TEM Layer Identification MoS2/MoTe2
June 1, 2022 – How many layers does a flake of MoS2 or MoTe2 have? For two-dimensional materials the exact number decides their properties, yet it is hard to read directly in the electron microscope. Researchers from Ulm University have now compared high-resolution TEM, electron energy-loss spectroscopy and 3D electron diffraction on freestanding flakes with one to four layers. HRTEM reliably singles out the monolayer, while momentum-resolved EELS and 3D electron diffraction can tell up to four layers apart; diffraction proved the least affected by contamination and the gentlest on the sample.
How many layers does a flake of a two-dimensional material have? The question sounds simple, yet the answer decides what the flake can do. Materials such as MoS2 and MoTe2 are crystalline solids that are periodic in the plane but only a few atoms thick, and this confinement gives them unique properties for novel electronic[1] and optical[2] applications as well as for spintronics.[3] Stacked on top of each other, they form tailored van der Waals heterostructures.[4]–[6]
In their bulk form these transition metal dichalcogenides (TMDs) are often semiconductors with an indirect band gap, which gradually turns into a direct band gap when the crystal is thinned down from a few layers to a monolayer.[7],[8] While bulk crystals and monolayers have been studied extensively, few reports deal with few-layer material. On the way from bulk to a few layers, quantum confinement and reduced dielectric screening become more and more important.[9]–[11] Few-layer MoS2, for example, behaves differently in photoluminescence and in electron energy-loss spectroscopy depending on the exact number of layers.[12]–[14] Any application of 2D materials and their heterostructures therefore needs precise control over the number of layers.[10],[15]–[17]
On a substrate, several methods can count layers, among them atomic force microscopy, Raman spectroscopy, X-ray photoelectron spectroscopy and optical contrast.[18]–[25] Transmission electron microscopy (TEM), however, is the indispensable tool for the structure of freestanding 2D TMDs, and it would be a great advantage to read the number of layers directly from the TEM data. Despite the broad range of TEM methods, this is not a trivial task.[26],[27] Researchers from the Electron Microscopy Group of Materials Science and the Institute of Experimental Physics at Ulm University have now examined it systematically. They chose MoS2 with its light sulfur atoms and MoTe2 with its heavy tellurium atoms as representatives of the hexagonal 2H TMDs, determined the number of layers of each flake beforehand, and then studied the same thicknesses as freestanding samples with high-resolution TEM (HRTEM), electron energy-loss spectroscopy (EELS) and three-dimensional electron diffraction (3D ED). How the samples were made and imaged is summarized in the box below.
Aberration-corrected HRTEM at low voltage locates atoms precisely. It can distinguish atomic columns of different elements and identifies single defects in monolayer TMDs.[29],[30] A monolayer can therefore be recognized unambiguously by its isolated point defects. The open question was whether the image can also tell two, three or four layers apart. A monolayer consists of three atomic planes: a plane of molybdenum atoms between two planes of chalcogen atoms. In a 2H multilayer, the chalcogen atoms of each layer sit directly above the metal atoms of the layer below and vice versa (Fig. 1a). The team compared experimental images with calculated ones for MoS2 and MoTe2 (Figs. 1b and 1c) and extracted intensity profiles along the atomic rows, marked by red and blue arrows.
In monolayer MoS2, the peaks of the profile belong to the single molybdenum atom (atomic number 42) and to the pair of sulfur atoms (2 × 16) in the neighbouring column. Their scattering powers differ little, so the peaks differ only slightly in height, in experiment and calculation alike. In a bilayer, every column holds the same mix of atoms, and the peak pairs become almost equal. The remaining contrast differences vanish in the image noise, so the contrast cannot reliably distinguish two to four layers of MoS2. In MoTe2 the situation is different: one tellurium atom (atomic number 52) already scatters more strongly than a molybdenum atom, and a column of two tellurium atoms is clearly brighter than a molybdenum column. With more layers, however, these differences disappear here too. The researchers also noticed a bright spot in the middle of the hexagons in MoTe2 that belongs to no atom and grows stronger with thickness. Multiple scattering in the periodic structure explains it: interference between the scattered waves creates additional features that are not atomic positions. Overall, the calculated images agreed quantitatively with the experiment; small deviations came from uncertainties in defocus and in the other aberrations and from local corrugation and strain. Other stacking types than 2H could still be told apart in the images.
