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

Learn More →

Semiclassical Boltzmann transport theory of few-layer black phosphorus in various phases

Semiclassical Boltzmann transport theory of few-layer black phosphorus in various phases Semiclassical Boltzmann transport theory of few-layer black phosphorus in various phases Sanghyun Park, Seungchan Woo, and Hongki Min Department of Physics and Astronomy, Seoul National University, Seoul 08826, Korea (Dated: April 4, 2019) Black phosphorus (BP), a two-dimensional (2D) van der Waals layered material composed of phosphorus atoms, has been one of the most actively studied 2D materials in recent years due to its tunable energy band gap (tunable even to a negative value) and its highly anisotropic electronic structure. Depending on the sign of the band gap tuning parameter, few-layer BP can be in a gapped insulator phase, gapless Dirac semimetal phase, or gapless semi-Dirac transition point between the two phases. Using the fully anisotropic multiband Boltzmann transport theory, we systematically study the dc conductivity of few-layer BP as a function of the carrier density and temperature by varying the band gap tuning parameter, and determine the characteristic density and temperature dependence corresponding to each phase. I. INTRODUCTION eralized to anisotropic multiband systems, we calculate the dc conductivity as a function of the carrier density and temperature for each phase. We determine that each Since the discovery of graphene [1, 2], which is a carbon phase shows the characteristic density and temperature allotrope of two-dimensional (2D) honeycomb lattice, 2D dependence, which can be used as a transport signature materials have been one of the most active research areas of BP in di erent phases. in condensed matter physics. Black phosphorus (BP) is The rest of this paper is organized as follows. In a 2D material with van der Waals layered structure com- Sec. II, we describe our model Hamiltonian and develop posed of phosphorus atoms, and it has recently attracted the Boltzmann transport theory in anisotropic multiband considerable attention [3, 4]. As a layered semiconductor systems. In Sec. III, we present the dc conductivity of BP in its natural form, BP has a tunable band gap, and ma- in each phase as a function of density at zero tempera- nipulation of its band gap through various methods has ture. In Sec. IV, we provide the temperature dependence been validated by multiple theoretical and experimental of dc conductivity at a xed density. We conclude our reports [5]. Notable examples of the band gap tuning paper in Sec. V with discussions on the dominant scat- include thickness change [6, 7], strain control [8], pres- tering source, the e ect of potential uctuations at low sure [9], electronic gating [10{12], and chemical doping densities, and the e ect of the parabolic term omitted in [13]. Some of the band gap manipulation methods [9, 13] the current model. demonstrated that the band gap can be tuned to zero, showing the semi-Dirac state with a combination of lin- ear and quadratic dispersions [14], which is also predicted II. METHODS in TiO =VO heterostructures [15, 16]. Furthermore, the 2 2 band gap can be inverted, leading to the Dirac semimetal A. Model phase [17{19]. Due to its anisotropic electronic band structure, BP shows many peculiar transport properties such as large By expanding the tight-binding lattice model of few- in-plane anisotropic transport [20, 21]. The e ects of layer BP [4, 30, 31], the corresponding low-energy e ec- temperature [10, 22, 23], the number of layers [10], and tive Hamiltonian can be obtained as [32{36] substrate [22] on the anisotropic transport properties of 2 2 BP have been studied experimentally. Furthermore, the ~ k " H = +  + ~v k  ; (1) x 0 y y transport properties of BP have been studied theoret- 2m 2 ically [24{29], demonstrating its anisotropic nature in energy- and temperature-dependent transport. However, where m is the e ective mass along the zigzag (x) di- there has been no systematic study on the anisotropic rection, v is the band velocity along the armchair (y) transport of BP in each phase, fully considering the direction, " is the size of the band gap (which will be anisotropy of the system and the interband scattering. used as a tuning parameter), and  and  are the Pauli x y In this study, we theoretically investigate the transport matrices. The eigenenergies of the Hamiltonian are given properties of BP in the gapped insulator phase, gapless 2 2 2 ~ k " x 2 2 2 by " =  + + ~ v k ; thus, the Hamilto- semi-Dirac transition point, and Dirac semimetal phase. 0 y 2m 2 Using the semiclassical Boltzmann transport theory gen- nian H has a direct band gap for " > 0, a semi-Dirac band touching point at (k ; k ) = (0; 0) for " = 0, or two x y g m j" j Dirac points at (k ; k ) = ( ; 0) for " < 0. The x y 2 g hmin@snu.ac.kr characteristic energy scales along the zigzag and arm- arXiv:1811.03903v2 [cond-mat.mes-hall] 3 Apr 2019 2 2 2 1.2 1.2 2 ~ k chair directions are given by " = and ~v k , re- 0  0 0 2m 1 1 1.5 spectively, where k = a and a is the lattice constant. 0.8 0.8 We introduce the dimensionless parameters  = and 2" 0.6 0.6 1 ~v k 0 0 c = , which represent a gap tuning parameter and " 0.4 0.4 0.5 the ratio of the characteristic energy scales along the 0.2 0.2 zigzag and armchair directions, respectively. Through- 0 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 out the paper, we use c = 1 and the spin degeneracy g = 2 for the calculation. We will discuss the e ect of higher-order terms omitted in Eq. (1) in Sec. V. 6 5 7 0 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 FIG. 2. (a)-(c) Calculated DOS and (d)-(e) the carrier den- sity as a function of Fermi energy for the (a), (c) insulator phase, (b), (e) semi-Dirac transition point, and (c), (f ) Dirac semimetal phase. Here,   is the band gap tuning pa- 2" rameter, and g = 2 and c = 1 are used for calculation. FIG. 1. (a)-(c) Energy dispersions and (d)-(f ) the corre- sponding Fermi surfaces of few-layer BP for the (a), (c) insu- lator phase, (b), (e) semi-Dirac transition point, and (c), (f ) a function of Fermi energy for each phase. At the Dirac semimetal phase. semi-Dirac transition point, the DOS is simply given by 1=2 D(")  " [Fig. 2(a)], and the carrier density (which Figure 1 shows the energy dispersion and the corre- is an energy integral of the DOS up to " ) is given by sponding Fermi surface of few-layer BP in each phase. 3=2 n  " [Fig. 2(d)]. (See Appendix A for the detailed Initially, few-layer BP without band gap tuning is in the derivations of the DOS and the carrier density.) In the gapped insulator phase, as shown in Fig. 1(a). As the gapped insulator phase, both DOS at " and carrier den- band gap " decreases (for example, upon applying a sity vanish for " < " =2, whereas for " > " =2, they F g F g perpendicular electric eld), eventually it vanishes and follow those of the semi-Dirac transition point as " in- the system is described by the semi-Dirac Hamiltonian creases [Figs. 2(b) and 2(e)]. In the Dirac semimetal in Eq. (1) with " = 0, as shown in Fig. 1(b). If the phase, when " is very small, the system resembles band gap decreases even further and becomes negative a typical 2D Dirac semimetal such as graphene; thus, (" < 0), band inversion occurs, which has been achieved g 1 D(")  " . As " increases and approaches " =2 near the F g experimentally using surface doping [17, 18] and external top of the inverted band, the band dispersion e ectively pressure [19] becomes hyperbolic paraboloid with a di erent sign in In the gapped insulator phase, the inherent anisotropy each direction in momentum space. Subsequently, a van of the system is less evident and the system at low Hove singularity occurs in the DOS, diverging logarith- densities resembles typical semiconductors with a dif- mically with D(")  log(jj") [37]. If " increases ferent e ective mass in each direction. At the semi- further, the DOS and the carrier density follow those of Dirac transition point, the energy dispersion becomes the semi-Dirac transition point with a discontinuous en- linear (quadratic) along the armchair (zigzag) direction, ergy derivative in the DOS at the van Hove singularity as shown in Fig. 1(e). At the Dirac semimetal phase, [Figs. 2(c) and 2(f )]. the anisotropy in the energy dispersion becomes more pronounced and the Fermi surface vastly changes its shape depending on the value of the Fermi energy " . Notably, as the energy dispersion and the Fermi surface For " < " =2, the Fermi surface becomes two distinct are anisotropic, and the Fermi energy can cross multiple F g lines, as shown in Fig. 1(f ), whereas for " > " =2, the bands, we cannot naively use the conventional Boltzmann F g two Fermi surfaces become joined completely, forming a transport theory assuming an isotropic single-band sys- closed line. At " = " =2, a van Hove singularity occurs tem. Thus, the anisotropic multiband Boltzmann trans- F g in the density of states (DOS), as explained below. port theory is necessary to calculate the dc conductivity Figure 2 shows the DOS and the carrier density as of such systems, as explained in