Tuning the properties of ultrathin free-standing TMPTs by controlling their stoichiometry

September 01, 2022 - Transition metal phosphorus trisulfides (TMPTs) are new inorganic materials with inherent magnetic properties. Scientists from the Universities of Ulm (Germany), Helmholtz-Zentrum Dresden-Rossendorf (Germany) and Aalto University (Finland) have presented a detailed analysis of the interaction of the electron beam (30-80 kV) with TMPTs and demonstrate the possibility of tuning the properties of ultrathin, free-standing TMPTs by controlling their stoichiometry.

Transition metal phosphorus trisulfides (TMPTs) are inorganic materials with inherent magnetic properties. Due to their layered structure, they can be stacked into ultrathin films that have different properties than their solid counterparts. Here we present a detailed analysis of the interaction of the electron beam (30-80 kV) of a fully aberration-corrected transmission electron microscope (AC-TEM) with free-standing TMPTs in a few layers. Structural changes related to chemical and electronic properties were systematically investigated using different AC-TEM methods such as electron diffraction (ED), energy dispersive X-ray spectroscopy (EDX) and time-dependent electron energy loss spectroscopy (EELS) on FePS3, MnPS3 and NiPS3. The results are compared using DFT calculations, which predict that the trigger threshold for sulfur (S) removal is significantly lower than that for phosphorus (P). We analyze the electronic structure of TMPTs with single vacancies and oxygen impurities and predict different electronic properties depending on the defect type. Therefore, our study demonstrates the possibility of controlling the properties of ultrathin freestanding TMPTs by controlling their stoichiometry.

Two-dimensional (2D) materials exhibit a variety of fascinating electronic, optical, and magnetic properties. (1-4) In addition, heterostructures with unique properties can be developed by combining individual layers. (5,6,10) A promising class of 2D materials with interesting properties are transition metal phosphorus trisulfides (TMPTs). (12,13) They exhibit a wide variety of band gaps, (15) ranging from 1.5 to 3.0 eV, (16,17) and the chemical elements that compose them are mostly abundant on Earth and nontoxic, (18) making them suitable candidates for optoelectronic applications. (20) In particular, TMPTs exhibit inherent magnetic properties. They are among the most recently studied antiferromagnetic 2D materials (21,23) because the magnetism persists down to the limit of a few layers, (25) suggesting that they can serve as test objects for studying the fundamental aspects of 2D magnetism. (18,26) In addition, the layered structure of TMPTs makes few-layer and bulk systems excellent candidates for intercalation by ions or molecules.(19) Consequently, they are also suitable materials for energy storage. (27-29) In particular, due to the high binding energy of Li on their surface, applications of TMPTs as anode materials in Li-ion batteries have already been predicted. (33)

Initial theoretical and experimental studies demonstrated the possibility of modifying the properties of these materials. For example, cooperative spin crossover and semiconductor-metal transitions have been observed in MnPS3 and MnPSe3 under high pressure. (34) At the same time, TMPTs have not yet been intensively analyzed experimentally to understand the defect-induced properties, although theoretical studies on single layers predicted intriguing defect-induced changes in the electronic and magnetic properties of the materials, e.g., the inclusion of spin-cleaved center hole states in MnPS3. (35) In addition, several studies (36,38) have reported stacking defects in various TMPTs and have shown that the properties (e.g., thermal conductivity) of the materials are strongly altered by these defects; (39) however, no detailed studies on the effects of these defects on multilayer systems can be found in the literature.

Transmission electron microscopy (TEM) is a particularly powerful tool to modify (40), probe (41-44), and individualize (45-49) the properties of free-standing nanosheets with high-energy electrons. This technique can be used to extract chemical and crystallographic information from the materials, which can be correlated even at the atomic level using aberration-corrected (AC) high-resolution TEM (HRTEM). (50-52) A necessary prerequisite is the preparation of pure, free-standing TMPT samples with few layers, which has recently been achieved by polymer-assisted mechanical exfoliation. (53) However, detailed experimental analysis and characterization of the interaction of high-energy electrons with the ultrathin TMPTs in the low-voltage TEM (30-80 kV) range to minimize damage (54) is still pending.

