Atomic imaging of two-dimensional materials

Two-dimensional (2D) systems such as graphene have been among the most studied materials in the past 10 years. Stable 2D materials are highly anisotropic crystals with strong covalent in-plane bonding and weak van der Waals interlayer bonding. The unusual properties of graphene, namely, its ballistic electron conductance and its high tensile strength, have made it one of the most studied materials of our time.^{Reference Novoselov, Falko, Colombo, Gellert, Schwab and Kim1,Reference Geim2} However, the absence of a bandgap in graphene motivates the search for alternative 2D materials. Hexagonal boron nitride (h-BN) and the chalcogenides of molybdenum or tungsten are the most prominent examples that have been synthesized and characterized in detail.^{Reference Han, Wu, Zhu, Watanabe and Taniguchi3–Reference Zeng, Yin, Huang, Li, He, Lu, Boey and Zhang5} A common feature of many 2D materials is a hexagonal lattice in which the atoms are arranged in a honeycomb structure.

Transmission electron microscopy (TEM) has turned out to be ideally suited for studying 2D materials. A large number of in situ experiments have been carried out in which the beam served not only for imaging but also for irradiation. The creation of point defects has been monitored in real time and with atomic resolution. The particular appeal of TEM studies of 2D materials is that the reconstruction of the lattice occurs in the 2D atomic plane and can, in many cases, be seen without any projection artifacts. Hence, a detailed picture of the atomic arrangement of defective materials becomes visible. Striking TEM images of point defects in graphene, recorded at a lateral resolution smaller than the bond length of carbon, leave almost no room for uncertain interpretation. Novel mechanisms of lattice reconstruction have been discovered in in situ irradiation experiments, such as transformations between different polygons.^{Reference Banhart, Kotakoski and Krasheninnikov6–Reference Robertson and Warner8}

In addition to a detailed understanding of radiation-induced defects and their influence on the properties of 2D sheets, the design of new structures has become feasible. Irradiation-induced structural changes can be achieved and characterized in real time and at the full lateral resolution of the instrument. On one hand, modern TEM instruments achieve an imaging resolution of better than 1 Å, which is less than the distance between atoms in solids. This provides the possibility of imaging individual atoms and their migration under electron irradiation. On the other hand, electron beams can be focused onto spots of less than 1 Å, which allows use of the beam as tweezers to manipulate materials at the single atom level.^{Reference Robertson, Allen, Wu, He, Olivier, Neethling, Kirkland and Warner9–Reference Rodríguez-Manzo, Cretu and Banhart11}

Detailed in situ electron irradiation experiments of 2D materials have been carried out,^{Reference Jin, Lin, Suenaga and Iijima12–Reference Meyer, Eder, Kurasch, Skakalova, Kotakoski, Park, Roth, Chuvilin, Eyhusen, Benner, Krasheninnikov and Kaiser18} and the transmission electron microscope has become a useful tool for studying the structural stability of materials and creating new nanostructures.^{Reference Liu, Xu, Wu, Zhang, Yu, Qiu, Hong, Jin, Li, Wang, Sun and Guo15,Reference Chuvilin, Meyer, Algara-Siller and Kaiser19–Reference Neng, Sun, Hu, Zhu, Lin, Tao, Bi, Jun and Dong27} Not only the formation but also the annealing of defects can be controlled in such irradiation experiments if high-temperature specimen stages are used. The reconstruction of vacancies in carbon nanostructures, which has turned out to be the basis for many surprising phenomena, requires temperatures above 200–300°C.^{Reference Banhart28,Reference Chen, Shi, Cai, Xu, Sun, Wu and Yu29} Although detailed pictures of radiation effects and defects in graphene have been obtained, the information available about defects in other 2D materials is still rather limited. Therefore, many inorganic 2D materials are still awaiting investigation, and the field is certainly ripe for new discoveries and surprises. In this article, the atomic structural evolution of 2D materials such as graphene, h-BN, and metal dichalcogenides under in situ electron irradiation and the mechanisms of defect reconstruction are reviewed.

Defects in graphene

Graphene, as the most prominent 2D material, was the first in which irradiation effects were studied. The rise of aberration-corrected TEM instruments allowed the imaging of each carbon atom in a graphene sheet and, hence, the achievement of a precise and unambiguous picture of the atomic arrangement in defects such as single or multiple vacancies.