Sec. II B. 3 B. Boltzmann transport theory in anisotropic obtain the following integral equation for the relaxation multiband systems time: (i) d 0 d k v (i) 0 0 (i) We use semiclassical Boltzmann transport theory to 1 = W   : (7) 0 0 0 kk k k (i) (2) calculate the density and temperature dependence of the 0 v dc conductivity of few-layer BP in each phase in the pres- This is a coupled integral equation relating the relax- ence of impurities, assuming elastic scattering (see Sec. V ation times at di erent angles in di erent bands, which for the limitation of the current approach). In the Boltz- correctly considers the anisotropy and multiple bands of mann transport theory, electron states are described by the system. Note that, for an isotropic single-band sys- the non-equilibrium distribution function f = f (r;k; t). (i) Its time rate of change is balanced out by the collision tem [ =  (") for a given energy " = " ], Eq. (7) is term, which represents the total scattering probability reduced to the well-known expression for the relaxation df df time given by [43] per unit time, i.e., = . dt dt coll d 0 We assume a spatially homogeneous system without 1 d k 0 0 = W (1 cos  ): (8) explicit time dependence in the distribution function, i.e., kk kk (2) f = f . Thus, the time derivative of the distribution df @f _ k function is given by = k  , whereas the collision The current density J induced by an electric eld E is dt @k term is given by thus given by d 0 X X df d k d k (i) (i) (j) 0 0 = W (f f ); (2) kk k k J = g (e)v f   E ; (9) k ij d k dt (2) (2) coll 2 2 where W 0 = n jV 0j (" " 0 ) is the transition kk imp kk k k ~ where  is the conductivity tensor given by ij rate from k to k for an elastic scattering with the impu- 0 d rity potential V and the impurity density n . In the X kk imp d k (i) (j) (j) 2 (0) _  = ge S (")v v  : (10) ij presence of a uniform electric eld E, ~k = (e)E, and k k k (2) to the leading order in E, We nd that the Hall conductivity (i 6= j) vanishes, thus (0) (0) df @f @f k k we consider only the diagonal part of the dc conductivity (e)E  = (e)E  v ; (3) dt ~@k @" k (i = j). (0) 1 @" k (0) (" ) where v = and f = f (" ) = e + 1 k k ~ @k is the Fermi{Dirac distribution function at equilibrium III. DENSITY DEPENDENCE OF DC with = and the chemical potential . Assume CONDUCTIVITY k T that, to the leading order in E, the non-equilibrium dis- (0) tribution function f is given by f  f (") + f at k k k Using the anisotropic multiband Boltzmann transport df df theory developed in Sec. II B, we calculate the dc conduc- energy " = " . Thus, from = , we obtain dt dt coll tivity of few-layer BP as a function of the carrier density d 0 or Fermi energy at zero temperature for each phase: the d k (0) (e)E  v S (") = W 0 (f f 0 ); (4) semi-Dirac transition point ( = 0), gapped insulator k kk k k (2) phase ( > 0), and Dirac semimetal phase ( < 0), all (0) of which can be expressed by Eq. (1). @f (") (0) where S (") = . If the Fermi energy crosses @" As for the impurity potential, we consider two types multiple energy bands, Eq. (4) is generalized to [38, 39] of impurity scattering: short-range impurities and long-range Coulomb impurities (or charged impurities). d 0 0 0 d k (0) (e)E  v S (") = W f f ; Short-range impurities originate from lattice defects, va- 0 0 kk k k (2) cancies, dislocations, etc., and their potential form is (5) 0 given by a constant in momentum space, V = V , kk short where and are band indices. as they are approximately represented by the delta func- We parameterize f in the following form [40{42]: tion in real space. For charged impurities distributed randomly in the background, the impurity potential is d 2 X 2e given by V 0 = in 2D, where (q) is the dielec- (i) (i) (i) (0) kk (q)jqj f = (e) E v  S ("); (6) k k tric function for q = k k . Within the Thomas{Fermi i=1 approximation, (q) can be approximated as (q) (i) (i) (i) (1 + q =jqj), where  is the background dielectric con- TF where E , v , and  are the electric eld, velocity, k k 2e and relaxation time, respectively, along the ith direction stant, q = D(" ) is the Thomas{Fermi wave vec- TF F (i) for each band. After matching each coecient in E , we tor, and D(" ) is the total DOS at the Fermi energy " F F 4 1 2 (including all the contributions from the bands crossing (g  1), V  q  D (" )  " ; thus, we obtain 0 F F TF F " and the spin degeneracy). The interaction strength for charged impurities can be characterized by an e ective 2 2 e 3 "  n ; (12a) xx ne structure constant = . Note that q / g . F 0 TF 0 ~v Thus, the screening strength for Coulomb impurities is "  n : (12b) yy also characterized by . At general screening strength, the power-law behavior is determined by the competition between the screening A. Semi-Dirac transition point wave vector and the momentum transfer. We present the numerically calculated power-law behavior for the semi- Dirac transition point and for the other phases in Fig. 6. 7 8 20 6 15 B. Insulator phase 4 10 2 5 6 6 14 0 0 0 5 5 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 4 4 3 3 3 3 8 2 2 2.5 2.5 6 1 1 0 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 2 1.5 4 1.5 2 1.4 1.4 4 0.5 1.2 1.2 1 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 1 1 0.8 0.8 0.6 0.6 FIG. 3. Calculated dc conductivities (a)-(c)  and (d)-(f ) xx 0.4 0.4 as a function of Fermi energy at the semi-Dirac transi- yy 0.2 0.2 tion point ( = 0) for (a), (d) short-range impurities, (b), 0 0 0 (d) charged impurities with = 1000, and (c), (f ) charged 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 2 2 2 ge k c impurities with = 1. Here,  = . 0 0 2~n imp FIG. 4. Calculated dc conductivities (a)-(c)  and (d)-(f ) xx as a function of Fermi energy in the insulator phase with yy First, let us consider the semi-Dirac transition point = 1 for (a), (d) short-range impurities, (b), (d) charged ( = 0). Figure 3 shows the Fermi energy depen- impurities with = 1000, and (c), (f ) charged impurities dence of dc conductivity at the semi-Dirac transition with = 1. point. The characteristic density or Fermi energy depen- dence of the dc conductivity can be understood as fol- (i) Figure 4 shows the Fermi energy dependence of the 1 2 lows. From Eq. (10) with   D (" )=V , we expect dc conductivity in the insulator phase ( > 0). In the (i) (i) (i) (i) (i) 2 2 2 D(" )[v ]   [v ] =V , where  and v ii F F F F F F F insulator phase, the power-law dependence of the dc con- are the relaxation time and velocity, respectively, at the ductivity at low densities becomes similar to that of 2D Fermi energy along the ith direction, and V is the angle- electron gas (2DEG) with a di erent e ective mass in averaged squared impurity potential at the Fermi energy. each direction. (See Appendix C 1 for detailed deriva- 1=2 At the semi-Dirac transition point, D(" )  " , and the tions.) (x) 1=2 Fermi velocity in each direction is given by v  " For short-range impurities, the power-law dependence F F (y) of the dc conductivity at low densities is given by and v  " , from which we can deduce the power-law F F behavior of the dc conductivity. (See Appendix A for the " ; (13a) xx detailed derivations of the power-law dependences.) " : (13b) yy For short-range impurities, V is a constant indepen- F dent of density; in this case, we obtain For charged impurities, in the strong screening limit, at 3 low densities, we obtain "  n ; (11a) xx 0 0 "  n : (11b) yy " ; (14a) xx For charged impurities, in the strong screening limit   " : (14b) yy F 5 Note that, as the Fermi energy or the carrier density For charged impurities, due to the dominant contribution increases, the power-law dependence becomes similar to from the diverging Thomas{Fermi wave vector q / TF that of the semi-Dirac transition point. D(" ), the conductivity is largely given by the square of the DOS as follows: [log (jj " )] ; (18a) xx F C. Dirac semimetal phase [log (jj " )] : (18b) yy F As the Fermi energy or the carrier density increases 8 10 25 7 further, the power-law dependence of the dc conductiv- 8 20 ity becomes similar to that of the semi-Dirac transition 6 15 point, as in the insulator phase. 