Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Profiling novel high-conductivity 2D semiconductors

Profiling novel high-conductivity 2D semiconductors Pro ling novel high-conductivity 2D semiconductors 1, 2 3, 4, 2 2 Thibault Sohier, Marco Gibertini, and Nicola Marzari nanomat/QMAT/CESAM and European Theoretical Spectroscopy Facility, University of Liege (Uliege), Belgium Theory and Simulation of Materials (THEOS), and National Centre for Computational Design and Discovery of Novel Materials (MARVEL), Ecole Polytechnique F ed erale de Lausanne, CH-1015 Lausanne, Switzerland Dipartimento di Fisica Informatica e Matematica, Universit a di Modena e Reggio Emilia, Via Campi 213/a, I-41125 Modena, Italy Department of Quantum Matter Physics, University of Geneva, CH-1211 Geneva, Switzerland (Dated: August 3, 2020) When complex mechanisms are involved, pinpointing high-performance materials within large databases is a major challenge in materials discovery. We focus here on phonon-limited conductivi- ties, and study 2D semiconductors doped by eld e ects. Using state-of-the-art density-functional perturbation theory and Boltzmann transport equation, we discuss 11 monolayers with outstanding transport properties. These materials are selected from a computational database of exfoliable ma- terials providing monolayers that are dynamically stable and that do not have more than 6 atoms per unit cell. We rst analyze electron-phonon scattering in two well-known systems: electron-doped InSe and hole-doped phosphorene. Both are single-valley systems with weak electron-phonon inter- actions, but they represent two distinct pathways to fast transport: a steep and deep isotropic valley for the former and strongly anisotropic electron-phonon physics for the latter. We identify similar features in the database and compute the conductivities of the relevant monolayers. This process yields several high-conductivity materials, some of them only very recently emerging in the litera- ture (GaSe, Bi SeTe , Bi Se , Sb SeTe ), others never discussed in this context (AlLiTe , BiClTe, 2 2 2 3 2 2 2 ClGaTe, AuI). Comparing these 11 monolayers in detail, we discuss how the strength and angular dependency of the electron-phonon scattering drives key di erences in the transport performance of materials despite similar valley structure. We also discuss the high conductivity of hole-doped WSe , and how this case study shows the limitations of a selection process that would be based on band properties alone. I. INTRODUCTION ductivity. We focus here on room-temperature phonon-limited electronic transport and search across 2D semiconduc- Two-dimensional (2D) semiconductors with excellent tors. First-principles methods have proven quite use- intrinsic transport properties would be bene cial to many ful and successful in predicting physics and properties in 1{3 applications . Some well-known 2D materials like 28 29,30 this regime and for 2D materials , provided one is transition-metal dichalcogenides (TMDs), phosphorene 19 31,32 aware of its limits . Although dedicated codes exist , or silicene have been extensively studied both experi- performing state-of-the-art rst-principles transport sim- 4,5 6,78 9{13 mentally (MoS , Si ) and theoretically (MoS , 2 2 ulations on large sets of materials remains nevertheless 13{16 17 P ). Other candidates have been proposed us- a challenge, and few works tackle more than one mate- ing approximate deformation potential models and the rial at a time. Yet, an overall panorama on the property Takagi formula, but it has been shown that such ap- landscape would be very valuable to materials design. proaches su er from limited reliability . The full poten- For example, in a previous work involving 5 materials , tial of 2D materials for future device applications could we have shown the importance of intervalley scattering, be much broader than the dozen materials currently while the authors of Ref. 33 have studied 4 hexagonal under extensive experimental investigation, since many elemental materials to show the bene ts of a sharp and more monolayers have been predicted to be either exfoli- deep single valley. 20{23 able from experimentally known layered materials In this work, we explore our portfolio of exfoliable 2D 24,25 or synthesizable . Such computational collections 23 34,35 materials taken from the Materials Cloud , limit- of prospective 2D materials would likely contain some ing ourselves to at most 6 atoms per unit cell. We se- promising candidates for electronic transport in various lect the most promising systems from a band structure contexts. analysis, and then study their transport properties using Finding the ideal material for a given device would re- an accurate and automated framework that we recently quire the cross-examination and co-optimization of many developed , nding several excellent 2D semiconductors. 26 28,36 di erent properties . Here, rather than focusing on one Compared to other methods , the approach of Ref. 13 particular application, we are interested in the physical has two main advantages in the context of this work: i) it features leading to good transport performance. We de- includes several tools to analyze band structure proper- scribe the tools and knowledge needed to explore large ties like valley structure or Fermi velocity and ii) the cal- databases with an informed, purposeful approach, and culation of conductivity is automated within the AiiDA 37,38 eventually identify the candidates with the highest con- framework , which is key to study many materials in arXiv:2007.16110v1 [cond-mat.mtrl-sci] 31 Jul 2020 2 a high-throughput fashion. These tools are available II. COMPUTATIONAL AND THEORETICAL FRAMEWORK to the community on the Archive section of the Materi- als Cloud , in support of the FAIR principles of open science and open data. A. Computational details High carrier densities are routinely obtained in mono- layer 2D semiconductors by eld-e ect doping, especially 40{42 using ionic-liquid gating . Here, we consider a xed First-principles calculations of structures, bands, 13 2 carrier density of n=p = 10 cm (electrons or holes). phonons and electron-phonon interactions are performed Electron-phonon interactions are computed in an elec- 51,52 with the Quantum ESPRESSO (QE) software, by trostatic framework including such doping , as well as combining density-functional theory (DFT) and density- screening from the induced free electrons or holes. This is functional perturbation theory (DFPT) within the gener- in contrast to other rst-principles studies, usually simu- alized gradient approximation as formulated by Perdew, lating electron-phonon interactions (EPI) in neutral ma- Burke, and Ernzerhof (PBE). 2D periodic-boundary terials, which would be valid only in the limit of vanish- conditions are applied and the electrostatics of a sym- ing carrier density. The capability to study explicitly the metric double-gate eld-e ect setup are simulated using high-doping regime is a valuable complement to this, and the approach described in Ref. 43. Open-boundary con- very relevant for monolayers operating in the degenerate ditions are important to properly describe polar-optical limit, when the chemical potential is close to or inside 54 55 phonons and screening . The eld-e ect setup is used the conduction or valence band (doping regimes are fur- to induce an electron or hole density with a default value ther discussed in App. A). In addition, its predictive ac- 13 2 of n=p = 10 cm . The simulation of this relatively curacy provides meaningful comparisons to experiments. high-doping regime is further discussed in App. A. Cal- Indeed, the high-doping regime is often used experimen- culations are managed using the AiiDA materials infor- tally to characterize the intrinsic properties of novel ma- 37,38 matics infrastructure . The AiiDA database contain- terials, because it allows to screen charged impurities. ing the provenance for the transport calculations and the tools necessary to reproduce this work are provided in the 35,39 In this study we nd excellent phonon-limited trans- Archive section of the Materials Cloud . The stan- port properties for electron-doped Bi SeTe , Bi Se , Bi- dard numerical setup for all ground state and phonon 2 2 2 3 ClTe, Sb SeTe , InSe, GaSe, AlLiTe and hole-doped calculations has 32 32 Monkhorst-Pack k-point grids to 2 2 2 phosphorene (P ), AuI, ClGaTe, and WSe , all show- sample the full Brillouin zone, 0.02 Ry cold smearing 4 2 ing mobilities in the range from few hundreds to few with SSSP pseudopotentials (eciency version 0.7) and thousands cm /Vs. Three of these 2D materials are energy cuto s recommended therein. PAW pseudopoten- 13{16 13 44 very well-known and studied (P , WSe , InSe ). tials were substituted because they are incompatible with 4 2 The Sb X and Bi X (X = Se, Te, S) compounds are the use of symmetry in the electron-phonon routines of 2 3 2 3 better known for the topological properties of their 3D QE. Other minor variations from the standard setup can 45,46 parents . Monolayers have been recently studied in be found in the AiiDA database provided. The use of 17,47,48 the context of phonon-limited transport , but only cold smearing allows k-point convergence while keeping within the approximate Tagaki formalism . The trans- a lower e ective temperature in the calculations, mak- port performance of monolayer Sb SeTe has been con- ing the free-carrier screening closer to what it would be 2 2 rmed experimentally . GaSe (along with InSe) was in real conditions. Spin-orbit interactions are included studied ab initio with a representation of electron- only for hole-doped WSe , where they play a signi cant phonon scattering that goes beyond deformation poten- role. For other materials, while spin-orbit interactions tial theory while this work was carried out. AlLiTe , Bi- may have an e ect on the band structure in general (e.g. ClTe, ClGaTe, and AuI are, to our knowledge, still new in on the band gap for BiSeTe ), they will not have large the context of 2D charge transport. Beyond the relatively consequences on the conductivity or mobility as long as large number of materials identi ed and the novelty of there is no signi cant energy splitting of valleys with op- some of them, this work most importantly provides visual posite spins (small variations may come from changes in and intuitive understanding of electron-phonon scatter- the e ective masses). The analysis starts from a database ing as well as a systematic and data-supported analysis of of band structures computed (non-self-consistently) in the features leading to good transport performance. We the neutral material on very ne electronic momenta con rm and extend previous remarks on the importance grids (about 90 by 90) to analyse the valley structure. of band properties such as number of valleys, e ective These bands are then recomputed with eld-e ect dop- masses or anisotropy; and show that these can be used ing for the selected materials. Phonon momenta are cho- for larger scale databases studies to shortlist prospective sen to include only relevant transitions, and the Boltz- conductors. On the other hand, we also show how di er- mann transport equation (BTE) is solved as described ences in the strength and angular dependency of electron- in Ref. 13. We improved the stability for the solution of phonon scattering still induce signi cant variations in the the BTE by using the velocity of each initial state as a conductivity of materials with similar band properties. ctitious electric eld direction [Eq. (11) of Ref. 13]. 3 B. Boltzmann transport equation WSe2 T = 300 K The solution of the BTE is brie y outlined here for 13 2 n/p = 10 cm convenience; more details can be found in Ref. 13. The (longitudinal) conductivity is computed as follows: 60 P4 holes electrons 1 dk @f (" ) 50 2 k = = 2e (v(k) u )  (k) Bi2SeTe2 (2) @" (1) Sb2SeTe2 where e is the Coulomb charge, and k; " ; v(k) repre- GaSe InSe sent electronic momenta (and band index implicitly), en- AlLiTe2 BiClTe ergies, and velocities, respectively. The unit vector u AuI ClGaTe Bi2Se3 points in the electric eld's direction and f is the Fermi- 0.0 0.5 1.0 1.5 2.0 2.5 3.0 Dirac distribution. Here  is an energy- and momentum- gap (eV) dependent variable that has the dimensions of time and solves the linearized BTE: FIG. 1. Computed conductivities and mobilities at room tem- 0 0 0 perature plotted against PBE gap for all the materials con- (1 f (k))v(k) u = P (1 f (k )) E kk sidered in this work. BTE is solved for a xed carrier density 13 2 of n=p = 10 cm (electrons or holes). 