Even if no atoms are removed, structural defects can be created by the electron beam. A Stone–Wales (SW) defect is an example that is formed by an in-plane 90° rotation of a C–C bond (Figure 1a–d). Four adjacent hexagons are transformed into two pentagons and two heptagons. The kinetic barrier for such a bond rotation is 9–10 eV, whereas the barrier of the reverse transformation is only 5 eV.^{Reference Li, Reich and Robertson30} Both barriers are less than the calculated displacement threshold energy of 18–22 eV,^{Reference Krasheninnikov and Banhart31,Reference Russo and Golovchenko32} such that both transformations can occur below the displacement threshold.^{Reference Meyer, Kisielowski, Erni, Rossell, Crommie and Zettl33} A remarkable and stable defect structure in graphene can be formed from six SW rotations to yield a closed-loop structure,^{Reference Robertson, Allen, Wu, He, Olivier, Neethling, Kirkland and Warner9} shown in Figure 1e–h. An aberration-corrected TEM (AC-TEM) image of this defect is shown in Figure 1h. No atoms are lost, with only a core of seven hexagons rotated by 30° to form a miniature grain boundary. The detailed mechanisms by which this structure is produced are not yet fully clear, but it is likely that the structure results from the relaxation of a more complex defect into this stable configuration. The schematic presented in Figure 1e–h is for illustrative purposes to help in understanding the final closed-loop structure. It has also been reported that bond rotations play an important role in other defect reconstructions, as discussed in the next subsection.

Figure 1.

Stone–Wales (SW) rotation and single-vacancy defects in graphene: (a–c) Atomic models showing the formation of a SW defect by an in-plane 90° rotation of a C–C bond. Atoms indicated in yellow show the rotated bond. (d) Aberration-corrected transmission electron microscopy (AC-TEM) image of a SW defect. (e–g) Atomic models showing the formation of a miniature grain boundary by a rotation of a seven-hexagon core. Yellow atoms indicate the rotated region. (h) AC-TEM image of a closed-loop defect. (i–k) Atomic models showing the formation of a reconstructed single vacancy. The yellow atom is the one removed, and the yellow bond indicates the reconstructed bond. (l) AC-TEM image of a reconstructed single-vacancy defect. For (d) and (l), red lines indicate heptagons, and yellow lines indicate pentagons. (a–d) Reproduced with permission from Reference 8. © 2013 Royal Society of Chemistry. (e–h) Reproduced with permission from Reference 9. © 2012 Nature Publishing Group. (i–l) Reproduced with permission from Reference 34. © 2013 American Chemical Society.

Traditionally, vacancy formation is thought to occur through direct sputtering of atoms from the lattice by high-energy electrons. However, in graphene, this process is much more complicated than the simple picture of knock-on damage. It was long thought that using electrons accelerated at 80 kV would not damage graphene because the maximum energy transferred to a carbon atom is below the theoretical displacement energy. However, recent work showed that vacancies are formed using an accelerating voltage of 80 kV.^{Reference Meyer, Eder, Kurasch, Skakalova, Kotakoski, Park, Roth, Chuvilin, Eyhusen, Benner, Krasheninnikov and Kaiser18} Surprisingly, the sputtering cross-section at 80 kV also depends on the beam current density.^{Reference Meyer, Eder, Kurasch, Skakalova, Kotakoski, Park, Roth, Chuvilin, Eyhusen, Benner, Krasheninnikov and Kaiser18} The origin of defect creation in graphene at 80 kV is likely due to the high mobility of adatoms on the surface of graphene, allowing them to interact with graphene within the intense electron beam irradiation, thereby disrupting the regular sp ² hexagonal network. Any perturbation to the pristine structure, such as SW rotations or self-interstitials, can generate bonds that are weaker and therefore atoms that can be sputtered at a lower energy to form vacancies. Calculations of sputtering cross-sections in graphene are also complicated by the fact that vacancies can be created and then easily quenched through the trapping of mobile surface C adatoms, allowing them to disappear before being registered in the image. Therefore, the total number of missing atoms counted at any one time can vary and might not reflect the total number of atoms sputtered in a specific time interval. Further work is needed to clarify this issue in more detail. One thing is clear: The damage mechanisms in graphene at energies below 100 kV are more complicated than simple knock-on damage from one electron hitting one carbon atom to cause ejection and vacancy formation.^{Reference Meyer, Eder, Kurasch, Skakalova, Kotakoski, Park, Roth, Chuvilin, Eyhusen, Benner, Krasheninnikov and Kaiser18}