4 10 2.08 1.08 2.1 2 5 2.06 1.06 1 0 0 1.9 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 2.04 1.04 1.8 2.02 1.02 1.7 5 5 14 1.6 2 1 4 4 1.5 0 2 4 0 2 4 0 2 4 10 10 10 10 10 10 10 10 10 3 3 2 2 1.2 1.08 2.05 1 1 2 2 1.06 1.15 1.95 0 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 1.04 1.1 1.9 1.02 1.85 1.05 1.8 FIG. 5. Calculated dc conductivities (a)-(c)  and (d)- xx (f )  as a function of Fermi energy in the Dirac semimetal yy 1.75 0 2 4 0 2 4 0 2 4 10 10 10 10 10 10 10 10 10 phase with  = 1 for (a), (d) short-range impurities, (b), (d) charged impurities with = 1000, and (c), (f ) charged impurities with = 1. FIG. 6. (a)-(c) d log  =d log " and (d)-(f ) xx F d log  =d log " as a function of for charged impurities yy F 0 Figure 5 shows the Fermi energy dependence of the dc in each phase. The red dashed lines represent the Fermi en- ergy exponents obtained in the strong screening limit. Here, conductivity in the Dirac semimetal phase ( < 0). In " = " for the semi-Dirac transition point, " = 1:01" for F 0 F 0 the Dirac semimetal phase, the power-law dependence of the gapped insulator phase, and " = 0:01" for the Dirac F 0 the dc conductivity at low densities becomes similar to phase are used for the calculation. that of graphene but with a di erent Fermi velocity in each direction. (See Appendix C 2 for detailed deriva- Figure 6 shows the evolution of the Fermi-energy power tions.) law of the dc conductivity as a function of the screen- For short-range impurities, the power-law dependence ing strength for each phase in the low carrier density of the dc conductivity at low densities is given by 0 limit. For the insulator phase and the semi-Dirac transi- " ; (15a) tion point, the Fermi-energy exponent decreases, whereas xx for the Dirac semimetal phase, it shows a non-monotonic " : (15b) yy behavior with a dip structure, which originates from the For charged impurities, in the strong screening limit, at interband-like scattering between two distinct Fermi sur- low densities, we obtain faces shown in Fig. 1(f ). As the screening strength in- creases, all the Fermi-energy exponents approach the cor- " ; (16a) xx F responding power law estimated in the strong screening limit. " : (16b) yy Near the van Hove singularity, "  " =2, the DOS F g diverges logarithmically [37] and it dominates the overall IV. TEMPERATURE DEPENDENCE OF DC CONDUCTIVITY power-law behavior of conductivity [39]. Therefore, for short-range impurities, the conductivity becomes We can apply the anisotropic multiband Boltzmann [ log (jj " )] ; (17a) xx F transport theory developed in Sec. II B to the dc conduc- [ log (jj " )] : (17b) tivity at nite temperature. In Eq. (10), the nite tem- yy F 6 3 15 25 perature a ects the conductivity through the Fermi dis- 2.5 tribution and the temperature-dependent screening for the charged impurity potential. At nite temperatures, the chemical potential of the system also deviates from 1.5 the Fermi energy " due to the broadening of the Fermi distribution function. From the invariance of carrier den- sity n with respect to temperature T , we obtain the 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 temperature dependence of the chemical potential (T ). For charged impurities, the nite temperature Thomas{ 2 1.1 2.5 3 2e @n Fermi screening wave vector is given by q (T ) = TF @ 1 2 2.5 for 2D systems. (See Appendix D for the detailed deriva- 0.9 1.5 2 tion of the temperature dependence of the chemical po- 0.8 tential and Thomas{Fermi wave vector). In this section, 0.7 1 1.5 0.6 we calculate the dc conductivity of few-layer BP as a 0.5 1 0.5 function of the temperature for each phase. The detailed 0.4 0 0.5 derivation of the temperature-dependent conductivity is 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 presented in Appendices E and F. FIG. 7. Calculated dc conductivities (a)-(c)  and (d)-(f ) xx as a function of the temperature at the semi-Dirac transi- yy A. Semi-Dirac transition point tion point ( = 0) for (a), (d) short-range impurities, (b), (d) charged impurities with = 1000, and (c), (f ) charged impu- rities with = 1. Here, if the temperature is normalized by From the power-law dependence of the DOS, D(")  0 1=2 T = " =k , the result is independent of " at the semi-Dirac F F B F " at the semi-Dirac transition point [Fig. 2(a)], we transition point. The blue dashed-dotted lines and red dashed can obtain the asymptotic behaviors of (T ) and q (T ) TF lines represent tting by the corresponding asymptotic form in a relatively straightforward manner. In the low- and [Eqs. (21) and (22)] in the low- and high-temperature limits, high-temperature limits, the chemical potential at the respectively. semi-Dirac transition point is given by >  T 1 (T  T ); F For charged impurities, the asymptotic behavior is given 12 T =   (19) by " 1 T (T  T ); 1 5 F 8 2( )( ) 2 2 1 + C (T  T ); xx F (T ) xx T whereas the Thomas{Fermi wave vector is given by =   (22a) (0) > xx D (T  T ); 8 xx F < 8 1 (T  T ); q (T ) 12 T TF F < T = 1 (20) 1 + C (T  T ); (T ) yy F yy 2 F q (0) TF 1 3 T = (22b) 2 (T  T ); 2 2 T  (0) : yy D (T  T ); yy F where is the Gamma function and  is the Dirichlet where C (D ) indicates the low- (high-) temperature co- ii ii eta function [44]. In a single-band system, q (T ) typ- TF ecients. In the strong screening limit, the coecients ically decreases with the temperature at high tempera- 2 2 become C = 0, D = , C = , and D = xx xx yy yy tures, whereas at the semi-Dirac transition point, q (T ) 6 4 TF log 2. As the screening strength decreases, the high- increases with the temperature due to the thermal exci- temperature coecients D remain positive, whereas the ii tation of carriers participating in the screening. low-temperature coecients C decrease and we expect ii Figure 7 shows the temperature dependence of the dc that the initially negative or vanishing C would even- ii conductivity at the semi-Dirac transition point, normal- tually become positive in the weak screening limit. (See ized by the zero-temperature conductivity value in each Appendix E for the detailed derivations of the coecients direction. For short-range impurities, we determine that C and D .) ii ii the asymptotic behavior is given by The temperature dependence in the high-temperature limit can be easily understood by replacing " with <  T 1 (T  T ); (T ) F xx 12 T F T in the Fermi energy dependence of dc conductivity = (21a) : T (0) [Eqs. (11) and (12)]. At high temperatures,  (T ) for xx yy log 2 (T  T ); short-range impurities decreases with the temperature, T =T <1 e (T  T ); showing a metallic behavior. Otherwise, the conductiv- (T ) yy = (21b) 2 ities increase with the temperature, showing an insulat- 1 1 T (0) : + (T  T ): yy F 1 5 2 T F ing behavior. Note that the high-temperature asymp- ( ) ( ) 2 2 7 totic form for charged impurities is obtained by consid- perature dependence of the chemical potential, Thomas{ ering the e ect of the energy averaging and that of the Fermi screening wave vector, and conductivity of the temperature-dependent screening separately. It correctly gapped 2DEG system.) predicts the temperature power-law dependence but not In the high-density limit, the temperature dependence the coecients in the asymptotic form, showing a dis- of dc conductivity in the insulator phase resembles that crepancy with the numerical result, as the e ect of tem- of the semi-Dirac transition point. perature cannot be simply separated into the energy av- eraging and the temperature-dependent screening at high temperatures. C. Dirac semimetal phase B. Insulator phase 1.1 15 0.9 25 120 400 10 0.8 1 3 0.7 2 5 80 5 0.5 0.6 60 200 0 0 0.5 0 0.5 0 0.5 40 0.4 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 0 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 1.1 15 0.9 8 50 140 0.8 1 3 0.7 0.5 1 0.6 0 0 0 0.5 0 0.5 0.5 0.4 0 40 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 0 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 FIG. 9. Calculated dc conductivities (a)-(c)  and (d)- xx (f )  in the low-density limit as a function of the temper- yy ature in the Dirac semimetal phase with  = 1 for (a), FIG. 8. Calculated dc conductivities (a)-(c)  and (d)-(f ) xx (d) short-range impurities, (b), (d) charged impurities with in the low-density limit as a function of the temperature yy = 1000, and (c), (f ) charged impurities with = 1. Here, 0 0 in the insulator phase with  = 1 for (a), (d) short-range " = 0:01" is used for the calculation. The blue dashed- F 0 impurities, (b), (d) charged impurities with = 1000, and dotted lines and red dashed lines represent tting by the cor- (c), (f ) charged impurities with = 1. Here, " = 1:1" is 0 F 0 responding asymptotic form [Eqs. (23) and (24)] in the low- used for the calculation. The blue dashed-dotted lines repre- and high-temperature limits, respectively. sent the result for the gapped 2DEG system (see Appendix F), and the red dashed lines represent power-law tting by the asymptotic form of the semi-Dirac transition point [Eqs. Figure 9 shows the calculated temperature-dependent (21) and (22)] in the high-temperature limit. conductivity in the Dirac semimetal phase, with the xed Fermi energy of " = 0:01" , which corresponds to the F 0 Figure 8 shows the temperature dependence of the dc low-density limit. At low densities, the Dirac semimetal conductivity in the insulator phase with the xed Fermi phase can be e ectively considered as graphene (with energy of " = 1:1" , which corresponds to the low- anisotropic velocities); thus, we can understand its F 0 density limit. At zero temperature, the insulator phase temperature-dependent conductivity behavior using the in the low-density limit can be e ectively considered as a result of graphene. (See Appendix F for the temperature gapped 2DEG (with anisotropic e ective masses). Simi- dependence of the chemical potential, Thomas{Fermi larly, at nite temperatures, the temperature-dependent screening wave vector, and conductivity of graphene.) conductivity of the insulator phase in the low-density For graphene with short-range impurities, the asymptotic limit resembles that of the gapped 2DEG system (blue form of the temperature-dependent conductivity becomes dash-dotted lines in Fig. 8), especially in the low- T =T temperature limit. In the high-temperature