Weiwei
Sun
*a,
Yunguo
Li
*b,
Baotian
Wang
cd,
Xue
Jiang
e,
Mikhail I.
Katsnelson
fg,
Pavel
Korzhavyi
hi,
Olle
Eriksson
a and
Igor
Di Marco
*a
aDepartment of Physics and Astronomy, Materials Theory, Uppsala University, Box 516, SE-75120 Uppsala, Sweden. E-mail: sun.weiwei@physics.uu.se; igor.dimarco@physics.uu.se
bDepartment of Earth Sciences, University College London, London WC1E 6BT, UK. E-mail: yunguo.li@ucl.ac.uk
cChina Spallation Neutron Source (CSNS), Institute of High Energy Physics (IHEP), Chinese Academy of Sciences (CAS), Dongguan 523803, China
dDongguan Institute of Neutron Science (DINS), Dongguan 523808, China
eKey Laboratory of Materials Modification by Laser, Ion and Electron Beams (Dalian University of Technology), Ministry of Education, Dalian 116024, China
fRadboud University of Nijmegen, Institute for Molecules and Materials, Heijendaalseweg 135, 6525 AJ Nijmegen, The Netherlands
gTheoretical Physics and Applied Mathematics Department, Ural Federal University, Mira Street 19, 620002 Ekaterinburg, Russia
hDepartment of Material Science and Engineering, KTH-Royal Institute of Technology, Stockholm SE-10044, Sweden
iInstitute of Metal Physics, Ural Division of the Russian Academy of Sciences, 620219 Ekaterinburg, Russia
First published on 2nd August 2016
The existence of BiXenes, a new family of 2D monolayers, is hereby predicted. Theoretically, BiXenes have 1H symmetry (Pm2) and can be formed from the 4d/5d binary carbides. As the name suggests, they are close relatives of MXenes, which instead have 1T symmetry (P
m1). The newly found BiXenes, as well as some new MXenes, are shown to have formation energies close to that of germanene, which suggests that these materials should be possible to be synthesised. Among them, we illustrate that 1H-Tc2C and 1T-Mo2C are dynamically stable at 0 K, while 1H-Mo2C, 1T-Tc2C, 1H-Os2C, and 1T-Rh2C are likely to be stabilised via strain or temperature. In addition, the nature of the chemical bonding is analysed, emphasizing that the covalency between the transition metal ions and carbon is much stronger in BiXenes than in MXenes. The emergence of BiXenes can not only open up a new era of conducting 2D monolayers, but also provide good candidates for carrier materials aimed at energy storage and spintronic devices that have already been unveiled in MXenes.
MXenes are produced by immersing selected MAX phase powders in hydrofluoric acid (HF). MAX phases include compounds of composition Mn+1AXn, where M is a transition metal element, A is an element mostly from IIIA and IVA columns, and X is carbon or nitrogen. The index n can be 1, 2 or 3, depending on the geometrical arrangement of atoms.15,16 In principle, many MCs can be intercalated by light ions to form 2D materials, providing several fascinating properties like superconductivity, good electrochemical performance and storage properties for hydrogen and lithium ions.13 Unfortunately, to date, existing 2D materials derived from MCs are limited to MXenes, and involve mostly a 3d M and a very limited number of heavier elements.17
We have recently reported18 that under ambient conditions bulk Ru2C can be stable in the space groups R3m and Rm. These structures are composed of layers that are ideal candidates as novel 2D systems. It is natural, therefore, to explore the structural and dynamical stability of 2D monolayers arising from Ru2C and neighbouring systems, Tc2C, Rh2C and Os2C, which are also likely to exist in a layered structure. To these compounds, one can add Mo2C, which was recently studied in a couple of experiments. First, a large surface of a superconducting Mo2C thin film of 10 Å thickness was synthesised at a high temperature.19 The preparation procedure was based on a modification of the method used for growing high-quality graphene, chemically-functionalized and defective reduced graphene oxide, or also for preparing MXenes. This approach may be used as a general strategy for fabricating monolayer carbides and other 2D crystals, including transition metal nitrides (MNs).20 Further, large-scale synthesis of 2D Mo2C/Mo2CTx flakes (where T is a surface termination group, e.g. –O –OH or –F) was also realized, via selective etching of Ga from Mo2Ga2C powders.17,21 Holding to these clues, it becomes crucial to understand the structural and dynamical stability of 2D monolayers arising from Mo, Tc, Ru, Rh, and Os binary carbides.