Spectroscopy offers a second route. The standard way of measuring thickness with EELS is the log-ratio method, which takes the relative thickness t/λ, the thickness in units of the inelastic mean free path, from the ratio between the total intensity of the spectrum and that of the zero-loss peak. It works well for relative thickness changes, but for absolute values it is expected to be accurate only to about 10 % or 2 nm, as found for samples 10–200 nm thick.[26] Still, it has been used for TMD flakes thinner than ten layers.[12],[14] The team therefore recorded spectra from freestanding areas 1 µm in diameter in one- to four-layer MoS2 and MoTe2 (Fig. 2). Thicker flakes did give higher values on average, but two problems ruled out counting layers this way. First, areas with the same number of layers gave widely different results, sometimes more different than areas with different numbers of layers, partly because of varying hydrocarbon contamination on the surface.[26] Second, the method overestimated the absolute thickness of both materials, and of MoTe2 even more than of MoS2, although both have very similar lattice constants and should come out alike. The increase per layer was at least of the same order as the known layer spacing of about 0.65 nm for MoS2 and 0.7 nm for MoTe2. The stronger contamination of MoTe2 explains part of the difference. The authors conclude that the log-ratio method is not a reliable way to obtain the number of layers in few-layer TMDs; its results even changed with minor experimental settings such as the energy dispersion of the spectrometer.
A more specific signature lies in the low-loss spectrum itself. The A and B exciton peaks near the band gap shift only very little with the number of layers,[12] too little to detect reliably even with a monochromated source. The team therefore turned to the two main features between 0 and 50 eV, commonly called the π and π + σ plasmon peaks.[5],[13] In few-layer 2D materials both peaks shift in energy and change in intensity with the number of layers,[12],[13] and so does their dispersion, the way the peak energy depends on the momentum q that the electron transfers. For a collective excitation in a perfectly two-dimensional electron gas, the energy rises with the square root of q, whereas a three-dimensional system shows a quadratic dispersion. To measure the dispersion, the researchers recorded so-called ω-q maps, which spread spectra for a whole range of momentum transfers across the detector in a single exposure. The box below describes the measurement.
For MoS2, the ω-q maps of one to four layers showed a clear trend (Fig. 3): both the π and the π + σ peak moved to higher energies with more layers and with larger momentum transfer. Square-root fits described the dispersion up to 0.15 Å−1, the largest momentum at which the signal was still good enough. The energy differences between one and four layers were largest at the highest recorded momentum transfers and, as calculations showed, shrink again at very high q, which makes intermediate momentum transfers the ideal choice for thickness measurements on MoS2. Around q = 0.1 Å−1, the peak positions for different layer numbers can be several electronvolts apart.
MoTe2 scatters more strongly, so its π + σ peak could be followed with very good signal up to 0.25 Å−1 (Fig. 4). Here, however, the dispersion did not follow the square-root law, the less so the larger the momentum and the more layers, a sign of the transition from two- to three-dimensional behaviour. Two and three layers had very similar peak positions over the whole momentum range, which makes MR-EELS thickness measurements on MoTe2 very challenging. The π + σ peaks were well separated only between 0.11 and 0.13 Å−1, and at around 0.2 Å−1 they converged with those of a flake more than 18 layers thick. To tell thin from thick MoTe2, spectra should therefore be taken at the smallest possible momentum, where the nearly flat dispersion of bulk material differs most from that of a few layers. Contamination, in addition, can change the measured dispersion considerably, a particular problem for MoTe2, which oxidizes and contaminates easily. Without checking cleanliness by another method such as HRTEM, a single MR-EELS measurement does not suffice; careful annealing or mechanical cleaning in vacuum can help.