Crystal structure and thickness analysis

We begin our study by analyzing the orientation, scale thickness, and quality of the transferred pristine material with few layers, which is known to be susceptible to oxidation. (53) Since MnPS3, FePS3, and NiPS3 are isostructural, all structural considerations presented below for MnPS3 apply to all three materials. In the bulk of MnPS3, the layers are stacked in a shifted manner and obey monoclinic symmetry (69) with ? = 107° (Figure 1a). [103] is the normal vector to the layer plane and is referred to here as normal incidence, since this is the observed orientation in all HRTEM results presented (Figure 1b). In this projection, the structure exhibits a rectangular centered pattern with pronounced hexagonal pseudosymmetry, and the unit cell contains 6 rows of atoms along the b vector (marked in Figure 1a) with different atomic compositions: rows 1 and 4 are composed of S and P atoms; rows 2, 3, 5, and 6 are composed of S and Mn atoms, resulting in a 1:2 row change shown in Figure 1b. In Figure 1b, a four-layer crystal with missing stacking faults shows the best agreement with the experimental image, and the good crystallinity of these thin samples can be seen. The line scans illustrate the 1:2 alternation mentioned above. An alternative TEM method for thickness determination (70) gives a thickness value of about 3 nm, confirming the results of the image simulation. Therefore, we conclude that the investigated pristine samples are few-layered systems (3-6 layers) with very good sample quality and without stacking errors.

Interaction between electrons and sample

Continuous irradiation of the sample leads to degradation of individual layers. Figure S5 shows an example of a four-layer sample, ranging from the original to the degraded layer, including areas with 1-3 layers. In this context, individual defects can be observed in the layers. Figure 2a and b show 80-kV HRTEM images of a NiPS3 single and double layer, respectively, showing distinct defects highlighted by white arrows. The contrast at the defect location in the double layer disappears, as can be seen from the inset in (Figure 2b), which shows a red line scan originating from the area marked by a red rectangle. The defect is a sulfur (S) gap (VS) in an atomic column where two S atoms of the two layers are stacked (S2).

Understanding the formation of individual defects paves the way to explore the response of these layer systems to electron irradiation in the TEM and to systematically change the properties of the system. Our TEM experiments show that sulfur or phosphorus vacancies are formed first, which then lead to larger defects or even holes in the layer. The formation energies for phosphorus and sulfur vacancies and the corresponding displacement energies (Td) in single-layer TMPTs were determined by ab initio molecular dynamics simulations (Figure 2c). Td is determined as the minimum kinetic energy that must be assigned to the recoil atom for it to leave its position in the crystal without immediate recombination.

In the case of normal displacement (i.e., the direction of the initial velocity vector is perpendicular to the surface), the recoiling sulfur atom pulls the nearest phosphorus atom behind it, resulting in the displacement of a phosphorus-sulfur dumbbell (S-P). A displacement deviating from the surface normal, e.g. when the electron beam deviates slightly from the normal direction, is required to remove individual sulfur atoms (S*, denotes a non-normal displacement).