To understand complex defect structures in graphene, it is important to start with an overview of the basic defects that form the building blocks of defects in graphene. The simplest type is a single vacancy (SV). To minimize its energy, the SV undergoes a Jahn–Teller distortion, which leads to the saturation of two of the three dangling bonds around the missing atom (Figure1i–l). The Jahn–Teller distortion lowers the overall energy of the system by removing electronic degeneracies. This leads to the formation of a pentagon and a nonagon^{Reference Meyer, Kisielowski, Erni, Rossell, Crommie and Zettl33,Reference Robertson, Montanari, He, Allen, Wu, Harrison, Kirkland and Warner34} (the nonagon has a dangling bond). The energy barriers among the three Jahn–Teller reconstructions are small, and therefore, it is expected that under an electron beam at 80 kV, the structure should easily rotate among these three configurations. This has the effect of averaging out the reconstruction in a TEM image, giving a symmetric appearance to the defect. Recent in situ TEM work showed that, in many cases, a Jahn–Teller reconstructed SV locks into a specific orientation and remains stable for up to one minute.^{Reference Robertson, Montanari, He, Allen, Wu, Harrison, Kirkland and Warner34} It is proposed that this is due to either hydrogen attaching to the dangling bond of the C atom or one of the atoms within the vacancy being N. Further work is under way to shed more light on this stable SV configuration. SVs can frequently be observed at 80 kV and are stable for periods of up to several minutes, but are quenched by the trapping of mobile adsorbates. Most of the time, these are C atoms, but sometimes, they can be other elements, such as N, Fe, or Si.^{Reference Lee, Zhou, Pennycook, Idrobo and Pantelides35–Reference Robertson, Montanari, He, Kim, Allen, Wu, Olivier, Neethling, Harrison, Kirkland and Warner37}

Divacancies (DVs) are also commonly observed and are able to reconstruct the lattice such that no dangling bonds occur. They appear in several configurations of slightly different energy, as shown in Figure 2. They are created either when two migrating SVs coalesce or when an atom next to an existing SV is displaced. Because of the absence of open bonds, a DV is energetically more favorable than two SVs.^{Reference Krasheninnikov, Lehtinen, Foster and Nieminen38} Under 80 kV electron beam irradiation, the DV oscillates between the three configurations shown in Figure 2, but with a higher percentage of time spent in the DVs with +1 and +2 bond rotations (atomic models in Figure 2e–f).

Figure 2.

Divacancy (DV) defects in graphene: (a–c) Atomic models showing the removal of two C atoms (indicated with yellow) and bond reconstruction to form the 5–8–5 (pentagon–octagon–pentagon) DV (indicated with yellow). (d–f) Atomic models showing how (d) a 5–8–5 DV transforms into (e) a 555–777 DV by one bond rotation (yellow atoms in panel d) and then into (f) a 5555–6–7777 DV by a second bond rotation (yellow atoms in panel e). (g–i) Aberration-corrected transmission electron microscope images of DV structures shown in corresponding atomic models above each panel. Red lines indicate an octagon in (g) and heptagons in (h) and (i); yellow lines indicate pentagons. Reproduced with permission from Reference 9. © 2012 Nature Publishing Group.

Larger and more complex defect configurations can be formed by removing more atoms. Generally, vacancies with an even number of missing atoms, where all bonds are saturated, are energetically favored over an odd number of vacancies with at least one open bond.^{Reference Kotakoski, Krasheninnikov and Nordlund39} In the extreme case, graphene with a large density of vacancies can appear as a 2D carbon glass, which can be described as a coherent quasi-amorphous membrane composed of sp ²-hybridized carbon atoms.^{Reference Meyer, Eder, Kurasch, Skakalova, Kotakoski, Park, Roth, Chuvilin, Eyhusen, Benner, Krasheninnikov and Kaiser18,Reference Kotakoski, Krasheninnikov, Kaiser and Meyer40,Reference Eder, Kotakoski, Kaiser and Meyer41}

Recent work on tetravacancies in graphene (i.e., four atoms missing) revealed that the most common structures under electron beam irradiation involve either linear defects along the armchair direction or bond rotations.^{Reference Robertson, Lee, He, Yoon, Kirkland and Warner61}Figure 3a–e shows atomic models of how four atoms can be removed from the graphene lattice, considering only nearest-neighbor types without any bond rotations. Of these structures, only the structures in Figure 3a and 3d have been observed. Figure 3f–g shows AC-TEM images of the two most common tetravacancy structures in graphene: (f) the armchair defect (cf. Figure 3d) and (g) the tetravacancy with three bond rotations, the formation of which is schematically shown in Figure 3h–m. As in the case of the DV structure, continual electron beam irradiation at 80 kV causes successive bond rotations that enable transitions to a variety of structures.^{Reference Kotakoski, Krasheninnikov, Kaiser and Meyer40}

Figure 3.

Tetravacancy defects in graphene: (a–e) Atomic models showing how the removal of four carbon atoms in close proximity leads to tetravacancies without any bond rotations. Red atoms indicate those to be removed. Sequences run horizontally. (f–g) Aberration-corrected transmission electron microscope images of the most common tetravacancy structures in graphene: (f) the armchair defect (cf. Figure 3d) and (g) the tetravacancy with three bond rotations. (h–m) Atomic models showing how the tetravacancy structure in panel g can be constructed. Red indicates atoms to be removed; yellow indicates bond rotations. Reproduced with permission from Reference 61. © 2014 American Chemical Society.

