1977; 13(1): 242–244. DOI:10.1109/TMAG.

[69] Heike S. and Hashizume T.: Atomic resolution noncontact atomic force/scanning tunneling microscopy using a 1 MHz quartz resonator*.* Applied Physics Letters. 2003;

[70] An T., Nishio T., Eguchi T., Ono M., Nomura A., Akiyama K., and Hasegawa Y.: Atomically resolved imaging by low-temperature frequency-modulation atomic force microscopy using a quartz length-extension resonator. Reviews Scientific Instruments.

[71] An T., Eguchi T., Akiyama K., and Hasegawa Y. Atomically-resolved imaging by frequency-modulation atomic force microscopy using a quartz length-extension

[72] Bolotov L., Tada T., Iitake M., Nishizawa M., and Kanayama T.: Measurements of electrostatic potential across p–n junctions on oxidized Si surfaces by scanning multimode tunneling spectroscopy. Japanese Journal of Applied Physics*.* 2011; 50:

[73] Giessible F. J. Forces and frequency shifts in atomic-resolution dynamic-force micro‐ scopy. Physical Reviews B. 1997; 56: 16010–16015. DOI:10.1103/PhysRevB.56.16010 [74] Giessible F. J. A direct method to calculate tip–sample forces from frequency shifts in frequency-modulation atomic force microscopy. Applied Physics Letters. 2001; 78: 123–

[75] Giessible F. J. Advances in atomic force microscopy. Reviews Modern Physics. 2003;75:

[76] Bolotov L., Tada T., Iitake M., Nishizawa M., and Kanayama T.: Electrostatic potential fluctuations on passivated Si surfaces measured by integrated AFM-STM. e-J. Surface

[77] Bolotov L., Fukuda K., Tada T., Matsukawa T., and Masahara M.: Spatial variation of the work function in nano-crystalline TiN films measured by dual-mode scanning

Science and Nanotechnology. 2011; 9: 117–120. DOI:10.1380/ejssnt.2011.117.

resonator. Applied Physics Letters. 2005; 87: 133114. DOI:10.1063/1.2061850

Physical Chemistry. 1994; 98: 8169–8172. DOI:10.1021/j100084a040

DOI:10.1063/1.1931822

326 Microscopy and Analysis

137. DOI:10.1038/30156

1977.1059351

1992. 335 p. DOI:10.1007/978-1-4757-2166-9

83: 3620–3623. DOI:10.1063/1.1623012

04DA04. DOI:10.1143/JJAP.50.04DA04.

949–983. DOI:10.1103/RevModPhys.75.949

125. DOI:10.1063/1.1335546

2008; 79: 033703–033707. DOI:10.1063/1.2830937


## **Electron Orbital Contribution in Distance‐Dependent STM Experiments**

Alexander N. Chaika

[88] Poborchii V., Tada T, and Kanayama T.: High-spatial-resolution Raman microscopy of stress in shallow-trench-isolated Si structures. Applied Physics Letters. 2006; 89:

[89] Poborchii V., Tada T, and Kanayama T.: Edge-enhanced Raman scattering in Si nanostripes. Applied Physics Letters. 2009; 94: 131907. DOI:10.1063/1.3110964

[90] Furuhashi M. and Taniguchi K.: Additional stress-induced band gap narrowing in a silicon die. Journal of Applied Physics. 2008; 103: 026103. DOI:10.1063/1.2833435 [91] Houard J., Vella A., Vurpillot F., and Deconihout B.: Optical near-field absorption at a metal tip far from plasmonic resonance. Physical Reviews B. 2010; 81: 125411. DOI:

[92] Bardeen J.: Tunnelling from a many-particle point of view. Physical Review Letters*.*

[93] Bono J. and Good Jr. R. H.: Theoretical discussion of the scanning tunneling microscope applied to a semiconductor surface. Surface Science. 1986; 175: 415–420. DOI:DOI:

[94] Fukuda K., Nishizawa M., Tada T., Bolotov L., Suzuki K., Satoh S., Arimoto H., and Kanayama T. Three-dimensional simulation of scanning tunneling microscopy for semiconductor carrier and impurity profiling. Japanese Journal of Applied Physics.

[95] Fukuda K., Nishizawa M., Tada T., Bolotov L., Suzuki K., Sato S., Arimoto H., and Kanayama T.: Simulation of light-illuminated STM measurements. In: Proceedings of the International Conference on Simulation of Semiconductor processes and Devices

[96] Sano N. Impurity-limited resistance and phase interference of localized impurities under quasi-one dimensional nano-structures. Journal of Applied Physics. 2015; 118:

[97] Advanced Algorithm and Systems. [Internet]. 2016. Available from: https:// www.aasri.jp/pub/spm/pdf/spm\_concept\_eng.pdf [Accessed 2016-02-10].

[98] Frauenheim T., Seifert G., Elsterner M., Hajnal Z., Jungnickel G., Porezag D., Suhai S., and Scholz R.: A self-consistent charge density-functional based tight-binding method for predictive materials simulations in physics, chemistry and biology. Physica Status Solidi (b). 2000; 217(1): 41–62. DOI:10.1002/(SICI)1521-3951(200001)217:1<41::AID-

[99] Card H. C. and Rhoderick E. H. Studies of tunnel MOS diodes I. Interface effects in silicon Schottky diodes. Journal of Physics D. 1971; 4: 1589–1605. DOI:

(SISPAD 2014); 9–11 September 2014; Yokohama, Japan; 2014. pp. 129–132

233505. DOI:10.1063/1.2400057

328 Microscopy and Analysis

10.1103/PhysRevB.81.125411

10.1016/0039-6028(86)90243-8

244302. DOI:10.1063/1.4938392

PSSB41>3.0.CO;2-V

10.1088/0022-3727/4/10/319

1961; 6(2): 57–59. DOI:10.1103/PhysRevLett.6.57

2014; 116: 023701. DOI:10.1063/1.4884876

Additional information is available at the end of the chapter

http://dx.doi.org/10.5772/63270

#### **Abstract**

Scanning tunneling microscopy (STM) is one of the most powerful techniques for the analysis of surface reconstructions at the atomic scale. It utilizes a sharp tip, which is brought close to the surface with a bias voltage applied between the tip and the sample. The value of the tunneling current, flowing between the tip and the sample, is determined by the structure of the surface and the tip, the bias voltage, and the tip‐ sample distance. By scanning the tip over the surface, a tunneling current map is produced, which reflects the local atomic and electronic structures. This chapter focuses on the role of the tip‐surface distance in ultrahigh vacuum STM experiments with atomic and subatomic resolution. At small distances, i.e., comparable with interatomic distances in solids, the interaction between the tip and the surface atoms can modify their electronic structure changing the symmetry of the atomically resolved STM images and producing unusual features at the subatomic scale. These features are related to changes of the relative contribution of different electron orbitals of the tip and the surface atoms at varying distances.

**Keywords:** atomic resolution, density of states, electron orbital, gap resistance, scan‐ ning tunneling microscopy

## **1. Introduction**

The invention of scanning probe microscopy (SPM) [1–3] allowed studying the surface structures with extremely high spatial resolution. The SPM methods use sharp tips, which are ultimate‐ ly ended with a single atom at the apex, for surface imaging [4–7], fabricating low‐dimension‐ al structures from individual atoms and molecules [8–14], studying the physical properties of the nanoobjects [15, 16], and getting the information about the chemical [17, 18] and magnetic

© 2016 The Author(s). Licensee InTech. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

order on surfaces [19–21]. SPM methods are utilized in the fields of physics, chemistry and biology for precise studies of organic and inorganic nanoobjects. The subatomic spatial resolution [22–33] has been achieved during the last decade in the atomic force microscopy (AFM) and scanning tunneling microscopy (STM) experiments.

STM is based on quantum tunneling of electrons from atoms present at the surface of a sample to the front atom of an STM tip (or vice versa) through the vacuum gap. Because of the exponential distance dependence [34], the tunneling current drops approximately by one order of magnitude with every 1 Å increase in the tip‐sample separation. Because of the strong distance dependence, more than 90% of the current can be localized on only two, closest to each other, tip and surface atoms. This provides unique spatial resolution of STM, reaching the picometer scale [15, 22–24] and allowing even direct imaging individual electron orbitals at certain tunneling parameters [22–28]. However, the ultimate orbital resolution could rarely be achieved in experiments because of simultaneous contribution of different electron states of the tip and the surface atoms and nonideal geometries of the tips used. Probing particular atomic orbitals with STM is further impeded by possible modifications of the tip and surface electronic structure at very small tunneling gaps (2.0–5.0 Å), generally required for atomically resolved STM imaging.

The role of different electron states and tip‐surface distance in high‐resolution STM imaging has been discussed since 1980s. Apparently, the first anomalous distance dependence of the STM contrast was reported in reference [35]. Afterward, the distance dependence has been studied in a number of works [22, 23, 27, 28, 36–46]. However, because of complexity of the theoretical calculations for realistic tip‐surface systems and experimental studies with precise control of the tunneling gap and the tip state, there is still no detailed knowledge about the role of particular electron orbitals in the distance‐dependent STM experiments. This chapter is focused on the experimental and theoretical studies of the role of the tip‐sample distance in STM experiments with atomic and subatomic spatial resolution which can pave the way to selective probing particular electron orbitals, controllable chemical analysis at the atomic scale, and surface imaging with picometer lateral resolution.