In addition, the atomic displacement cross section was estimated using the McKinley-Feshbach formalism. (50,71-73) For an accurate estimation of the influence of lattice vibrations on the effective cross section, the Debye temperature model (?D) was used, which is described in the main article J. Phys. Chem. C 2022, 126, 36, 15446. Figure 2d shows the effective cross sections for knock-on displacements of S and P atoms in FePS3, MnPS3, and NiPS3. (50,71,74,75) For comparison, the knock-on cross section for graphene is also shown, indicating a much higher threshold. Overall, the displacement threshold for sulfur is significantly lower than for phosphorus. Therefore, it can be predicted that in the accelerating voltage range of 50-300 kV, more S than P should be removed by elastic interaction. It is well known that inelastic scattering plays an important role in 2D insulators and semiconductors. (54a,75,76a,) It has been shown that such additional channels for defect generation are important for the formation of single S gaps in the MoS2 single layer, (54a,76b) where damage generation has been observed at electron energies well below the knock-on threshold. Therefore, considering the inelastic effects and the fact that S is three times more abundant in the initial structure, S is expected to be readily removed by high-energy electrons (20-300 kV) in single-layer TMPTs. In systems with few layers, strong structural changes are also expected due to electron interactions, which will be investigated experimentally in the following sections. Indeed, it may become impossible to image 2d materials without beam-sample interactions.(76c)

To work out the effects of defects on the electronic properties, the atomic structure of the original material was first calculated, and the optimized lattice constants (Table 1) were found to agree very well with experimental values and previous theoretical reports. (18) We then calculated the spin-polarized density of states (DOS), which is shown in Figure 3. Our results show that TMPT single layers have a semiconductor character with a band gap in the range of 2.0-2.3 eV. Although the DFT band gap should not be directly compared with the optical gap measured in experiment (77), our band gap calculated at the PBE + U level for the FePS3 single layer agrees with the experimental value of 2.2 eV. (17)

We also considered 3 types of point defects (Figure 3): (i) sulfur and (ii) phosphorus vacancies (VS and VP) and a complex vacancy corresponding to (iii) sulfur and phosphorus (VSP) with missing atoms. The presence of sulfur vacancies introduces new spin-polarized states near the conduction and valence band edges, mainly carried by S-p orbitals. The exception is MnPS3, where the defect states are formed as part of the band continuum. In the case of a single P vacancy, new localized states are formed 0.2-0.7 eV above the conduction band edge. Defect-associated states are much more pronounced in the case of VSP because the electron density is increased around the Fermi level, especially for FePS3.

Electron beam excited chemical and structural changes analyzed with EELs and EDX

Here FePS3 is presented in detail, and unless otherwise noted, analogous behavior was found for the corresponding Mn- and Ni-containing components. Figure 4 shows an 80 kV HAADF image and elemental maps of irradiated and unirradiated samples. Disorder and cluster formation is observed in the irradiated regions; see the magnified area in Figures 4a and S12. Comparison of the atomic fractions in the unirradiated and irradiated regions shows that mainly S is removed, which agrees well with our theoretical results (Figure 4b,c).

To further facilitate comparison with experimental results, we modeled phases with intermediate stoichiometry that could potentially form in the low-sulfur material. The phases were generated by uniformly propagating vacancies in the original structure, simulating experimental conditions under prolonged electron irradiation. At high concentrations of S vacancies (>50%), the optimized structures of the phases showed similar clustering of metal atoms as our experimental data. Such nonstoichiometric phases can be stabilized in the metal-rich boundary region, as previously shown for transition metal dichalcogenides. (78) The electronic structures were also analyzed for different concentrations of S-features. As expected, the number of center hole states increases with defect concentration. However, we found that the position of the states depends on the distribution of the defects. It is expected that the states eventually fill the entire bandgap region, suggesting that the irradiated samples may exhibit hopping-type conductivity.