One-dimensional line defects such as dislocations have also been observed in graphene under electron irradiation.^{Reference Warner, Margine, Mukai, Robertson, Giustino and Kirkland42–Reference Lehtinen, Kurasch, Krasheninnikov and Kaiser44} Because of graphene’s two-dimensional nature, only dislocations of the edge type appear.^{Reference Yazyev and Louie45} They can be constructed by the removal of a zigzag chain of carbon atoms.^{Reference Jeong, Ihm and Lee46} The role of the dislocation core is played by a pentagon–heptagon pair that appears at the end of the strip. Dislocations can migrate within the graphene sheet either by bond rotations along the glide plane (glide) or by the addition or removal of two carbon atoms (creep). The strain induced by the addition of dislocations causes out-of-plane distortions in the graphene sheet, buckling the structure. Electron beam irradiation can cause these ripples to propagate within graphene sheets.

Intense irradiation removes a large number of atoms from a small area, causing the formation of a hole with unsaturated bonds around its circumference.^{Reference Xu, Xie and Sun23–Reference Xu, Yin, Xie, He, Wang and Sun25,Reference Girit, Meyer, Erni, Rossell, Kisielowski, Yang, Park, Crommie, Cohen, Louie and Zettl47–Reference Warner, Rummeli, Ge, Gemming, Montanari, Harrison, Buchner and Briggs49} The edges of the holes can be zigzag, armchair, or reconstructed lines.^{Reference Girit, Meyer, Erni, Rossell, Kisielowski, Yang, Park, Crommie, Cohen, Louie and Zettl47,Reference Kim, Coh, Kisielowski, Crommie, Louie, Cohen and Zettl50–Reference Koskinen, Malola and Häkkinen53} Defective edges also appear^{Reference Liu, Suenaga, Harris and Iijima54} because of the sustained removal of atoms from perfect edges. Furthermore, upon exposure of two adjacent holes to electron irradiation, the bridge between the edges often becomes narrower, such that only a single chain of carbon atoms is left before the bridge breaks.^{Reference Chuvilin, Meyer, Algara-Siller and Kaiser19,Reference Jin, Lan, Peng, Suenaga and Iijima20,Reference Song, Schneider, Xu, Pandraud, Dekker and Zandbergen48,Reference Cretu, Botello-Mendez, Janowska, Pham-Huu, Charlier and Banhart55} Further types of defects in graphene have been described in specialized reviews.^{Reference Banhart, Kotakoski and Krasheninnikov6,Reference Robertson and Warner8}

Previous experiments on electron irradiation^{Reference Xu, Yin, Xie, He, Wang and Sun25,Reference Banhart28} also indicated that a large fraction of irradiation-induced damage in graphene can anneal in situ, especially at temperatures higher than 300°C, where SVs become mobile. Two important mechanisms govern defect annealing:^{Reference Krasheninnikov, Nordlund and Keinonen56} the saturation of dangling bonds by the formation of non-hexagonal rings and the migration of carbon interstitials or vacancies, followed by Frenkel pair recombination (i.e., a recombination of interstitial and vacancy).

As described earlier, the formation of pentagon–nonagon rings after atom removal is a simple reconstruction that saturates two of three dangling bonds.^{Reference Robertson, Montanari, He, Allen, Wu, Harrison, Kirkland and Warner34} The reconstruction causes strain in the surrounding lattice. At the same time, non-hexagonal rings induce local curvature in the graphene sheet, creating hillocks or trenches to accommodate strain.^{Reference Warner, Fan, Robertson, He, Yoon and Lee57,Reference Robertson, He, Kirkland and Warner58} In addition, the formation of pentagons causes positive curvature in graphene and leads to bowl-shaped structures or even closed spherical cages.^{Reference Chuvilin, Kaiser, Bichoutskaia, Besley and Khlobystov59}

It is worth noting that bond rotation can reconstruct most types of defects in graphene. The barrier of bond rotation is smaller than that for displacement, so bond rotation occurs at a much higher frequency under electron irradiation. The transformation among three different DV configurations^{Reference Robertson and Warner8,Reference Kotakoski, Mangler and Meyer60} is shown in Figure 2d–f. A bond rotation of the yellow bond in Figure 2d leads to transformation from a 5–8–5 (a DV configuration with 5–8–5 membered rings, as shown in Figure 2d) to a 555–777 configuration (another DV configuration with three 5-membered rings and 7-membered rings, as shown in Figure 2e). Another rotation in Figure 2e results in a transformation to a DV of the 5555–6–7777 type (a complex DV configuration, as shown in Figure 2f). Similarly, bond rotations can also govern the transformation between different configurations of a tetravacancy^{Reference Robertson, Lee, He, Yoon, Kirkland and Warner61} or induce the migration of vacancies^{Reference Kotakoski, Mangler and Meyer60} or grain boundaries.^{Reference Kurasch, Kotakoski, Lehtinen, Skákalová, Smet, Krill, Krasheninnikov and Kaiser62}