limit, the < 1 e (T  T ); (T ) gp power-law behavior of the temperature-dependent con- 2 = (23) 1 1 T (0) : gp + (T  T ); ductivity for the insulator phase becomes similar to that 2 16 log 2 T of the semi-Dirac transition point [Eqs. (21) and (22)], because thermally excited carriers above the gap con- whereas for charged impurities in the strong screening tribute to the conductivity. (See Appendix F for the tem- limit, the asymptotic form of the temperature-dependent 8 conductivity becomes tial (T ) [Fig. 13(b) in the Appendix], shifting the central point of the energy averaging. T In the high-density limit, the temperature dependence 1 (T  T ); (T ) 3 T gp F of dc conductivity in the Dirac semimetal phase resembles =   (24) (0)  T gp that of the semi-Dirac transition point. (T  T ): 6 T Similar to the result of the semi-Dirac transition point, V. DISCUSSION AND CONCLUSION the high-temperature asymptotic form for charged im- purities correctly captures the temperature power-law dependence (but not the exact coecient value, as dis- When we consider both short-range and charged impu- cussed in Sec. IV A). rities, assuming that each scattering mechanism is inde- pendent, the total scattering rate is obtained by adding 1.2 1.5 2.5 their scattering rates in accordance with Matthiessen's 1.1 rule. Note that the scattering mechanism with a higher scattering rate (or equivalently a lower conductivity) 1.5 dominates the resulting conductivity. From the obtained 0.9 Fermi-energy power-law dependence of dc conductivity, 0.8 0.5 we can determine the dominant scattering mechanism. At the semi-Dirac transition point, we can observe from 0.7 0.5 0 1 2 0 1 2 0 1 2 Eq. (11) and Figs. 6(a) and (d) that, for both  and xx , the Fermi-energy power law for short-range impuri- yy 1.4 1.6 3 ties is always smaller than that of charged impurities. 1.4 1.3 This indicates that, at low densities, charged impuri- 2.5 1.2 ties are dominant over short-range impurities, whereas 1.2 at high densities, short-range impurities are dominant 1.1 0.8 over charged impurities. In the insulator phase, at low 1.5 densities, the system can be approximated as a 2DEG 0.6 and the Fermi-energy power laws for short-range and 0.9 0.4 0 1 2 0 1 2 0 1 2 charged impurities are almost comparable (except in the no-screening limit) as shown in Eq. (13) and Figs. 6(b) and (e). At high densities, the power-law dependence FIG. 10. Calculated dc conductivities (a){(c)  and (d){ xx follows that of the semi-Dirac transition point; thus, (f )  immediately below the van Hove singularity point as a yy function of the temperature in the Dirac semimetal phase with short-range impurities dominate over charged impuri- = 1 for (a), (d) short-range impurities, (b), (d) charged ties. In the Dirac semimetal phase, at low densities, the impurities with = 1000, and (c), (f ) charged impurities 0 Fermi-energy power law for short-range impurities is al- with = 1. Here, " = 0:9" is used for the calculation. 0 F 0 ways smaller than that of charged impurities as shown in Eq. (15) and Figs. 6(c) and (f ); thus, charged impurities are dominant over short-range impurities as in the case Figure 10 shows the temperature dependence of the dc of graphene. At high densities, short-range impurities conductivity in the Dirac semimetal phase immediately become dominant over charged impurities, following the below the van Hove singularity point, exhibiting a non- trend of the semi-Dirac transition point. Note that, near monotonic behavior with temperature. As explained ear- the van Hove singularities, charged impurities are highly lier, the temperature dependence of the dc conductivity is (0) screened due to the enhanced DOS, and thus, short-range determined by the energy averaging with S (") broad- impurities are dominant over charged impurities [39]. ened by temperature, and by the temperature-dependent screening for charged impurities. Thus, if the Fermi en- Our analysis is based on the semiclassical Boltzmann transport theory, which is known to be valid in the high- ergy is near the van Hove singularity, the distance be- density limit. At low densities, the e ect of potential tween the Fermi energy and the van Hove singularity sets an important energy scale for the temperature de- uctuations induced by spatially inhomogeneous impu- rities becomes important, which is not captured by our pendence, k T  jj" jj" jj. For charged impurities, B 1 F g the conductivity rst increases, showing a peak at T , approach assuming a spatially homogeneous system. At ch the semi-Dirac transition point or in the Dirac semimetal and thereafter decreases, showing a dip at T  0:5T corresponding to the minimum of q (T ), mainly follow- phase, the potential uctuation is expected to result in TF a minimum conductivity [45{47]. In the insulator phase, ing the temperature dependence of the screening wave vector q (T ) [Fig. 13(e) in the Appendix]. For short- if the band gap is suciently large, the e ect of the po- TF tential uctuation might be limited. The interplay of the range impurities, the conductivity rst decreases, show- ing a dip at T , and thereafter increases, showing a peak impurity potential uctuation, temperature, and band sh gap would be an interesting future research direction. at T  0:25T . These dips and peaks are from the temperature-dependent evolution of the chemical poten- Finally, we wish to mention the additional parabolic 9 2 2 ~ k \interband" scattering between these two surfaces. The term  omitted in Eq. (1) along the armchair (y) 2m maximum value of  is thus given by direction beyond the lowest order [48]. This term could a ect the dc conductivity, especially at high densities arccos ( 6= 0 and jj < r); cr 2m v above the crossover Fermi energy " = , where (r) = (A3) ( < 0 and jj  r); max the e ective Hamiltonian in Eq. (1) is no longer valid. (otherwise); For example, at the semi-Dirac transition point with cr "  " , the parabolic term becomes dominant over where  2 [ (r);  (r)]. This coordinate transfor- max max the linear term along the armchair direction; thus, xx mation changes the Hamiltonian into the following form: and  will follow those of (anisotropic) 2DEG. yy In summary, we calculate the dc conductivity of few- 0 e H = " r : (A4) layer BP as a function of the density and temperature us- 0 e 0 ing the anisotropic multiband Boltzmann transport the- ory, which is essential when the e ect of anisotropic en- In the transformed coordinates, the energy dispersion is ergy dispersion or interband scattering becomes impor- given by " (r) = " r and the corresponding eigenstates tant. We nd that the dc conductivities in the Boltzmann are given by limit show characteristic density and temperature depen- dence in each phase, which could be used as a signature j+i = ; (A5a) of the tunable electronic structure of BP in transport measurements. 1 1 ji = : (A5b) The Jacobian J corresponding to this transformation ACKNOWLEDGMENTS is given by @k @k This work was supported by the National Research x x k r @r @ 0 J = = p  J (r; ): (A6) Foundation of Korea (NRF) grant funded by the Ko- @k @k y y 2c r cos @r @ rea government (MSIT) (No. 2018R1A2B6007837) and Creative-Pioneering Researchers Program through Seoul (i) Note that, for the + band, the band velocity v = National University (SNU). @" 1 +;k can be expressed as ~ @k (x) v = 2 v cos  r cos  ; (A7a) Appendix A: Eigenstates and density of states (y) v = v c sin ; (A7b) In this section, we provide a detailed explanation on where v = . ~k the model Hamiltonian of few-layer black phosphorus 0 The DOS at the semi-Dirac transition point ( = 0) (BP), and its various properties including density of at the energy " > 0 can be obtained analytically as states (DOS). In the model Hamiltonian given by Eq. (1) in the main text, the exact values of m and v depend d k D(") = g (" " ) on the number of layers and the gap tuning parameter. +;k (2) We introduce the normalization constants k  a and Z Z 2 2 1 ~ k J (r; ) "  ; thus, the Hamiltonian becomes = 2g dr d (" " r) 2m (2) ~ ~ p   1 0 k ick + y 2 x 2 H = " ; (A1) 2gk 2K (1=2) " 0 0 ~ ~ k + ick +  0 = ; (A8) c" " 0 0 where k = k=k , c = ~v k =" , and   . To avoid where g is the spin degeneracy, and the factor 2 0 0 0 0 2" diculties associated with anisotropic dispersion, we con- originates from the duplicate parts of the Fermi sur- sider the following coordinate transformation with faces parameterized by = 1. Here, K (k) = 1 2 2n [(2n 1)!!=(2n)!!] k is the complete elliptic in- n=0 tegral of the rst kind with K (1=2)  1:854 [44]. Note k ! k (r cos  ) ; x 0 (A2) that the Thomas{Fermi wave vector is determined by the k ! r sin ; DOS at the Fermi energy " given by p 1 2 2 2e 4g k 2K (1=2) " 0 0 F where = 1 represents each half of the Fermi sur- q = D(" ) = ; (A9) TF F faces. This Fermi surface splitting is especially useful for  c " the  < 0 case where there are two distinct Fermi sur- faces (see Fig. 1(f ) in the main text), accounting for the where = is the e ective ne structure constant. ~v 0 10 The carrier density is thus given by energy dispersion. To consider the anisotropy of the en- ergy dispersion, we express the multiband anisotropic Boltzmann equation in Eq. (7) using the transformed co- 4g 2K (1=2) " n = d"D(") = n ; (A10) 2 ordinates in Eq. (A6) as follows: 3 c " 0 0 2 1 3 3 where n = k . Note that "  n and D(" )  n . 