The conversion process (Figure 5a) is analyzed using ELNES in the corresponding core loss EELS edges. For this purpose, EELS signals with constant low electron dose (dose rate of 15 ? 103 e-/(s-nm2)) were recorded over time, and the corresponding changes in the ELNES of the materials were identified, and the evolution of the spectrum over the recording time is shown in Figure 5. Figure 5b shows the Fe M2.3 edge of FePS3 with a few layers for different acquisition times. The M2,3 edge of first-row transition metals was used to identify the oxidation state of the corresponding transition metal in a compound. (58) The spectrum of a divalent iron (Fe2+) is characterized by a prominent pre-peak and a broad maximum at about 57 eV. (58) When changing from divalent to trivalent iron (Fe3+), the main peak shifts to the blue and the intensity of the pre-peak decreases. (58) The EELS signal from the M2,3 edge of untreated FePS3 with few layers corresponds to a spectrum of mainly divalent iron, which is in good agreement with the proposed valence of iron in this material. (80) With increasing exposure time, a blue shift of the main peak and a strong decrease in the intensity of the pre-peak in the spectrum is observed. The decrease in the pre-peak is more evident in Figure 5c, where the maximum for each spectrum is aligned at 56.9 eV. The intensity ratio between the main peak and the pre-peak w_M/w_p (Figure 5d) changes from 1.7 (unirradiated) to about 2.8 (irradiated), indicating a transition from divalent to trivalent iron. (58) However, an absolute determination of the concentration from divalent to trivalent iron is not possible because the derived universal curve (58) was proposed only for valence-calibrated iron oxides in the high-spin configuration, and a general unambiguous quantification of valence using the intensity ratio is difficult. (57) Figure 5e shows the core loss EELS of the Fe L2.3 edge, and the integral ratio of the white line regions (L3 to L2) is shown in Figure 5f. The blue shift of the main L3 peak and the increase of the w_(L_3 )/w_(L_2 ) ratio for the irradiated sample demonstrate the gradual change of valence (55,56,59,60) from Fe2+ to Fe3+ and confirm the results of the M2,3 edge. In the case of the L2,3-P-S edge, the beginnings of the P and S edges are highlighted by a clear peak at about 132 eV (I) and a small peak at slightly higher energy 138.5 eV (II), respectively (see Figure 5g). The former peak agrees well with previous XPS results (23) and can be attributed to a combined peak from P-2p1/2 and P-2p3/2 transitions to 3s ?* due to S-P bonding. (23,83) The same is true for the peak observed at 162 eV in the original structure, which can be attributed to the S-2p1/2 and S-2p3/2 transitions. (23,83) With increasing time and interaction with the electron beam, the P peak at 132 eV decreases (i.e., the P-S bond is broken) and finally two peaks at 138.5 eV (II) and 146.5 eV (III) appear. Moreover, in the normalized spectra, the peaks originating from the phosphorus contributions are more prominent compared to the sulfur contributions. This is in good agreement with the theoretical analysis, which suggests that sulfur atoms are mainly removed from the sample under the electron beam. Figure 5h,i shows the raw signal and the extracted signal of the O-K edge taken sequentially with the other spectra shown. As the exposure time increases, the oxygen edge becomes more pronounced, indicating a higher oxygen concentration in the illuminated region. The source of the oxygen is due to two main factors (i) residual surface contamination and (i) insufficient vacuum in the TEM chamber. This is consistent with the EDX results, which show a high oxygen content in the illuminated region compared to the unmodified platelet, as shown in Figure 4.

The emerging peaks at 138.5 and 146.5 eV at the L2.3 P edge fit well with previous reports of P-O bonding in oxidized P-containing materials (86) and suggest that the accumulated oxygen is bound to the residual phosphorus. Also, the combination of the EELS experiments (Figure 5), showing a shift in valence from Fe2+ to Fe3+, with the EDX results (Figure 4) confirm the binding between O and Fe. Nevertheless, a clear identification of the local chemical (i.e., stoichiometric) and structural composition of the oxidized structure after electron beam interaction is not possible. Similar results are obtained for 30- and 60-kV irradiation experiments reported for MnPS3 in the main article J. Phys. Chem. C 2022, 126, 36, 15446.

Resource: Köster, J., Storm, A., Ghorbani-Asl, M., Kretschmer, S., Gorelik, T. E., Krasheninnikov, A. V., & Kaiser, U. (2022). Structural and chemical modifications of few-layer transition metal phosphorous trisulfides by electron irradiation. The Journal of Physical Chemistry C, 126(36), 15446-15455.