## **2. Role of the electron orbitals and tip-surface distance in STM experiments: theory**

#### **2.1. Spatial resolution with different atomic orbitals at the tip apex**

One of the critical issues of high‐resolution STM studies is the lack of reliable information about the tip apex structure, which complicates theoretical calculations and comparison with experimental results. To simplify the calculations, Tersoff and Hamann considered a spheri‐ cally symmetric *s*‐wave STM tip [47, 48]. In this case, the tunneling current is proportional to the local density of states (DOS) of the surface at the position of the tip, integrated in the energy range defined by the bias voltage. The STM images, calculated using the Tersoff‐Hamann model, reflect the surface DOS distribution. This theory has provided a good agreement between the simulated and experimental images for numerous atomically resolved STM studies.

order on surfaces [19–21]. SPM methods are utilized in the fields of physics, chemistry and biology for precise studies of organic and inorganic nanoobjects. The subatomic spatial resolution [22–33] has been achieved during the last decade in the atomic force microscopy (AFM) and

STM is based on quantum tunneling of electrons from atoms present at the surface of a sample to the front atom of an STM tip (or vice versa) through the vacuum gap. Because of the exponential distance dependence [34], the tunneling current drops approximately by one order of magnitude with every 1 Å increase in the tip‐sample separation. Because of the strong distance dependence, more than 90% of the current can be localized on only two, closest to each other, tip and surface atoms. This provides unique spatial resolution of STM, reaching the picometer scale [15, 22–24] and allowing even direct imaging individual electron orbitals at certain tunneling parameters [22–28]. However, the ultimate orbital resolution could rarely be achieved in experiments because of simultaneous contribution of different electron states of the tip and the surface atoms and nonideal geometries of the tips used. Probing particular atomic orbitals with STM is further impeded by possible modifications of the tip and surface electronic structure at very small tunneling gaps (2.0–5.0 Å), generally required for atomically

The role of different electron states and tip‐surface distance in high‐resolution STM imaging has been discussed since 1980s. Apparently, the first anomalous distance dependence of the STM contrast was reported in reference [35]. Afterward, the distance dependence has been studied in a number of works [22, 23, 27, 28, 36–46]. However, because of complexity of the theoretical calculations for realistic tip‐surface systems and experimental studies with precise control of the tunneling gap and the tip state, there is still no detailed knowledge about the role of particular electron orbitals in the distance‐dependent STM experiments. This chapter is focused on the experimental and theoretical studies of the role of the tip‐sample distance in STM experiments with atomic and subatomic spatial resolution which can pave the way to selective probing particular electron orbitals, controllable chemical analysis at the atomic scale,

scanning tunneling microscopy (STM) experiments.

and surface imaging with picometer lateral resolution.

**2. Role of the electron orbitals and tip-surface distance in STM**

One of the critical issues of high‐resolution STM studies is the lack of reliable information about the tip apex structure, which complicates theoretical calculations and comparison with experimental results. To simplify the calculations, Tersoff and Hamann considered a spheri‐ cally symmetric *s*‐wave STM tip [47, 48]. In this case, the tunneling current is proportional to the local density of states (DOS) of the surface at the position of the tip, integrated in the energy range defined by the bias voltage. The STM images, calculated using the Tersoff‐Hamann model, reflect the surface DOS distribution. This theory has provided a good agreement

**2.1. Spatial resolution with different atomic orbitals at the tip apex**

resolved STM imaging.

330 Microscopy and Analysis

**experiments: theory**

The ultimate lateral resolution *R* within the Tersoff‐Hamann approach [47] is defined by the formula *R* = *d* ⋅2*k* <sup>−</sup><sup>1</sup> , where *d* is the distance between the interacting tip and surface atoms and *k*‐1 ≈ 1 Å. The limit of the lateral resolution for a tip‐sample distance of 4.5 Å should be about 3 Å. The theory [47] predicts an enhancement of the spatial resolution with decreasing tip‐ sample distance but cannot explain the large atomic corrugations in STM images of metal surfaces [49–51] and the experimentally observed sub‐Ångström lateral resolution [7, 22–24], even at tunneling gaps in the range of 2.0–2.5 Å. The former can be explained either by the tip and surface atom relaxations [37] or by decisive contribution of higher momentum (*l*) electron states with nonzero momentum projections on the *z*‐axis (*m* ≠ 0 states) [52–56]. For example, calculations by Tersoff and Lang for different tip atoms (Mo, Na, Ca, Si, and C) suggest that atomic corrugations on the same surface can vary by one order of magnitude depending on the relative contribution of the electron states with different values of *l* and *m* [52].

To get more general description of the tip structure and to clarify the role of particular electron orbitals in STM imaging, Chen introduced the so‐called derivative rule [54], where the tunneling matrix elements, corresponding to individual orbital contributions, are proportional to the *z* derivatives of the surface atom wave functions at the center of the tip apex atom. The total tunneling current can be calculated by summing up all individual contributions of the electron orbitals with different values of *l* and *m* [54]. According to the theory [54, 55], *dz 2* and *pz* tip electron states can be responsible for enhanced atomic resolution and large corrugations on metallic surfaces. For the tips having *m* = 0 electron states at the apex, the simulated STM images should correspond to the surface DOS distribution, similar to the *s*‐wave tip. This assumes that the best STM tips should possess either *dz 2* or *pz* electron states at the apex. At the same time, the *m* ≠ 0 tip states (e.g., *dxz*, *dyz*, *dxy*, and *dx 2 –y2* ) can provide enhanced atomic corrugations and intra‐atomic effects at small distances [53–56]. For example, *dxz*, *dyz*, *dxy*, and *dx 2 –y2* electron orbitals at the apex can produce twofold and fourfold split subatomic features, respectively, instead of a single atom at very small tunneling gaps [56]. The dependence of atomic corrugations on the tunneling gap resistance, calculated for different electron orbitals by Sacks [56], suggested that *dz 2* and *m* ≠ 0 *d*‐orbitals can produce enhanced corrugations in comparison to the *s‐*state. Furthermore, the corrugations can substantially increase with decreasing gap resistance.

The improvement of the lateral resolution of the SPM to the picometer scale [7, 22–24, 57–60], observation of orbital channels in tunneling conductance [61], and subatomic contrast in SPM experiments [22–32] led to the development of new theories accounting for the distance dependence of the electron orbital contribution [62, 63], energy‐dependent combinations of the tip orbitals [64–68], and effects of inter‐ and intra‐atomic interference of the electron orbitals in the tunneling junction [69–73]. Recently, the revised Chen's derivative rule, accounting for the orbital interference effects, has been proposed by Mandi and Palotas [74], who considered the realistic electronic structure and arbitrary spatial orientation of the tip minimizing the computational efforts. Their results showed that the electronic interference effects have a considerable effect on the STM images. As an example, in certain cases a tip with a mixture of s and *pz* electron states can provide even higher spatial resolution than pure *pz*‐orbital tip [74].

#### **2.2. Electronic structure of realistic tips at small tunneling gaps**

Although the developed theories can provide a correlation between experimental and theoretical images in certain cases, they rarely take into account possible interaction‐induced changes of the tip orbital structure at very small tunneling gaps. It has been demonstrated recently [62–64, 75] that electronic structure of the transition metal tips and relative weights of different *d*‐orbitals in the tip's DOS near *E*<sup>F</sup> strongly depend on the tip element, tip cluster size, and local environment around the tip. In particular, it has been shown that *dz 2* ‐orbital is more sensitive to the local environment and the tip cluster size than *m* ≠ 0 *d*‐orbitals (*dxz*, *dyz*, *dxy,* and *dx 2 ‐y 2* ) [63]. The electronic structure of different transition metal tips can nonequally depend on the tip‐surface distance [63, 75]; therefore, determination of the most suitable tip for particular surface can be crucial for reaching highest spatial resolution in the experiments. Stimulated by the subatomic resolution AFM experiments [33], Wright and Solares calculated

**Figure 1.** (a–c) The d*ρ* = 0.08 eV/Å3 isosurface of the change in electron density for the (a) He–W[011], (b) He–W[111], and (c) He–W[001] systems. The helium atom is positioned directly below the front atom of the tip, with a gap of 4.0 Å for the W[011] tip and 3.5 Å for the W[111] and W[001] tips. (d–f) Constant‐height slices through the electron density distribution. The relative sizes and positions of the first two layers of the tips' tungsten atoms are drawn as black cir‐ cles. Reproduced from Reference [75] with permission of IOP.

the electronic structure of the tungsten tips with different crystallographic orientations in proximity to different surface clusters [75–77]. The calculations showed that the experimental SPM images with subatomic resolution [22–24, 33] were, most probably, related to the tip electronic structure, which could become asymmetric at small tip‐sample distances because of the tip‐sample interaction. The stability of the tungsten tips was found to be dependent on the crystallographic orientation of the apex. In particular, W[001] tip did not substantially relax even at the tip‐sample distance of 1.50 Å, while [011]‐ and [111]‐oriented tungsten tips were stable only at gaps above 2.25 and 2.50 Å, respectively. **Figure 1(a)**–**(c)** shows the isosurfaces of the change in the electron density calculated for the [011]‐, [111]‐, and [001]‐oriented tungsten tips interacting with the helium atom positioned directly below the front atom of the tips [75]. **Figure 1(d)**–**(f)** illustrates that essentially asymmetric charge density distribution is observed near the front atom of the W[011] and W[001] tips. These areas of increased charge density (dark red color) reveal the twofold and fourfold symmetry representatives of each tip's crystallographic orientation. However, for the W[111] tip, the areas of increased density do not reveal the threefold symmetry proposed from the crystal symmetry consideration [33]. These results suggest that the highest spatial resolution can be achieved in STM experiments using W[111] tips which have symmetrical charge density distribution along the tip axis even at relatively small distances from the surface. In contrast, W[001] and W[011] tips can produce subatomic features at small tip‐sample distances [22–24] but can be more suitable for scanning tunneling spectroscopy (STS) experiments demanding high tip stability.

#### **2.3. Role of atomic relaxations at small tip-surface distances**

considerable effect on the STM images. As an example, in certain cases a tip with a mixture of s and *pz* electron states can provide even higher spatial resolution than pure *pz*‐orbital tip [74].