0 F F Figure 2 in the main text shows the calculated DOS and the carrier density for each phase. Appendix B: Density dependence of dc conductivity in black phosphorus In this section, we derive the dc conductivity at zero temperature for 2D multiband systems with anisotropic Z Z (i) 1  (r ) max 0 0 J (r ;  ) v (i) 0 0 (i) 0 0 1 = dr d W 0 0 0 kk k k (i) 0 (2) 0  (r ) 0 max v Z Z 1 0 0 2 0 0 0 d k r 2 (i) (i) (i) 0 0 2 0 = dr p n jV j F (" r " r )  d 0 0 0 0 0 imp 0 0 kk kk k kk k 0 0 (2) 2c r cos   ~ 0 r 2 0 2 k d r 0 0 (i) (i) (i) 0 2 = n p jV j F  d  ; (B1) 0 0 0 0 0 imp kk kk k kk k ~ 2c" 2 2 r cos where = 1 represents each half of the Fermi surfaces, Since Eq. (B3) holds for both = 1, we can rewrite it (i) (i) (i) as d = v =v , and F = [1 + cos(  )] is 0 0 0 0 kk k k kk the square of the wave function overlap between k and (i) (i) (i) (i) 0 0 1 = w ~ () ~ () w ~ (;  ) ~ ( ) (B5a) k states in the same conduction (or valence) band. Let 11 1 11 1 2 r " 1 2 2 0 0 us de ne  = , V = , and = n V  , 0 2 0 0 0 imp 0 2c" k  (r) ~ d 0 0 0 (i) (i) (i) (i) 0 0 + w ~ () ~ () w ~ (;  ) ~ ( ); then we have 11 1 11 1 Z 0 0 0 X d (i) (i) (i) (i) d r 0 0 1 = w ~ () ~ () w ~ (;  ) ~ ( ) (B5b) 1 = 11 1 11 1 0 2 2 2 r cos 0 (i) (i) (i) (i) 0 0 0 0 (i) (i) (i) + w ~ () ~ () w ~ (;  ) ~ ( ): jV j F  ~ d  ~ ; (B2) 11 1 11 1 0 0 0 0 0 kk kk k kk k 0 0 Now, let us discretize  to  (n = 1; 2; ; N ) with (i) (i) (i) (i) where V = V =V and  ~ =  = . Here, 0 0 (i) (i) 0 0 kk kk k k R an interval  = 2 (r)=N . Thus, for  ~ =  ~ ( ), max n n 0 0 d represents an integration over  (r) <  < max r we have (r). Thus, Eq. (B2) becomes max (i) (i) (i) n (i) nn 1 = P  ~ P  ~ n1 11 n 1 0 0 (i) (i) (i) (i) 0 0 1 = w ~ () ~ () w ~ (;  ) ~ ( ) ; 0 0 0 0 (i) (i) (i) n (i) nn 2 + P  ~ P  ~ ; (B6a) r 11 0 n1 11 n 1 (B3) (i) 0 (i) (i) n (i) nn where 1 = P  ~ P  ~ 11 0 n1 11 n 1 d r 0 0 (i) 0 2 (i) (i) (i) n (i) nn w ~ () = p jV j F ; 0 0 0 + P  ~ P  ~ ; (B6b) 11 0 kk kk 0 n1 11 n 1 2 2 r cos (B4a) (i) (i) (i)nn 0 0 r where P 0 = w ~ ( ) is an N -vector and P = (i) (i) 0 0 0 2 n w ~ (;  ) = p jV j F d : 0 0 0 0 kk kk kk (i) 2 r cos w ~ ( ;  ) is an N  N matrix, which correlates 0 n n (B4b) the di erent -dependent relaxation times for a given 11 ( , ) combination. Note that Eq. (B6) shares the basic minimum point of the conduction band, we have structure with the multiband scattering formula [38, 39] 2 2 (which accounts for the scattering between each half of k " c k x 0 y "(k) = " + the Fermi surface) and the anisotropic scattering formula k 2 k 0 0 [42] (which accounts for the scattering between di erent 2 2 2 2 ~ k ~ k 0 y and  points). Furthermore, Eq. (B6) is a 2N  2N + ; (C1) 2m 2m x y matrix equation with two independent basis indices ( , ), i.e., index for each half of the Fermi surfaces and 2 2 2 2 ~ k ~ k 0 0 where m = and m = . the -discretization index n. x y 2 2" c " 0 0 For comparison, we rst consider a 2DEG with an Thus, the dc conductivity at zero temperature is given isotropic energy dispersion given by by 2 2 X 2 ~ k d k (i) (j) (j) "(k) = : (C2) = ge (" " )v v ij k F k k k 2m (2) Z Z (i) (j) (j) 1 0 2 As the system is isotropic, we can readily calculate the k r(" r " )v v 0 F 2 k k k = ge dr d conductivity of each case using the Einstein relation 2(2) c r cos 0 r Z Z (i) (j) (j) 1 0 X  = e D(" )D; (C3) iso F r(r r )v ~ v ~  ~ k k k =  dr p ; (B7) 2 r cos 0 r 2 where D = is the di usion constant and D(") = gm is the DOS for the isotropic 2DEG. The relaxation (i) (i) 2 2 2 2~ where  = ge  v  , r = " =" and v ~ = v =v . 0 0 0 F F 0 0 0 k k time at the Fermi energy  is given by Thus, from Eq. (A7), we have 2 0 Z 1 2n d k imp 2 0 = jV 0j (" " )(1 cos  ) xx kk F 2 (x) 2 ~ (2) = 2 r cos  r cos   ~ (); F F F 0 Z F 2 2n m d imp 2 0 = jV 0j (1 cos  ) (B8a) ~ 2~ (2) d r sin yy F 2 (y) 2n m imp = c p  ~ (): (B8b) V ; (C4) i2DEG 2 2 2 r cos 0 F ~ 2~ where V 0 is the angle-dependent potential on the Fermi Note that  , v ,  , and  are the density-independent 0 0 0 0 R 2 2 0 surface and V  jV 0j (1 cos  ) is the normalization constants in units of time, velocity, DOS, i2DEG 0 2 angle-averaged square of the impurity potential. and conductivity, respectively. Therefore, the dc conductivity of the isotropic 2DEG is given by Appendix C: Low-density approximate models for 2 2 gm v ~ 2~ 2 F the insulator phase and Dirac semimetal phase = e iso 2 2 2~ 2 2n mV imp i2DEG 2 2 2 ge ~ ~ k In this section, we derive the dc conductivity of low- 2 2 density approximate models for the insulator phase and 2n V 2m imp i2DEG Dirac semimetal phase. Note that the only anisotropy ge ~" considered in these models is the anisotropy in the ef- = ; (C5) 2n mV imp i2DEG fective mass or velocity with the same power-law depen- dence in momentum. 2 2 ~ k ~k F F where v = and " = . F F 2 m 2m Subsequently, let us consider the Fermi energy depen- dence of the dc conductivity using the Einstein relation 1. Insulator phase at low densities in Eq. (C3). For short-range impurities, V is a con- i2DEG stant independent of " ; thus, we have For the insulator phase, as well as the Dirac semimetal phase discussed later, the DOS and carrier density do " : (C6) not follow the simple power-law behavior. Therefore, 2 2 we utilize approximate models to understand the asymp- Here, we used v  k  " . For charged impurities in F F 2 2 totic behavior of dc conductivity at low densities. When the strong screening limit, V  q  D (" ) is i2DEG TF j" j > j" j but the carrier density is suciently small, F g also a constant; thus, the system can be approximated as a two-dimensional electron gas (2DEG). From the series expansion at the   " : (C7) F 12 For the anisotropic 2DEG with di erent e ective the x-direction, the conductivity  becomes xx masses in each direction, we introduce the following co- p Z 2 h i ge m m ordinate transformation [(k ; k ) ! (k; )]: x y (x) x y =  v xx F 2~ 2 m ge m m x x y ~ 2~ k ! k cos ; x = 2~ 2n m m V imp x y a2DEG r (C8) m 2 ~ d 2 2 k ! k sin ; k cos m F mm 2 p ge ~" m m m x y = : (C13) which gives the Jacobian dk dk = kdkd. The x y 2 2n V m m imp a2DEG (i) 1 @" band velocity v = can be expressed as ~ @k Similarly, when the electric eld and the current density are along the y-direction, the conductivity  becomes yy ~k (x) v = p cos ; p Z mm 2 2 h i ge m m x y (y) (C9) =  v yy F ~k F (y) 2 2~ 2 v = sin : mm ge ~" m = : (C14) 2n V m m imp y a2DEG Subsequently, the energy dispersion becomes isotropic in the transformed coordinates; thus, the DOS is given Therefore, the dc conductivities for the anisotropic case by are modi ed as =  ; (C15a) g m m xx iso x y D(") = : (C10) 2~ m =  : (C15b) yy iso The relaxation time of the anisotropic 2DEG for k at the Fermi energy can be obtained by solving the cou- Thus, for short-range impurities or charged-impurities in the strong screening limit, the Fermi energy dependence pled integral equation [Eq. (7) in the main text]. For short-range impurities or charged impurities in the strong of the dc conductivities for the anisotropic 2DEG follows 0 that of the isotropic 2DEG given by Eqs. (C6) and (C7). screening limit, the scattering potential V = V is kk 0 independent of the angle, thus it can be shown that Note that, as the Fermi energy or the carrier density (i) (i) (i) increases, the insulator phase can no longer be approxi- =    . Then the coupled equation can be k k mated by a 2DEG model, and the energy dispersion fol- simpli ed as lows that of the semi-Dirac transition point. Therefore, the power-law dependence eventually follows that of the (i) 2 0 1 2n d k v imp 0 2 k semi-Dirac transition point. 0 0 = jV j (" " ) 1 kk k k (i) 2 (i) ~ (2) m m 2n x y imp = V ; (C11) a2DEG 2 2. Dirac semimetal phase at low densities ~ 2~ (i) 2 v For the Dirac semimetal phase ( < 0), the series 2 d 2 2 where V  jV j 1 = jV j is the 0 0 (i) a2DEG (2) expansion at one of the band touching points gives angle-averaged square of the impurity potential for the (i) anisotropic 2DEG. Note that  is independent of the H (k) = 2 k  + ck  (C16) x x y y direction i. ~ (v k  + v k  ) ; x x x y y y Therefore, the conductivity of the anisotropic 2DEG is given by 2 " c" 0 0 where v = and v = . x y ~k ~k 0 0 For comparison, we rst consider an isotropic 2D Dirac d k 2 (i) (j) (j) = ge (" "(k))v v  semimetal with the Hamiltonian given by ij F (2) p Z 2 2 H (k) = ~v (k  + k  ) : (C17) ge m m x y d x x y y (i) (j) =  v v ; (C12) F F 2~ 2 The DOS is thus given by where  is the relaxation time at the Fermi energy. F gk g" D(") = = : (C18) When the electric eld and the current density are along 2 2 2~v 2~ v 13 The relaxation time at the Fermi energy  is given by the relaxation time of the anisotropic Dirac semimetals Z given by 2 0 1 2n d k imp 2 0 = jV 0j F 0(" " )(1 cos  ) kk kk F 2 Z (i) ~ (2) 2 0 1 2n d k v imp 0 0 0 0 2 = jV j F (" " ) 1 0 kk kk k k (i) 2 (i) 2n k d imp F ~ (2) 2 0 0 = jV 0j F ( )(1 cos  ) ~ 2~v 2 2n gvk imp = V ; (C26) 2n k agp imp F ~ 2~v v = V ; (C19) x y igp ~ 2~v (i) R 2 0 1 0 0 2 v jV j where F ( ) = (1 + cos  ) is the square of the wave 2 d 2 0 0 0 2 where V  jV j F ( ) 1 = 0 (i) R agp (2) 4 0 0 2 v 2 2 0 0 k function overlap and V  jV 0j F ( )(1cos  ) igp 0 2 is the angle-averaged square of the impurity potential is the angle-averaged square of the impurity potential. (i) for the anisotropic graphene. Note that  for the Therefore, the dc conductivity of the isotropic Dirac anisotropic Dirac semimetal is also independent of the semimetal is given by direction i. Therefore, the conductivity of the anisotropic Dirac gk v ~ 2~v = e iso semimetal is given by 2~v 2 2n k V imp igp 2 2 d k ge ~v 2 (i) (j) (j) = ge (" "(k))v v  (C27) = : (C20) ij F (2) 4n V imp igp ge vk d (i) (j) Subsequently, let us consider the Fermi energy depen- =  v v : F F 2~v v 2 x y 0 dence of the dc conductivity using the Einstein relation in Eq. (C3). For short-range impurities, V is a constant igp When the electric eld and the current density are along independent of " ; thus, we have the x-direction, the conductivity  becomes xx " ; (C21) 2 h i ge vk d (x) =  v whereas for charged impurities in the strong screening xx F 2~v v 2 2 2 x y 2 2 0 limit, V  q  D (" )  " ; thus, igp Z TF F 2 2 ge vk ~ 2~v v d x y 2 2 2 = v cos " : (C22)  2 F 2~v v 2n vkV 2 x y imp agp 0 2 2 2 2 2 Note that, even in the weak screening limit, V  k  ge ~v v igp F = : (C28) 2 2 2 " , and in general,   " for charged impurities. 