We hereby predict the existence of a new family of 2D materials having 1H symmetry and involving two layers of heavy M atoms sandwiching a graphene-like interlayer of C atoms. Due to the fact that they originate from binary carbides and have the same chemical composition of MXenes, we name these systems BiXenes. BiXenes differ from MXenes not only for their symmetry, but also for the nature of the M components. While MXenes are mostly composed of light M elements, e.g. 3d elements, BiXenes contain 4d/5d elements, which results in wider d bands. This is likely to bring out different functionalities like magnetic or superconducting properties.22 As for several known 2D monolayers, distinctive geometry and compositions are of great importance for improving mechanical stability and related properties.23 The metallic character makes BiXenes also potentially interesting as transparent conductors to be used in modern technology, e.g. in touch screens.24 The likely high thermal conductivity may be potentially interesting for electronics where the problem of heat elimination is one of the crucial obstacles for industrial exploitation. In any case, these are only a few exemplary features, since the emergence of BiXenes, analogously to that of any new class of 2D materials, offers a truly exciting opportunity for future discoveries.
For 2D materials, a supercell was constructed including a vacuum layer with a thickness of at least 15 Å. Full optimisation was carried out in the in-plane directions. The formation energies Eform with respect to the bulk were evaluated as:
Eform = E2D(M2C) − Ebulk(n[M2C])/n | (1) |
The density functional perturbation theory (DFPT) implemented in the PHONOPY code33 was employed to examine the dynamical stability. To ensure a reasonable convergence, we used an energy cutoff of 700 eV with a precision of 1 × 10−8 eV. The force constants in the bulk were calculated for an isotropic 5 × 5 × 5 supercell whose BZ was sampled with 5 × 5 × 5 k-points. Those in the 2D materials were instead obtained for an isotropic 5 × 5 × 1 supercell whose BZ was sampled with 5 × 5 × 1 k-points. For Mo2C, it was necessary to extend the 2D supercell to 6 × 6 × 1. The Fermi–Dirac (FD) broadening scheme34 for an electronic temperature σ was used in the self-consistency cycle to smear out the abrupt change of the Fermi–Dirac statistics in the ground state.
The 3D structures are composed of layers of different symmetries as basic building blocks. As shown in panel (e), in the Pm1 bulk structure one can identify layers with 1T symmetry with inversion (point group D3d). This corresponds to the 2D structure of MXenes, mentioned above. As shown in panel (f), instead, from the P63/mmc and R3m bulk structures, one can extract 2D systems with the 1H symmetry without inversion (point group D3h). This corresponds to the 2D structure of MoS2 but has never been found before for carbides or nitrides.8 Moreover, there is a fundamental feature which makes BiXene different from the transition metal dichalcogenides MX2 (X = S, Se, Te). This feature is that metal and non-metal atoms are swapped. The different arrangement leads to very different electronic structures and related properties, as we will see in section 2. Finally, it is worth stressing that both the 1H and 1T structures exhibit hexagonal lattices, while their main geometrical difference concerns the M layers, which are overlapping in the top view in BiXene (see Fig. 1).