Although the developed theories can provide a correlation between experimental and theoretical images in certain cases, they rarely take into account possible interaction‐induced changes of the tip orbital structure at very small tunneling gaps. It has been demonstrated recently [62–64, 75] that electronic structure of the transition metal tips and relative weights of different *d*‐orbitals in the tip's DOS near *E*<sup>F</sup> strongly depend on the tip element, tip cluster

more sensitive to the local environment and the tip cluster size than *m* ≠ 0 *d*‐orbitals (*dxz*, *dyz*,

depend on the tip‐surface distance [63, 75]; therefore, determination of the most suitable tip for particular surface can be crucial for reaching highest spatial resolution in the experiments. Stimulated by the subatomic resolution AFM experiments [33], Wright and Solares calculated

) [63]. The electronic structure of different transition metal tips can nonequally

isosurface of the change in electron density for the (a) He–W[011], (b) He–W[111],

and (c) He–W[001] systems. The helium atom is positioned directly below the front atom of the tip, with a gap of 4.0 Å for the W[011] tip and 3.5 Å for the W[111] and W[001] tips. (d–f) Constant‐height slices through the electron density distribution. The relative sizes and positions of the first two layers of the tips' tungsten atoms are drawn as black cir‐

*2*

‐orbital is

size, and local environment around the tip. In particular, it has been shown that *dz*

**2.2. Electronic structure of realistic tips at small tunneling gaps**

*dxy,* and *dx*

332 Microscopy and Analysis

*2 ‐y 2*

**Figure 1.** (a–c) The d*ρ* = 0.08 eV/Å3

cles. Reproduced from Reference [75] with permission of IOP.

An enhancement of the lateral resolution and atomic corrugations with decreasing tip‐surface distance was anticipated from the theories [47, 56] omitting possible relaxations of the interacting tip and surface atoms. Additionally, the topographic contrast in STM experiments can be enhanced at small distances because of the tip and surface atom relaxations reported for the first time almost three decades ago [49]. The role of elastic effects in high‐resolution STM experiments on close‐packed metal surfaces was thoroughly studied in Ref. [37]. To evaluate possible corrugation enhancement related to atomic relaxations, the simulations were performed on Cu(111) and Al(111) surfaces at different tip configurations and tunneling gaps [37]. The relaxations were found to be substantially stronger for Al(111) because of the higher elasticity of the aluminum surface. As a result, the atomic corrugations in the simulated images of Al(111) increased almost by one order of magnitude (from 10 to 70 pm) with decreasing distance from 600 to 400 pm. The calculations [37] explain the anomalously high corrugation amplitudes observed in numerous STM experiments on metal surfaces. The large outward relaxations of the surface atoms under typical tunneling conditions can lead to atomic corrugations which can hardly be explained from the electron DOS and simplified theoretical considerations [47, 56]. The corrugation enhancement was proved on different metal surfaces; therefore, one can anticipate the validity of the work [37] for other metals, although the effect cannot be so well pronounced as calculated for aluminum surface scanned by an aluminum‐ terminated tip.

#### **2.4. Reduction in tunneling current channels with decreasing tip‐surface distance**

First observations of the tunneling current channels associated with different electron orbitals were reported almost two decades ago [61]. However, their role in STM imaging at different gaps is still not clear. Recent theoretical studies showed that relative contributions of different electron orbitals can be substantially modified because of tip‐sample interaction changing the tip and surface DOS near *E*<sup>F</sup> [62–64, 75, 78–80]. For example, **Figure 1** demonstrates the distinct asymmetry of the [001]‐ and [011]‐oriented tungsten probes appearing only at small gaps when the overlap of the orbitals of the interacting tungsten and helium atoms takes place [75]. **Figure 2** shows the theoretical calculations demonstrating a drastic reduction of the DOS near *E*<sup>F</sup> associated with the Si(111)7×7 surface atom interacting with the [001]‐oriented tungsten tip [79]. The interaction modifies the electronic structure of the surface atom at distances below 4.75 Å, as shown in **Figure 2(d)**. The calculations predict a nonmonotonous dependence of the interaction forces [**Figure 2(b)**] and the tunneling current [**Figure 2(c)**] at these distances. The calculations reveal the exponential current–distance dependence at large distances and drop of tunneling conductance at a distance of 4.75 Å [**Figure 2(c)**]. **Figure 2(b)** and **(c)** demonstrates a correlation between the upward adatom displacement, the decrease of the conductance, and the formation of the chemical bond between the apex atom and the surface adatom. At large tip‐sample distances, there is no significant vertical displacement of the adatom [**Figure 2(c)**]. Large upward relaxations of the adatom are observed at the distances of around 5 Å. At the same distances, the short‐range chemical force between the tip and the sample changes rapidly

**Figure 2.** (a) Isosurfaces of electron charge density (the isovalue is 0.08 *e*/Å3 ) integrated in the energy range from *E*<sup>F</sup> to *E*<sup>F</sup> – 0.4 eV, for different tunneling gaps. The probe is placed over the corner adatom in the Si(111)7×7 faulted unit cell. (b) Evolution of the quantum conductance *G* (right axis) and the atomic force (left axis) as a function of the tip‐sample distance. (c) The quantum conductance (left axis) and the vertical adatom displacement (right axis) as a function of the distance. (d) PDOS of the silicon corner adatom as a function of the tip‐sample distance. Reproduced from Reference [78] with permission of APS.

[**Figure 2(b)**]. Then, the vertical adatom displacement decreases until it reaches its initial value at a tip‐sample distance of 3 Å. The electron DOS isosurfaces calculated for different tip‐sample distances [**Figure 2(a)**] show the typical charge density distribution of the Si(111)7×7 surface at 6.0 Å, a charge transfer from the adatom to neighboring surface atoms at 4.75 Å, the suppression of the adatom dangling bond at 4.5 Å, and the chemical bond formation at 3.0 Å that prevents the electron transport between the tip and the surface at small distances. The W[001] tip electronic structure remained practically unchanged in this tip‐sample distance range. Note that calculations of Jelinek et al. [78] were carried out at low voltages, correspond‐ ing to probing the *pz*‐states dominating in the Si(111)7×7 surface DOS near *E*F. Therefore, it can be assumed that the overlap of the silicon *pz*‐orbital with the tungsten tip orbitals is responsible for the strong modification of the tunneling current as shown in **Figure 2(d)**. Furthermore, one can expect similar phenomena in other systems where the electron transport occurs through *pz* electron orbitals.

**2.4. Reduction in tunneling current channels with decreasing tip‐surface distance**

**Figure 2.** (a) Isosurfaces of electron charge density (the isovalue is 0.08 *e*/Å3

[78] with permission of APS.

334 Microscopy and Analysis

*E*<sup>F</sup> – 0.4 eV, for different tunneling gaps. The probe is placed over the corner adatom in the Si(111)7×7 faulted unit cell. (b) Evolution of the quantum conductance *G* (right axis) and the atomic force (left axis) as a function of the tip‐sample distance. (c) The quantum conductance (left axis) and the vertical adatom displacement (right axis) as a function of the distance. (d) PDOS of the silicon corner adatom as a function of the tip‐sample distance. Reproduced from Reference

) integrated in the energy range from *E*<sup>F</sup> to

First observations of the tunneling current channels associated with different electron orbitals were reported almost two decades ago [61]. However, their role in STM imaging at different gaps is still not clear. Recent theoretical studies showed that relative contributions of different electron orbitals can be substantially modified because of tip‐sample interaction changing the tip and surface DOS near *E*<sup>F</sup> [62–64, 75, 78–80]. For example, **Figure 1** demonstrates the distinct asymmetry of the [001]‐ and [011]‐oriented tungsten probes appearing only at small gaps when the overlap of the orbitals of the interacting tungsten and helium atoms takes place [75]. **Figure 2** shows the theoretical calculations demonstrating a drastic reduction of the DOS near *E*<sup>F</sup> associated with the Si(111)7×7 surface atom interacting with the [001]‐oriented tungsten tip [79]. The interaction modifies the electronic structure of the surface atom at distances below 4.75 Å, as shown in **Figure 2(d)**. The calculations predict a nonmonotonous dependence of the interaction forces [**Figure 2(b)**] and the tunneling current [**Figure 2(c)**] at these distances. The calculations reveal the exponential current–distance dependence at large distances and drop of tunneling conductance at a distance of 4.75 Å [**Figure 2(c)**]. **Figure 2(b)** and **(c)** demonstrates a correlation between the upward adatom displacement, the decrease of the conductance, and the formation of the chemical bond between the apex atom and the surface adatom. At large tip‐sample distances, there is no significant vertical displacement of the adatom [**Figure 2(c)**]. Large upward relaxations of the adatom are observed at the distances of around 5 Å. At the same distances, the short‐range chemical force between the tip and the sample changes rapidly

Recent theoretical calculations of the tunneling current between an STM tip and a single Cu/Co atom adsorbed on a Cu(001) surface [62] have shown that conductance in these systems can be decomposed into several orbital contributions. The tunneling probabilities of these individual channels can provide the information about the modifications of the adatom's DOS caused by the tip‐adatom interaction. The calculations revealed that the *dz 2* ‐orbitals of the interacting atoms in such systems are especially sensitive to the tip‐adatom distance because they start to overlap at larger distances. The DOS associated with the *dx 2 –y2* ‐orbitals starts to decrease at smaller tunneling gaps and the total reduction at distances between 4.0 and 2.5 Å for this orbital is smaller in comparison with that of the *dxz‐*, *dyz‐*, and *dz 2* ‐orbitals. At tip‐sample distances exceeding 3*.*0 Å, the channels contributing to the most of the tunneling current were generally related to hybrids of *s‐* and *dz* 2 ‐orbitals. The contribution of *m* ≠ 0 electron states increased with decreasing distances.

## **3. Electron orbital resolution in distance‐dependent STM experiments**

The first STM images with the atomic orbital resolution were obtained at the beginning of 1980s. For example, typical images of the Si(111)7×7 surface [5] correspond to direct visuali‐ zation of the *pz* surface orbitals. Similar selective visualization of the atomic orbitals at certain tunneling conditions was achieved on other semiconducting surfaces [17]. The control of the orbital contribution on multicomponent metal surfaces is more difficult. Up to date, several STM studies demonstrating subatomic, electron orbital contrast have been published. In several cases, the observed features were related to direct visualization of the electronic structure of the tip atom by more localized surface atomic orbitals [22–28]. However, some STM studies revealed asymmetric subatomic features associated with the *m*≠ 0 electron orbitals of the surface atoms [30, 81]. The STM studies presented in this section can be important for optimizing the tunneling parameters for high‐resolution STM imaging and shed light on the gap resistance dependence of atomically resolved STM images and chemical contrast observed on metallic and semiconducting surfaces.