Most types of defects can act as potential doping sites when they trap other atoms.^{Reference Wang, Wang, Cheng, Li, Yao, Zhang, Dong, Wang, Schwingenschlögl, Yang and Zhang63,Reference Zhao, Deng, Bachmatiuk, Sandeep, Popov, Eckert and Rümmeli64} The effects of foreign atoms on the electronic properties of graphene depend on the charge transfer between the atom and the graphene, which is governed by the type of bonding.^{Reference Banhart65} Various bonding configurations (such as on top of a carbon atom or on top of the center of a hexagon) are possible. In situ observation of the migration of atoms as a function of temperature allows the determination of which types of bonding can occur.^{Reference Robertson, Montanari, He, Kim, Allen, Wu, Olivier, Neethling, Harrison, Kirkland and Warner37,Reference Gan, Sun and Banhart66–Reference Wang, Li, Cheng, Wang, Yao, Schwingenschlögl, Zhang and Yang72} Because of the locally increased reactivity of the π-electron system,^{Reference Cretu, Krasheninnikov, Rodriguez-Manzo, Sun, Banhart and Nieminen70,Reference Erni, Rossell, Nguyen, Blankenburg, Passerone, Hartel, Alem, Erickson, Gannett and Zettl73,Reference Warner, Liu, He, Robertson and Suenaga74} reconstructed defects serve as reactive sites even if all covalent bonds are saturated. At the same time, the strain field around reconstructed defects can lead to an attractive interaction over a scale of 1–2 nm between the defect and a foreign atom migrating on the surface.^{Reference Cretu, Krasheninnikov, Rodriguez-Manzo, Sun, Banhart and Nieminen70} As shown in Figure 4, a tungsten atom can jump back and forth between two defective sites at a distance of 0.5–1.4 nm. A biaxial strain of 1% in the graphene lattice increases the adsorption energy for transition-metal adatoms by 0.06–0.16 eV, depending on the metal species and the adsorption site.

Figure 4.

Jumping of a W atom on the surface of few-layer graphene at 480°C: (a–d) Repeated jumping between two trapping sites (denoted by white arrows and 1 and 2) at a distance of 1.5 nm. (e) The 555-777 defect naturally produces a strain field in its vicinity. The bonds are colored according to an increase (blue) or decrease (red) in the bond length (given in picometers). The letters indicate all nonequivalent position of W atoms in the configurations. The red lines outline the (10 × 10) simulation cell. (f) Adsorption energy of a W adatom near a 555–777 defect as a function of the separation between the adatom and the middle of the defect (circled atom in [e]) obtained from a 10 × 10 simulation cell (red) and a 7 × 7 supercell (blue). The letters indicate W adatom position in (e). The insets show the positions of the adatoms in the lowest energy configurations. The energy drop (approximately 2 eV) results from the stretched carbon–carbon bonds at the defects where a higher electron density is available for forming new bonds with the adatom. (g) Energy of a W adatom in various positions on a graphene sheet with two 555–777 defects. The potential well for the adatom can explain the attraction between the adatom and pinning centers, which is seen in transmission electron microscopy observations as a repeated jumping of the W atom between two trapping sites. Reproduced with permission from Reference 70. © 2010 American Physical Society.

Foreign atoms can also be incorporated into graphene as substitutional impurities. In this case, the impurity atom replaces one or two carbon atoms. However, most foreign atoms are located slightly off the layer because of the greater bond lengths between carbon and other atoms.^{Reference Krasheninnikov, Lehtinen, Foster, Pyykkö and Nieminen75} Substitutional dopants are expected to be very stable because of their strong covalent bonding with the graphene lattice; nevertheless, they can migrate under electron irradiation, especially when the foreign atoms are located at graphene edges.^{Reference Wang, Li, Cheng, Wang, Yao, Schwingenschlögl, Zhang and Yang72}Figure 5 shows the jumping of a Fe atom between two sites.^{Reference Robertson, Montanari, He, Kim, Allen, Wu, Olivier, Neethling, Harrison, Kirkland and Warner37}Figure 5a–b shows the Fe atom moving between different graphene sublattice sites, whereas Figure 5e–f shows the rotation of the dopant position through a change in the lattice site.^{Reference Robertson, Montanari, He, Kim, Allen, Wu, Olivier, Neethling, Harrison, Kirkland and Warner37}Figure 5i–l shows an initial vacancy in graphene that traps a mobile Fe atom to form a dopant. The Fe atom then switches configurations from bonding to four nearest carbon atoms to bonding to three, induced by electron beam irradiation at 80 kV.^{Reference Robertson, Montanari, He, Kim, Allen, Wu, Olivier, Neethling, Harrison, Kirkland and Warner37} Similar behavior has been observed for silicon atoms.^{Reference Zhou, Kapetanakis, Prange, Pantelides, Pennycook and Idrobo90}

Figure 5.