The relative energies shown in the top panel in Table 1 indicate that either 1H or 1T is more energetically favoured. In the framework of PBE, all 1H-M2Cs are favoured, with the exception of 1T-Rh2C. Inclusion of vdW interactions makes also 1T-Os2C more favourable than the corresponding 1H structure. We notice that for the early elements Tc and Mo, results are only barely affected by including vdW interactions, while for the other systems the corrections are larger. We are not sure that this trend is physical, since it may be related to the particular vdW implementation used in VASP, whose accuracy increases with the number of valence electrons.35
PBE | vdW | |||||||||
---|---|---|---|---|---|---|---|---|---|---|
Mo2C | Tc2C | Ru2C | Rh2C | Os2C | Mo2C | Tc2C | Ru2C | Rh2C | Os2C | |
E Rtot (1H) | 0 | 0 | 0 | 0.25 | 0 | 0 | 0 | 0 | 0.24 | 0.01 |
E Rtot (1T) | 0.19 | 0.25 | 0.08 | 0 | 0.11 | 0.24 | 0.27 | 0.49 | 0 | 0 |
E form (1H) | 2.06 | 2.82 | 2.19 | 1.58 | 2.57 | 2.39 | 2.91 | 2.57 | 1.90 | 2.92 |
E form (1T) | 2.69 | 2.17 | 1.81 | 1.38 | 1.81 | 3.03 | 2.52 | 2.56 | 1.66 | 2.06 |
Most of the 2D materials are exfoliated slices of powders (bulk)/thin films or can be prepared by designed methods such as chemical vapour deposition (CVD).19 Their feasibility can first be evaluated through their formation energies with respect to the bulk, i.e. Eform from eqn (1). The values for the 2D materials addressed in this work are reported in the bottom panel of Table 1. In general, most of the formation energies appearing in Table 1 are smaller than the values of the primary (intra-layer) chemical bonds, suggesting that the interaction between the M layers is sufficiently weak for possible exfoliation. In fact, we emphasise that, if normalised per atom, the values reported in Table 1 are comparable to (or smaller than) the formation energies of germanene and silicene.36 This trend is confirmed by the inclusion of vdW interactions. As expected, the latter make the bulk energetically more convenient, increasing the formation energies of the monolayers. Despite this small increase, the formation energies are still below 1 eV per atom. Overall, all these 2D compounds seem possible to be realised, provided that they are dynamically stable (see below). Mo2C stands out as the only material for which the 1H structure seems easier to realise than the 1T structure, with a difference in the formation energies of about 0.6 eV. Finally, an interesting speculation based on Table 1 concerns the dependence of Eform on the atomic number. For M-ions heavier than Tc, the 1T structure seems more favourable, while the opposite holds for lighter elements. The Tc ion in Tc2C, corresponding to a half-filled 4d-shell, seems in fact to act as a transition element, although more compounds should be investigated to support this analysis.
Furthermore one has to consider that under experimental conditions 2D materials are often grown on top of a substrate, such as Si or SiO2. The substrate may induce an effective strain on the monolayer, which can in turn modify its dynamical stability. Eventually, an appropriate substrate can be selected to remove small pockets of imaginary modes. In addition to the role of the substrate, one has to consider that effects due to a finite temperature can further improve the dynamical stability with respect to our zero temperature calculations (see below for a more extensive discussion). Therefore, it is worth considering not only the fully stable monolayers but also those monolayers characterised by small imaginary pockets, which are likely to be removed by strain and temperature. In total, we can thus identify 6 (quasi) stable monolayers, including 1T-Mo2C, 1T-Tc2C and 1T-Rh2C for the MXene structure, and 1H-Mo2C, 1H-Tc2C and 1H-Os2C for the BiXene structure.
The vibrational properties of these 6 structures are discussed by means of the phonon site-projected density of state (PSPDOS), reported in the right panel of Fig. 2. A general feature for all materials except 1T-Rh2C is the presence of a big gap between modes projected over M and C species, which is related to their large mass ratio. For 1T-Rh2C, the unusually wide phonon band of the C atoms extends from 18 THz to 8 THz, where it overlaps with the M band. The absence of a gap will be further discussed in the next subsection. In the low frequency region of all materials one can also notice an overlap of M and C modes, corresponding to co-vibrations. This overlap is weak but noticeable, especially for 1T-Tc2C, at around 4 THz. Concerning the general differences between the 1T and 1H structures (to be inferred from both Mo2C and Tc2C), they appear to be not too marked. Once can notice that in 1H a marked peak characterises the top of the M band, and also that the end of the C band is not as high in energy as in 1T. Finally, we note that the phonon spectrum of Os2C is slightly different from the other ones, as the M band is narrower and no multiple peaks can be resolved. This may be a consequence of the fact that Os is a 5d element, i.e. much heavier than the 4d elements involved in the other materials. However, it may also be connected to a chemical bonding of a different nature, as arising from the more delocalised 5d electrons. This issue will be addressed more in detail in the next subsection.