4n V imp agp F F For the anisotropic Dirac semimetals with di erent ve- Similarly, when the electric eld and the current density locities in each direction, we introduce the following co- are along the y-direction, the conductivity  becomes ordinate transformation [(k ; k ) ! (k; )]: yy x y 2 2 h i ge vk d (y) k ! k cos ; =  v yy F x F 2~v v 2 (C23) x y k ! k sin ; 2 2 2 ge ~v = : (C29) 2 2 4n V v imp agp which gives the Jacobian dk dk = kdkd. The x y v v x y Therefore, the dc conductivities for the anisotropic case (i) 1 @" band velocity v = can be expressed as k ~ @k are modi ed as (x) v = v cos ; x x =  ; (C30a) xx iso (C24) (y) v = v sin : =  : (C30b) yy iso Subsequently, the energy dispersion becomes isotropic in the transformed coordinates; thus, the DOS is given Thus, for short-range impurities or charged-impurities in by the strong screening limit, the Fermi energy dependence gvk g" of the dc conductivities for the anisotropic graphene fol- D(") = = : (C25) 2~v v 2~ v v lows that of the isotropic graphene given by Eqs. (C21) x y x y and (C22). Similarly, using the same assumptions which were used Near the van Hove singularities, where the energy dis- in Eq. (C11), for short-range impurities or charged im- 2 2 c k 2 y persion can be expanded as "(k)="  jjk + , the purities in the strong screening limit, we can calculate 0 2jj 14 DOS diverges logarithmically [37], dominating the over- Thus, we have all power-law behavior of the conductivity. Therefore, Z q (T ) TF for short-range impurities, the conductivity becomes = d"D(") (D3) q (0) 2D(" ) TF F [ log(jj " )] ; (C31a) xx F 1 1 + : [ log(jj " )] : (C31b) 1 + cosh (" ) 1 + cosh (" + ) yy F For a given T , the chemical potential is calculated us- For the charged impurities near the van Hove singulari- ing the density invariance in Eq. (D2), and subsequently, ties, the conductivity becomes q (T ) is obtained from the above relation. TF When the DOS is given by a simple power law with [log(jj " )] ; (C32a) xx F respect to energy, we can analytically obtain the temper- [log(jj " )] : (C32b) yy F ature dependence of the chemical potential and Thomas{ Fermi wave vector, and their asymptotic behaviors at low Note that, as the Fermi energy or the carrier density and high temperatures. increases, the power-law dependence of the dc conductiv- Consider a gapless electron{hole system with a DOS ity follows that of the semi-Dirac transition point, as in given by D(") = C j"j ("), where C is a constant the gapped insulator case. and (") is a step function. Using the results from the Supplemental Materials in Ref. [42], we can obtain Appendix D: Temperature dependence of chemical < 1 (T  T ); 12 T potential and Thomas{Fermi wave vector in black = (D4) " : 1 T phosphorus (T  T ); 2( 1)( +1) T where T = " =k is the Fermi temperature,  is the F F B In this section, we derive the temperature-dependent Dirichlet eta function, and is the gamma function [44]. chemical potential and Thomas{Fermi wave vector of For the temperature-dependent Thomas{Fermi wave vec- few-layer BP. When the temperature is nite, the chemi- tor q (T ), we obtain TF cal potential  deviates from the Fermi energy " due to the broadening of the Fermi distribution function (0) (") 1 ( 1) (T  T ); f ("; ) = e + 1 where = . As the q (T ) TF 6 T k T =   (D5) charge carrier density n does not vary under the temper- q (0) > T TF 2( 1)( ) (T  T ); ature change, we have For few-layer BP at the semi-Dirac transition point, (0) n = d"D(")f ("; ) 2 the DOS is given by D(") / " ; thus, = . Thus, we have 1 h i (0) (0) = d"D(") f ("; ) + f ("; ) >  T 1 (T  T ); 0 F 12 T F 1 =   (D6) " > 1 T d"D("): (D1) (T  T ); 1 5 ( ) ( ) 1 2 2 and Thus, the carrier density measured from the charge neu- tral point, n  nj nj , is given by 2 =0 1 (T  T ); q (T ) 12 T TF F 1 h i =   1 (D7) (0) (0) q (0) TF 1 3 T n = d"D(") f ("; ) f (";) 2 (T  T ); 2 2 T where q (0) = q is given by Eq. (A9). TF TF d"D("); (D2) Figure 11, Figure 12, and Figure 13 show the calcu- lated temperature dependence of the chemical potential where the rst and second lines represent the carrier den- (T ) and Thomas{Fermi wave vector q (T ) in various TF sity evaluated at the nite and zero temperatures, respec- phases of BP using Eqs. (D2) and (D3), respectively. tively. Here, we used f ("; ) = 1f (";). By solving this equality in terms of , we can calculate the chemical potential of the system for a given temperature T . See Appendix E: Temperature dependence of dc the Supplemental Material in [42] for the simpli ed cases. conductivity at the semi-Dirac transition point Subsequently, consider the temperature-dependent Thomas{Fermi wave vector q (T ). Note that, in 3D, TF Using Eq. (10) in the main text, we can general- 2 2 2e 2e @n q (0) = D(" ) and at nite T , q (T ) = . ize the conductivity tensor at zero temperature to that TF F TF @ 15 1 2.2 1 1 1 0.9 0.9 0.8 0.8 0.8 1.8 0.8 0.7 0.6 0.6 1.6 0.6 0.7 0.5 1.4 0.4 0.4 0.6 0.4 1.2 0.3 0.2 0.2 0.5 0.2 0.4 0.8 0 0 0.1 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 0.3 0.6 0 1 2 3 4 0 1 2 3 4 2.5 2.5 2.4 2.2 FIG. 11. Calculated temperature dependence of (a) chem- 2 2 ical potential and (b) Thomas{Fermi wave vector for the 1.8 1.5 1.5 semi-Dirac transition point ( = 0). Here, the black solid 1.6 lines represent the numerical result, and the red/blue dashed 1.4 1 1 1.2 lines represent the high-/low-temperature asymptotic forms 0.5 0.5 in Eqs. (D6) and (D7). If the chemical potential and temper- 0.8 ature are normalized by " and T , respectively, the result is F F 0 0 0.6 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 independent of " at the semi-Dirac transition point. FIG. 12. Calculated temperature dependence of (a)-(c) chemical potential and (d)-(f ) Thomas{Fermi wave vector for the gapped insulator phase with  = 1 at (a), (d) " = 1:01" , F 0 (b), (e) " = 1:1" , and (c), (f ) " = 1:5" . F 0 F 0 (0) 1 " at nite temperature. For f (") = z e + 1 , (0) @f (") (0) where z = e is the fugacity, S (") = = @" 1 " z e (0) (0) f (") 1 f (") = . Thus, the conduc- 1 " 2 (z e +1) tivity at nite temperature is given by 2 (0) d k @f (" ) k (i) (j) (j) (T ) = ge v v ij k k k (2) @" Z Z 1 0 2 1 " r X 0 k r z e (i) (j) (j) 2 0 = ge dr d p v v k k k 1 " r 2 2c r cos   (z e + 1) 0 r Z Z 1 0 1 " r X 0 d r " z e 0 (i) (j) (j) =  dr p v ~ v ~  ~ : (E1) k k k 1 " r 2 2 (z e + 1) 2 r cos 0 r Thus, from Eq. (A7), we have Z Z 1 0 1 " r X 0 p " z e d 2 (x) (T ) = 2 dr r cos  r cos  ~ (); (E2a) xx 0 1 " r 2 (z e + 1) 2 0 r Z Z 1 0 2 1 " r X 0 " z e d r sin 2 (y) (T ) = c  dr p  ~ (): (E2b) yy 0 1 " r 2 (z e + 1) 2 2 r cos 0 r To derive the asymptotic behaviors of  (T ) at low and is the energy-independent correction term from the ii (i) high temperatures, assume that the relaxation time can temperature-dependent screening e ect with g (0) = 1. be decomposed into energy- and temperature-dependent (i) T For short-range impurities, g = 1. For charged (i) (i) (i) T (i) T parts as  ("; T ) =  (")g where g T T F F 1.2 1.1 1.1 0.9 0.9 For charged impurities in the strong screening limit, 0.8 0.8 0.8 (i) 0.7 A = , which is two times the low-temperature co- 0.6 0.7 0.6 0.6 ecient in Eq. (D7), and = . Thus, we obtain 0.4 0.5 12 2 0.5 0.4 0.2 0.4 0.3 0 0.2 0.3 1 (T  T ); 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 F (T ) xx = (E5a) (0) : (T  T ); xx 6 T 6 1.6 1.5 1.4 1 (T  T ); 5  (T ) yy 1.4 4 T 1.3 = (E5b) 4  (0) : yy log 2 (T  T ): 1.2 1.2 3 1.1 As the screening strength decreases, the low-temperature 0.9 coecient in Eq. (E3) increases, because the screening 0.8 0.8 (i) coecient A decreases whereas the other part remains 0 0.6 0.7 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 positive. FIG. 13. Calculated temperature dependence of (a)-(c) Appendix F: Temperature dependence of dc chemical potential and (d)-(f ) Thomas{Fermi wave vector conductivity in the low-density approximate models for the Dirac semimetal phase with  = 1 at (a), (d) for