Migration of Fe atoms under electron irradiation: (a–b) Jumping of an Fe atom between adjacent single vacancy sites. Scale bar indicates 0.5 nm and (c–d) corresponding atomic models. Blue atom indicates the location of the Fe atom. (e–f) Changing of an Fe atom between two divacancy sites. Scale bar indicates 0.5 nm and (g–h) corresponding atomic models. Blue atom indicates the location of the Fe atom. Red circles show the location of the dopant atom in the previous frame. (i–l) Time series of images (10 s between images) showing an Fe atom hopping into a vacancy in graphene and switching its bonding configuration. Reproduced with permission from Reference 37. © 2013 American Chemical Society.

Defect formation in graphene has been demonstrated to have spatial control to within 10 nm using in situ electron beam irradiation within an AC-TEM. By controlling the beam current density and exposure time, it is possible to produce either small MVs and DVs or large complex vacancy structures.^{Reference Robertson, Allen, Wu, He, Olivier, Neethling, Kirkland and Warner9} The small MVs and DVs act as traps for mobile metal atoms, such as Fe, to form single atom dopants covalently bonded in the lattice.^{Reference Robertson, Montanari, He, Kim, Allen, Wu, Olivier, Neethling, Harrison, Kirkland and Warner37} This demonstrates how in situ electron beam irradiation can be used to engineer the doping of graphene with 10 nm precision.

Defects in other inorganic 2D materials

Structurally related 2D inorganic materials such as h-BN and transition metal dichalcogenides (TMDs) have attracted much interest recently because of their remarkable properties, which differ considerably from those of graphene. These 2D materials are alternatives to graphene and show promise for applications in electronics, optoelectronics, catalyses, or energy storage devices. Despite the technological potential of inorganic 2D materials, electron irradiation inside the TEM has been applied in only a few studies. The effects of irradiation in inorganic 2D materials are much more complex than those in graphene because of their complex layered structures with several atomic layers per monolayer in some of these 2D materials, and the fact that more than one element is involved, leaving a larger variety of possible defect structures. Although data on the responses of such inorganic 2D materials to electron irradiation are scarce, we analyze the available experimental results here and try to draw conclusions on the general behavior of inorganic 2D materials under irradiation.

h-BN layers have the closest structural similarity to graphene. In situ, as well as ex situ, experiments in AC-TEM instruments have shown that boron atoms are displaced more often than nitrogen atoms,^{Reference Jin, Lin, Suenaga and Iijima12,Reference Meyer, Chuvilin, Algara-Siller, Biskupek and Kaiser16,Reference Alem, Erni, Kisielowski, Rossell, Gannett and Zettl76,Reference Kotakoski, Jin, Lehtinen, Suenaga and Krasheninnikov77} because the threshold energy for knock-on displacements of B atoms (74 keV) is smaller than that of N atoms (84 keV) in defect-free h-BN monolayer sheets.^{Reference Zobelli, Gloter, Ewels, Seifert and Colliex78} Single boron vacancies, but not single nitrogen vacancies, have been created under irradiation at an electron energy of 80 keV.^{Reference Meyer, Chuvilin, Algara-Siller, Biskupek and Kaiser16,Reference Alem, Erni, Kisielowski, Rossell, Gannett and Zettl76} It is interesting that the selective sputtering of B atoms has also been observed at an electron energy of 120 keV,^{Reference Jin, Lin, Suenaga and Iijima12} although both species should be displaced, such that other mechanisms might also be involved.

Because B–B and N–N bonds are less favorable than B–N bonds, SW defects or reconstructed vacancies were not expected in h-BN. Most recently, however, pentagon–heptagon (5–7) and square–octagon (4–8) defects were observed by TEM.^{Reference Cretu, Lin and Suenaga79,Reference Gibb, Alem, Chen, Erickson, Ciston, Gautam, Linck and Zettl80} Cretu et al. observed the in situ formation of 4–8 defects at high temperature (∼1000 K) under 60 keV electrons.^{Reference Cretu, Lin and Suenaga79} Surprisingly, the 4–8 defects appeared to be stable with homoelemental B–B and N–N bonds, although these bonds should be energetically less favorable than B–N bonds. Apparently, this compensates the energy from possible open bonds and the strain energy in the reconstructed defect. Further studies, however, are necessary to clarify this issue. The defects could also be mobilized along the armchair direction by the gliding of atomic rows, induced by high temperature and electron irradiation.^{Reference Cretu, Lin and Suenaga79} As more defect structures are found, further studies will be required to fully understand the formation and evolution of multivacancies in h-BN sheets under electron irradiation.

In situ electron irradiation with a higher dose was found to give rise to the formation of larger triangular holes in BN because unsaturated nitrogen atoms at the edges could also be displaced by 80 keV electrons, resulting in the growth of the vacancy.^{Reference Meyer, Chuvilin, Algara-Siller, Biskupek and Kaiser16,Reference Alem, Erni, Kisielowski, Rossell, Gannett and Zettl76,Reference Warner, Rümmeli, Bachmatiuk and Büchner81} Although holes with other shapes have also been observed in few-layer h-BN sheets,^{Reference Warner, Rümmeli, Bachmatiuk and Büchner81} triangular holes dominate in monolayer h-BN under 80 keV irradiation. It is interesting to note that the internal edges of the triangular holes show a nitrogen-terminated zigzag configuration because of the higher threshold energies of N atoms.