As we stressed previously, the quasi-stable structures can in principle be stabilised by the presence of a substrate or by effects due to a finite temperature. In the perspective of practical applications, it is important to analyse these factors more in detail. Among the quasi-stable structures, 1H-Mo2C is the most interesting system to address, since it offers the lowest formation energy compatible with the 1H symmetry (see Table 1). Moreover, 1T-Mo2C has already been identified as a very promising candidate for a high-performance thermoelectric material, and its functionalized 2D layer could find many applications within power generation, energy storage devices, and catalysis.39,40 This suggests that its counterpart 1H-Mo2C may have equally interesting properties. Therefore, we choose 1H-Mo2C as a test example to investigate the effects of strain and temperature on the dynamical stability.
In Fig. 3 we report the phonon spectra of 1H-Mo2C under various conditions. For practical purposes, strain is considered within a range of 3%, compatible with recent work on analogous systems.41,42 Given that GGA often tends to underestimate the binding, the exchange–correlation functional was switched from GGA to LDA for these calculations. The phonon spectrum with no strain at zero temperature, shown in panel (a) of Fig. 3, is very similar to the corresponding spectrum of Fig. 2. As expected, the imaginary modes are slightly enhanced and also the upper phonon branches (mainly due to C) are upshifted, revealing that LDA strengthens the Mo–C bonds. These effects are further emphasized if a biaxial compressive strain (3%) is applied, as illustrated in panel (b) of Fig. 3. Instead, applying a tensile strain (3%), leads to a strong downward shift of the upper phonon branches, as illustrated in panel (c) of Fig. 3. Moreover, the imaginary modes are strongly enhanced with respect to both the unstrained system and the system with a compressive strain. So, a tensile strain (3%) is the worst case scenario that will be considered below for investigating the effects due to a finite temperature.
Our calculations provide a model of the monolayers at zero temperature, while real life experiments are performed at a finite temperature. Ideally, a finite temperature should affect the lattice system both directly, through phonon–phonon interactions, and indirectly, through the electronic system. The latter can in principle be handled in Kohn–Sham DFT by considering a Fermi–Dirac distribution of the Kohn–Sham eigenvalues as well as a temperature dependent exchange–correlation functional. Unfortunately, there are no widespread functionals with this capability,43 so we have to limit our analysis for considering thermal smearing in a Fermi–Dirac distribution, as routinely discussed in the literature, e.g. for Al hydrides44 or Nb/Mo nitrides.45,46 The phonon spectrum of 1H-Mo2C under a 3% tensile strain and for an effective smearing (temperature) of σ = 0.4 eV is reported in panel (d) of Fig. 3. There are no signs of negative phonon branches, so under these conditions the monolayer can be considered as dynamically stable. The value σ = 0.4 eV should not be compared to experimental temperatures directly, since we are still neglecting the temperature dependence of the exchange–correlation functional as well as the coupling with the lattice.44 Instead, it is much more relevant to compare our σ to values obtained for other systems that have already been synthesised. For example, the monolayer of 2H-TaSe2 exists in the experiment at room temperature,47 despite being predicted in theory to be dynamically stable for a thermal smearing of at least σ = 0.68 eV.48 Our σ, for the worst case scenario, is well below 0.68 eV, and therefore we can expect that the predicted monolayers are likely to be stable under experimental conditions. This conclusion is even stronger if we consider the less unstable cases, corresponding to panels (a) and (b) of Fig. 3, which are expected to become stable for a σ as small as 0.1 or 0.2 eV.