the insulator phase and Dirac semimetal phase " = 0:01" , (b), (e) " = 0:9" , and (c), (f ) " = 1:1" . F 0 F 0 F 0 In this section, we present the temperature dependence T (i) T of the chemical potential, Thomas{Fermi wave vector, Coulomb impurities, we expect g  1 A T T F F and conductivity of the low-density approximate models at low temperatures, whereas at high temperatures, the for the insulator phase and Dirac semimetal phase, which energy averaging typically dominates over the screen- are the gapped 2DEG and graphene, respectively. ing contribution and we can assume g  1. Sup- pose the following power-law dependence: D(")  " , (i)  (i) v (")  " , and  (")  " . Subsequently, by 1. Insulator phase rewriting Eq. (E1) as an energy-integral form, we obtain the asymptotic power-law behavior of the temperature- We introduce the gapped 2DEG model system with dependent conductivity at low and high temperatures as the energy dispersion given by "(k) = "  + (k=k ) 0 0 h i 2 with  > 0, to account for the thermal excitation behav- (i) 1 + ( ) A (T  T ); (T ) ii 6 T ior involving the band gap between the valence and con- (0) > duction bands, similar to the insulator phase. Note that ii ( + 1)() (T  T ); the e ects of the di erence between the e ective mass (E3) of each direction are canceled out by zero-temperature normalization. where  = 1 + 2 + . (See the Supplemental Material Figure 14 shows the calculated dc conductivities as a of Ref. [42] for the detailed derivation of the power-law function of the temperature for the gapped 2DEG system analysis of the temperature-dependent dc conductivity.) in the low-density limit with " = 1:1" along with the F 0 For the semi-Dirac transition point ( = 0), = . result of the insulator phase with the same Fermi energy For short-range impurities, g = 1 and from the (see also Fig. 8 in the main text). At low temperatures, 1 the calculated results of temperature-dependent conduc- energy dependence of the relaxation time, = . Thus, tivity in the insulator phase show a similar behavior as from Eq. (E2), the asymptotic behavior is given by that of the low-density approximate model. However, as T the temperature increases, the discrepancy between the 1 (T  T ); (T ) xx 12 T two results increases, and in the high-temperature limit, = (E4a) (0) : xx log 2 (T  T ); the conductivity becomes similar to that of the semi- Dirac transition point. T =T <1 e (T  T ); (T ) yy = (E4b) 1 1 T (0) : + (T  T ): yy 1 5 F 2 T 8 F ( ) ( ) 2 2 2. Dirac semimetal Phase Here, the extra terms in  (T )= (0) were obtained yy yy through the next-order expansion of the temperature cor- For graphene (which is an approximate model for the rections. Dirac semimetal phase in the low-density limit) from 17 30 120 400 whereas those for the Thomas{Fermi wave vector are 25 100 given by 1 3 20 80 250 2 0.5  T 1 1 (T  T ); q (T ) F TF 15 60 200 6 T 0 0 = (F2) 0 0.5 0 0.5 q (0) : TF 2 log 2 (T  T ): 10 40 F 5 20 As shown in Figs. 13(a) and 13(d), the result of 0 0 0 the low-density approximate model and the numeri- 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 cally calculated result of the Dirac semimetal phase in the low-density limit are consistent with each other. FIG. 14. Calculated dc conductivities as a function of the For short-range impurities, the asymptotic form of the temperature for the gapped 2DEG system in the low-density temperature-dependent conductivity becomes [Eq. (E4) limit with  = 1 for (a) short-range impurities, (b) charged with = 0] impurities with = 1000, and (c) charged impurities with = 1. Here, " = 1:1" is used for the calculation. The red 0 F 0 T =T < F 1 e (T  T ); (T ) dashed lines and blue dashed-dotted lines represent the con- gp = (F3) 1 1 T ductivity of the insulator phase (with the same Fermi energy) : (0) gp + (T  T ); 2 16 log 2 T ins ins F and  , respectively. xx yy whereas for charged impurities in the strong screening limit, [Eq. (E4) with = 2] Eqs. (D4) and (D5) with = 2, the low{ and high{ 8 temperature asymptotic behaviors for chemical potential < 1 (T  T ); (T ) gp 3 T are given by = (F4) (0) > gp (T  T ): 6 T 8 F 1 (T  T ); 6 T =   (F1) " > 1 T (T  T ); 4 log 2 T [1] A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. [9] Z. J. Xiang, G. J. Ye, C. Shang, B. Lei, N. Z. Wang, Novoselov and A. K. Geim, The electronic properties of K. S. Yang, D. Y. Liu, F. B. Meng, X. G. Luo, L. J. graphene, Rev. Mod. Phys. 81, 109 (2009). Zou, Z. Sun, Y. Zhang and X. H. Chen, Pressure-Induced [2] S. Das Sarma, S. Adam, E. H. Hwang and E. Rossi, Elec- Electronic Transition in Black Phosphorus, Phys. Rev. tronic transport in two-dimensional graphene, Rev. Mod. Lett. 115, 186403 (2015). Phys. 83, 407 (2011). [10] B. Deng, V. Tran, Y. Xie, H. Jiang, C. Li, Q. Guo, [3] G. Rui, S. Zdenek and P. Martin, Black Phosphorus Re- X. Wang, H. Tian, S. J. Koester, H. Wang, J. J. Cha, discovered: From Bulk Material to Monolayers, Angew. Q. Xia, L. Yang and F. Xia, Ecient electrical control of Chem. Int. Ed. 56, 8052 (2017). thin- lm black phosphorus bandgap, Nature Communi- [4] A. Chaves, W. Ji, J. Maassen, T. Dumitric a and T. Low, cations 8, 14474 (2017). Theoretical Overview of Black Phosphorus, in P. Avouris, [11] Y. Liu, Z. Qiu, A. Carvalho, Y. Bao, H. Xu, S. J. R. Tan, T. Low and T. F. Heinz, eds., 2D Materials: Properties W. Liu, A. H. Castro Neto, K. P. Loh and J. Lu, Gate- and Devices, 381{412, Cambridge University Press, Cam- Tunable Giant Stark E ect in Few-Layer Black Phospho- bridge (2017). rus, Nano Lett. 17, 1970 (2017). [5] R. Rold an and A. Castellanos-Gomez, A new bandgap [12] L. L. Li, B. Partoens and F. M. Peeters, Tuning the elec- tuning knob, Nature Photonics 11, 407 (2017). tronic properties of gated multilayer phosphorene: A self- [6] J. Yang, R. Xu, J. Pei, Y. W. Myint, F. Wang, Z. Wang, consistent tight-binding study, Phys. Rev. B 97, 155424 S. Zhang, Z. Yu and Y. Lu, Optical tuning of exciton and (2018). trion emissions in monolayer phosphorene, Light: Science [13] J. Kim, S. S. Baik, S. H. Ryu, Y. Sohn, S. Park, B.-G. & Applications 4, e312 (2015). Park, J. Denlinger, Y. Yi, H. J. Choi and K. S. Kim, [7] D. akr, C. Sevik and F. M. Peeters, Signi cant e ect of Observation of tunable band gap and anisotropic Dirac stacking on the electronic and optical properties of few- semimetal state in black phosphorus, Science 349, 723 layer black phosphorus, Phys. Rev. B 92, 165406 (2015). (2015). [8] J. Quereda, P. San-Jose, V. Parente, L. Vaquero-Garzon, [14] S. Banerjee, R. R. P. Singh, V. Pardo and W. E. Pickett, A. J. Molina-Mendoza, N. Agra t, G. Rubio-Bollinger, Tight-Binding Modeling and Low-Energy Behavior of the F. Guinea, R. Rold an and A. Castellanos-Gomez, Strong Semi-Dirac Point, Phys. Rev. Lett. 103, 016402 (2009). Modulation of Optical Properties in Black Phosphorus [15] V. Pardo and W. E. Pickett, Half-Metallic Semi-Dirac- through Strain-Engineered Rippling, Nano Lett. 16, 2931 Point Generated by Quantum Con nement in TiO =VO 2 2 (2016). Nanostructures, Phys. Rev. Lett. 102, 166803 (2009). 18 [16] V. Pardo and W. E. Pickett, Metal-insulator transition [31] D. J. P. de Sousa, L. V. de Castro, D. R. da Costa, through a semi-Dirac point in oxide nanostructures: VO J. M. Pereira and T. Low, Multilayered black phospho- (001) layers con ned within TiO , Phys. Rev. B 81, rus: From a tight-binding to a continuum description, 035111 (2010). Phys. Rev. B 96, 155427 (2017). [17] J. Kim, S. S. Baik, S. W. Jung, Y. Sohn, S. H. Ryu, [32] S. S. Baik, K. S. Kim, Y. Yi and H. J. Choi, Emergence of H. J. Choi, B.-J. Yang and K. S. Kim, Two-Dimensional Two-Dimensional Massless Dirac Fermions, Chiral Pseu- Dirac Fermions Protected by Space-Time Inversion Sym- dospins, and Berrys Phase in Potassium Doped Few- metry in Black Phosphorus, Phys. Rev. Lett. 119, 226801 Layer Black Phosphorus, Nano Lett. 15, 7788 (2015). (2017). [33] H. Doh and H. J. Choi, Dirac-semimetal phase diagram [18] N. Ehlen, A. Sanna, B. V. Senkovskiy, L. Petaccia, A. V. of two-dimensional black phosphorus, 2D Materials 4, Fedorov, G. Profeta and A. Gruneis,  Direct observation 025071 (2017). of a surface resonance state and surface band inversion [34] G. Montambaux, F. Pi echon, J.-N. Fuchs and M. O. Go- control in black phosphorus, Phys. Rev. B 97, 045143 erbig, Merging of Dirac points in a two-dimensional crys- (2018). tal, Phys. Rev. B 80, 153412 (2009). [19] P. Di Pietro, M. Mitrano, S. Caramazza, F. Capitani, [35] G. Montambaux, F. Pi echon, J.-N. Fuchs and M. O. Go- S. Lupi, P. Postorino, F. Ripanti, B. Joseph, N. Ehlen, erbig, A universal Hamiltonian for motion and merging of A. Gruneis,  A. Sanna, G. Profeta, P. Dore and A. Peruc- Dirac points in a two-dimensional crystal, The European chi, Emergent Dirac carriers across a pressure-induced Physical Journal B 72, 509 (2009). Lifshitz transition in black phosphorus, Phys. Rev. B 98, [36] R. de Gail, J.