## **3.1. Tip orbitals resolved using** *pz* **states of the Si(111)7×7 surface atoms**

SPM imaging with subatomic resolution was first claimed by Giessibl et al. [31] who reported the AFM images of the Si(111)7×7 reconstruction demonstrating a regular twofold splitting of the surface atomic features. This was explained by direct visualization of the two atomic orbitals of an Si[001] tip atom by the *pz*‐orbitals of the surface atoms. This is schematically shown in **Figure 3(b)**. Later studies [28] showed that qualitatively the same asymmetric features can be resolved with STM using a silicon‐terminated tungsten tip. **Figure 3(a)** demonstrates the change of the contrast in the STM experiments [28] with decreasing tunneling gap resistance. Note that precise structure of the tip apex, responsible for the subatomic contrast in both SPM experiments [28, 31], was unknown because of the tip preparation procedure. Therefore, the origin of the subatomic features in the Si(111)7×7 SPM images was disputed in a number of papers [82–87]. Alternative explanations based on the feedback loop

**Figure 3.** (a) 7.1×7.1 nm2 STM images of the Si(111)7×7 surface measured at different bias voltages and tunneling cur‐ rents (shown on each frame) with the silicon-terminated tungsten tip [28]. The fast scanning direction was from left to right. (b) Side view of a [001]‐oriented silicon tip over the Si(111)7×7 surface; right panel depicts a tip bending opposite to the scan direction. (c,d) Zooms of the double features indicated by squares on images measured at *U* = –0.3 V, *I* = 80 pA (c) and *U* = –0.1 V, *I* = 80 pA (d).

artifacts [83], presence of a carbon atom at the apex [86], and visualization of the surface atoms' backbonds [87] were proposed. Nevertheless, independent theoretical calculations [31, 32, 82, 84] support the possibility of direct visualization of the asymmetric charge distribution around the [001]‐oriented Si tip atom [**Figure 3(b)**] at small (2.5–4.0 Å) tip‐surface separations. This is in agreement with the gap‐resistance–dependent STM experiments [28]. **Figure 3(a)** demon‐ strates that the twofold splitting of the adatoms becomes discernible only at small bias voltages (small distances), while at large negative voltages a (7×7) pattern with with well‐resolved adatoms, rest atoms, corner holes, and single-atom defects are observed, which is only possible with extremely sharp single atom–terminated tips [88]. The effect can hardly be explained by the formation of a two‐atom terminated apex and visualization of the surface atom backbonds because deep corner holes, rest atoms, and single atom defects on the surface were simulta‐ neously resolved even at low gap resistances. The double features become sharper and change their appearance from symmetric to asymmetric with decreasing gap resistance, as shown in **Figure 3 (c)** and **(d)**. The shape of the double features could be changed both by increasing tunneling current at fixed voltage and decreasing voltage at the same current [28]. The asymmetry of the double features at small gaps can be explained by the relaxation of the apex atom, as shown in **Figure 3(b)**. The distance between the two maxima of the double features decreased from 2.7 to 2.2 Å with decreasing gap resistance from 7.5 to 2.5 GΩ and then increased to 2.75 Å with further decreasing gap resistance from 2.5 to 1.25 GΩ. The decrease of the distance between the subatomic maxima is in line with the theoretical spatial distribution of the two *sp3* dangling bonds at the [001]‐oriented silicon apex [32]. The increase of the splitting at smaller distances can be related to tip‐sample interaction modifying the electronic structure of the apex atom and inducing lateral relaxations of the tip and surface atoms in near‐to‐contact regime.

**3.1. Tip orbitals resolved using** *pz* **states of the Si(111)7×7 surface atoms**

**Figure 3.** (a) 7.1×7.1 nm2

336 Microscopy and Analysis

pA (c) and *U* = –0.1 V, *I* = 80 pA (d).

SPM imaging with subatomic resolution was first claimed by Giessibl et al. [31] who reported the AFM images of the Si(111)7×7 reconstruction demonstrating a regular twofold splitting of the surface atomic features. This was explained by direct visualization of the two atomic orbitals of an Si[001] tip atom by the *pz*‐orbitals of the surface atoms. This is schematically shown in **Figure 3(b)**. Later studies [28] showed that qualitatively the same asymmetric features can be resolved with STM using a silicon‐terminated tungsten tip. **Figure 3(a)** demonstrates the change of the contrast in the STM experiments [28] with decreasing tunneling gap resistance. Note that precise structure of the tip apex, responsible for the subatomic contrast in both SPM experiments [28, 31], was unknown because of the tip preparation procedure. Therefore, the origin of the subatomic features in the Si(111)7×7 SPM images was disputed in a number of papers [82–87]. Alternative explanations based on the feedback loop

STM images of the Si(111)7×7 surface measured at different bias voltages and tunneling cur‐

rents (shown on each frame) with the silicon-terminated tungsten tip [28]. The fast scanning direction was from left to right. (b) Side view of a [001]‐oriented silicon tip over the Si(111)7×7 surface; right panel depicts a tip bending opposite to the scan direction. (c,d) Zooms of the double features indicated by squares on images measured at *U* = –0.3 V, *I* = 80 Herz et al. applied a Co6Fe3Sm tip for high‐resolution STM experiments on the Si(111)7×7 surface [25]. They utilized a dynamic‐STM mode with an oscillating probe to reduce the lateral forces in tip‐sample contact and increase the tip stability at extremely small tunneling gaps. At some tunneling parameters, the adatoms were resolved as extremely sharp spherically symmetric features surrounded by lower lying crescents. These asymmetric features, observed at small gaps, were explained by a convolution of the *pz*‐orbitals of the surface atoms and the *fz 3* ‐orbital of an Sm atom at the apex tilted to the surface normal by 37°. The validity of this interpretation was supported by the theoretical calculations in an assumption of the tilted Sm *fz 3* tip orbital [25].

## **3.2.** *dyz* **electron orbital of a MnNi tip resolved in STM experiments on the Cu(014)–O surface**

**Figure 4(a)** shows the STM image of the Cu(014)–O surface measured using a polycrystalline MnNi tip [26, 27]. The image demonstrates regular twofold splitting of the copper atomic features along the [1–10] direction. The experimental image displays the 7.2±0.2 Å wide terraces, step edges along the [100] direction and an additional fine structure within the terraces. The image reveals a single atom defect proving the sharpness of the MnNi tip. The doubling of atomic features was observed in rare experiments at small negative bias voltages between -30 and -50 mV in a very narrow range of the tunneling currents. That was explained by the distance dependence of the Cu(014)–O STM images and strong dependence of the tip's electronic structure on the crystallographic orientations of the apex [26, 27].

**Figure 4(b)** shows the partial density of electron states (PDOS) associated with the Ni and Mn atoms at the apexes of the [001]‐ and [111]‐oriented MnNi tips. The tight binding (TB) calcu‐ lations demonstrate domination of the *dyz*‐orbital near the Fermi level only for the [111]‐ oriented MnNi tip terminated by an Mn atom [**Figure 4(b)**]. The density functional theory (DFT) calculations shown in **Figure 4(c)** demonstrate that only certain configurations of the Mn‐terminated MnNi[111] tips could produce regular doubling with almost symmetric twofold split features [27]. The regular pattern of the double features reproducing the shape of the Mn atom *dyz* electron orbital could be resolved when the Cu(014)–O surface *dz 2* ‐orbitals and the tip *dyz*‐orbital provided major contribution to the tunneling current. This is schemati‐ cally shown in **Figure 4(a)**.

**Figure 4.** (a) Regular doubling of atomic features in STM images of the Cu(014)–O surface measured with a MnNi tip (bottom) and schematic model of the *dyz* tip orbital scanning over *dz 2* surface orbitals (top). Eight instead of four fea‐ tures are resolved along the [1–10] direction within each terrace. The image was taken at *U* = -30 mV and *I* = 80 pA. (b) PDOS associated with the *d*‐orbitals of the Ni‐terminated MnNi[001] tip (top) and Mn‐terminated MnNi[111] tip (bot‐ tom). (c) Models (left) and calculated electron density isosurfaces (right) for different MnNi[111] tip configurations. The isosurfaces display the electron density at 2.4 × 10-3 electrons per Å3 in the *E*<sup>F</sup> ± 0.22 eV energy range. The apex atom is indicated by circles. Reproduced from Reference [27] with permission of APS.

#### **3.3. Distance dependence of the W[001] tip orbital contribution in STM experiments on graphite**

The selection of the tip orbital responsible for high‐resolution STM imaging has been demonstrated in references [22–24]. The electron orbitals of the graphite surface atoms were used to study the relative contribution of the [001]‐oriented single crystalline tungsten tip orbitals at different distances (**Figure 5**). To avoid apex contamination, W[001] tips were cleaned by flash heating and ion sputtering in ultra-high vacuum (UHV) before the distancedependent experiments. The transmission electron microscopy (TEM) studies proved the fabrication of the [001]‐oriented nanopyramids with well‐defined structure at the apexes [24].

by the distance dependence of the Cu(014)–O STM images and strong dependence of the tip's

**Figure 4(b)** shows the partial density of electron states (PDOS) associated with the Ni and Mn atoms at the apexes of the [001]‐ and [111]‐oriented MnNi tips. The tight binding (TB) calcu‐ lations demonstrate domination of the *dyz*‐orbital near the Fermi level only for the [111]‐ oriented MnNi tip terminated by an Mn atom [**Figure 4(b)**]. The density functional theory (DFT) calculations shown in **Figure 4(c)** demonstrate that only certain configurations of the Mn‐terminated MnNi[111] tips could produce regular doubling with almost symmetric twofold split features [27]. The regular pattern of the double features reproducing the shape

and the tip *dyz*‐orbital provided major contribution to the tunneling current. This is schemati‐

**Figure 4.** (a) Regular doubling of atomic features in STM images of the Cu(014)–O surface measured with a MnNi tip

tures are resolved along the [1–10] direction within each terrace. The image was taken at *U* = -30 mV and *I* = 80 pA. (b) PDOS associated with the *d*‐orbitals of the Ni‐terminated MnNi[001] tip (top) and Mn‐terminated MnNi[111] tip (bot‐ tom). (c) Models (left) and calculated electron density isosurfaces (right) for different MnNi[111] tip configurations.