Another interesting experiment involved the introduction of C atoms into BN sheets by electron irradiation such that BCN sheets were created.^{Reference Wei, Wang, Bando and Golberg82,Reference Krivanek, Chisholm, Nicolosi, Pennycook, Corbin, Dellby, Murfitt, Own, Szilagyi, Oxley, Pantelides and Pennycook83} Wei et al. used paraffin wax as a carbon source to create BCN sheets in a transmission electron microscope operating at 300 kV.^{Reference Wei, Wang, Bando and Golberg82} Under electron irradiation, carbon atoms were split from hydrocarbon molecules. The electron energy was sufficiently high to knock both B and N atoms from the h-BN lattice. These vacancies were immediately filled by C atoms. The behavior of BCN sheets under electron irradiation shows features of the behaviors of both graphene and h-BN because triangular vacancy aggregates as well as irregular holes appear.^{Reference Ci, Song, Jin, Jariwala, Wu, Li, Srivastava, Wang, Storr, Balicas, Liu and Ajayan84} Electron-irradiation-assisted doping appears to indicate a pathway to the design and synthesis of in-plane heterostructures. The upscaling of in situ electron irradiation studies (e.g., by exposing large areas to electron irradiation in a dedicated facility) could be used to fabricate 2D heterostructures, such as BCN with pre-defined electronic properties, on an industrial level.

TMDs represent another interesting class of 2D materials. Here, a unit cell consists of three atomic layers. Early experiments^{Reference Hansen, Ramasse, Kisielowski, Brorson, Johnson, Topsøe and Helveg85,Reference Liu, Suenaga, Wang, Shi, Okunishi and Iijima86} showed that chalcogenides exposed to electron irradiation are most likely deformed through the removal of sulfur atoms by knock-on displacements. Recent experiments on MoS₂^{Reference Komsa, Kotakoski, Kurasch, Lehtinen, Kaiser and Krasheninnikov13,Reference Qiu, Xu, Wang, Ren, Nan, Ni, Chen, Yuan, Miao, Song, Long, Shi, Sun, Wang and Wang17} showed that both single and double sulfur vacancies can be generated under irradiation of 80 keV electrons. Figure 6a shows a high-resolution TEM image of a single S vacancy in a monolayer of MoS₂.^{Reference Qiu, Xu, Wang, Ren, Nan, Ni, Chen, Yuan, Miao, Song, Long, Shi, Sun, Wang and Wang17} The types of defects can be distinguished by analyzing the intensity profile and comparing it with the intensity profiles of simulated vacancy images.^{Reference Qiu, Xu, Wang, Ren, Nan, Ni, Chen, Yuan, Miao, Song, Long, Shi, Sun, Wang and Wang17} Further studies of defect evolution under electron irradiation with different exposure times demonstrated that the average vacancy generation rate of single S vacancies was 10¹⁰ cm^–2s^–1 when the beam current density was several amperes per square centimeter, in agreement with theoretical predictions.^{Reference Komsa, Kotakoski, Kurasch, Lehtinen, Kaiser and Krasheninnikov13,Reference Qiu, Xu, Wang, Ren, Nan, Ni, Chen, Yuan, Miao, Song, Long, Shi, Sun, Wang and Wang17} Similar to graphene and h-BN, vacancies can be filled with other atomic species that are deliberately introduced onto the specimen.^{Reference Komsa, Kotakoski, Kurasch, Lehtinen, Kaiser and Krasheninnikov13} It is interesting that the structure can also be transformed between semiconducting (2H) and metallic (1T) phases in single MoS₂ sheets under electron irradiation. This transformation can be achieved even when no substitutional atoms are present and, moreover, at an electron energy of 60 keV, where no beam-irradiation-induced displacements are expected.^{Reference Lin, Dumcenco, Huang and Suenaga87}

Figure 6.

Electron irradiation damage of monolayer MoS₂ under 80 keV electron irradiation: (a) Atomic structure and high-resolution transmission electron microscope image of a single sulfur vacancy. The inset is the atomic model, where the blue atoms are molybdenum atoms and the others are sulfur atoms. (b) Sulfur vacancies (denoted by arrows) diffuse and agglomerate and then form an extended line defect (highlighted by dotted lines). The inset is the same area before electron irradiation. (c) Holes (denoted by yellow arrows) without a distinct shape formed under continuous irradiation. (d) Growth of holes, where the connecting regions between the adjacent holes are narrowed into nanoribbons (denoted by red arrow), which are robust under 80 keV irradiation. (e) The ribbons exhibit a different crystalline structure than MoS₂ ribbons. (a–b) Reproduced with permission from Reference 17. © 2013 Nature Publishing Group. (c–e) Reproduced with permission from Reference 15. © 2013 Nature Publishing Group.