The third method looks at the sample in reciprocal space. When a crystal becomes very thin in one direction, its diffraction spots stretch out along that direction; in a 2D material they turn into continuous lines of intensity, the so-called relrods. At normal incidence, the diffraction pattern of a 2D crystal shows the discrete Bragg spots of the [0001] orientation (Fig. 5a); the relrods extend perpendicular to this plane (Fig. 5b). The notation follows the nomenclature introduced in earlier work of the group.[27] The relrods are not thin lines but have a complex shape and broaden at higher L, best seen for the 10L rod, because freestanding 2D layers are intrinsically wavy. The team analysed three types of relrods, 10L, 11L and 20L, and calculated their intensity profiles in the kinematic approximation for one to eight layers of MoS2 and MoTe2; the data for one to four layers of MoS2 had already been presented before.[27]
The results are summarized in Figs. 6 and 7. The different spacings between the outer atomic planes, 3.173 Å between the two sulfur planes in monolayer MoS2 and 3.491 Å between the tellurium planes in MoTe2, and the different scattering factors change the profiles slightly. The two strongest peaks of the monolayer 10L rod, at 0.3 Å−1 in MoS2 (Fig. 6a), move closer to the centre in MoTe2 (Fig. 7a), and the central peak at L = 0 differs in height between the two materials. The symmetry follows the scheme established for MoS2:[27] crystals with an odd number of layers have relrods that are symmetric about the base plane L = 0, while crystals with an even number of layers show asymmetric profiles, clearly seen for bilayers (Figs. 6b and 7b) and four layers (Figs. 6d and 7d). The researchers turned these differences into two numbers: an asymmetry factor and the width of the intensity oscillations along the rod. Plotted against each other, they form a detectability map (Fig. 8), described in the box below. Well-separated points mark layer numbers that can be distinguished in an experiment, points close together mark those that cannot: from about five layers on, MoS2 and MoTe2 crystals give practically the same signature.
Measured 10L relrods of one to four layers of MoTe2 confirmed the simulations (Fig. 9). For the monolayer, the overall intensity distribution, the widths and the positions of the peaks agreed well (Fig. 9a). The neighbouring rods 10L and 01L had the same shape but differed in scale by about 20 %, a known effect of dynamical scattering in non-centrosymmetric structures. Such differences within Friedel pairs identify a monolayer from a single normal-incidence diffraction pattern, but cannot distinguish higher layer numbers. For the bilayer, peak widths and positions matched, while the relative intensity at L = 0 deviated (Fig. 9b); the three-layer rod was symmetric (Fig. 9c), and the four-layer rod asymmetric, as expected (Fig. 9d). In all crystals the measured profiles deviated systematically from the calculation: in the monolayer, the central peak was stronger than predicted, which significant corrugation of the layer explains; in two to four layers it was weaker. Despite these differences, the overall pattern is easy to recognize, and the combination of asymmetry factor and peak widths "can serve as an unambiguous fingerprint for a MoTe2 crystal with a given number of layers in the range of 1–4," the authors write. The largest differences appear close to L = 0: the central part of the 10L rod within ±0.2 Å−1 would suffice, so a short tilt series of ±30° is enough once the expected profiles are known.
Comparing the three methods, the researchers weighed how many layers each can assign, how unambiguous the assignment is, the effort, the sensitivity to contamination and the electron dose. HRTEM under optimal conditions identifies monolayers of both materials unambiguously, by chalcogen vacancies or by the contrast between metal and chalcogen columns, but from bilayers on the differences fall below the noise. It requires precise image calculations, exact knowledge of the aberrations, atomically clean areas, which is particularly hard for MoTe2, and the highest electron dose of the three methods. The log-ratio method is quick and low-dose but unreliable, mainly because of contamination. Momentum-resolved EELS is far more sensitive, yet a direct assignment needs reference measurements or simulations for every thickness, the technique is laborious, contamination affects it as well, and the weak signal at higher momentum calls for a relatively high dose. Electron diffraction suffers much less from contamination, needs a much lower dose and is therefore the least damaging method: no beam damage was seen even when several data sets were taken from the same area.