0 0 fv(k) u  (k) v(k ) u  (k )g E E (2) compute the transport properties of the most promising P 0 being the probability for an electron in state k to kk candidates. Electron-phonon scattering is analyzed in be scattered into state k . In the following we consider each monolayer. The sux \-e" or \-h" is attached to only phonon-induced scattering, so that: the materials' formula to indicate if either electron dop- 2 1 ing or hole doping is considered. P = jg j fn (" " ~! ) kk+q kk+q; q; k+q k q; ~ N +(n + 1)(" " + ~! )g: q; k+q k q; A. Prototypical high-conductivity materials: InSe (3) and P where q; ; ~! ; n are phonon momenta, mode index, q; q; We start by analyzing key features of two well-known energy and occupations, and g are the electron- kk+q; semiconductors with excellent transport performance, phonon coupling matrix elements. InSe on the electron side (InSe-e) and phosphorene on the hole side (P -h). In the following we will use repeatedly III. HIGH-CONDUCTIVITY 2D two di erent plots to visualize, understand and compare SEMICONDUCTORS the details of the transport properties of 2D materials, reported rst in Fig. 2 for InSe and P . On the left, a \velocity plot" shows the band structure in a format that We study in careful detail eleven monolayers obtained is relevant for transport. For each electronic state k (each from our database of 256 easily exfoliable materials with black dot) on the ne k-point grid used for solving the at most 6 atoms per unit cell , as available on the Ma- 34,35 BTE and within a certain energy range from the band terials Cloud and in the supplementary material of edge (set as the origin of the y-axis), the energy " is Ref. 23. The corresponding conductivity and mobility, plotted against the norm of the velocity jvj = j r " j, k k as a function of the PBE gap, are shown in Fig. 1. At given in atomic Rydberg units (ARU) . The spread in- this doping level a conductivity of a few e =h is already 4,8 dicates the anisotropy of the valley. The red scale of the considered good for a semiconductor, while the best 2 58,59 background is proportional to the derivative of the Fermi- graphene devices yield values around 500 e =h . The Dirac occupation at 300 K, thus highlighting the states materials presented in this work are in the intermedi- participating to transport with a non-vanishing contri- ary orders of magnitude, with conductivities of 10  100 bution to Eq. (1). The Fermi level is computed at room e =h, and mobilities in the same range as bulk silicon 0 2 13 2 2 temperature (i.e. such that f (" ; T )dk=(2 ) = 10 (400 cm /Vs for holes, 1400 cm /Vs for electrons). k In the following, we detail the exploration process cm for electrons, and similarly for holes). On the right that led to those materials. First, two well-known high- side, a \scattering plot" shows where and how easily a conductivity 2D semiconductors are analyzed to identify certain initial state can be scattered. The initial state representative band features. We then search for those at k , indicated by a black square, is chosen to be at in 23,34 features within the Materials Cloud database and the Fermi level (and in the transport direction, when rel- conductivity (e /h) mobility (cm /(Vs)) 4 evant). The grey shading shows the morphology of the direction is maximized. However, the DOS here is rela- valley(s) considered for transport. The red color scale tively larger, and the Fermi level stays closer to the band represents an e ective coupling constant g which ac- edge, where velocities are lower. The good performance eff counts for all phonon modes, their occupation at 300 K, of phosphorene thus relies on a somewhat more fragile and energy conservation, since this quantity can be mean- balance. ingfully compared between materials. In practice, g An essential feature to maximize both i) and ii) is to eff is de ned as the square root of the sum of the interpo- work with single valley materials. Single valley band lated jg j n ( or n + 1) over all  and q structures lead to a lower DOS, higher Fermi level and k ;k +q; q; q; in in that ful ll energy conservation during phonon absorption higher velocities. Even more compelling, one avoids in- (or emission) when the nal state at k + q falls into a tervalley scattering, which is usually stronger than in- in certain zone (the valley being tessellated into triangles). travalley scattering and hinders transport . The value given at the top of each scattering plot is the In a related work , several promising materials for lifetime of the initial state computed within the momen- ultra-short transistor devices have been identi ed. In tum relaxation time approximation: that situation, one optimizes the performance of the ma- terial in the ballistic limit and the role of electron-phonon 0 0 0 1 1 f (k ) v(k ) v(k) scattering is secondary, since the channels are usually be- = P 0  1 (4) kk 0 2 (k) 1 f (k) v(k) low the scattering length. Thus, maximizing both the velocity in the transport direction and the DOS is ex- Everything is computed at 300 K, with the correspond- tremely bene cial, and anisotropic and multivalley ma- ing chemical potential in f . terials are interesting. When electron-phonon scattering is turned on, multivalley band structures are not ideal, We now discuss qualitatively the features allowing while anisotropy can remain relevant. electron-doped InSe and hole-doped phosphorene to max- imize the conductivity in Eq. (1): this is a sum over electronic states weighted by the derivative of the Fermi- B. More like InSe: Steep and deep single-valleys Dirac occupation. The weight e ectively selects states around the chemical potential within an energy range As shown in Fig. 2, InSe has a sharp and deep single that scales with temperature, as represented in the left panels of Fig. 2. The rest of the integrand can be electron valley, with carrier velocities close to 6 ARU at 13 2 the Fermi level when n = 10 cm (roughly 6 times separated in two contributions: i) one from scattering (electron-phonon here) through the scattering lifetime  ; larger than the Fermi velocity in graphene,  1 ARU). Steep isotropic valleys are obviously characterized by low ii) and one from the velocity of the carriers v projected along the transport direction u . DOS, which means that the Fermi level quickly shifts away from the band edge with increasing carrier density. The scattering contribution i) is inversely proportional to the strength of the EPI and the amount of states avail- This allows to reach higher carrier velocities, but it often means that the Fermi level reaches other valleys in the able for scattering (as apparent in Eq. (4)), i.e. the band structure. In that case, the aforementioned bene ts density of states (DOS) within a certain energy win- dow around the initial states, depending on the energy of the single valley structure are lost. If the Fermi level is close to the edges of the next valleys, intervalley scat- of the phonons involved. We note, however, that the DOS contribution is essentially compensated when doing tering is activated and the new states populated are not helping to conduct due to their low velocity. Note that, the integral over electronic states to get the conductiv- ity, Eq. (1). To maximize  , the quantity to focus on is in a similar spirit, the suppression of intervalley scatter- ing via valley-engineering (e.g. through strain) has been thus the strength of the EPI, which should of course be minimized. Weak EPIs is a feature shared by InSe and proposed to enhance the transport properties of multi- valley materials such as arsenene. Thus, a steep valley phosphorene. There are di erent ways to maximize the velocity con- is not sucient to have good transport performance: it must also be deep. More precisely, higher-energy val- tribution ii); in general, one needs to maximize the ve- leys need to be far enough for the material to operate locity in the direction of transport. In the case of InSe, the velocity is isotropic and very large, with a small ef- e ectively as a single valley material for a doping of 10 cm . fective mass. Accordingly, the DOS is low and the Fermi level reaches high energies above the valley edge, to- We look for such valleys among the exfoliable mate- wards higher velocities, eventually saturating at a maxi- rials in our study set. We use band structures com- mal value in case of non-parabolic materials. The bene ts puted on very ne grids in the neutral materials and of a such a steep and deep single valley is that current select materials with positive phonons, a limited gap carrying states around the chemical potential have high E < 2:5 eV, a single valley (within 100 meV of the velocities, as also pointed out in Ref. 33. In the contrast- Fermi level), and a maximum Fermi velocity jv j > 6 max ing case of phosphorene, the valley is highly anisotropic. ARU. This allows us to identify BiClTe-e, AlLiTe -e, Provided one chooses the low e ective mass direction for GaSe-e, Bi SeTe -e, Bi STe -e, Bi Se -e, and Sb SeTe - 2 2 2 2 2 3 2 2 transport, the projection of the velocity in the transport e as promising candidates. Electron-phonon scattering 5 InSe = 61.201 fs 0.30 0.06 0.25 0.04 0.20 0.02 0.00 0.15 0.02 0.10 0.04 0.05 0.06 0.00 0 0 2 4 6 0.10 0.05 0.00 0.05 0.10 |v| (ARU) k (bohr ) = 204.02 fs 0.00 0.15 0.05 0.10 0.10 0.05 50 0.00 0.15 0.05 0.20 0.10 0.25 0.15 2 4 6 0.2 0.1 0.0 0.1 0.2 |v| (ARU) k (bohr ) FIG. 2. Transport properties of InSe and phosphorene (P ), two well-known good conductors. On the left, the \velocity plot" shows, for each electronic state k in the valley, the energy " from the band edge plotted against the norm of the velocity jvj = j r " j, in atomic Rydberg units (ARU). The color scale of the background is proportional to the derivative of the k k Fermi-Dirac distribution, which appears in Eq. (1). On the right, the \scattering plot" shows where an initial state (black square) can be scattered within the valley (grey shade). The color scale represents the e ective coupling constant g , which eff accounts for all phonon modes, their occupation at 300 K, and energy conservation. At the top,  is the scattering time of the initial state computed within the momentum relaxation time approximation, Eq. (4). is studied for these monolayers (except Bi STe -e, a bit higher velocities. Finally, AlLiTe -e ( = 18 e =h, see 2 2 2 redundant with respect to the other two members of Fig. 3) and BiClTe-e ( = 16 e =h) have slightly larger the Bi X family). The results are shown in Fig. 3 for velocities but stronger EPIs than InSe-e, making them 2 3 Bi SeTe -e and AlLiTe -e, and in the appendix for the slightly less conductive. Note that all these systems with 2 2 2 rest. The general similarity of the plots indicates at a sharp valleys are electron-doped, which may be rational- glance why these materials have similar transport proper- ized by considering that conduction bands are more often ties. The carrier velocities at the Fermi level, the strength made of delocalized non-bonding states which tend to be of EPI, and the scattering times are of similar orders of more dispersive. Despite the valleys being very similar magnitude. A closer look reveals the reasons underlying (isotropic with jv j  6  7 ARU), the conductivity max their precise ranking. BiSeTe -e, BiSe -e, and Sb SeTe - varies from 14 to 42 e =h. This points to the importance 2 3 2 2 e are very similar in terms of chemistry, band structure, of accounting for the details of EPIs to rank materials and phonons. BiSeTe -e (see Fig. 3) displays the high- accurately. est conductivity ( = 42 e =h), combining some of the We also note that the selection process does not guar- weakest EPI with the largest velocities. It is followed by antee to nd all the best conductors, as demonstrated Sb SeTe -e ( = 30 e =h) with similar EPI but slightly 2 2 by the example of WSe -h: TMDs were not selected smaller velocities. Bi Se -e ( = 14 e =h) also has a very 2 3 because they are not single-valley (also their Fermi ve- sharp valley, but the EPI is 2 to 3 times stronger. As locities would be too small). However, thanks to strong could be expected, electron-phonon scattering in GaSe-e spin-orbit interactions, the hole side of TMDs can be con- 2 2 ( = 25 e =h) is similar to InSe-e ( = 20 e =h), with sidered to be in the steep and deep single valley category. scattering times di ering by only  10%. GaSe-e owes Indeed, as is well known, the hole valleys associated to its higher conductivity mostly to a sharper valley and opposite spin textures split very strongly in energy, of the Energy (from edge) (eV) Energy (from edge) (eV) k (bohr ) k (bohr ) g (meV) g (meV) eff eff 6 Bi SeTe 2 2 = 110.16 fs 0.06 0.30 0.04 120 0.25 0.02 0.20 0.00 0.15 0.02 0.10 0.04 0.05 0.06 0.00 0 0 2 4 6 0.10 0.05 0.00 0.05 0.10 |v| (ARU) k (bohr ) AlLiTe = 48.003 fs 0.4 0.075 0.050 0.3 0.025 0.2 0.000 0.025 0.1 0.050 0.075 0.0 0 0 2 4 6 8 0.10 0.05 0.00 0.05 0.10 |v| (ARU) k (bohr ) 2 2 FIG. 3. Transport properties of Bi SeTe ( = 42 e =h) and AlLiTe ( = 18 e =h), showing the velocity (left) and scattering 2 2 2 (plots) as described in Fig. 2. Both monolayers belong to a category of materials with steep and deep single valleys in the conduction band. Similar plots are given in App. B for other monolayers in this category: BiClTe-e, GaSe-e, Bi STe -e, 2 2 Bi Se -e, and Sb SeTe -e. 2 3 2 2 WSe = 575.58 fs 0.00 0.6 0.05 0.10 0.4 0.15 0.2 0.20 0.25 0.0 0.30 1 2 3 4 0.0 0.2 0.4 0.6 0.8 1.0 |v| (ARU) k (bohr ) FIG. 4. Transport properties of WSe -h ( = 81 e =h), showing the velocity (left) and scattering (plots) as described in Fig. 2. Thanks to spin-orbit interactions, weak spin- ip scattering, and the e ective absence of any valley at , spins travel in parallel channels not connected to each other. order of 100 meV, making the lower valleys irrelevant for inating a potential intervalley scattering channel. Fur- transport. One e ectively obtains two valleys at K and thermore, as shown in Fig. 4, the intervalley scattering, K' with opposite spin textures. WSe -h, at least within associated with spin- ip EPIs, is weak (< 10 meV) com- our computational framework, also has the advantage pared to intravalley, spin-conserving EPIs ( 50 meV). that the edge of