The temperature effects on the electronic system alone seem sufficient to stabilize the predicted monolayers, but we should not forget that the phonon spectra are also directly affected by the temperature in the form of phonon–phonon interactions. These are not included in our calculations, since they are beyond the harmonic approximation. Anharmonic effects are often responsible for stabilizing some phases at high temperature and can be taken into account by more sophisticated techniques such as the self-consistent ab initio lattice dynamics method (SCAILD).49–52 While this analysis is too demanding to be performed now, we wish to address this issue in future studies.
The two structures of Tc2C show the highest number of states at Ef, which is related to four bands crossing the Fermi energy in this case, while there are only three for the other materials (see also next subsection). Another major feature noticeable in the DOS is that the states around Ef have mainly M character, and there is no significant hybridization with the C derived states. The most relevant hybridized M–C states lay instead at high binding energy, several eV below the Fermi energy. Moreover, for 1T-Rh2C and 1H-Os2C, no pronounced antibonding peak can be observed, due to the shift of Ef induced by filling the d shell. In 1H-Os2C, hybridized Os–C states spread over a wider range, which is partially due to the broader 5d band with respect to the 4d band. This is also noticeable through the high energy tail of the DOS, which in 1H-Os2C goes down to −8.5 eV, i.e. about 1 eV more than for the other systems. Moreover, although it is not visible from the bottom panel of Fig. 4, the tail does not suddenly stop at −8.5 eV but slowly decays at higher binding energies. This does not happen for the other systems. Finally, there seem to be no obvious differences between the DOS in the 1T and 1H structures, except that the latter appears to be characterised by slightly broader peaks.
Another interesting feature noticeable from the bottom panel of Fig. 4 is that occasionally a pseudo-gap is formed at about −4 eV. In one case, i.e. for 1T-Mo2C, this pseudo-gap becomes a real gap in the spectrum. This transition from a pseudo-gap to a full gap is related to the formation of unique inter-layered electrostatic interactions, and can be better analysed through the electronic localisation function (ELF), shown in the top panel of Fig. 4.
Before discussing the ELF of 1T-Mo2C, we first notice that the ELF of the 1H structures shows a chemical bonding featured with some covalency. The orientations of the ellipsoids of the electron clouds of the C layer are noticeably different. In 1H-Os2C they prefer to align parallel to the C axis, while in 1H-Mo2C and 1H-Tc2C the favoured direction is perpendicular to the C axis. Although the overlap of the electron clouds belonging to the M and C layers can be observed in all three cases, it is particularly strong for 1H-Os2C, where bridge-like channels emerge. It can also be inferred that the in-plane C–C π bonds are weaker in the case of 1H-Os2C, and this is different from the other 2D materials.
The 1T structures show conversely more localised charges around the C atoms, which points to the fact that they are more ionic than their 1H counterparts. The ELF of 1T-Mo2C stands out as significantly different from the others, due to the presence of the point charges at the surface (M layers). These point charges produce a unique inter-layered electrostatic interaction with the C layers, which induces the formation of the gap at −4 eV observed in the bottom panel of Fig. 4. Another interesting feature is that the ELF of 1T-Rh2C shows a strongly ionic Rh–C bond, where the states at the surface are almost empty. In general, the stronger the inter-layer bonds, the higher the frequency of the C modes in the phonon spectra. Therefore, the ELF data explain the extension of the C bands in the phonon spectra of Fig. 2.
We can then analyse what happens by removing an electron from the half-filled 4d shell. This situation corresponds to a transition from Tc2C to Mo2C. In 1T-Mo2C we note that the 1st band forms the overlapping “saddles”, while the 2nd band forms hexagons with the holes from the 3rd band. These features are consistent with a simple evolution of the FS of 1T-Tc2C, which has one electron more. Instead, the evolution of the two 1H structures is not so straightforward. In the FS of 1H-Mo2C, there is one less band than in 1H-Tc2C, and the “dog-bone” sheets evolve into smaller holes. Moreover, for 1H-Mo2C, in contrast to 1H-Tc2C, the FS sheets formed by the 2nd band shrink, and additionally there is one less band missing in the FS, also mentioned above. Finally, notice that only 1H-Mo2C shows the coexistence of well-defined hole and electron pockets (see the 1st and 2nd bands in the corresponding electronic structure of Fig. 5). 1T-Mo2C and 1H-Os2C also show pockets of both characters, but the electronic ones are extremely small.