The study thus gives microscopists a practical guide for a basic question that precedes almost every experiment on few-layer TMDs. "3D ED is less affected by sample contamination and beam damage, and is therefore the most feasible method for thickness determination," the authors conclude. Momentum-resolved EELS and 3D ED can both identify up to four layers, whereas HRTEM at its best singles out the monolayer, which in turn reveals the atomic structure that the other two methods cannot. The proposed scheme for collecting and processing 3D ED data can be used to determine the thickness of all 2H TMDs, so the number of layers can be read directly from the very flake under study, combined with the atomic-resolution imaging that the Cc/Cs-corrected SALVE microscope at Ulm University provides.
Resource:
Köster, J., Storm, A., Gorelik, T. E., Mohn, M. J., Port, F., Gonçalves, M. R., Kaiser, U. (2022).
Evaluation of TEM methods for their signature of the number of layers in mono- and few-layer TMDs as exemplified by MoS2 and MoTe2.
Micron 160, 103303, 103303.
https://doi.org/10.1016/j.micron.2022.103303
Download this news article as PDF – for reading on paper, taking notes or filing.
-
Chang, Y.-M., Lin, C.-Y., Lin, Y.-F., Tsukagoshi, K., 2016. Two-dimensional MoTe2 materials: From synthesis, identification, and charge transport to electronics applications. Jpn. J. Appl. Phys. 55, 1102A1. https://doi.org/10.7567/jjap.55.1102a1
-
Liu, T., Xiang, D., Zheng, Y., Wang, Y., Wang, X., Wang, L., He, J., Liu, L., Chen, W., 2018. Nonvolatile and programmable photodoping in MoTe2 for photoresist-free complementary electronic devices. Adv. Mater. 30, 1804470. https://doi.org/10.1002/adma.201804470
-
Sanikop, R., Sudakar, C., 2020. Tailoring magnetically active defect sites in MoS2 nanosheets for spintronics applications. ACS Appl. Nano Mater. 3 (1), 576–587. https://doi.org/10.1021/acsanm.9b02121
-
Pezeshki, A., Shokouh, S.H.H., Nazari, T., Oh, K., Im, S., 2016. Electric and photovoltaic behavior of a few-layer α-MoTe2 /MoS2 dichalcogenide heterojunction. Adv. Mater. 28, 3216–3222 https://doi.org/10.1002/adma.201504090
-
Mohn, M.J., Hambach, R., Wachsmuth, P., Giorgetti, C., Kaiser, U., 2018. Dielectric properties of graphene/MoS2 heterostructures from ab initio calculations and electron energy-loss experiments. Phys. Rev. B 97, 235410. https://doi.org/10.1103/PhysRevB.97.235410
-
Duong, N.T., Lee, J., Bang, S., Park, C., Lim, S.C., Jeong, M.S., 2019. Modulating the functions of MoS2/MoTe2 van der Waals heterostructure via thickness variation. ACS Nano 13 (4), 4478–4485. https://doi.org/10.1021/acsnano.9b00014
-
Lezama, I.G., Arora, A., Ubaldini, A., Barreteau, C., Giannini, E., Potemski, M., Morpurgo, A.F., 2015. Indirect-to-direct band gap crossover in few-layer MoTe2. Nano Lett. 15, 2336–2342. https://doi.org/10.1021/nl5045007
-