the valence band at is quite low, elim- This implies that the two valleys are e ectively decou- Energy (from edge) (eV) Energy (from edge) (eV) Energy (from edge) (eV) k (bohr ) k (bohr ) k (bohr ) g (meV) g (meV) g (meV) eff eff eff 7 pled as far as phonon-limited transport is concerned and the monoatomic nature of P , eliminating all Born e ec- transport is similar to the single valley case, except op- tive charge and piezoelectric couplings), one can observe posite spins travel in separate channels located at dif- in P a predominance of \side-scattering": states in the ferent points in the Brillouin zone. Comparing with the direction of transport are mostly scattered to states on isotropic single-valley materials discussed above, WSe - the sides of the valley with velocities perpendicular to the h has velocities half as small but EPIs at least 3 times direction of transport. As can be seen from Eq. (4), this weaker, leading to the largest conductivity ( = 81 e =h). leads to a longer scattering times, since the term involv- This high value is due to the (e ective) absence of the ing a scalar product of the velocities vanishes. The con- valley. Indeed, opposite spins are degenerate at , and tribution of the DOS from the integral of the conductivity if the valley were accessible for scattering, intervalley usually cancels out with the integral over available scat- scattering from both K and K' would be quite strong. In tering states. This is not the case here because we bene t addition, as shown in Ref. 61, multi-valley occupations from the larger weight of high velocity states in the con- enhances the intravalley coupling to the homopolar op- ductivity integral while having less dense, low-velocity tical phonon mode by making free-carrier screening in- states in the scattering integral. Thus, \side scattering" ecient. At any rate, the extraordinarily weak EPI in de nitely brings the performance of P -h from great to WSe -h is tied to subtle spin-orbit and screening e ects exceptional. ClGaTe-h is similar to P in terms of veloc- 2 4 that are dicult to predict and highlights the limitations ities, but the EPI is stronger and, importantly, mostly of simple selection processes based solely on the band backscattering. The scattering time is consequently 10 structure. times larger. AuI-h has lower velocities and higher EPI. While targeting the band features of phosphorene has allowed us to nd other excellent candidates, we see that the essence of its exceptional performance comes C. More like phosphorene: anisotropic single valleys from something much harder to identify at the level of a database: the anisotropy of its electron-phonon inter- actions. Thus, phosphorene is at the same time both a In contrast with the fairly isotropic band structures of prototype and an example of the limitations of selection InSe, the case of phosphorene showed that anisotropy can processes based on band structures. also lead to good performance. So, we look for similar materials in our database, having stable phonons, single valleys and small band gap at the PBE level (E < 2:5 IV. CONCLUSIONS eV). This time we lter materials keeping materials that max combine high velocity ratios ( < 1:7) and a decent min maximum velocity (v > 2:0 ARU) at the Fermi level. State-of-the-art density-functional perturbation theory max These search criteria allow us to identify AuI-h, and and the Boltzmann transport equation are used to study ClGaTe-h as well as the electron-side of P -e as promising 4 the outstanding transport properties of several 2D semi- candidates. The electron side of phosphorene is a similar, conductors. Focusing on conductivity at a xed den- 13 2 less pronounced version of the hole side, and was already sity of n=p = 10 cm , the present results o er a studied in our previous work . Thus, we focus here on complementary perspective with respect to most rst- 2 2 ClGaTe-h ( = 13 e =h) and AuI-h ( = 14 e =h). principles calculations valid in the zero carrier density The results of the electron-phonon calculations are limit. We provide a detailed analysis of electron-phonon plotted in Fig. 5. The spread of the velocity traces in- scattering in two well-known high-conductivity systems: dicates high anisotropy, the velocity varying by a factor electron-doped InSe and hole-doped phosphorene. While 2 at a given energy. The high-velocity direction prefer- they share some features {like weak EPI and a single- able for transport is x for ClGaTe and x + y for AuI. valley electronic structure{ they exemplify two di erent The anisotropy allows one to bene t from many high ve- strategies to maximize the conductivity. InSe's high- locity states, thanks to a atter band in the direction velocity, isotropic valley can be exploited thanks to the perpendicular to transport. However, if the band is too fact that the next valleys are much higher in energy. at and the DOS too high, the Fermi level stays close Phosphorene, instead, owes its excellent transport perfor- to the band edges and the velocities are too low. Here, mance to the anisotropy of both its band structure and velocities of the order of 3  4 ARU ensure good trans- electron-phonon scattering. Analyzing the band proper- port properties. While ClGaTe-h and AuI-h share the ties of around  150 small stable semiconductors with 6 anisotropic character of phosphorene, their conductivity atoms or less in the unit-cell, from the Materials Cloud, remains lower, which in part re ects the fact that the we identify systems with band features similar to either good transport properties of anisotropic materials rely InSe or phosphorene. We nd large phonon-limited con- on a more fragile balance of features. ductivities for electron-doped Bi SeTe , Bi Se , BiClTe, 2 2 2 3 A closer look at the scattering plots in Fig. 5, con- Sb SeTe , AlLiTe, and GaSe, as well as hole-doped AuI, 2 2 trasted with phosphorene in Fig. 2 further reveals the ClGaTe, and WSe . These results con rm that the band reasons behind phosphorene's superior conductivity. In structure landscape plays an important role in determin- addition to generally weaker EPIs (which is partly due to ing transport and shows that seeking peculiar features 8 ClGaTe = 38.744 fs 0.00 0.15 0.05 0.10 0.10 0.05 0.15 0.00 0.20 0.05 0.10 0.25 0.15 0.30 1 2 3 4 5 0.2 0.1 0.0 0.1 0.2 |v| (ARU) k (bohr ) AuI = 34.346 fs 0.00 0.10 0.05 0.05 0.10 0.00 0.15 0.05 0.20 0.10 0 1 2 3 4 0.15 0.10 0.05 0.00 0.05 0.10 0.15 |v| (ARU) k (bohr ) FIG. 5. Transport properties of ClGaTe and AuI showing the velocity (left) and scattering (right) plots as described in Fig. 2. Anisotropic valleys imply there is one optimal transport direction, corresponding to higher velocities. Initial states (marked with a black square) are chosen to be in that direction. in the electronic structure does lead to high-performance was awarded by PRACE on Marconi at Cineca, Italy materials. Nevertheless, we also show how the details of (project id. 2016163963). Computational resources have the strength and angular dependency of electron-phonon been provided by the Consortium des quipements de Cal- scattering play a critical role in ranking those materials cul Intensif (CCI), funded by the Fonds de la Recherche with respect to each other. Scienti que de Belgique (F.R.S.-FNRS) under Grant No. 2.5020.11 and by the Walloon Region. T.S. acknowledges support from the University of Liege under Special Funds ACKNOWLEDGEMENTS: for Research, IPD-STEMA Programme. M.G. acknowl- edges support by the Italian Ministry for University and Research through the Levi-Montalcini program and by The authors are grateful to Davide Campi for shar- the Swiss National Science Foundation (SNSF) through ing the initial band structures. This work has been in the Ambizione program (grant PZ00P2 174056). part supported by NCCR MARVEL. Simulation time 1 4 S. Z. Butler et al., Progress, challenges, and opportunities B. Radisavljevic and A. Kis, Mobility engineering and a in two-dimensional materials beyond graphene., ACS nano metal{insulator transition in monolayer MoS , Nature Ma- 7, 2898 (2013). terials 12, 815 (2013). 2 5 D. Akinwande, N. Petrone, and J. Hone, Two-dimensional B. Radisavljevic, A. Radenovic, J. Brivio, V. Giacometti, exible nanoelectronics , Nature Communications 5, 5678 and A. Kis, Single-layer MoS transistors , Nature nan- (2014). otechnology 6, 147 (2011). 3 6 M. Chhowalla, D. Jena, and H. Zhang, Two-dimensional X. Li, Z. Yu, X. Xiong, T. Li, T. Gao, R. Wang, R. Huang, semiconductors for transistors , Nature Reviews Materials and Y. Wu, High-speed black phosphorus eld-e ect tran- 1, 16052 (2016). sistors approaching ballistic limit , Science Advances 5 Energy (from edge) (eV) Energy (from edge) (eV) 1 k (bohr ) k (bohr ) g (meV) g (meV) eff eff 9 (2019). Scienti c Reports 7, 5179 (2017). 7 23 L. Li et al., Black phosphorus eld-e ect transistors., Nat. N. Mounet et al., Two-dimensional materials from high- Nanotechnol. 9, 372 (2014). throughput computational exfoliation of experimentally L. Tao, E. Cinquanta, D. Chiappe, C. Grazianetti, M. Fan- known compounds , Nature Nanotechnology 13, 246 (2018). ciulli, M. Dubey, A. Molle, and D. Akinwande, Silicene S. Haastrup et al., The Computational 2D Materials eld-e ect transistors operating at room temperature , Na- Database: high-throughput modeling and discovery of ture Nanotechnology 10, 227 (2015). atomically thin crystals , 2D Materials 5, 042002 (2018). 9 25 X. Li, J. T. Mullen, Z. Jin, K. M. Borysenko, M. Buon- J. Zhou et al., 2DMatPedia, an open computational giorno Nardelli, and K. W. Kim, Intrinsic electrical trans- database of two-dimensional materials from top-down and port properties of monolayer silicene and MoS from rst bottom-up approaches , Scienti c Data 6, 86 (2019). principles , Physical Review B - Condensed Matter and Ma- C. Klinkert, A. Szab o, C. Stieger, D. Campi, N. Marzari, terials Physics 87, 115418 (2013). and M. Luisier, 2-D Materials for Ultrascaled Field-E ect W. Li, Electrical transport limited by electron-phonon cou- Transistors: One Hundred Candidates under the Ab Initio pling from Boltzmann transport equation: An ab initio Microscope , ACS Nano 14, 8605 (2020). study of Si, Al, and MoS , Physical Review B 92, 075405 2 F. Giustino, Electron-phonon interactions from rst prin- (2015). ciples , Reviews of Modern Physics 89, 015003 (2017). K. Kaasbjerg, K. S. Thygesen, and K. W. Jacobsen, S. Ponc e, W. Li, S. Reichardt, and F. Giustino, First- Phonon-limited mobility in n-type single-layer MoS from principles calculations of charge carrier mobility and con- rst principles , Physical Review B - Condensed Matter ductivity in bulk semiconductors and two-dimensional ma- and Materials Physics 85, 115317 (2012). terials , Reports on Progress in Physics 83, 036501 (2020). T. Gunst, T. Markussen, K. Stokbro, and M. Brandbyge, C.-H. Park, N. Bonini, T. Sohier, G. Samsonidze, B. Kozin- First-principles method for electron-phonon coupling and sky, M. Calandra, F. Mauri, and N. Marzari, Electron- electron mobility: Applications to two-dimensional materi- Phonon Interactions and the Intrinsic Electrical Resistivity als , Physical Review B 93, 035414 (2016). of Graphene., Nano letters 14, 1113 (2014). T. Sohier, D. Campi, N. Marzari, and M. Gibertini, Mo- T. Sohier, M. Calandra, C.-H. C.-H. Park, N. Bonini, bility of 2D materials from rst principles in an accurate N. Marzari, and F. Mauri, Phonon-limited resistivity of and automated framework , Physical Review Materials 2, graphene by rst-principles calculations: Electron-phonon 114010 (2018). interactions, strain-induced gauge eld, and Boltzmann A. N. Rudenko, S. Brener, and M. I. Katsnelson, Intrinsic equation , Physical Review B 90, 125414 (2014). Charge Carrier Mobility in Single-Layer Black Phospho- S. Ponc e, E. Margine, C. Verdi, and F. Giustino, EPW: rus , Physical Review Letters 116, 246401 (2016). Electron{phonon coupling, transport and superconducting Y. Trushkov and V. Perebeinos, Phonon-limited carrier properties using maximally localized Wannier functions , mobility in monolayer black phosphorus , Physical Review Computer Physics Communications 209, 116 (2016). B 95, 075436 (2017). J.-J. Zhou, J. Park, I.-T. Lu, I. Maliyov, X. Tong, and G. Gaddemane, W. G. Vandenberghe, M. L. Van De Put, M. Bernardi, Perturbo: a software package for ab initio S. Chen, S. Tiwari, E. Chen, and M. V. Fischetti, Theoreti- electron-phonon interactions, charge transport and ultra- cal studies of electronic transport in monolayer and bilayer fast dynamics , arXiv:2002.02045 (2020). phosphorene: A critical overview , Physical Review B 98 L. Cheng, C. Zhang, and Y. Liu, The Optimal Electronic (2018). Structure for High-Mobility 2D Semiconductors: Excep- B. Wang, X. Niu, Y. Ouyang, Q. Zhou, and J. Wang, Ul- tionally High Hole Mobility in 2D Antimony , Journal of trathin Semiconducting Bi Te S and Bi Te Se with High 2 2 2 2 the American Chemical Society 141, 16296 (2019). Electron Mobilities , The Journal of Physical Chemistry https://www.materialscloud.org/discover/2dstructures/. Letters 9, 487 (2018). L. Talirz et al., Materials Cloud, a platform for open com- S. Takagi, A. Toriumi, M. Iwase, and H. Tango, On the putational science , arXiv:2003.12510 (2020). universality of inversion layer mobility in Si MOSFET's: G. Brunin et al., Phonon-limited electron mobility in Part I-e ects of substrate impurity concentration , IEEE Si, GaAs and GaP using plane waves and Bloch states , Transactions on Electron Devices 41, 2357 (1994). arXiv:2002.00630 (2020). G. Gaddemane, S. Gopalan, M. Van de Put, and M. V. G. Pizzi, A. Cepellotti, R. Sabatini, N. Marzari, and Fischetti, Limitations of ab initio methods to predict the B. Kozinsky, AiiDA: automated interactive infrastructure electronic-transport properties of two-dimensional materi- and database for computational science , Computational als: The computational example of 2H-phase transition Materials Science 111, 218 (2016). metal dichalcogenides , arXiv:1912.03795 (2019). S. P. Huber et al., AiiDA 1.0, a scalable computational in- G. Cheon, K.