Under continuous irradiation, S vacancies diffuse and agglomerate and then form an extended line defect that can be described as a row of SVs (Figure 6b). After prolonged exposure to the electron beam, the lines expand and span over two neighboring rows of the S sublattice to form wider defect lines.^{Reference Komsa, Kurasch, Lehtinen, Kaiser and Krasheninnikov14} Additional irradiation gives rise to the formation of large holes in MoS₂ sheets, as shown in Figure 6c, but the holes have no regular shape. During growth of the holes, the bridges between adjacent holes narrowed to nanoribbons (Figure 6d). The ribbons showed high crystalline perfection, unlike any other known molybdenum disulfide ribbons (Figure 6e). It is interesting to note that the ribbons were much more robust under electron irradiation than large MoS₂ sheets.^{Reference Liu, Xu, Wu, Zhang, Yu, Qiu, Hong, Jin, Li, Wang, Sun and Guo15} Through a combination of first-principles calculations and TEM imaging simulations, it was shown that such <1-nm ribbons were Mo₅S₄ crystals.^{Reference Liu, Xu, Wu, Zhang, Yu, Qiu, Hong, Jin, Li, Wang, Sun and Guo15} Further calculations indicated that the ribbons were semiconductors with an indirect band gap of 0.77 eV. Recently, based on further experimental evidence, Lin et al. offered new insight into ribbon properties and structure.^{Reference Lin, Cretu, Zhou, Suenaga, Prasai, Bolotin, Cuong, Otani, Okada, Lupini, Idrobo, Caudel, Burger, Ghimire, Yan, Mandrus, Pennycook and Pantelides26}In situ electrical measurements had demonstrated that the ribbons were metallic, which was inconsistent with previous work. After further analysis, Lin et al. concluded that a model with the MoS stoichiometry more closely matched the experimental observations. They also fabricated WS, MoSe, and WSe nanowires by 80 keV electron irradiation.^{Reference Lin, Cretu, Zhou, Suenaga, Prasai, Bolotin, Cuong, Otani, Okada, Lupini, Idrobo, Caudel, Burger, Ghimire, Yan, Mandrus, Pennycook and Pantelides26}

To our knowledge, no electron irradiation studies of other 2D materials have been published to date. Most recently, the responses of van der Waals heterostructures (such as stacked graphene and MoS₂ layers) to electron irradiation were studied by two groups.^{Reference Zan, Ramasse, Jalil, Georgiou, Bangert and Novoselov88,Reference Algara-Siller, Kurasch, Sedighi, Lehtinen and Kaiser89} The results indicated that irradiation damage could be lessened by placing a MoS₂ layer on a single layer of graphene. Furthermore, the damage of single-layer MoS₂ under 60 keV electron irradiation could also be reduced by encapsulation of the single-layer MoS₂ between graphene layers.^{Reference Zan, Ramasse, Jalil, Georgiou, Bangert and Novoselov88} Sandwiched graphene/MoS₂/graphene structures were also found to be protected from radiation damage by 80 keV electrons.^{Reference Algara-Siller, Kurasch, Sedighi, Lehtinen and Kaiser89} It was estimated that the total vacancy formation rate was reduced by a factor of 600 in the sandwiched configuration^{Reference Algara-Siller, Kurasch, Sedighi, Lehtinen and Kaiser89} compared to that in a freestanding MoS₂ layer.

Conclusions and outlook

The transmission electron microscope projects the structure of monoatomic sheets directly onto the image plane. No loss of information in the third dimension occurs because the third dimension either does not exist (graphene) or consists of only two or three atomic planes (chalcogenides). As a result, the formation of defects, even at the single-atom level, can be observed in situ. Similar studies, in which the atomic coordinates of a particular defect configuration can be determined with high accuracy, have never been possible in three-dimensional materials.

Since aberration-corrected electron microscopy became available, structural defects in graphene have been studied in detail, and a large variety of defect configurations are known. Other 2D materials, however, are just now attracting the interest of electron microscopists, so it can be expected that these materials will be studied in more detail in the near future. In situ electron microscopy is indispensable since it allows both defect generation by electron irradiation and defect observation at the same time. Common defect configurations have already been observed, such that the understanding of defect formation and annealing will allow for the identification of more general principles regarding reconstruction in hexagonal 2D materials. The availability of electron beam spot sizes smaller than the distance between atoms in solids allows the structure of 2D materials to be tailored with atomic precision. Because defects in hexagonal monoatomic lattices show unique reconstruction behavior, the removal of atoms allows for the creation of new morphologies and, accordingly, new types of nanoparticles in one, two, and three dimensions.