**3.3. Distance dependence of the W[001] tip orbital contribution in STM experiments on**

The selection of the tip orbital responsible for high‐resolution STM imaging has been demonstrated in references [22–24]. The electron orbitals of the graphite surface atoms were used to study the relative contribution of the [001]‐oriented single crystalline tungsten tip

*2*

surface orbitals (top). Eight instead of four fea‐

in the *E*<sup>F</sup> ± 0.22 eV energy range. The apex

(bottom) and schematic model of the *dyz* tip orbital scanning over *dz*

The isosurfaces display the electron density at 2.4 × 10-3 electrons per Å3

atom is indicated by circles. Reproduced from Reference [27] with permission of APS.

*2* ‐orbitals

of the Mn atom *dyz* electron orbital could be resolved when the Cu(014)–O surface *dz*

cally shown in **Figure 4(a)**.

338 Microscopy and Analysis

**graphite**

electronic structure on the crystallographic orientations of the apex [26, 27].

**Figure 5.** (a) Left: schematic model of a [001]‐oriented W tip over a graphite (0001) surface. Right: 6 × 6 Å<sup>2</sup> (top) and 1.8 × 1.8 Å<sup>2</sup> (bottom) STM images measured with the W[001] tips at *U* = 23 mV and *I* = 2.7 nA (left panels), *U* = –35 mV and *I* = 7.4 nA (central panels), *U* = –100 mV and *I* = 1.7 nA (right panels). (b) PDOS associated with different *d*-orbitals of the W[001] tip atom interacting with the graphite (0001) surface. The distances between interacting tip and surface atom nuclei are indicated on each panel. (c) Gap resistance dependence of graphite (0001) STM images measured using a W[001] tip at *U* = –35 mV. Tunneling currents are indicated on each frame. Panel (b) is reproduced from reference [23] with permission of Elsevier.

**Figure 5(c)** shows a gap resistance dependence measured with a W[001] tip at a bias voltage of −35 mV. At smaller currents (larger distances) a hexagonal pattern is observed. The atomic features become sharper with increasing tunneling current (decreasing distance). With further increase in the tunneling current, the symmetric features are transformed into two, three, and fourfold split subatomic features at currents between 5.3 and 9.1 nA. Similar subatomic features were earlier observed in AFM experiments with polycrystalline tungsten probes [33] and ascribed to the orbital structure of the tungsten tips with three different crystallographic orientations. The gap resistance dependence measured with the unchanged W[001] tip [**Figure 5(c)**] demonstrates that these images reproduce the electronic structure of the same tip atom modified with decreasing tip‐sample distance. The two and fourfold split features in **Figure 5** cannot correspond to the threefold symmetrical graphite surface. At the same time, the observed two and threefold split features cannot be related to the tip atom backbonding because of the symmetry of the W[001] tip used in experiments. The subatomic features can only be explained by a direct visualization of the tip atom's electronic structure modified by the tip‐sample interaction at small tunneling gaps. According to the tunneling currents, the observed transformation of the asymmetric features took place in the narrow range of tip‐ sample distances of the order of 0.1 Å. The actual distance could be slightly above this value because of the tip and surface atom relaxations at small distances [37].

At certain tunneling parameters (voltages and gap resistances), the shapes of the subatomic features in the STM experiments reproduced the electron density distribution associated with the *dz 2 ‐*, *dxz‐,* and *dxy‐*orbitals in the tungsten atom. This can be explained by a major contribution of particular electron orbitals of the tip atom at certain tip‐sample distances [**Figure 5(a)**]. According to **Figure 5(c)**, the relative contribution of the *dxz‐* and *dxy*‐orbitals increases with decreasing gap resistance. PDOS calculations for the fully relaxed "W[001] tip–graphite" system [**Figure 5(b)**] demonstrate a suppression of the tip *dz 2* electron orbital near *E*<sup>F</sup> at distances below 2.5 Å because of the overlap with the carbon orbitals. The domination of the tip *dxy* electron orbital near *E*F is observed in a narrow range of the tip‐surface distances between 2.2 and 2.5 Å, which is in agreement with the current values in **Figure 5(c)**. The surface imaging by the tungsten *dz 2* ‐orbital can be realized at tunneling gaps between 2.5 and 4.0 Å, while *dxy* electron orbital can yield a maximum contribution to the tunneling current at distances between 2.2 and 2.5 Å. The appearance of the asymmetric features is related to the interaction‐ induced modification of the PDOS associated with different *d‐*orbitals of the tungsten tip atom.

The DFT calculations [**Figure 5(b)**] and distance‐dependent STM experiments [**Figure 5(c)**] show the correlation between the spatial distribution of the atomic orbitals (in particular, extension in the *z*‐direction) and the order of their suppression with decreasing tip‐sample distance. The suppression of the further protruded in the *z*‐direction electron orbitals with *m* = 0 can increase the contribution of the *m* ≠ 0 electron states at small gaps. F**igure 5** demonstrates that picoscale spatial resolution and even direct imaging of the transition metal *d*‐orbitals using the *p*‐orbitals of light elements can experimentally be achieved.

#### **3.4.** *dxz***‐orbitals of the surface atoms resolved using tungsten tips**

STM images shown in **Figures 3**–**5** correspond to probing the tip electronic structure by the surface atomic orbitals. Similar STM experiments revealing the tip atom's subatomic structure with decreasing tunneling gap have recently been reported for several tip‐sample systems [89– 92]. **Figure 6(a)** shows the STM image of the Cu(014)–O surface measured with a polycrystal‐ line tungsten tip [30]. This image reveals one bright atomic row within the four‐atom‐wide terraces [see the model in **Figure 4(a)**]. The atomic features within the well‐resolved copper row have two maxima separated by approximately 1 Å [**Figure 6(b)**] reproducing the electron density distribution in the Cu *dxz* atomic orbital. According to the TB calculations [**Fig‐ ure 6(c)**], the DOS associated with the *dxz*‐orbitals of the surface copper atoms is maximal for the fourth (down‐step) copper row of each terrace suggesting that visualization of one atomic row within the terraces can be achieved if the surface *dxz*‐orbitals yield the largest contribution to the tunneling current. Similar to the STM experiments with the W[001] tips [22–24], this can be reached at small bias voltages and small tip‐sample distances, where the *dxz*‐orbital channel dominates in the tunneling current. The STM image with one well‐resolved row with double features (**Figure 6(a)**) can be explained by imaging the *dxz*‐orbitals of the Cu(014)–O surface using the *dz 2* ‐orbital of a tungsten tip atom.

fourfold split subatomic features at currents between 5.3 and 9.1 nA. Similar subatomic features were earlier observed in AFM experiments with polycrystalline tungsten probes [33] and ascribed to the orbital structure of the tungsten tips with three different crystallographic orientations. The gap resistance dependence measured with the unchanged W[001] tip [**Figure 5(c)**] demonstrates that these images reproduce the electronic structure of the same tip atom modified with decreasing tip‐sample distance. The two and fourfold split features in **Figure 5** cannot correspond to the threefold symmetrical graphite surface. At the same time, the observed two and threefold split features cannot be related to the tip atom backbonding because of the symmetry of the W[001] tip used in experiments. The subatomic features can only be explained by a direct visualization of the tip atom's electronic structure modified by the tip‐sample interaction at small tunneling gaps. According to the tunneling currents, the observed transformation of the asymmetric features took place in the narrow range of tip‐ sample distances of the order of 0.1 Å. The actual distance could be slightly above this value

At certain tunneling parameters (voltages and gap resistances), the shapes of the subatomic features in the STM experiments reproduced the electron density distribution associated with

distances below 2.5 Å because of the overlap with the carbon orbitals. The domination of the tip *dxy* electron orbital near *E*F is observed in a narrow range of the tip‐surface distances between 2.2 and 2.5 Å, which is in agreement with the current values in **Figure 5(c)**. The surface imaging

electron orbital can yield a maximum contribution to the tunneling current at distances between 2.2 and 2.5 Å. The appearance of the asymmetric features is related to the interaction‐ induced modification of the PDOS associated with different *d‐*orbitals of the tungsten tip atom.

The DFT calculations [**Figure 5(b)**] and distance‐dependent STM experiments [**Figure 5(c)**] show the correlation between the spatial distribution of the atomic orbitals (in particular, extension in the *z*‐direction) and the order of their suppression with decreasing tip‐sample distance. The suppression of the further protruded in the *z*‐direction electron orbitals with *m* = 0 can increase the contribution of the *m* ≠ 0 electron states at small gaps. F**igure 5** demonstrates that picoscale spatial resolution and even direct imaging of the transition metal *d*‐orbitals using

STM images shown in **Figures 3**–**5** correspond to probing the tip electronic structure by the surface atomic orbitals. Similar STM experiments revealing the tip atom's subatomic structure with decreasing tunneling gap have recently been reported for several tip‐sample systems [89– 92]. **Figure 6(a)** shows the STM image of the Cu(014)–O surface measured with a polycrystal‐ line tungsten tip [30]. This image reveals one bright atomic row within the four‐atom‐wide

*‐*, *dxz‐,* and *dxy‐*orbitals in the tungsten atom. This can be explained by a major contribution of particular electron orbitals of the tip atom at certain tip‐sample distances [**Figure 5(a)**]. According to **Figure 5(c)**, the relative contribution of the *dxz‐* and *dxy*‐orbitals increases with decreasing gap resistance. PDOS calculations for the fully relaxed "W[001] tip–graphite"

*2*

‐orbital can be realized at tunneling gaps between 2.5 and 4.0 Å, while *dxy*

electron orbital near *E*<sup>F</sup> at

because of the tip and surface atom relaxations at small distances [37].

system [**Figure 5(b)**] demonstrate a suppression of the tip *dz*

the *p*‐orbitals of light elements can experimentally be achieved.