-A. N. Duerloo, A. D. Sendek, C. Porter, frastructure for automated reproducible work ows and data Y. Chen, and E. J. Reed, Data Mining for New Two- provenance , arXiv:2003.12476 (2020). and One-Dimensional Weakly Bonded Solids and Lattice- https://doi.org/10.24435/materialscloud:fr-r0. Commensurate Heterostructures , Nano Letters 17, 1915 Y. Zhang, J. Ye, Y. Matsuhashi, and Y. Iwasa, Ambipo- (2017). lar MoS Thin Flake Transistors , Nano Letters 12, 1136 M. Ashton, J. Paul, S. B. Sinnott, and R. G. Hennig, (2012). Topology-Scaling Identi cation of Layered Solids and Sta- D. Braga, I. Guti errez Lezama, H. Berger, and A. F. Mor- ble Exfoliated 2D Materials , Physical Review Letters 118, purgo, Quantitative Determination of the Band Gap of 106101 (2017). WS with Ambipolar Ionic Liquid-Gated Transistors , Nano K. Choudhary, I. Kalish, R. Beams, and F. Tavazza, High- Letters 12, 5218 (2012). throughput Identi cation and Characterization of Two- dimensional Materials using Density functional theory , 10 42 59 J. T. Ye, Y. J. Zhang, R. Akashi, M. S. Bahramy, R. Arita, L. Banszerus et al., Extraordinary high room-temperature and Y. Iwasa, Superconducting Dome in a Gate-Tuned carrier mobility in graphene-WSe heterostructures , Band Insulator , Science 338, 1193 (2012). arXiv:1909.09523 (2019). 43 60 T. Sohier, M. Calandra, and F. Mauri, Density functional Atomic Rydberg units are the most convenient units for ve- perturbation theory for gated two-dimensional heterostruc- locity in materials. Typical Fermi velocities in the database tures: Theoretical developments and application to exu- {including graphene{ are on the order of 1 ARU, which ral phonons in graphene , Physical Review B 96, 075448 corresponds to  10 m/s. (2017). T. Sohier, E. Ponomarev, M. Gibertini, H. Berger, W. Li, S. Ponc e, and F. Giustino, Dimensional Crossover N. Marzari, N. Ubrig, and A. F. Morpurgo, Enhanced in the Carrier Mobility of Two-Dimensional Semiconduc- Electron-Phonon Interaction in Multivalley Materials , tors: The Case of InSe , Nano Letters 19, 1774 (2019). Physical Review X 9, 031019 (2019). 45 62 W. Zhang, R. Yu, H. J. Zhang, X. Dai, and Z. Fang, First- N. Ma and D. Jena, Charge scattering and mobility in principles studies of the three-dimensional strong topologi- atomically thin semiconductors , Physical Review X 4, cal insulators Bi Te , Bi Se and Sb Te , New Journal of 011043 (2014). 2 3 2 3 2 3 Physics 12, 065013 (2010). C. Verdi, F. Caruso, and F. Giustino, Origin of the S. Jafarpisheh, A. Ju, K. Janen, T. Taniguchi, K. Watan- crossover from polarons to Fermi liquids in transition metal abe, C. Stampfer, and B. Beschoten, Reducing the Impact oxides , Nature Communications 8, 15769 (2017). of Bulk Doping on Transport Properties of Bi-Based 3D P. Bogulsawski and J. Mycielski, Is the deformation poten- Topological Insulators , Physica Status Solidi (B) Basic Re- tial in semiconductors screened by free carriers? , Journal search , pssb.202000021 (2020). of Physics C: Solid State Physics 10, 2413 (1977). 47 65 Y. Liu, Y. Xu, Y. Ji, and H. Zhang, Monolayer P. Boguslawski, Screening of the deformation potential by Bi Se Te : novel two dimensional semiconductors with free electrons in the multivalley conduction band , Journal 2 3x x excellent stability and high electron mobility , Physical of Physics C: Solid State Physics 10, L417 (1977). Chemistry Chemical Physics 22, 9685 (2020). P. Boguslawski and J. Mycielski, Free-electron screening of C. Tang, L. Zhang, C. Zhang, J. Macleod, K. Ostrikov, short-range scattering potentials in semiconductors , Jour- and A. Du, Highly stable two-dimensional gold selenide nal of Physics C: Solid State Physics 13, 1019 (1980). with large in-plane anisotropy and ultrahigh carrier mo- M. Verstraete and X. Gonze, Smearing scheme for nite- bility , Nanoscale Horizons 5, 366 (2020). temperature electronic-structure calculations , Physical Re- H. Qu, W. Zhou, S. Guo, Z. Li, Y. Wang, and S. Zhang, view B 65, 035111 (2001). Ballistic Quantum Transport of Sub10 nm 2D Sb Te Se J. Ma, D. Xu, R. Hu, and X. Luo, Examining two- 2 2 Transistors , Advanced Electronic Materials 5, 1900813 dimensional Fr ohlich model and enhancing the electron (2019). mobility of monolayer InSe by dielectric engineering , Jour- J. Chen, X. Tan, P. Lin, B. Sa, J. Zhou, Y. Zhang, C. Wen, nal of Applied Physics 128, 035107 (2020). and Z. Sun, Comprehensive understanding of intrinsic mo- bility in the monolayers of III-VI group 2D materials , Physical Chemistry Chemical Physics 21, 21898 (2019). P. Giannozzi et al., QUANTUM ESPRESSO: a modular and open-source software project for quantum simulations of materials., Journal of Physics: Condensed Matter 21, 395502 (2009). P. Giannozzi et al., Advanced capabilities for materials modelling with Quantum ESPRESSO , Journal of Physics Condensed Matter 29, 465901 (2017). J. P. Perdew, K. Burke, and M. Ernzerhof, Generalized Gradient Approximation Made Simple , Physical Review Letters 77, 3865 (1996). T. Sohier, M. Gibertini, M. Calandra, F. Mauri, and N. Marzari, Breakdown of Optical Phonons' Splitting in Two-Dimensional Materials , Nano Letters 17, 3758 (2017). T. Sohier, M. Calandra, and F. Mauri, Density-functional calculation of static screening in two-dimensional materi- als: The long-wavelength dielectric function of graphene , Physical Review B 91, 165428 (2015). N. Marzari, D. Vanderbilt, A. De Vita, and M. C. Payne, Thermal contraction and disordering of the al(110) sur- face , Physical Review Letters (1999). G. Prandini, A. Marrazzo, I. E. Castelli, N. Mounet, and N. Marzari, Precision and eciency in solid-state pseu- dopotential calculations , npj Computational Materials 4, 72 (2018). L. Wang et al., One-dimensional electrical contact to a two- dimensional material., Science 342, 614 (2013). 11 Appendix A: Comments on doping-dependent phonon interactions. If the strength of EPIs are con- transport performance stant, simple models using quadratic bands and elastic scattering show that the mobility is roughly constant up to fairly large dopings and then decreases. Consider- The mobility is a typical gure of merit for \trans- ing inelastic scattering increases mobility at small dop- port performance": It depends on doping (i.e. the carrier ing because there are no more available states for phonon density induced by eld e ects), and the relevant dop- emission close to the band edge. In any case, maximiz- ing range might vary with the application. Most rst- ing  implies maximizing  . However, for the ma- 13 0 principles computations of mobility are done in the zero terials considered here, EPIs are likely to be at least carrier density limit  = lim . Here, instead, we 0 n!0 partially doping-dependent via free-carrier screening. In- focus on 2D semiconductors with high conductivity  at deed, those are all single-valley materials, implying that 13 2 a xed carrier density of n=p = 10 cm . Of course, scattering is dominated by momenta smaller than the size by optimizing  we also maximize the mobility  = en=p of the Fermi surface, where free carriers screening is e- at this particular density, denoted with  . It should be 13 cient. If all EPI were sensitive to free-carrier screening, highlighted that, in general,  6=  , and the variation 0 13 we would see an opposite trend, with mobility increas- of the mobility in between these two doping regimes is ing as a function of doping , as free carriers screen the not obvious to predict. scattering sources.  might be then signi cantly lower than the  computed here. In practice, the magnitude The low (but nite) doping regime is very challeng- of this trend will depend on which EPIs are sensitive to ing to simulate realistically. The chemical potential is screening and how strong the bare EPIs are. below the band edge and depends strongly on temper- The electron-doped materials studied here (InSe, ature. One would ideally run one simulation per tem- GaSe, Bi SeTe , Sb SeTe ) have strong Born e ec- 2 2 2 2 perature, with an electronic smearing corresponding to tive charges and Fr ohlich couplings, which is screening- this temperature and an accordingly dense grid of k- sensitive and sharply increases at small momenta. For points. Unfortunately, even room temperature corre- InSe, we compute   490 cm =Vs, compared to sponds to a very low electronic smearing compared to 100 cm =Vs computed in Refs. 44 and 68 including standard DFT and DFPT calculations, resulting in very polar e ects and  = 488 cm =Vs when the Fr ohlich dense grids of k-points and prohibitively expensive cal- coupling is suppressed (as due to screening). Similar culations (especially when studying many materials as trends are expected for the other electron-doped materi- in this work). This issue is usually circumvented by als in this work. Whether  or  is more relevant to 0 13 simulating the neutral system and computing only  . a certain operating doping range depends on the critical Sometimes, the most obvious consequence of doping, carrier density at which free-carrier screening becomes ef- i.e. screening form free carriers, is added as an analyt- 62,63 cient. In 2D and assuming a constant density of states ical post-processing correction , but which scattering h i F C k T sources should be screened by free carriers, and how, has B per area D, one can derive n = k TD ln 1 + e 64{66 been debated . Moreover, eld e ects and screen- where k T is the thermal energy,  the chemical po- B F ing can have non-trivial consequences, as demonstrated tential (entering the Fermi-Dirac distribution) and " the 61 43 in TMDs or in graphene . In this work we choose bottom of the conduction band. If we estimate the onset to fully account for doping in the calculations, but at of free-carrier screening as  " > k T (when the F C B relatively high carrier density. This allows us to per- occupations are not dominated by the tail of the Fermi- form more realistic calculations, easier to converge, with 11 2 Dirac distribution), we obtain n > 5 10 cm for all a chemical potential within the band and a well-de ned the electron-doped materials studied here. Fermi surface. We use a smearing that is large compared The hole-doped materials studied here have weaker to room temperature in order to have accurate results Born e ective charges, but that is not to say that the with a ordable k-point grids, but the \cold" nature of remaining EPIs are not sensitive to screening; neverthe- the smearing allows to gets closer to room tempera- less, smaller variations of the mobility are to be expected. ture conditions. The e ects of smearing can be signif- icant at low doping, when the chemical potential is in the gap; at the high doping levels considered here, how- Appendix B: Additional electron-phonon scattering ever, we expect the calculations to be representative of data room-temperature conditions. Note that an alternative consistent approach to smear a nite temperature Fermi- Fig. 6 shows the transport properties of GaSe-e, Dirac distribution has been put forward , but it is not Bi Se -e, Sb SeTe -e, and BiClTe-e. 2 3 2 2 implemented in our computational framework. Fig. 7 shows the velocity plot for Bi STe -e, for which 2 2 To predict the behavior of the mobility as doping de- phonon-limited transport was not computed, but is ex- creases, one needs to account for the variation of electron- pected to be similar to Bi SeTe -e. 2 2 12 GaSe = 71.619 fs 0.075 0.30 0.050 0.25 0.025 0.20 0.000 0.15 0.025 0.10 0.050 0.05 0.075 0.00 0 0 2 4 6 8 0.10 0.05 0.00 0.05 0.10 |v| (ARU) k (bohr ) Bi Se 2 3 = 37.318 fs 0.06 0.30 0.04 0.25 0.02 0.20 0.00 0.15 0.02 0.10 0.04 0.05 0.06 0.00 0 0 2 4 6 0.10 0.05 0.00 0.05 0.10 |v| (ARU) k (bohr ) Sb SeTe 2 2 = 90.739 fs 0.35 0.075 0.30 0.050 0.25 0.025 0.20 0.000 0.15 0.025 0.10 0.050 0.05 25 0.075 0.00 0 0 1 2 3 4 5 6 0.10 0.05 0.00 0.05 0.10 |v| (ARU) k (bohr ) BiClTe = 44.621 fs 0.30 0.06 0.25 0.04 0.20 0.02 125 0.00 0.15 0.02 0.10 0.04 0.05 0.06 0.00 0 0 2 4 6 0.10 0.05 0.00 0.05 0.10 |v| (ARU) k (bohr ) FIG. 6. Velocity and scattering plots of GaSe-e, Bi Se -e, Sb SeTe -e, and BiClTe-e, in order of decreasing conductivity. 2 3 2 2 Energy (from edge) (eV) Energy (from edge) (eV) Energy (from edge) (eV) Energy (from edge) (eV) 1 k (bohr ) k (bohr ) 1 1 k (bohr ) k (bohr ) y y g (meV) g (meV) g (meV) g (meV) eff eff eff eff 13 Bi STe 2 2 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 0 1 2 3 4 5 6 7 |v| (ARU) FIG. 7. Velocity plot of neutral Bi STe -e. Electron-phonon scattering was not computed but it is expected to be similar to 2 2 Bi SeTe -e. 2 2 Energy (from edge) (eV) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Condensed Matter arXiv (Cornell University)

Profiling novel high-conductivity 2D semiconductors

Condensed Matter , Volume 2020 (2007) – Jul 31, 2020

Loading next page...