After this analysis, we can investigate the cleavage energy Ecl due to the interlayer coupling strength, which is a way to explore how feasible the exfoliation process is.66 The cleavage energies of the predicted monolayers with respect to the separation distance of a fracture in the bulk material are reported in Fig. 6. 1H-Os2C stands out as the monolayer having the highest cleavage energy and can be achieved by overcoming a barrier of at least 0.81 J m−2. This value has to be compared with analogous values found in the literature. Theoretical calculations show67 that the cleavage energy of ReSe2 is around 1.12 J m−2, and the corresponding monolayer has been successfully exfoliated in experiment.68 Another relevant term of comparison is the experimentally estimated Ecl of graphite, which is around 0.37 J m−2.69 The cleavage energies of 1T-Mo2C and 1T-Rh2C are comparable to this value, while those of 1H-Mo2C and 1T-Tc2C are slightly larger, but still smaller than 0.6 J m−2. Among these monolayers, 1H-Tc2C is the easiest to exfoliate, with an Ecl of about 0.2 J m−2. Overall, the values reported in Fig. 6 are much smaller than those obtained for other materials that are considered suitable for mechanical exfoliation, such as Ca2N, TcSe2 or TcS2.67,70
![]() | ||
Fig. 6 Cleavage energy Ecl in J m−2 as a function of the separation distance d for a fracture in bulk materials for the six (quasi) stable monolayers. |
An alternative process to synthesise the three predicted monolayers of the MXene family is via etching from related MAX phases. In fact, 1T-Mo2C has been already realised experimentally from the non-standard MAX phase Mo2Ga2C.17,21 Electronic structure calculations of the MAX phases M2XC, where M = Tc, Rh and X = Al, Si are reported in the ESI.† Among those, Tc2AlC and Rh2AlC have negative formation energies, but only the latter is dynamically stable and can therefore be used to create 1T-Rh2C. In order to find a corresponding MAX phase for 1T-Tc2C, we also performed calculations for Tc2Ga2C but this structure was also found to be unstable.
In addition to the discovery of BiXene, in this work we also identify three systems in the MXene family, i.e. 1T-Tc2C, 1T-Rh2C, and 1T-Mo2C, which was also recently synthesised.17,21 These systems enrich the small group of MXenes originating from heavy transition metal carbides and are likely to lead to several interesting properties, which only future theoretical and experimental analyses will unveil. Moreover, the discovery of new pathways to MXenes through binary MCs or MNs will be very useful to overcome the limitation that most MXenes so far required to be produced from MAX phases including Al as an A element, which made many predicted monolayers not accessible experimentally.17
In practice, we suggest that these materials may be realised via mechanical exfoliation from their corresponding bulk structures. This seems immediately doable for 1T-Mo2C, 1H-Mo2C, and 1H-Tc2C, while 1H-Os2C, 1T-Rh2C and 1T-Tc2C require first that a suitable metastable bulk-like structure is realised in experiments. In contrast, 1T-Rh2C seems also possible to be synthesised from the MAX phase Rh2AlC.
Finally, the new M2C monolayers are analogous to MX2 in terms of structures (1T and 1H), but differ from the arrangement of the metal and non-metal sites, which is likely to lead to different chemical and physical properties. Our future research will focus on exploring the formation of BiXenes from other carbides and nitrides, possibly exploring all the transition metal elements that can form MAX phases. Furthermore, we will investigate in more detail the role of the temperature in stabilising the group of quasi-stable M2C monolayers, by means of more sophisticated techniques to describe thermally induced anharmonic effects.49–52
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/C6NR03602C |
This journal is © The Royal Society of Chemistry 2016 |