**3.4.** *dxz***‐orbitals of the surface atoms resolved using tungsten tips**

the *dz 2*

340 Microscopy and Analysis

by the tungsten *dz*

*2*

**Figure 6.** (a) A 3.4×2.1 nm<sup>2</sup> STM image of the Cu(014)–O surface illustrating the asymmetry of the atomic features asso‐ ciated with the down‐step copper row resolved using a polycrystalline W tip at *U* = –5 mV and *I* = 0.1 nA [30]. (b) A 2.8×2.8 Å<sup>2</sup> STM image of a subatomic feature reproducing the shape of the Cu *dxz* electron orbital. (c) PDOS associated with the Cu *dxz*‐orbitals and O pz‐orbitals of the Cu(014)–O surface. The model of the surface is shown in **Figure 4(a)**.

#### **3.5. STM imaging of graphite (0001) using a [111]‐oriented diamond tip**

Conductive diamond tips, generally considered as AFM probes, can provide carbon atomic orbitals for imaging with picometer lateral resolution (**Figure 5**) and high apex stability at small tunneling gaps. It has been demonstrated that high spatial resolution can be achieved in STM experiments with the boron‐doped single crystal diamond tips [93]. **Figure 7(a)**–**(d)** displays a comparison of the spatial resolution achieved with the [111]‐oriented diamond and [001]‐ oriented tungsten probes. Note that image in **Figure 7(b)** was measured with the W[001] tip used for the high-resolution experiments on graphite (**Figure 5**) and surfaces with nontrivial atomic and electronic structures [94, 95]. The images taken with the diamond and W[001] probes [**Figure 7(a)** and **(b)**] reveal two sublattices corresponding to nonequivalent α and β atoms in the honeycomb lattice. However, the hollow sites are substantially deeper and individual surface atoms are better resolved in the image measured with the diamond probe, as the cross‐sections in **Figure 7(c)** and **(d)** illustrate. DFT calculations [93] showed that the DOS at *E*F is larger by approximately 25% for β atoms [**Figure 7(f)**] and the difference decreases for the DOS integrated over a wider range of the electron energies. That explains the two slightly nonequivalent sublattices in **Figure 7(a)** and **(b)**. The smaller height difference between the features corresponding to α and β atoms for the image measured at *U* = –0.4 V [**Fig‐ ure 7(b)**] is in agreement with the decreasing difference in the DOS at larger energies.

**Figure 7(g)** and **(h)** shows the experimental STM image resolved with the diamond tip and the calculated charge density map corresponding to the surface electron states near *E*F, respec‐ tively. Both images demonstrate different contrast on α and β atoms. The DOS shown in **Figure 7(e, f)** and the charge density map presented in **Figure 7(g)** were calculated at a tip‐ sample distance of 4.5 Å where the atomic and electronic structures of the tip and surface atoms are not substantially modified by the interaction [93]. The excellent agreement between experimental and theoretical images suggests that the highest resolution was achieved with the diamond probe at tunneling gaps of 3.5–4.5 Å where the tip did not strongly interact with the surface. This result shows the advantages of the oriented single crystal diamond probes: their structure is stable and well defined while high lateral and vertical resolution can be achieved at larger tunneling gaps comparing to typical tip‐sample distances used in experi‐ ments with transition metal tips.

**Figure 7.** (a, b) 18 × 9 Å2 atomically resolved STM images of graphite (0001) measured with a diamond probe at *U* = –50 mV and *I* = 0.1 nA (a) and a W[001] probe at *U* = –0.4 V and *I* = 0.18 nA (b). (c, d) Cross sections 1–2 (c) and 3–4 (d) of the images in panels (a) and (b), respectively. (e, f) Total DOS associated with the α and β atoms of a graphite (0001) surface. (g, h) Comparison of the 9 × 9 Å2 calculated electron density distribution map in the energy range from *E*F – 0.2 eV to *E*<sup>F</sup> (g) and the experimental STM image measured with the diamond probe at *U* = –50 mV and *I* = 0.8 nA (h). Both images show nonequivalence of the α and β atoms in accordance with the DOS shown in panels (e) and (f). Repro‐ duced from Reference [93] with permission of IOP.

Calculations of the partial DOS associated with the electron orbitals of the diamond probe and graphite surface atoms at different tunneling gaps (**Figure 8**) show that electronic structure of the interacting atoms is modified at tip‐sample distances below 3.0 Å. The overlap of the tip and surface atomic orbitals leads to decrease in the PDOS associated with the *pz*‐orbital of the graphite surface atoms when the diamond tip atom is positioned directly above the surface atom [**Figure 7**(**a**, right panel)]. The suppression of the surface atom's *pz*‐orbital at small tunneling gaps (*d* < 2.5 Å) is in qualitative agreement with the results of the theoretical calculations for the Si(111)7×7 surface interacting with the W[001] tip shown in **Figure 2** [78]. If the diamond tip atom is positioned above the hollow site (left panels in **Figure 8**), the overlap of the front tip atom and the surface orbitals does not take place even at small distances. Therefore, the change of the PDOS of the tip and surface atoms is minor even at the tunneling gaps of 1.5 Å. According to the calculations [93], the electronic structure of the diamond probe is defined by a mixture of the carbon *s* and *p*‐states with domination of the *px‐* and *py*‐orbitals, responsible for the STM imaging (**Figure 7**).

the features corresponding to α and β atoms for the image measured at *U* = –0.4 V [**Fig‐**

**Figure 7(g)** and **(h)** shows the experimental STM image resolved with the diamond tip and the calculated charge density map corresponding to the surface electron states near *E*F, respec‐ tively. Both images demonstrate different contrast on α and β atoms. The DOS shown in **Figure 7(e, f)** and the charge density map presented in **Figure 7(g)** were calculated at a tip‐ sample distance of 4.5 Å where the atomic and electronic structures of the tip and surface atoms are not substantially modified by the interaction [93]. The excellent agreement between experimental and theoretical images suggests that the highest resolution was achieved with the diamond probe at tunneling gaps of 3.5–4.5 Å where the tip did not strongly interact with the surface. This result shows the advantages of the oriented single crystal diamond probes: their structure is stable and well defined while high lateral and vertical resolution can be achieved at larger tunneling gaps comparing to typical tip‐sample distances used in experi‐

**Figure 7.** (a, b) 18 × 9 Å2 atomically resolved STM images of graphite (0001) measured with a diamond probe at *U* = –50 mV and *I* = 0.1 nA (a) and a W[001] probe at *U* = –0.4 V and *I* = 0.18 nA (b). (c, d) Cross sections 1–2 (c) and 3–4 (d) of the images in panels (a) and (b), respectively. (e, f) Total DOS associated with the α and β atoms of a graphite (0001) surface. (g, h) Comparison of the 9 × 9 Å2 calculated electron density distribution map in the energy range from *E*F – 0.2 eV to *E*<sup>F</sup> (g) and the experimental STM image measured with the diamond probe at *U* = –50 mV and *I* = 0.8 nA (h). Both images show nonequivalence of the α and β atoms in accordance with the DOS shown in panels (e) and (f). Repro‐

Calculations of the partial DOS associated with the electron orbitals of the diamond probe and graphite surface atoms at different tunneling gaps (**Figure 8**) show that electronic structure of the interacting atoms is modified at tip‐sample distances below 3.0 Å. The overlap of the tip

**ure 7(b)**] is in agreement with the decreasing difference in the DOS at larger energies.

ments with transition metal tips.

342 Microscopy and Analysis

duced from Reference [93] with permission of IOP.

**Figure 8.** Partial DOS of the β atom of a graphite (0001) surface closest to the diamond tip (a) and the tip apex atom (b) at different tunneling gaps and lateral positions of the tip. Reproduced from Reference [93] with permission of IOP.

#### **3.6. STM experiments with functionalized tips terminated by a light element atom**

The experiments with the diamond tip (**Figure 7**) show that an enhancement of the spatial resolution can be achieved using conductive tips having molecules or light element atom at the apex. During the recent years, a number of high‐resolution STM studies performed with the conductive tips functionalized by different light elements have been reported [57–59, 89, 96–103]. These studies generally demonstrated an enhanced contrast in the STM experiments with the light element–terminated probes, especially, at small tip‐surface distances. For example, the lateral resolution at the level of intramolecular chemical bonds has been reached in STM experiments using molecule‐terminated probes [58, 59]. In the recent theoretical work [101], the high resolution in experiments with the functionalized probes was explained by significant apex atom relaxations toward the local minima of the interaction potential at small distances. However, the tip functionalization procedures are generally not capable of produc‐ ing stable apexes with well‐defined structure. These tips could produce substantial noise [58] and asymmetric subatomic features [89] at small tunneling gaps. Besides, STM images measured with the molecule‐terminated tips can strongly depend on the precise orientation of the molecule at the apex and the bias voltage [102, 103] that can complicate the interpretation of the atomically resolved STM data.