 
/lp/arxiv-cornell-university/profiling-novel-high-conductivity-2d-semiconductors-Rc3C5lhbUg
ISSN
2053-1583
eISSN
ARCH-3331
DOI
10.1088/2053-1583/abc5d0
Publisher site
See Article on Publisher Site

Abstract

Pro ling novel high-conductivity 2D semiconductors 1, 2 3, 4, 2 2 Thibault Sohier, Marco Gibertini, and Nicola Marzari nanomat/QMAT/CESAM and European Theoretical Spectroscopy Facility, University of Liege (Uliege), Belgium Theory and Simulation of Materials (THEOS), and National Centre for Computational Design and Discovery of Novel Materials (MARVEL), Ecole Polytechnique F ed erale de Lausanne, CH-1015 Lausanne, Switzerland Dipartimento di Fisica Informatica e Matematica, Universit a di Modena e Reggio Emilia, Via Campi 213/a, I-41125 Modena, Italy Department of Quantum Matter Physics, University of Geneva, CH-1211 Geneva, Switzerland (Dated: August 3, 2020) When complex mechanisms are involved, pinpointing high-performance materials within large databases is a major challenge in materials discovery. We focus here on phonon-limited conductivi- ties, and study 2D semiconductors doped by eld e ects. Using state-of-the-art density-functional perturbation theory and Boltzmann transport equation, we discuss 11 monolayers with outstanding transport properties. These materials are selected from a computational database of exfoliable ma- terials providing monolayers that are dynamically stable and that do not have more than 6 atoms per unit cell. We rst analyze electron-phonon scattering in two well-known systems: electron-doped InSe and hole-doped phosphorene. Both are single-valley systems with weak electron-phonon inter- actions, but they represent two distinct pathways to fast transport: a steep and deep isotropic valley for the former and strongly anisotropic electron-phonon physics for the latter. We identify similar features in the database and compute the conductivities of the relevant monolayers. This process yields several high-conductivity materials, some of them only very recently emerging in the litera- ture (GaSe, Bi SeTe , Bi Se , Sb SeTe ), others never discussed in this context (AlLiTe , BiClTe, 2 2 2 3 2 2 2 ClGaTe, AuI). Comparing these 11 monolayers in detail, we discuss how the strength and angular dependency of the electron-phonon scattering drives key di erences in the transport performance of materials despite similar valley structure. We also discuss the high conductivity of hole-doped WSe , and how this case study shows the limitations of a selection process that would be based on band properties alone. I. INTRODUCTION ductivity. We focus here on room-temperature phonon-limited electronic transport and search across 2D semiconduc- Two-dimensional (2D) semiconductors with excellent tors. First-principles methods have proven quite use- intrinsic transport properties would be bene cial to many ful and successful in predicting physics and properties in 1{3 applications . Some well-known 2D materials like 28 29,30 this regime and for 2D materials , provided one is transition-metal dichalcogenides (TMDs), phosphorene 19 31,32 aware of its limits . Although dedicated codes exist , or silicene have been extensively studied both experi- performing state-of-the-art rst-principles transport sim- 4,5 6,78 9{13 mentally (MoS , Si ) and theoretically (MoS , 2 2 ulations on large sets of materials remains nevertheless 13{16 17 P ). Other candidates have been proposed us- a challenge, and few works tackle more than one mate- ing approximate deformation potential models and the rial at a time. Yet, an overall panorama on the property Takagi formula, but it has been shown that such ap- landscape would be very valuable to materials design. proaches su er from limited reliability . The full poten- For example, in a previous work involving 5 materials , tial of 2D materials for future device applications could we have shown the importance of intervalley scattering, be much broader than the dozen materials currently while the authors of Ref. 33 have studied 4 hexagonal under extensive experimental investigation, since many elemental materials to show the bene ts of a sharp and more monolayers have been predicted to be either exfoli- deep single valley. 20{23 able from experimentally known layered materials In this work, we explore our portfolio of exfoliable 2D 24,25 or synthesizable . Such computational collections 23 34,35 materials taken from the Materials Cloud , limit- of prospective 2D materials would likely contain some ing ourselves to at most 6 atoms per unit cell. We se- promising candidates for electronic transport in various lect the most promising systems from a band structure contexts. analysis, and then study their transport properties using Finding the ideal material for a given device would re- an accurate and automated framework that we recently quire the cross-examination and co-optimization of many developed , nding several excellent 2D semiconductors. 26 28,36 di erent properties . Here, rather than focusing on one Compared to other methods , the approach of Ref. 13 particular application, we are interested in the physical has two main advantages in the context of this work: i) it features leading to good transport performance. We de- includes several tools to analyze band structure proper- scribe the tools and knowledge needed to explore large ties like valley structure or Fermi velocity and ii) the cal- databases with an informed, purposeful approach, and culation of conductivity is automated within the AiiDA 37,38 eventually identify the candidates with the highest con- framework , which is key to study many materials in arXiv:2007.16110v1 [cond-mat.mtrl-sci] 31 Jul 2020 2 a high-throughput fashion. These tools are available II. COMPUTATIONAL AND THEORETICAL FRAMEWORK to the community on the Archive section of the Materi- als Cloud , in support of the FAIR principles of open science and open data. A. Computational details High carrier densities are routinely obtained in mono- layer 2D semiconductors by eld-e ect doping, especially 40{42 using ionic-liquid gating . Here, we consider a xed First-principles calculations of structures, bands, 13 2 carrier density of n=p = 10 cm (electrons or holes). phonons and electron-phonon interactions are performed Electron-phonon interactions are computed in an elec- 51,52 with the Quantum ESPRESSO (QE) software, by trostatic framework including such doping , as well as combining density-functional theory (DFT) and density- screening from the induced free electrons or holes. This is functional perturbation theory (DFPT) within the gener- in contrast to other rst-principles studies, usually simu- alized gradient approximation as formulated by Perdew, lating electron-phonon interactions (EPI) in neutral ma- Burke, and Ernzerhof (PBE). 2D periodic-boundary terials, which would be valid only in the limit of vanish- conditions are applied and the electrostatics of a sym- ing carrier density. The capability to study explicitly the metric double-gate eld-e ect setup are simulated using high-doping regime is a valuable complement to this, and the approach described in Ref. 43. Open-boundary con- very relevant for monolayers operating in the degenerate ditions are important to properly describe polar-optical limit, when the chemical potential is close to or inside 54 55 phonons and screening . The eld-e ect setup is used the conduction or valence band (doping regimes are fur- to induce an electron or hole density with a default value ther discussed in App. A). In addition, its predictive ac- 13 2 of n=p = 10 cm . The simulation of this relatively curacy provides meaningful comparisons to experiments. high-doping regime is further discussed in App. A. Cal- Indeed, the high-doping regime is often used experimen- culations are managed using the AiiDA materials infor- tally to characterize the intrinsic properties of novel ma- 37,38 matics infrastructure . The AiiDA database contain- terials, because it allows to screen charged impurities. ing the provenance for the transport calculations and the tools necessary to reproduce this work are provided in the 35,39 In this study we nd excellent phonon-limited trans- Archive section of the Materials Cloud . The stan- port properties for electron-doped Bi SeTe , Bi Se , Bi- dard numerical setup for all ground state and phonon 2 2 2 3 ClTe, Sb SeTe , InSe, GaSe, AlLiTe and hole-doped calculations has 32 32 Monkhorst-Pack k-point grids to 2 2 2 phosphorene (P ), AuI, ClGaTe, and WSe , all show- sample the full Brillouin zone, 0.02 Ry cold smearing 4 2 ing mobilities in the range from few hundreds to few with SSSP pseudopotentials (eciency version 0.7) and thousands cm /Vs. Three of these 2D materials are energy cuto s recommended therein. PAW pseudopoten- 13{16 13 44 very well-known and studied (P , WSe , InSe ). tials were substituted because they are incompatible with 4 2 The Sb X and Bi X (X = Se, Te, S) compounds are the use of symmetry in the electron-phonon routines of 2 3 2 3 better known for the topological properties of their 3D QE. Other minor variations from the standard setup can 45,46 parents . Monolayers have been recently studied in be found in the AiiDA database provided. The use of 17,47,48 the context of phonon-limited transport , but only cold smearing allows k-point convergence while keeping within the approximate Tagaki formalism . The trans- a lower e ective temperature in the calculations, mak- port performance of monolayer Sb SeTe has been con- ing the free-carrier screening closer to what it would be 2 2 rmed experimentally . GaSe (along with InSe) was in real conditions. Spin-orbit interactions are included studied ab initio with a representation of electron- only for hole-doped WSe , where they play a signi cant phonon scattering that goes beyond deformation poten- role. For other materials, while spin-orbit interactions tial theory while this work was carried out. AlLiTe , Bi- may have an e ect on the band structure in general (e.g. ClTe, ClGaTe, and AuI are, to our knowledge, still new in on the band gap for BiSeTe ), they will not have large the context of 2D charge transport. Beyond the relatively consequences on the conductivity or mobility as long as large number of materials identi ed and the novelty of there is no signi cant energy splitting of valleys with op- some of them, this work most importantly provides visual posite spins (small variations may come from changes in and intuitive understanding of electron-phonon scatter- the e ective masses). The analysis starts from a database ing as well as a systematic and data-supported analysis of of band structures computed (non-self-consistently) in the features leading to good transport performance. We the neutral material on very ne electronic momenta con rm and extend previous remarks on the importance grids (about 90 by 90) to analyse the valley structure. of band properties such as number of valleys, e ective These bands are then recomputed with eld-e ect dop- masses or anisotropy; and show that these can be used ing for the selected materials. Phonon momenta are cho- for larger scale databases studies to shortlist prospective sen to include only relevant transitions, and the Boltz- conductors. On the other hand, we also show how di er- mann transport equation (BTE) is solved as described ences in the strength and angular dependency of electron- in Ref. 13. We improved the stability for the solution of phonon scattering still induce signi cant variations in the the BTE by using the velocity of each initial state as a conductivity of materials with similar band properties. ctitious electric eld direction [Eq. (11) of Ref. 13]. 3 B. Boltzmann transport equation WSe2 T = 300 K The solution of the BTE is brie y outlined here for 13 2 n/p = 10 cm convenience; more details can be found in Ref. 13. The (longitudinal) conductivity is computed as follows: 60 P4 holes electrons 1 dk @f (" ) 50 2 k = = 2e (v(k) u )  (k) Bi2SeTe2 (2) @" (1) Sb2SeTe2 where e is the Coulomb charge, and k; " ; v(k) repre- GaSe InSe sent electronic momenta (and band index implicitly), en- AlLiTe2 BiClTe ergies, and velocities, respectively. The unit vector u AuI ClGaTe Bi2Se3 points in the electric eld's direction and f is the Fermi- 0.0 0.5 1.0 1.5 2.0 2.5 3.0 Dirac distribution. Here  is an energy- and momentum- gap (eV) dependent variable that has the dimensions of time and solves the linearized BTE: FIG. 1. Computed conductivities and mobilities at room tem- 0 0 0 perature plotted against PBE gap for all the materials con- (1 f (k))v(k) u = P (1 f (k )) E kk sidered in this work. BTE is solved for a xed carrier density 13 2 of n=p = 10 cm (electrons or holes). 