Gusakova, J., Wang, X., Shiau, L.L., Krivosheeva, A., Shaposhnikov, V., Borisenko, V., Gusakov, V., Tay, B.K., 2017. Electronic properties of bulk and monolayer TMDs: theoretical study within DFT framework. Phys. Status Solidi A 214, 1700218. https://doi.org/10.1002/pssa.201700218
-
Li, J., Kolekar, S., Ghorbani-Asl, M., Lehnert, T., Biskupek, J., Kaiser, U., Krasheninnikov, A.V., Batzill, M., 2021. Layer-dependent band gaps of platinum dichalcogenides. ACS Nano 15, 13249. https://doi.org/10.1021/acsnano.1c02971
-
Lee, H., Deshmukh, S., Wen, J., Costa, V.Z., Schuder, J.S., Sanchez, M., Ichimura, A.S., Pop, E., Wang, B., Newaz, A.K.M., 2019. Layer-dependent interfacial transport and optoelectrical properties of MoS2 on ultraflat metals. ACS Appl. Mater. Interfaces 11, 31543–31550. https://doi.org/10.1021/acsami.9b09868
-
Liu, Y., Hu, X., Wang, T., Liu, D., 2019. Reduced binding energy and layer-dependent exciton dynamics in monolayer and multilayer WS2. ACS Nano 13, 14416–14425. https://doi.org/10.1021/acsnano.9b08004
-
Nerl, H.C., Winther, K.T., Hage, F.S., Thygesen, K.S., Houben, L., Backes, C., Coleman, J. N., Ramasse, Q.M., Nicolosi, V., 2017. Probing the local nature of excitons and plasmons in few-layer MoS2. Npj 2D Mater. Appl. 1, 2. https://doi.org/10.1038/s41699-017-0003-9
-
Moynihan, E., Rost, S., O’Connell, E., Ramasse, Q., Friedrich, C., Bangert, U., 2020. Plasmons in MoS2 studied via experimental and theoretical correlation of energy loss spectra. J. Microsc. 279, 256–264. https://doi.org/10.1111/jmi.12900
-
Pelaez-Fernandez, M., Lin, Y.-C., Suenaga, K., Arenal, R., 2021. Optoelectronic properties of atomically thin MoxW(1–x)S2 nanoflakes probed by spatially-resolved monochromated EELS. Nanomaterials 11, 3218. https://doi.org/10.3390/nano11123218
-
Lin, M.-W., Kravchenko, I.I., Fowlkes, J., Li, X., Puretzky, A.A., Rouleau, C.M., Geohegan, D.B., Xiao, K., 2016. Thickness-dependent charge transport in few-layer MoS2 field-effect transistors. Nanotechnology 27, 165203. https://doi.org/10.1088/0957-4484/27/16/165203
-
Pallecchi, I., Manca, N., Patil, B., Pellegrino, L., Marre, D., 2020. Review on thermoelectric properties of transition metal dichalcogenides. Nano Futures 4, 032008. https://doi.org/10.1088/2399-1984/ab92f4
-
Pawlik, A.S., Aswartham, S., Morozov, I., Knupfer, M., Büchner, B., Efremov, D.V., Koitzsch, A., 2018. Thickness dependent electronic structure of exfoliated mono-and few-layer 1 T′ − MoTe2. Phys. Rev. Mater. 2, 104004 https://doi.org/10.1103/PhysRevMaterials.2.104004
-
Shearer, C.J., Slattery, A.D., Stapleton, A.J., Shapter, J.G., Gibson, C.T., 2016. Accurate thickness measurement of graphene. Nanotechnology 27, 125704. https://doi.org/10.1088/0957-4484/27/12/125704
-
Xiao, Y., Zheng, W., Yuan, B., Wen, C., Lanza, M., 2021. Highly accurate thickness determination of 2D materials. Cryst. Res. Technol. 56, 2100056. https://doi.org/10.1002/crat.202100056
-
Grzeszczyk, M., Gołasa, K., Zinkiewicz, M., Nogajewski, K., Molas, M.R., Potemski, M., Wysmołek, A., Babinski, A., 2016. Raman scattering of few-layers MoTe2. 2D Mater. 3, 025010 https://doi.org/10.1088/2053-1583/3/2/025010