#### **3.7. STM imaging of the random bond length distortions in graphene using a W[111] tip**

**Figure 9** demonstrates the picometer lateral resolution achieved in high‐resolution STM studies of graphene/SiC(001) system [7] using a [111]‐oriented single crystalline tungsten tip. Theoretical calculations shown in **Figure 1** suggest that the front atom of the W[111] tip should possess symmetric charge distribution along the tip axis even at small tip‐sample distances. Although the W[111] tips were found to be the least stable from three possible low‐index crystallographic orientations [75], they can produce higher spatial resolution in STM experi‐ ments without possible tip structure effects which can be observed with the W[001] tips (**Figure 5**). Indeed, although high spatial resolution was achieved on the quasi‐freestanding graphene grown on SiC(001) with the stable W[001] tips [24], the best resolution was obtained with the W[111] tip [7]. **Figure 9(a)** demonstrates the rippled morphology of graphene on SiC(001) with the lateral and vertical dimensions of the ripples of about 30–50 and 1 Å, respectively. **Figure 9(b)** and **(c)** demonstrates random picoscale distortions of the carbon bond lengths, which are very close to the values calculated for the freestanding graphene monolayer [104]. The contrast in **Figure 9(b)** and **(c)** was adjusted to enhance the picometer scale bond length distribution. The picometer lateral resolution obtained in the RT STM experiments (**Figure 9**) corresponds to the best standards of the SPM and can be compared with the resolution achieved in the recent AFM experiments [96].

**Figure 9.** STM images of the graphene synthesized on SiC(001) demonstrating atomic scale rippling (a) and random picoscale distortions of the carbon bond lengths in the graphene lattice (b, c). A distorted hexagon is overlaid on the image on panel (c); the size of the hexagon sides is indicated for clarity. The STM images were measured at *U* = 22 mV and *I* = 70 pA (a); *U* = 22 mV and *I*=65 pA (b, c).

## **4. Conclusions**

significant apex atom relaxations toward the local minima of the interaction potential at small distances. However, the tip functionalization procedures are generally not capable of produc‐ ing stable apexes with well‐defined structure. These tips could produce substantial noise [58] and asymmetric subatomic features [89] at small tunneling gaps. Besides, STM images measured with the molecule‐terminated tips can strongly depend on the precise orientation of the molecule at the apex and the bias voltage [102, 103] that can complicate the interpretation

**3.7. STM imaging of the random bond length distortions in graphene using a W[111] tip**

**Figure 9** demonstrates the picometer lateral resolution achieved in high‐resolution STM studies of graphene/SiC(001) system [7] using a [111]‐oriented single crystalline tungsten tip. Theoretical calculations shown in **Figure 1** suggest that the front atom of the W[111] tip should possess symmetric charge distribution along the tip axis even at small tip‐sample distances. Although the W[111] tips were found to be the least stable from three possible low‐index crystallographic orientations [75], they can produce higher spatial resolution in STM experi‐ ments without possible tip structure effects which can be observed with the W[001] tips (**Figure 5**). Indeed, although high spatial resolution was achieved on the quasi‐freestanding graphene grown on SiC(001) with the stable W[001] tips [24], the best resolution was obtained with the W[111] tip [7]. **Figure 9(a)** demonstrates the rippled morphology of graphene on SiC(001) with the lateral and vertical dimensions of the ripples of about 30–50 and 1 Å, respectively. **Figure 9(b)** and **(c)** demonstrates random picoscale distortions of the carbon bond lengths, which are very close to the values calculated for the freestanding graphene monolayer [104]. The contrast in **Figure 9(b)** and **(c)** was adjusted to enhance the picometer scale bond length distribution. The picometer lateral resolution obtained in the RT STM experiments (**Figure 9**) corresponds to the best standards of the SPM and can be compared with the

**Figure 9.** STM images of the graphene synthesized on SiC(001) demonstrating atomic scale rippling (a) and random picoscale distortions of the carbon bond lengths in the graphene lattice (b, c). A distorted hexagon is overlaid on the image on panel (c); the size of the hexagon sides is indicated for clarity. The STM images were measured at *U* = 22 mV

of the atomically resolved STM data.

344 Microscopy and Analysis

resolution achieved in the recent AFM experiments [96].

and *I* = 70 pA (a); *U* = 22 mV and *I*=65 pA (b, c).

The spatial resolution on the level of individual electron orbitals corresponds to the ultimate resolution of the SPM. During the last decade, a number of SPM studies demonstrating selective visualization of individual electron orbitals and subatomic contrast have been published. The experimental and theoretical works conducted during the recent years demonstrate that selective imaging of the surface electron orbitals can only be achieved at finely adjusted bias voltages and tip‐sample distances. Therefore, for the development of the electron orbital imaging capability, the stability of the tip‐sample separation should be maintained at the level of 1 pm or below. Distance‐dependent STM imaging with electron orbital resolution can lead to further improvement of the lateral resolution down to the picometer scale and development of the chemical‐selective imaging of complex multicompo‐ nent surfaces.

## **Acknowledgements**

This work was supported by the Russian Academy of Sciences, Russian Foundation for Basic Research (grantNos. 14‐02‐01234 and 14‐02‐00949), and Marie Curie International Incoming Fellowship project within the Seventh European Community Framework Programme. The author is grateful to S. N. Molotkov, S. I. Bozhko, S. S. Nazin, V. N. Semenov, A. M. Ionov, V. Yu. Aristov, M. G. Lazarev, N. N. Orlova, A. N. Myagkov, K. N. Eltsov, A. N. Klimov, V. M. Shevlyuga, I. V. Shvets, S. A. Krasnikov, O. Lübben, and B. E. Murphy for help and fruitful discussions.

## **Author details**

Alexander N. Chaika

Address all correspondence to: chaika@issp.ac.ru

Institute of Solid State Physics, Russian Academy of Sciences, Chernogolovka, Russia

## **References**


[18] Schmid M, Stadler H, Varga P. Direct observation of surface chemical order by scanning tunneling microscopy. Phys. Rev. Lett. 1993;70:1441–1444. doi:10.1103/PhysRevLett. 70.1441.

[3] Binnig G, Quate CF, Gerber C. Atomic force microscope. Phys. Rev. Lett. 1986;56:930–

[4] Binnig G, Rohrer H, Gerber C, Weibel E. Surface studies by scanning tunneling microscopy. Phys. Rev. Lett. 1982;49: 57–61. doi:10.1103/PhysRevLett.49.57.

[5] Binnig G, Rohrer H, Gerber C, Weibel E. (7×7) reconstruction on Si(111) resolved in real

[6] Binnig G, Rohrer H, Gerber C, Weibel E. (111) facets as the origin of reconstructed Au(110) surfaces, Surf. Sci. 1983;131:L379–L384. doi:10.1016/0039‐6028(83)90112‐7. [7] Chaika AN, Molodtsova OV, Zakharov AA, Marchenko D, Sanchez‐Barriga J, Vary‐ khalov A, et al. Continuous wafer‐scale graphene on cubic‐SiC(001). Nano Res.

[8] Eigler DM, Schweizer EK. Positioning single atoms with a scanning tunneling micro‐

[9] Crommie MF, Lutz CP, Eigler DM. Confinement of electrons to quantum corrals on a

[10] Stroscio JA, Celotta RJ. Controlling the dynamics of a single atom in lateral atom

[11] Walsh MA, Hersam MC. Atomic‐scale templates patterned by ultrahigh vacuum scanning tunnelling microscopy on silicon. Annu. Rev. Phys. Chem. 2009;60:193–216.

[12] Khajetoorians AA, Wiebe J, Chilian B, Wiesendanger R. Realizing all‐spin–based logic operations atom by atom. Science 2012;332:1062–1064. doi:10.1126/science.

[13] Khajetoorians AA, Wiebe J, Chilian B, Lounis S, Blügel S, Wiesendanger R. Atom‐by‐ atom engineering and magnetometry of tailored nanomagnets. Nat. Phys. 2012;8:497–

[14] Krasnikov SA, Lübben O, Murphy BE, Bozhko SI, Chaika AN, Sergeeva NN, et al. Writing with atoms: oxygen adatoms on the MoO2/Mo(110) surface. Nano Res.

[15] Gawronski H, Mehlhorn M, Morgenstern K. Imaging phonon excitation with atomic

[16] Stipe BC, Rezaei MA, Ho W. Single‐molecule vibrational spectroscopy and microscopy.

[17] Feenstra RM, Stroscio JA, Tersoff J, Fein AP. Atom‐selective imaging of the GaAs(110) surface. Phys. Rev. Lett. 1987;58:1192–1195. doi:10.1103/PhysRevLett.58.1192.

resolution. Science 2008;319:930–933. doi:10.1126/science.1152473.

Science 1998;280:1732–1735. doi:10.1126/science.280.5370.1732.

metal surface. Science 1993;262:218–220. doi:10.1126/science.262.5131.218.

manipulation. Science 2004;306:242–247. doi:10.1126/science.1102370.

space. Phys. Rev. Lett. 1983;50:120–123. doi:10.1103/PhysRevLett.50.120.

933. doi:10.1103/PhysRevLett.56.930.

346 Microscopy and Analysis

2013;6:562–570. doi:10.1007/s12274‐013‐0331‐9.

doi:10.1146/annurev.physchem.040808.090314.

2013;6:929–937. doi:10.1007/s12274‐013‐0370‐2.

503. doi:10.1038/nphys2299.

scope. Nature 1990;344:524–526. doi:10.1038/344524a0.


[44] Mönig H, Todorovic M, Baykara MZ, Schwendemann TC, Rodrigo L, Altman EI, et al. Understanding scanning tunneling microscopy contrast mechanisms on metal oxides: a case study. ACS Nano 2013;7:10233–10244. doi:10.1021/nn4045358.

[31] Giessibl FJ, Hembacher S, Bielefeldt H, Mannhart J. Subatomic features on the silicon (111)‐(7×7) surface observed by atomic force microscopy. Science 2000;289:422–425. doi:

[32] Giessibl FJ, Bielefeldt H, Hembacher S, Mannhart J. Imaging of atomic orbitals with the atomic force microscope—experiments and simulations. Ann. Phys. 2001;10:887–910.

[33] Hembacher S, Giessibl FJ, Mannhart J. Force microscopy with light‐atom probes.

[34] Fowler RH, Nordheim L. Electron Emission in Intense Electric Fields. Proc. R. Soc.