0 0 fv(k) u  (k) v(k ) u  (k )g E E (2) compute the transport properties of the most promising P 0 being the probability for an electron in state k to kk candidates. Electron-phonon scattering is analyzed in be scattered into state k . In the following we consider each monolayer. The sux \-e" or \-h" is attached to only phonon-induced scattering, so that: the materials' formula to indicate if either electron dop- 2 1 ing or hole doping is considered. P = jg j fn (" " ~! ) kk+q kk+q; q; k+q k q; ~ N +(n + 1)(" " + ~! )g: q; k+q k q; A. Prototypical high-conductivity materials: InSe (3) and P where q; ; ~! ; n are phonon momenta, mode index, q; q; We start by analyzing key features of two well-known energy and occupations, and g are the electron- kk+q; semiconductors with excellent transport performance, phonon coupling matrix elements. InSe on the electron side (InSe-e) and phosphorene on the hole side (P -h). In the following we will use repeatedly III. HIGH-CONDUCTIVITY 2D two di erent plots to visualize, understand and compare SEMICONDUCTORS the details of the transport properties of 2D materials, reported rst in Fig. 2 for InSe and P . On the left, a \velocity plot" shows the band structure in a format that We study in careful detail eleven monolayers obtained is relevant for transport. For each electronic state k (each from our database of 256 easily exfoliable materials with black dot) on the ne k-point grid used for solving the at most 6 atoms per unit cell , as available on the Ma- 34,35 BTE and within a certain energy range from the band terials Cloud and in the supplementary material of edge (set as the origin of the y-axis), the energy " is Ref. 23. The corresponding conductivity and mobility, plotted against the norm of the velocity jvj = j r " j, k k as a function of the PBE gap, are shown in Fig. 1. At given in atomic Rydberg units (ARU) . The spread in- this doping level a conductivity of a few e =h is already 4,8 dicates the anisotropy of the valley. The red scale of the considered good for a semiconductor, while the best 2 58,59 background is proportional to the derivative of the Fermi- graphene devices yield values around 500 e =h . The Dirac occupation at 300 K, thus highlighting the states materials presented in this work are in the intermedi- participating to transport with a non-vanishing contri- ary orders of magnitude, with conductivities of 10  100 bution to Eq. (1). The Fermi level is computed at room e =h, and mobilities in the same range as bulk silicon 0 2 13 2 2 temperature (i.e. such that f (" ; T )dk=(2 ) = 10 (400 cm /Vs for holes, 1400 cm /Vs for electrons). k In the following, we detail the exploration process cm for electrons, and similarly for holes). On the right that led to those materials. First, two well-known high- side, a \scattering plot" shows where and how easily a conductivity 2D semiconductors are analyzed to identify certain initial state can be scattered. The initial state representative band features. We then search for those at k , indicated by a black square, is chosen to be at in 23,34 features within the Materials Cloud database and the Fermi level (and in the transport direction, when rel- conductivity (e /h) mobility (cm /(Vs)) 4 evant). The grey shading shows the morphology of the direction is maximized. However, the DOS here is rela- valley(s) considered for transport. The red color scale tively larger, and the Fermi level stays closer to the band represents an e ective coupling constant g which ac- edge, where velocities are lower. The good performance eff counts for all phonon modes, their occupation at 300 K, of phosphorene thus relies on a somewhat more fragile and energy conservation, since this quantity can be mean- balance. ingfully compared between materials. In practice, g An essential feature to maximize both i) and ii) is to eff is de ned as the square root of the sum of the interpo- work with single valley materials. Single valley band lated jg j n ( or n + 1) over all  and q structures lead to a lower DOS, higher Fermi level and k ;k +q; q; q; in in that ful ll energy conservation during phonon absorption higher velocities. Even more compelling, one avoids in- (or emission) when the nal state at k + q falls into a tervalley scattering, which is usually stronger than in- in certain zone (the valley being tessellated into triangles). travalley scattering and hinders transport . The value given at the top of each scattering plot is the In a related work , several promising materials for lifetime of the initial state computed within the momen- ultra-short transistor devices have been identi ed. In tum relaxation time approximation: that situation, one optimizes the performance of the ma- terial in the ballistic limit and the role of electron-phonon 0 0 0 1 1 f (k ) v(k ) v(k) scattering is secondary, since the channels are usually be- = P 0  1 (4) kk 0 2 (k) 1 f (k) v(k) low the scattering length. Thus, maximizing both the velocity in the transport direction and the DOS is ex- Everything is computed at 300 K, with the correspond- tremely bene cial, and anisotropic and multivalley ma- ing chemical potential in f . terials are interesting. When electron-phonon scattering is turned on, multivalley band structures are not ideal, We now discuss qualitatively the features allowing while anisotropy can remain relevant. electron-doped InSe and hole-doped phosphorene to max- imize the conductivity in Eq. (1): this is a sum over electronic states weighted by the derivative of the Fermi- B. More like InSe: Steep and deep single-valleys Dirac occupation. The weight e ectively selects states around the chemical potential within an energy range As shown in Fig. 2, InSe has a sharp and deep single that scales with temperature, as represented in the left panels of Fig. 2. The rest of the integrand can be electron valley, with carrier velocities close to 6 ARU at 13 2 the Fermi level when n = 10 cm (roughly 6 times separated in two contributions: i) one from scattering (electron-phonon here) through the scattering lifetime  ; larger than the Fermi velocity in graphene,  1 ARU). Steep isotropic valleys are obviously characterized by low ii) and one from the velocity of the carriers v projected along the transport direction u . DOS, which means that the Fermi level quickly shifts away from the band edge with increasing carrier density. The scattering contribution i) is inversely proportional to the strength of the EPI and the amount of states avail- This allows to reach higher carrier velocities, but it often means that the Fermi level reaches other valleys in the able for scattering (as apparent in Eq. (4)), i.e. the band structure. In that case, the aforementioned bene ts density of states (DOS) within a certain energy win- dow around the initial states, depending on the energy of the single valley structure are lost. If the Fermi level is close to the edges of the next valleys, intervalley scat- of the phonons involved. We note, however, that the DOS contribution is essentially compensated when doing tering is activated and the new states populated are not helping to conduct due to their low velocity. Note that, the integral over electronic states to get the conductiv- ity, Eq. (1). To maximize  , the quantity to focus on is in a similar spirit, the suppression of intervalley scatter- ing via valley-engineering (e.g. through strain) has been thus the strength of the EPI, which should of course be minimized. Weak EPIs is a feature shared by InSe and proposed to enhance the transport properties of multi- valley materials such as arsenene. Thus, a steep valley phosphorene. There are di erent ways to maximize the velocity con- is not sucient to have good transport performance: it must also be deep. More precisely, higher-energy val- tribution ii); in general, one needs to maximize the ve- leys need to be far enough for the material to operate locity in the direction of transport. In the case of InSe, the velocity is isotropic and very large, with a small ef- e ectively as a single valley material for a doping of 10 cm . fective mass. Accordingly, the DOS is low and the Fermi level reaches high energies above the valley edge, to- We look for such valleys among the exfoliable mate- wards higher velocities, eventually saturating at a maxi- rials in our study set. We use band structures com- mal value in case of non-parabolic materials. The bene ts puted on very ne grids in the neutral materials and of a such a steep and deep single valley is that current select materials with positive phonons, a limited gap carrying states around the chemical potential have high E < 2:5 eV, a single valley (within 100 meV of the velocities, as also pointed out in Ref. 33. In the contrast- Fermi level), and a maximum Fermi velocity jv j > 6 max ing case of phosphorene, the valley is highly anisotropic. ARU. This allows us to identify BiClTe-e, AlLiTe -e, Provided one chooses the low e ective mass direction for GaSe-e, Bi SeTe -e, Bi STe -e, Bi Se -e, and Sb SeTe - 2 2 2 2 2 3 2 2 transport, the projection of the velocity in the transport e as promising candidates. Electron-phonon scattering 5 InSe = 61.201 fs 0.30 0.06 0.25 0.04 0.20 0.02 0.00 0.15 0.02 0.10 0.04 0.05 0.06 0.00 0 0 2 4 6 0.10 0.05 0.00 0.05 0.10 |v| (ARU) k (bohr ) = 204.02 fs 0.00 0.15 0.05 0.10 0.10 0.05 50 0.00 0.15 0.05 0.20 0.10 0.25 0.15 2 4 6 0.2 0.1 0.0 0.1 0.2 |v| (ARU) k (bohr ) FIG. 2. Transport properties of InSe and phosphorene (P ), two well-known good conductors. On the left, the \velocity plot" shows, for each electronic state k in the valley, the energy " from the band edge plotted against the norm of the velocity jvj = j r " j, in atomic Rydberg units (ARU). The color scale of the background is proportional to the derivative of the k k Fermi-Dirac distribution, which appears in Eq. (1). On the right, the \scattering plot" shows where an initial state (black square) can be scattered within the valley (grey shade). The color scale represents the e ective coupling constant g , which eff accounts for all phonon modes, their occupation at 300 K, and energy conservation. At the top,  is the scattering time of the initial state computed within the momentum relaxation time approximation, Eq. (4). is studied for these monolayers (except Bi STe -e, a bit higher velocities. Finally, AlLiTe -e ( = 18 e =h, see 2 2 2 redundant with respect to the other two members of Fig. 3) and BiClTe-e ( = 16 e =h) have slightly larger the Bi X family). The results are shown in Fig. 3 for velocities but stronger EPIs than InSe-e, making them 2 3 Bi SeTe -e and AlLiTe -e, and in the appendix for the slightly less conductive. Note that all these systems with 2 2 2 rest. The general similarity of the plots indicates at a sharp valleys are electron-doped, which may be rational- glance why these materials have similar transport proper- ized by considering that conduction bands are more often ties. The carrier velocities at the Fermi level, the strength made of delocalized non-bonding states which tend to be of EPI, and the scattering times are of similar orders of more dispersive. Despite the valleys being very similar magnitude. A closer look reveals the reasons underlying (isotropic with jv j  6  7 ARU), the conductivity max their precise ranking. BiSeTe -e, BiSe -e, and Sb SeTe - varies from 14 to 42 e =h. This points to the importance 2 3 2 2 e are very similar in terms of chemistry, band structure, of accounting for the details of EPIs to rank materials and phonons. BiSeTe -e (see Fig. 3) displays the high- accurately. est conductivity ( = 42 e =h), combining some of the We also note that the selection process does not guar- weakest EPI with the largest velocities. It is followed by antee to nd all the best conductors, as demonstrated Sb SeTe -e ( = 30 e =h) with similar EPI but slightly 2 2 by the example of WSe -h: TMDs were not selected smaller velocities. Bi Se -e ( = 14 e =h) also has a very 2 3 because they are not single-valley (also their Fermi ve- sharp valley, but the EPI is 2 to 3 times stronger. As locities would be too small). However, thanks to strong could be expected, electron-phonon scattering in GaSe-e spin-orbit interactions, the hole side of TMDs can be con- 2 2 ( = 25 e =h) is similar to InSe-e ( = 20 e =h), with sidered to be in the steep and deep single valley category. scattering times di ering by only  10%. GaSe-e owes Indeed, as is well known, the hole valleys associated to its higher conductivity mostly to a sharper valley and opposite spin textures split very strongly in energy, of the Energy (from edge) (eV) Energy (from edge) (eV) k (bohr ) k (bohr ) g (meV) g (meV) eff eff 6 Bi SeTe 2 2 = 110.16 fs 0.06 0.30 0.04 120 0.25 0.02 0.20 0.00 0.15 0.02 0.10 0.04 0.05 0.06 0.00 0 0 2 4 6 0.10 0.05 0.00 0.05 0.10 |v| (ARU) k (bohr ) AlLiTe = 48.003 fs 0.4 0.075 0.050 0.3 0.025 0.2 0.000 0.025 0.1 0.050 0.075 0.0 0 0 2 4 6 8 0.10 0.05 0.00 0.05 0.10 |v| (ARU) k (bohr ) 2 2 FIG. 3. Transport properties of Bi SeTe ( = 42 e =h) and AlLiTe ( = 18 e =h), showing the velocity (left) and scattering 2 2 2 (plots) as described in Fig. 2. Both monolayers belong to a category of materials with steep and deep single valleys in the conduction band. Similar plots are given in App. B for other monolayers in this category: BiClTe-e, GaSe-e, Bi STe -e, 2 2 Bi Se -e, and Sb SeTe -e. 2 3 2 2 WSe = 575.58 fs 0.00 0.6 0.05 0.10 0.4 0.15 0.2 0.20 0.25 0.0 0.30 1 2 3 4 0.0 0.2 0.4 0.6 0.8 1.0 |v| (ARU) k (bohr ) FIG. 4. Transport properties of WSe -h ( = 81 e =h), showing the velocity (left) and scattering (plots) as described in Fig. 2. Thanks to spin-orbit interactions, weak spin- ip scattering, and the e ective absence of any valley at , spins travel in parallel channels not connected to each other. order of 100 meV, making the lower valleys irrelevant for inating a potential intervalley scattering channel. Fur- transport. One e ectively obtains two valleys at K and thermore, as shown in Fig. 4, the intervalley scattering, K' with opposite spin textures. WSe -h, at least within associated with spin- ip EPIs, is weak (< 10 meV) com- our computational framework, also has the