-
Liang, F., Xu, H., Wu, X., Wang, C., Luo, C., Zhang, J., 2018. Raman spectroscopy characterization of two-dimensional materials. Chin. Phys. B 27 (3), 037802. https://doi.org/10.1088/1674-1056/27/3/037802
-
Zemlyanov, D.Y., Jespersen, M., Zakharov, D.N., Hu, J., Paul, R., Kumar, A., Pacley, S., Galvin, N., Saenz, D., Smith, K.C., Fisher, T.S., Voevodin, A.A., 2018. Versatile technique for assessing thickness of 2D layered materials by XPS. Nanotechnology 29, 115705. https://doi.org/10.1088/1361-6528/aaa6ef
-
Hutzler, A., Matthus, C.D., Dolle, C., Rommel, M., Jank, M.P.M., Spiecker, E., Frey, L., 2019. Large-area layer counting of two-dimensional materials evaluating the wavelength shift in visible-reflectance spectroscopy. J. Phys. Chem. C. 123, 9192–9201. https://doi.org/10.1021/acs.jpcc.9b00957
-
Niu, Y., Gonzalez-Abad, S., Frisenda, R., Marauhn, P., Drüppel, M., Gant, P., Schmidt, R., Taghavi, N.S., Barcons, D., Molina-Mendoza, A.J., de Vasconcellos, S.M., Bratschitsch, R., Perez De Lara, D., Rohlfing, M., Castellanos-Gomez, A., 2018. Thickness-dependent differential reflectance spectra of monolayer and few-layer MoS2, MoSe2, WS2 and WSe2. Nanomaterials 8, 725. https://doi.org/10.3390/nano8090725
-
Ottaviano, L., Palleschi, S., Perrozzi, F., D’Olimpio, G., Priante, F., Donarelli, M., Benassi, P., Nardone, M., Gonchigsuren, M., Gombosuren, M., Lucia, A., Moccia, G., Cacioppo, O.A., 2017. Mechanical exfoliation and layer number identification of MoS2 revisited. 2D Mater. 4, 045013 https://doi.org/10.1088/2053-1583/aa8764
-
Heo, Y.-U., 2020. Comparative study on the specimen thickness measurement using EELS and CBED methods. Appl. Microsc. 50, 8. https://doi.org/10.1186/s42649-020-00029-4
-
Gorelik, T.E., Nergis, B., Schöner, T., Köster, J., Kaiser, U., 2021. 3D electron diffraction of mono- and few-layer MoS2. Micron 146, 103071. https://doi.org/10.1016/j.micron.2021.103071
-
Linck, M., Hartel, P., Uhlemann, S., Kahl, F., Müller, H., Zach, J., Haider, M., Niestadt, M., Bischoff, M., Biskupek, J., Lee, Z., Lehner, T., Börrnert, F., Rose, H., Kaiser, U., 2016. Chromatic aberration correction for atomic resolution TEM imaging from 20 to 80 kV. Phys. Rev. Lett. 117, 076101. https://doi.org/10.1103/physrevlett.117.076101
-
Algara-Siller, G., Kurasch, S., Sedighi, M., Lehtinen, O., Kaiser, U., 2013. The pristine atomic structure of MoS2 monolayer protected from electron radiation damage by graphene. Appl. Phys. Lett. 103, 203107. https://doi.org/10.1063/1.4830036
-
Lehnert, T., Lehtinen, O., Algara–Siller, G., Kaiser, U., 2017. Electron radiation damage mechanisms in 2D MoSe2. Appl. Phys. Lett. 110 (3), 033106 https://doi.org/10.1063/1.4973809
-
Gorelik, T.E., Habermehl, S., Shubin, A.A., Gruene, T., Yoshida, K., Oleynikov, P., Kaiser, U., Schmidt, M.U., 2021. Crystal structure of copper perchlorophthalocyanine analysed by 3D electron diffraction. Acta Cryst. B 77, 662–675. https://doi.org/10.1107/s2052520621006806