[35] Bryant A, Smith DPE, Binnig G, Harrison WA, Quate CF. Anomalous distance de‐ pendence in scanning tunneling microscopy. Appl. Phys. Lett. 1986;49:936–938. doi:

[36] Jurczyszyn L, Mingo N, Flores F. Influence of the atomic and electronic structure of the tip on STM images and STS spectra. Surf. Sci. 1998;402–404:459–463. doi:10.1016/S0039‐

[37] Hofer WA, Garcia‐Lekue A, Brune H. The role of surface elasticity in giant corrugations observed by scanning tunneling microscopes. Chem. Phys. Lett. 2004;397:354–359. doi:

[38] Bode M, Pascal R, Wiesendanger R. Distance‐dependent STM‐study of the W(110)/C‐

[39] Wiesendanger R, Bode M, Pascal R, Allers W, Schwarz UD. Issues of atomic‐resolution structure and chemical analysis by scanning probe microscopy and spectroscopy. J.

[40] Klijn J, Sacharow L, Meyer C, Blugel S, Morgenstern M, Wiesendanger R. STM measurements on the InAs(110) surface directly compared with surface electronic structure calculations. Phys. Rev. B 2003;68:205327. doi:10.1103/PhysRevB.68.205327.

[41] Calleja F, Arnau A, Hinarejos JJ, Vazquez de Parga AL, Hofer WA, Echenique PM, Miranda R. Contrast reversal and shape changes of atomic adsorbates measured with scanning tunneling microscopy. Phys. Rev. Lett. 2004;92:206101. doi:10.1103/PhysRev‐

[42] Blanco JM, González C, Jelínek P, Ortega J, Flores F, Pérez R, et al. Origin of contrast in STM images of oxygen on Pd(111) and its dependence on tip structure and tunneling

[43] Woolcot T, Teobaldi G, Pang CL, Beglitis NS, Fisher AJ, Hofer WA, Thornton G. Scanning tunneling microscopy contrast mechanisms for TiO2. Phys. Rev. Lett.

parameters. Phys. Rev. B 2005;71:113402. doi:10.1103/PhysRevB.71.113402.

2012;109:156105. doi:10.1103/PhysRevLett.109.156105.

R(15×3) surface. Z. Phys. B 1996;101:103–107. doi:10.1007/s002570050187.

Vac. Sci. Technol. A 1996;14:1161–1167. doi:10.1116/1.580259.

doi:10.1002/1521‐3889(200111)10:11/12<887::AID‐ANDP887=3.0.CO;2‐B.

Science 2004;305:380–383. doi:10.1126/science.1099730.

Lond. A 1928;119:173–181. doi:10.1098/rspa.1928.0091.

10.1126/science.289.5478.422.

348 Microscopy and Analysis

10.1063/1.97489.

6028(97)00971‐0.

Lett.92.206101.

10.1016/j.cplett.2004.08.110.


[71] Hagelaar JHA, Flipse CFJ, Cerda JI. Modeling realistic tip structures: scanning tunnel‐ ing microscopy of NO adsorption on Rh(111). Phys. Rev. B 2008;78:161405. doi:10.1103/ PhysRevB.78.161405.

[58] Weiss C, Wagner C, Kleimann C, Rohlfing M, Tautz FS, Temirov R. Imaging Pauli repulsion in scanning tunneling microscopy. Phys. Rev. Lett. 2010;105:086103. doi:

[59] Gross L, Moll N, Mohn F, Curioni A, Meyer G, Hanke F, Persson M. High‐resolution molecular orbital imaging using a *p*‐wave STM tip. Phys. Rev. Lett. 2011;107:086101.

[60] Gross L, Mohn F, Moll N, Schuler B, Criado A, Guitian E, Pena D, Gourdon A, Meyer G. Bond‐order discrimination by atomic force microscopy. Science 2012;337:1326–1329.

[61] Scheer E, Agrait N, Cuevas JC, Yeyati AL, Ludoph B, Martin‐Rodero A, et al. The signature of chemical valence in the electrical conduction through a single‐atom

[62] Polok M, Fedorov DV, Bagrets A, Zahn P, Mertig I. Evaluation of conduction eigen‐ channels of an adatom probed by an STM tip. Phys. Rev. B 2011;83:245426. doi:10.1103/

[63] Choi H, Longo RC, Huang M, Randall JN, Wallace RM, Cho K. A density‐functional theory study of tip electronic structures in scanning tunneling microscopy. Nanotech‐

[64] Suominen I, Nieminen J, Markiewicz RS, Bansil A. Effect of orbital symmetry of the tip on scanning tunneling spectra of Bi2Sr2CaCu2O8+δ. Phys. Rev. B 2011;84:014528. doi:

[65] Palotas K, Mandi G, Hofer WA. Three‐dimensional Wentzel‐Kramers‐Brillouin approach for the simulation of scanning tunneling microscopy and spectroscopy. Front.

[66] Palotas K, Mandi G, Szunyogh L. Orbital‐dependent electron tunneling within the atom superposition approach: theory and application to W(110). Phys. Rev. B 2012;86:235415.

[67] Mandi G, Palotas K. STM contrast inversion of the Fe(110) surface. Appl. Surf. Sci.

[68] da Silva Neto EH, Aynajian P, Baumbach RE, Bauer ED, Mydosh J, Ono S, Yazdani A. Detection of electronic nematicity using scanning tunneling microscopy. Phys. Rev. B

[69] Jurczyszyn L, Stankiewicz B. Interorbital interference in STM tip during electron tunneling in tip–sample system: influence on STM images. Prog. Surf. Sci. 2003;74:185–

[70] Jurczyszyn L, Stankiewicz B. The role of interorbital interference in the formation of STS spectra. Appl. Surf. Sci. 2005;242:70–81. doi:10.1016/j.apsusc.2004.07.066.

10.1103/PhysRevLett.105.086103.

350 Microscopy and Analysis

doi:10.1103/PhysRevLett.107.086101.

contact. Nature 1998;394:154–157. doi:10.1038/28112.

nology 2013;24:105201. doi:10.1088/0957‐4484/24/10/105201.

Phys. 2014;9:711–747. doi:10.1007/s11467‐013‐0354‐4.

2014;304:65–72. doi:0.1016/j.apsusc.2014.02.143.

2013;87:161117. doi:10.1103/PhysRevB.87.161117.

200. doi:10.1016/j.progsurf.2003.08.014.

doi:10.1126/science.1225621.

PhysRevB.83.245426.

10.1103/PhysRevB.84.014528.

doi:10.1103/PhysRevB.86.235415


[96] Gross L, Mohn F, Moll N, Schuler B, Criado A, Guitian E, et al. Bond‐order discrimi‐ nation by atomic force microscopy. Science 2012;337:1326–1329. doi:10.1126/science. 1225621.

[84] Huang M, Cuma M, Liu F. Seeing the atomic orbital: first‐principles study of the effect of tip termination on atomic force microscopy. Phys. Rev. Lett. 2003;90:256101. doi:

[85] Chen CJ. Possibility of imaging lateral profiles of individual tetrahedral hybrid orbitals in real space. Nanotechnology 2006;17: S195–S200. doi:10.1088/0957‐4484/17/7/S16. [86] Campbellova A, Ondracek M, Pou P, Perez R, Klapetek P, Jelınek P. "Sub‐atomic" resolution of non‐contact atomic force microscope images induced by a heterogeneous tip structure: a density functional theory study. Nanotechnology 2011;22:295710. doi:

[87] Sweetman A, Rahe P, Moriarty P. Unique determination of "subatomic" contrast by imaging covalent backbonding. Nano Lett. 2014;14:2265–2270. doi:10.1021/nl4041803.

[88] Wang YL, Gao HJ, Guo HM, Liu HW, Batyrev IG, McMahon WE, Zhang SB. Tip size effect on the appearance of a STM image for complex surfaces: theory versus experi‐ ment for Si(111)‐(7×7). Phys. Rev. B 2004;70:073312. doi:10.1103/PhysRevB.70.073312.

[89] Li Z, Schouteden K, Iancu V, Janssens E, Lievens P, van Haesendonck C, Cerdá JI. Chemicaly modified STM tips for atomic‐resolution imaging of ultrathin NaCl films.

[90] Emmrich M, Huber F, Pielmeier F, Welker J, Hofmann T, Schneiderbauer M, et al. Subatomic resolution force microscopy reveals internal structure and adsorption sites

[91] Hofmann T, Pielmeier F, Giessibl FJ. Chemical and crystallographic characterization of the tip apex in scanning probe microscopy. Phys. Rev. Lett. 2014;112:066101. doi:

[92] Hofmann T, Pielmeier F, Giessibl FJ. Erratum: Chemical and crystallographic charac‐ terization of the tip apex in scanning probe microscopy [Phys. Rev. Lett. 112, 066101

[93] Grushko V, Lubben O, Chaika AN, Novikov N, Mitskevich E, Chepugov A, et al. Atomically resolved STM imaging with a diamond tip: simulation and experiment.

[94] Chaika AN, Semenov VN, Glebovskiy VG, Bozhko SI. Scanning tunneling microscopy with single crystal W[001] tips: high resolution studies of Si(557)5×5 surface. Appl.

[95] Yashina LV, Püttner R, Volykhov AA, Stojanov P, Riley J, Vassiliev SY, et al. Atomic geometry and electron structure of the GaTe(1 0 ‐2) surface. Phys. Rev. B 2012;85:075409.

(2014)]. Phys. Rev. Lett. 2015;115:109901. doi:10.1103/PhysRevLett.115.109901.

Nanotechnology 2014;25:025706. doi:10.1088/09574484/25/2/025706

Phys. Lett. 2009;95:173107. doi:10.1063/1.3254240.

of small iron clusters. Science 2015;348:308–311. doi:10.1126/science.aaa5329.

Nano Res. 2015;8:2223–2230. doi:10.1007/s12274‐015‐0733‐y.

10.1103/PhysRevLett.90.256101.

352 Microscopy and Analysis

10.1088/0957‐4484/22/29/295710.

10.1103/PhysRevLett.112.066101.

doi:10.1103/PhysRevB.85.075409.