advantage pared to intravalley, spin-conserving EPIs ( 50 meV). that the edge of the valence band at is quite low, elim- This implies that the two valleys are e ectively decou- Energy (from edge) (eV) Energy (from edge) (eV) Energy (from edge) (eV) k (bohr ) k (bohr ) k (bohr ) g (meV) g (meV) g (meV) eff eff eff 7 pled as far as phonon-limited transport is concerned and the monoatomic nature of P , eliminating all Born e ec- transport is similar to the single valley case, except op- tive charge and piezoelectric couplings), one can observe posite spins travel in separate channels located at dif- in P a predominance of \side-scattering": states in the ferent points in the Brillouin zone. Comparing with the direction of transport are mostly scattered to states on isotropic single-valley materials discussed above, WSe - the sides of the valley with velocities perpendicular to the h has velocities half as small but EPIs at least 3 times direction of transport. As can be seen from Eq. (4), this weaker, leading to the largest conductivity ( = 81 e =h). leads to a longer scattering times, since the term involv- This high value is due to the (e ective) absence of the ing a scalar product of the velocities vanishes. The con- valley. Indeed, opposite spins are degenerate at , and tribution of the DOS from the integral of the conductivity if the valley were accessible for scattering, intervalley usually cancels out with the integral over available scat- scattering from both K and K' would be quite strong. In tering states. This is not the case here because we bene t addition, as shown in Ref. 61, multi-valley occupations from the larger weight of high velocity states in the con- enhances the intravalley coupling to the homopolar op- ductivity integral while having less dense, low-velocity tical phonon mode by making free-carrier screening in- states in the scattering integral. Thus, \side scattering" ecient. At any rate, the extraordinarily weak EPI in de nitely brings the performance of P -h from great to WSe -h is tied to subtle spin-orbit and screening e ects exceptional. ClGaTe-h is similar to P in terms of veloc- 2 4 that are dicult to predict and highlights the limitations ities, but the EPI is stronger and, importantly, mostly of simple selection processes based solely on the band backscattering. The scattering time is consequently 10 structure. times larger. AuI-h has lower velocities and higher EPI. While targeting the band features of phosphorene has allowed us to nd other excellent candidates, we see that the essence of its exceptional performance comes C. More like phosphorene: anisotropic single valleys from something much harder to identify at the level of a database: the anisotropy of its electron-phonon inter- actions. Thus, phosphorene is at the same time both a In contrast with the fairly isotropic band structures of prototype and an example of the limitations of selection InSe, the case of phosphorene showed that anisotropy can processes based on band structures. also lead to good performance. So, we look for similar materials in our database, having stable phonons, single valleys and small band gap at the PBE level (E < 2:5 IV. CONCLUSIONS eV). This time we lter materials keeping materials that max combine high velocity ratios ( < 1:7) and a decent min maximum velocity (v > 2:0 ARU) at the Fermi level. State-of-the-art density-functional perturbation theory max These search criteria allow us to identify AuI-h, and and the Boltzmann transport equation are used to study ClGaTe-h as well as the electron-side of P -e as promising 4 the outstanding transport properties of several 2D semi- candidates. The electron side of phosphorene is a similar, conductors. Focusing on conductivity at a xed den- 13 2 less pronounced version of the hole side, and was already sity of n=p = 10 cm , the present results o er a studied in our previous work . Thus, we focus here on complementary perspective with respect to most rst- 2 2 ClGaTe-h ( = 13 e =h) and AuI-h ( = 14 e =h). principles calculations valid in the zero carrier density The results of the electron-phonon calculations are limit. We provide a detailed analysis of electron-phonon plotted in Fig. 5. The spread of the velocity traces in- scattering in two well-known high-conductivity systems: dicates high anisotropy, the velocity varying by a factor electron-doped InSe and hole-doped phosphorene. While 2 at a given energy. The high-velocity direction prefer- they share some features {like weak EPI and a single- able for transport is x for ClGaTe and x + y for AuI. valley electronic structure{ they exemplify two di erent The anisotropy allows one to bene t from many high ve- strategies to maximize the conductivity. InSe's high- locity states, thanks to a atter band in the direction velocity, isotropic valley can be exploited thanks to the perpendicular to transport. However, if the band is too fact that the next valleys are much higher in energy. at and the DOS too high, the Fermi level stays close Phosphorene, instead, owes its excellent transport perfor- to the band edges and the velocities are too low. Here, mance to the anisotropy of both its band structure and velocities of the order of 3  4 ARU ensure good trans- electron-phonon scattering. Analyzing the band proper- port properties. While ClGaTe-h and AuI-h share the ties of around  150 small stable semiconductors with 6 anisotropic character of phosphorene, their conductivity atoms or less in the unit-cell, from the Materials Cloud, remains lower, which in part re ects the fact that the we identify systems with band features similar to either good transport properties of anisotropic materials rely InSe or phosphorene. We nd large phonon-limited con- on a more fragile balance of features. ductivities for electron-doped Bi SeTe , Bi Se , BiClTe, 2 2 2 3 A closer look at the scattering plots in Fig. 5, con- Sb SeTe , AlLiTe, and GaSe, as well as hole-doped AuI, 2 2 trasted with phosphorene in Fig. 2 further reveals the ClGaTe, and WSe . These results con rm that the band reasons behind phosphorene's superior conductivity. In structure landscape plays an important role in determin- addition to generally weaker EPIs (which is partly due to ing transport and shows that seeking peculiar features 8 ClGaTe = 38.744 fs 0.00 0.15 0.05 0.10 0.10 0.05 0.15 0.00 0.20 0.05 0.10 0.25 0.15 0.30 1 2 3 4 5 0.2 0.1 0.0 0.1 0.2 |v| (ARU) k (bohr ) AuI = 34.346 fs 0.00 0.10 0.05 0.05 0.10 0.00 0.15 0.05 0.20 0.10 0 1 2 3 4 0.15 0.10 0.05 0.00 0.05 0.10 0.15 |v| (ARU) k (bohr ) FIG. 5. Transport properties of ClGaTe and AuI showing the velocity (left) and scattering (right) plots as described in Fig. 2. Anisotropic valleys imply there is one optimal transport direction, corresponding to higher velocities. Initial states (marked with a black square) are chosen to be in that direction. in the electronic structure does lead to high-performance was awarded by PRACE on Marconi at Cineca, Italy materials. Nevertheless, we also show how the details of (project id. 2016163963). Computational resources have the strength and angular dependency of electron-phonon been provided by the Consortium des quipements de Cal- scattering play a critical role in ranking those materials cul Intensif (CCI), funded by the Fonds de la Recherche with respect to each other. Scienti que de Belgique (F.R.S.-FNRS) under Grant No. 2.5020.11 and by the Walloon Region. T.S. acknowledges support from the University of Liege under Special Funds ACKNOWLEDGEMENTS: for Research, IPD-STEMA Programme. M.G. acknowl- edges support by the Italian Ministry for University and Research through the Levi-Montalcini program and by The authors are grateful to Davide Campi for shar- the Swiss National Science Foundation (SNSF) through ing the initial band structures. This work has been in the Ambizione program (grant PZ00P2 174056). part supported by NCCR MARVEL. Simulation time 1 4 S. Z. Butler et al., Progress, challenges, and opportunities B. Radisavljevic and A. Kis, Mobility engineering and a in two-dimensional materials beyond graphene., ACS nano metal{insulator transition in monolayer MoS , Nature Ma- 7, 2898 (2013). terials 12, 815 (2013). 2 5 D. Akinwande, N. Petrone, and J. Hone, Two-dimensional B. Radisavljevic, A. Radenovic, J. Brivio, V. Giacometti, exible nanoelectronics , Nature Communications 5, 5678 and A. Kis, Single-layer MoS transistors , Nature nan- (2014). otechnology 6, 147 (2011). 3 6 M. Chhowalla, D. Jena, and H. Zhang, Two-dimensional X. Li, Z. Yu, X. Xiong, T. Li, T. Gao, R. Wang, R. Huang, semiconductors for transistors , Nature Reviews Materials and Y. Wu, High-speed black phosphorus eld-e ect tran- 1, 16052 (2016). sistors approaching ballistic limit , Science Advances 5 Energy (from edge) (eV) Energy (from edge) (eV) 1 k (bohr ) k (bohr ) g (meV) g (meV) eff eff 9 (2019). Scienti c Reports 7, 5179 (2017). 7 23 L. Li et al., Black phosphorus eld-e ect transistors., Nat. N. Mounet et al., Two-dimensional materials from high- Nanotechnol. 9, 372 (2014). throughput computational exfoliation of experimentally L. Tao, E. Cinquanta, D. Chiappe, C. Grazianetti, M. Fan- known compounds , Nature Nanotechnology 13, 246 (2018). ciulli, M. Dubey, A. Molle, and D. Akinwande, Silicene S. Haastrup et al., The Computational 2D Materials eld-e ect transistors operating at room temperature , Na- Database: high-throughput modeling and discovery of ture Nanotechnology 10, 227 (2015). atomically thin crystals , 2D Materials 5, 042002 (2018). 9 25 X. Li, J. T. Mullen, Z. Jin, K. M. Borysenko, M. Buon- J. Zhou et al., 2DMatPedia, an open computational giorno Nardelli, and K. W. Kim, Intrinsic electrical trans- database of two-dimensional materials from top-down and port properties of monolayer silicene and MoS from rst bottom-up approaches , Scienti c Data 6, 86 (2019). principles , Physical Review B - Condensed Matter and Ma- C. Klinkert, A. Szab o, C. Stieger, D. Campi, N. Marzari, terials Physics 87, 115418 (2013). and M. Luisier, 2-D Materials for Ultrascaled Field-E ect W. Li, Electrical transport limited by electron-phonon cou- Transistors: One Hundred Candidates under the Ab Initio pling from Boltzmann transport equation: An ab initio Microscope , ACS Nano 14, 8605 (2020). study of Si, Al, and MoS , Physical Review B 92, 075405 2 F. Giustino, Electron-phonon interactions from rst prin- (2015). ciples , Reviews of Modern Physics 89, 015003 (2017). K. Kaasbjerg, K. S. Thygesen, and K. W. Jacobsen, S. Ponc e, W. Li, S. Reichardt, and F. Giustino, First- Phonon-limited mobility in n-type single-layer MoS from principles calculations of charge carrier mobility and con- rst principles , Physical Review B - Condensed Matter ductivity in bulk semiconductors and two-dimensional ma- and Materials Physics 85, 115317 (2012). terials , Reports on Progress in Physics 83, 036501 (2020). T. Gunst, T. Markussen, K. Stokbro, and M. Brandbyge, C.-H. Park, N. Bonini, T. Sohier, G. Samsonidze, B. Kozin- First-principles method for electron-phonon coupling and sky, M. Calandra, F. Mauri, and N. Marzari, Electron- electron mobility: Applications to two-dimensional materi- Phonon Interactions and the Intrinsic Electrical Resistivity als , Physical Review B 93, 035414 (2016). of Graphene., Nano letters 14, 1113 (2014). T. Sohier, D. Campi, N. Marzari, and M. Gibertini, Mo- T. Sohier, M. Calandra, C.-H. C.-H. Park, N. Bonini, bility of 2D materials from rst principles in an accurate N. Marzari, and F. Mauri, Phonon-limited resistivity of and automated framework , Physical Review Materials 2, graphene by rst-principles calculations: Electron-phonon 114010 (2018). interactions, strain-induced gauge eld, and Boltzmann A. N. Rudenko, S. Brener, and M. I. Katsnelson, Intrinsic equation , Physical Review B 90, 125414 (2014). Charge Carrier Mobility in Single-Layer Black Phospho- S. Ponc e, E. Margine, C. Verdi, and F. Giustino, EPW: rus , Physical Review Letters 116, 246401 (2016). Electron{phonon coupling, transport and superconducting Y. Trushkov and V. Perebeinos, Phonon-limited carrier properties using maximally localized Wannier functions , mobility in monolayer black phosphorus , Physical Review Computer Physics Communications 209, 116 (2016). B 95, 075436 (2017). J.-J. Zhou, J. Park, I.-T. Lu, I. Maliyov, X. Tong, and G. Gaddemane, W. G. Vandenberghe, M. L. Van De Put, M. Bernardi, Perturbo: a software package for ab initio S. Chen, S. Tiwari, E. Chen, and M. V. Fischetti, Theoreti- electron-phonon interactions, charge transport and ultra- cal studies of electronic transport in monolayer and bilayer fast dynamics , arXiv:2002.02045 (2020). phosphorene: A critical overview , Physical Review B 98 L. Cheng, C. Zhang, and Y. Liu, The Optimal Electronic (2018). Structure for High-Mobility 2D Semiconductors: Excep- B. Wang, X. Niu, Y. Ouyang, Q. Zhou, and J. Wang, Ul- tionally High Hole Mobility in 2D Antimony , Journal of trathin Semiconducting Bi Te S and Bi Te Se with High 2 2 2 2 the American Chemical Society 141, 16296 (2019). Electron Mobilities , The Journal of Physical Chemistry https://www.materialscloud.org/discover/2dstructures/. Letters 9, 487 (2018). L. Talirz et al., Materials Cloud, a platform for open com- S. Takagi, A. Toriumi, M. Iwase, and H. Tango, On the putational science , arXiv:2003.12510 (2020). universality of inversion layer mobility in Si MOSFET's: G. Brunin et al., Phonon-limited electron mobility in Part I-e ects of substrate impurity concentration , IEEE Si, GaAs and GaP using plane waves and Bloch states , Transactions on Electron Devices 41, 2357 (1994). arXiv:2002.00630 (2020). G. Gaddemane, S. Gopalan, M. Van de Put, and M. V. G. Pizzi, A. Cepellotti, R. Sabatini, N. Marzari, and Fischetti, Limitations of ab initio methods to predict the B. Kozinsky, AiiDA: automated interactive infrastructure electronic-transport properties of two-dimensional materi- and database for computational science , Computational als: The computational example of 2H-phase transition Materials Science 111, 218 (2016). metal dichalcogenides , arXiv:1912.03795 (2019). S. P. Huber et al., AiiDA 1.0, a scalable computational in- G. Cheon, K.-A. N. Duerloo, A. D. Sendek, C. Porter, frastructure for automated reproducible work ows and data Y. Chen, and E. J. Reed, Data Mining for New Two- provenance , arXiv:2003.12476 (2020). and One-Dimensional Weakly Bonded Solids and Lattice- https://doi.org/10.24435/materialscloud:fr-r0. Commensurate Heterostructures , Nano Letters 17, 1915 Y. Zhang, J. Ye, Y. Matsuhashi, and Y. Iwasa, Ambipo- (2017). lar MoS Thin Flake Transistors , Nano Letters 12, 1136 M. Ashton, J. Paul, S. B. Sinnott, and R. G. Hennig, (2012). Topology-Scaling Identi cation of Layered Solids and Sta- D. Braga, I. Guti errez Lezama, H. Berger, and A. F. Mor- ble Exfoliated 2D Materials , Physical Review Letters 118, purgo, Quantitative Determination of the Band Gap of 106101 (2017). WS with Ambipolar Ionic Liquid-Gated Transistors , Nano K. Choudhary, I. Kalish, R. Beams, and F. Tavazza, High- Letters 12, 5218 (2012). throughput Identi cation and Characterization of Two- dimensional Materials using Density functional theory , 10 42 59 J. T. Ye, Y. J. Zhang, R. Akashi, M. S. Bahramy, R. Arita, L. Banszerus et al., Extraordinary high room-temperature and Y. Iwasa, Superconducting Dome in a Gate-Tuned carrier mobility in graphene-WSe heterostructures , Band Insulator , Science 338, 1193 (2012). arXiv:1909.09523 (2019). 43 60 T. Sohier, M. Calandra, and F. Mauri, Density functional Atomic Rydberg units are the most convenient units for ve- perturbation theory for gated two-dimensional heterostruc- locity in materials. Typical Fermi velocities in the database tures: Theoretical developments and application to exu- {including graphene{ are on the order of 1 ARU, which ral phonons in graphene , Physical Review B 96, 075448 corresponds to  10 m/s. (2017). T. Sohier, E. Ponomarev, M. Gibertini, H. Berger, W. Li, S. Ponc e, and F. Giustino, Dimensional Crossover N. Marzari, N. Ubrig, and A. F. Morpurgo, Enhanced in the Carrier Mobility of Two-Dimensional Semiconduc- Electron-Phonon Interaction in Multivalley Materials , tors: The Case of InSe , Nano Letters 19, 1774 (2019). Physical Review X 9, 031019 (2019). 45 62 W. Zhang, R. Yu, H. J. Zhang, X. Dai, and Z. Fang, First- N. Ma and D. Jena, Charge scattering and mobility in principles studies of the three-dimensional strong topologi- atomically thin semiconductors , Physical Review X 4, cal insulators Bi Te , Bi Se and Sb Te , New Journal of 011043 (2014). 2 3 2 3 2 3 Physics 12, 065013 (2010). C. Verdi, F. Caruso, and F. Giustino, Origin of the S. Jafarpisheh, A. Ju, K. Janen, T. Taniguchi, K. Watan- crossover from polarons to Fermi liquids in transition metal abe, C. Stampfer, and B. Beschoten, Reducing the Impact oxides , Nature Communications 8, 15769 (2017). of Bulk Doping on Transport Properties of Bi-Based 3D P. Bogulsawski and J. Mycielski, Is the deformation poten- Topological Insulators , Physica Status Solidi (B) Basic Re- tial in semiconductors screened by free carriers? , Journal search , pssb.202000021 (2020). of Physics C: Solid State Physics 10, 2413 (1977). 47 65 Y. Liu, Y. Xu, Y. Ji, and H. Zhang, Monolayer P. Boguslawski, Screening of the deformation potential by Bi Se Te : novel two dimensional semiconductors with free electrons in the multivalley conduction band , Journal 2 3x x excellent stability and high electron mobility , Physical of Physics C: Solid State Physics 10, L417 (1977). Chemistry Chemical Physics 22, 9685 (2020). P. Boguslawski and J. Mycielski, Free-electron screening of C. Tang, L. Zhang, C. Zhang, J. Macleod, K. Ostrikov, short-range scattering potentials in semiconductors , Jour- and A. Du, Highly stable two-dimensional gold selenide nal of Physics C: Solid State Physics 13, 1019 (1980). with large in-plane anisotropy and ultrahigh carrier mo- M. Verstraete and X. Gonze, Smearing scheme for nite- bility , Nanoscale Horizons 5, 366 (2020). temperature electronic-structure calculations , Physical Re- H. Qu, W. Zhou, S. Guo, Z. Li, Y. Wang, and S. Zhang, view B 65, 035111 (2001). Ballistic Quantum Transport of Sub10 nm 2D Sb Te Se J. Ma, D. Xu, R. Hu, and X. Luo, Examining two- 2 2 Transistors , Advanced Electronic Materials 5, 1900813 dimensional Fr ohlich model and enhancing the electron (2019). mobility of monolayer InSe by dielectric engineering , Jour- J. Chen, X. Tan, P. Lin, B. Sa, J. Zhou, Y. Zhang, C. Wen, nal of Applied Physics 128, 035107 (2020). and Z. Sun, Comprehensive understanding of intrinsic mo- bility in the monolayers of III-VI group 2D materials , Physical Chemistry Chemical Physics 21, 21898 (2019). P. Giannozzi et al., QUANTUM ESPRESSO: a modular and open-source software project for quantum simulations of materials., Journal of Physics: Condensed Matter 21, 395502 (2009). P. Giannozzi et al., Advanced capabilities for materials modelling with Quantum ESPRESSO , Journal of Physics Condensed Matter 29, 465901 (2017). J. P. Perdew, K. Burke, and M. Ernzerhof, Generalized Gradient Approximation Made Simple , Physical Review Letters 77, 3865 (1996). T. Sohier, M. Gibertini, M. Calandra, F. Mauri, and N. Marzari, Breakdown of Optical Phonons' Splitting in Two-Dimensional Materials , Nano Letters 17, 3758 (2017). T. Sohier, M. Calandra, and F. Mauri, Density-functional calculation of static screening in two-dimensional materi- als: The long-wavelength dielectric function of graphene , Physical Review B 91, 165428 (2015). N. Marzari, D. Vanderbilt, A. De Vita, and M. C. Payne, Thermal contraction and disordering of the al(110) sur- face , Physical Review Letters (1999). G. Prandini, A. Marrazzo, I. E. Castelli, N. Mounet, and N. Marzari, Precision and eciency in solid-state pseu- dopotential calculations , npj Computational Materials 4, 72 (2018). L. Wang et al., One-dimensional electrical contact to a two- dimensional material., Science 342, 614 (2013). 11 Appendix A: Comments on doping-dependent phonon interactions. If the strength of EPIs are con- transport performance stant, simple models using quadratic bands and elastic scattering show that the mobility is roughly constant up to fairly large dopings and then decreases. Consider- The mobility is a typical gure of merit for \trans- ing inelastic scattering increases mobility at small dop- port performance": It depends on doping (i.e. the carrier ing because there are no more available states for phonon density induced by eld e ects), and the relevant dop- emission close to the band edge. In any case, maximiz- ing range might vary with the application. Most rst- ing  implies maximizing  . However, for the ma- 13 0 principles computations of mobility are done in the zero terials considered here, EPIs are likely to be at least carrier density limit  = lim . Here, instead, we 0 n!0 partially doping-dependent via free-carrier screening. In- focus on 2D semiconductors with high conductivity  at deed, those are all single-valley materials, implying that 13 2 a xed carrier density of n=p = 10 cm . Of course, scattering is dominated by momenta smaller than the size by optimizing  we also maximize the mobility  = en=p of the Fermi surface, where free carriers screening is e- at this particular density, denoted with  . It should be 13 cient. If all EPI were sensitive to free-carrier screening, highlighted that, in general,  6=  , and the variation 0 13 we would see an opposite trend, with mobility increas- of the mobility in between these two doping regimes is ing as a function of doping , as free carriers screen the not obvious to predict. scattering sources.  might be then signi cantly lower than the  computed here. In practice, the magnitude The low (but nite) doping regime is very challeng- of this trend will depend on which EPIs are sensitive to ing to simulate realistically. The chemical potential is screening and how strong the bare EPIs are. below the band edge and depends strongly on temper- The electron-doped materials studied here (InSe, ature. One would ideally run one simulation per tem- GaSe, Bi SeTe , Sb SeTe ) have strong Born e ec- 2 2 2 2 perature, with an electronic smearing corresponding to tive charges and Fr ohlich couplings, which is screening- this temperature and an accordingly dense grid of k- sensitive and sharply increases at small momenta. For points. Unfortunately, even room temperature corre- InSe, we compute   490 cm =Vs, compared to sponds to a very low electronic smearing compared to 100 cm =Vs computed in Refs. 44 and 68 including standard DFT and DFPT calculations, resulting in very polar e ects and  = 488 cm =Vs when the Fr ohlich dense grids of k-points and prohibitively expensive cal- coupling is suppressed (as due to screening). Similar culations (especially when studying many materials as trends are expected for the other electron-doped materi- in this work). This issue is usually circumvented by als in this work. Whether  or  is more relevant to 0 13 simulating the neutral system and computing only  . a certain operating doping range depends on the critical Sometimes, the most obvious consequence of doping, carrier density at which free-carrier screening becomes ef- i.e. screening form free carriers, is added as an analyt- 62,63 cient. In 2D and assuming a constant density of states ical post-processing correction , but which scattering h i F C k T sources should be screened by free carriers, and how, has B per area D, one can derive n = k TD ln 1 + e 64{66 been debated . Moreover, eld e ects and screen- where k T is the thermal energy,  the chemical po- B F ing can have non-trivial consequences, as demonstrated tential (entering the Fermi-Dirac distribution) and " the 61 43 in TMDs or in graphene . In this work we choose bottom of the conduction band. If we estimate the onset to fully account for doping in the calculations, but at of free-carrier screening as  " > k T (when the F C B relatively high carrier density. This allows us to per- occupations are not dominated by the tail of the Fermi- form more realistic calculations, easier to converge, with 11 2 Dirac distribution), we obtain n > 5 10 cm for all a chemical potential within the band and a well-de ned the electron-doped materials studied here. Fermi surface. We use a smearing that is large compared The hole-doped materials studied here have weaker to room temperature in order to have accurate results Born e ective charges, but that is not to say that the with a ordable k-point grids, but the \cold" nature of remaining EPIs are not sensitive to screening; neverthe- the smearing allows to gets closer to room tempera- less, smaller variations of the mobility are to be expected. ture conditions. The e ects of smearing can be signif- icant at low doping, when the chemical potential is in the gap; at the high doping levels considered here, how- Appendix B: Additional electron-phonon scattering ever, we expect the calculations to be representative of data room-temperature conditions. Note that an alternative consistent approach to smear a nite temperature Fermi- Fig. 6 shows the transport properties of GaSe-e, Dirac distribution has been put forward , but it is not Bi Se -e, Sb SeTe -e, and BiClTe-e. 2 3 2 2 implemented in our computational framework. Fig. 7 shows the velocity plot for Bi STe -e, for which 2 2 To predict the behavior of the mobility as doping de- phonon-limited transport was not computed, but is ex- creases, one needs to account for the variation of electron- pected to be similar to Bi SeTe -e. 2 2 12 GaSe = 71.619 fs 0.075 0.30 0.050 0.25 0.025 0.20 0.000 0.15 0.025 0.10 0.050 0.05 0.075 0.00 0 0 2 4 6 8 0.10 0.05 0.00 0.05 0.10 |v| (ARU) k (bohr ) Bi Se 2 3 = 37.318 fs 0.06 0.30 0.04 0.25 0.02 0.20 0.00 0.15 0.02 0.10 0.04 0.05 0.06 0.00 0 0 2 4 6 0.10 0.05 0.00 0.05 0.10 |v| (ARU) k (bohr ) Sb SeTe 2 2 = 90.739 fs 0.35 0.075 0.30 0.050 0.25 0.025 0.20 0.000 0.15 0.025 0.10 0.050 0.05 25 0.075 0.00 0 0 1 2 3 4 5 6 0.10 0.05 0.00 0.05 0.10 |v| (ARU) k (bohr ) BiClTe = 44.621 fs 0.30 0.06 0.25 0.04 0.20 0.02 125 0.00 0.15 0.02 0.10 0.04 0.05 0.06 0.00 0 0 2 4 6 0.10 0.05 0.00 0.05 0.10 |v| (ARU) k (bohr ) FIG. 6. Velocity and scattering plots of GaSe-e, Bi Se -e, Sb SeTe -e, and BiClTe-e, in order of decreasing conductivity. 2 3 2 2 Energy (from edge) (eV) Energy (from edge) (eV) Energy (from edge) (eV) Energy (from edge) (eV) 1 k (bohr ) k (bohr ) 1 1 k (bohr ) k (bohr ) y y g (meV) g (meV) g (meV) g (meV) eff eff eff eff 13 Bi STe 2 2 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 0 1 2 3 4 5 6 7 |v| (ARU) FIG. 7. Velocity plot of neutral Bi STe -e. Electron-phonon scattering was not computed but it is expected to be similar to 2 2 Bi SeTe -e. 2 2 Energy (from edge) (eV)

Journal

Condensed MatterarXiv (Cornell University)

Published: Jul 31, 2020

There are no references for this article.