-N. Fuchs, M. Goerbig, F. Pichon 165111 (2018). and G. Montambaux, Manipulation of Dirac points in [20] F. Xia, H. Wang and Y. Jia, Rediscovering black phos- graphene-like crystals, Physica B: Condensed Matter phorus as an anisotropic layered material for optoelec- 407, 1948 , proceedings of the International Workshop tronics and electronics, Nature Communications 5, 4458 on Electronic Crystals (ECRYS-2011) (2012). (2014). [37] M. P. Marder, Condensed Matter Physics, 2nd edition, [21] A. Mishchenko, Y. Cao, G. L. Yu, C. R. Woods, R. V. John Wiley & Sons (2010). Gorbachev, K. S. Novoselov, A. K. Geim and L. S. Lev- [38] E. D. Siggia and P. C. Kwok, Properties of Electrons itov, Nonlocal Response and Anamorphosis: The Case in Semiconductor Inversion Layers with Many Occupied of Few-Layer Black Phosphorus, Nano Lett. 15, 6991 Electric Subbands. I. Screening and Impurity Scattering, (2015). Phys. Rev. B 2, 1024 (1970). [22] T. Li, Z. Zhang, X. Li, M. Huang, S. Li, S. Li and Y. Wu, [39] S. Woo, E. H. Hwang and H. Min, Large negative dif- High eld transport of high performance black phos- ferential transconductance in multilayer graphene: the phorus transistors, Applied Physics Letters 110, 163507 role of intersubband scattering, 2D Materials 4, 025090 (2017). (2017). [23] Y. Y. Illarionov, M. Waltl, G. Rzepa, J.-S. Kim, S. Kim, [40] J. Schliemann and D. Loss, Anisotropic transport in a A. Dodabalapur, D. Akinwande and T. Grasser, Long- two-dimensional electron gas in the presence of spin-orbit Term Stability and Reliability of Black Phosphorus Field- coupling, Phys. Rev. B 68, 165311 (2003). E ect Transistors, ACS Nano 10, 9543 (2016). [41] K. Vyb  orny,  A. A. Kovalev, J. Sinova and T. Jungwirth, [24] J. Qiao, X. Kong, Z.-X. Hu, F. Yang and W. Ji, High- Semiclassical framework for the calculation of transport mobility transport anisotropy and linear dichroism in anisotropies, Phys. Rev. B 79, 045427 (2009). few-layer black phosphorus, Nature Communications 5, [42] S. Park, S. Woo, E. J. Mele and H. Min, Semiclassical 4475 (2014). Boltzmann transport theory for multi-Weyl semimetals, [25] Y. Liu, T. Low and P. P. Ruden, Mobility anisotropy in Phys. Rev. B 95, 161113 (2017). monolayer black phosphorus due to scattering by charged [43] N. W. Ashcroft and N. D. Mermin, Solid State Physics, impurities, Phys. Rev. B 93, 165402 (2016). Brooks Cole (1976). [26] Y. Liu and P. P. Ruden, Temperature-dependent [44] G. B. Arfken, H. J. Weber and F. E. Harris, Mathematical anisotropic charge-carrier mobility limited by ionized im- Methods for Physicists, Academic Press, 7th ed. (2012). purity scattering in thin-layer black phosphorus, Phys. [45] S. Adam, E. H. Hwang, V. M. Galitski and S. Das Sarma, Rev. B 95, 165446 (2017). A self-consistent theory for graphene transport, Proceed- [27] F. W. Han, W. Xu, L. L. Li, C. Zhang, H. M. Dong and ings of the National Academy of Sciences 104, 18392 F. M. Peeters, Electronic and transport properties of n- (2007). type monolayer black phosphorus at low temperatures, [46] Q. Li, E. H. Hwang and S. Das Sarma, Disorder-induced Phys. Rev. B 95, 115436 (2017). temperature-dependent transport in graphene: Puddles, [28] M. Zare, B. Z. Rameshti, F. G. Ghamsari and R. As- impurities, activation, and di usion, Phys. Rev. B 84, gari, Thermoelectric transport in monolayer phospho- 115442 (2011). rene, Phys. Rev. B 95, 045422 (2017). [47] N. Ramakrishnan, M. Milletari and S. Adam, Trans- [29] P. Adroguer, D. Carpentier, G. Montambaux and port and magnetotransport in three-dimensional Weyl E. Orignac, Di usion of Dirac fermions across a topo- semimetals, Phys. Rev. B 92, 245120 (2015). logical merging transition in two dimensions, Phys. Rev. [48] J. Jang, S. Ahn and H. Min, Optical conductivity of B 93, 125113 (2016). black phosphorus with a tunable electronic structure, [30] A. N. Rudenko and M. I. Katsnelson, Quasiparticle band arXiv:1811.07529 (2018). structure and tight-binding model for single- and bilayer black phosphorus, Phys. Rev. B 89, 201408 (2014). http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Condensed Matter arXiv (Cornell University)

Semiclassical Boltzmann transport theory of few-layer black phosphorus in various phases

Condensed Matter , Volume 2018 (1811) – Nov 9, 2018

Loading next page...
 
/lp/arxiv-cornell-university/semiclassical-boltzmann-transport-theory-of-few-layer-black-phosphorus-vKR0T5gzkm
ISSN
2053-1583
eISSN
ARCH-3331
DOI
10.1088/2053-1583/ab01c3
Publisher site
See Article on Publisher Site

Abstract

Semiclassical Boltzmann transport theory of few-layer black phosphorus in various phases Sanghyun Park, Seungchan Woo, and Hongki Min Department of Physics and Astronomy, Seoul National University, Seoul 08826, Korea (Dated: April 4, 2019) Black phosphorus (BP), a two-dimensional (2D) van der Waals layered material composed of phosphorus atoms, has been one of the most actively studied 2D materials in recent years due to its tunable energy band gap (tunable even to a negative value) and its highly anisotropic electronic structure. Depending on the sign of the band gap tuning parameter, few-layer BP can be in a gapped insulator phase, gapless Dirac semimetal phase, or gapless semi-Dirac transition point between the two phases. Using the fully anisotropic multiband Boltzmann transport theory, we systematically study the dc conductivity of few-layer BP as a function of the carrier density and temperature by varying the band gap tuning parameter, and determine the characteristic density and temperature dependence corresponding to each phase. I. INTRODUCTION eralized to anisotropic multiband systems, we calculate the dc conductivity as a function of the carrier density and temperature for each phase. We determine that each Since the discovery of graphene [1, 2], which is a carbon phase shows the characteristic density and temperature allotrope of two-dimensional (2D) honeycomb lattice, 2D dependence, which can be used as a transport signature materials have been one of the most active research areas of BP in di erent phases. in condensed matter physics. Black phosphorus (BP) is The rest of this paper is organized as follows. In a 2D material with van der Waals layered structure com- Sec. II, we describe our model Hamiltonian and develop posed of phosphorus atoms, and it has recently attracted the Boltzmann transport theory in anisotropic multiband considerable attention [3, 4]. As a layered semiconductor systems. In Sec. III, we present the dc conductivity of BP in its natural form, BP has a tunable band gap, and ma- in each phase as a function of density at zero tempera- nipulation of its band gap through various methods has ture. In Sec. IV, we provide the temperature dependence been validated by multiple theoretical and experimental of dc conductivity at a xed density. We conclude our reports [5]. Notable examples of the band gap tuning paper in Sec. V with discussions on the dominant scat- include thickness change [6, 7], strain control [8], pres- tering source, the e ect of potential uctuations at low sure [9], electronic gating [10{12], and chemical doping densities, and the e ect of the parabolic term omitted in [13]. Some of the band gap manipulation methods [9, 13] the current model. demonstrated that the band gap can be tuned to zero, showing the semi-Dirac state with a combination of lin- ear and quadratic dispersions [14], which is also predicted II. METHODS in TiO =VO heterostructures [15, 16]. Furthermore, the 2 2 band gap can be inverted, leading to the Dirac semimetal A. Model phase [17{19]. Due to its anisotropic electronic band structure, BP shows many peculiar transport properties such as large By expanding the tight-binding lattice model of few- in-plane anisotropic transport [20, 21]. The e ects of layer BP [4, 30, 31], the corresponding low-energy e ec- temperature [10, 22, 23], the number of layers [10], and tive Hamiltonian can be obtained as [32{36] substrate [22] on the anisotropic transport properties of 2 2 BP have been studied experimentally. Furthermore, the ~ k " H = +  + ~v k  ; (1) x 0 y y transport properties of BP have been studied theoret- 2m 2 ically [24{29], demonstrating its anisotropic nature in energy- and temperature-dependent transport. However, where m is the e ective mass along the zigzag (x) di- there has been no systematic study on the anisotropic rection, v is the band velocity along the armchair (y) transport of BP in each phase, fully considering the direction, " is the size of the band gap (which will be anisotropy of the system and the interband scattering. used as a tuning parameter), and  and  are the Pauli x y In this study, we theoretically investigate the transport matrices. The eigenenergies of the Hamiltonian are given properties of BP in the gapped insulator phase, gapless 2 2 2 ~ k " x 2 2 2 by " =  + + ~ v k ; thus, the Hamilto- semi-Dirac transition point, and Dirac semimetal phase. 0 y 2m 2 Using the semiclassical Boltzmann transport theory gen- nian H has a direct band gap for " > 0, a semi-Dirac band touching point at (k ; k ) = (0; 0) for " = 0, or two x y g m j" j Dirac points at (k ; k ) = ( ; 0) for " < 0. The x y 2 g hmin@snu.ac.kr characteristic energy scales along the zigzag and arm- arXiv:1811.03903v2 [cond-mat.mes-hall] 3 Apr 2019 2 2 2 1.2 1.2 2 ~ k chair directions are given by " = and ~v k , re- 0  0 0 2m 1 1 1.5 spectively, where k = a and a is the lattice constant. 0.8 0.8 We introduce the dimensionless parameters  = and 2" 0.6 0.6 1 ~v k 0 0 c = , which represent a gap tuning parameter and " 0.4 0.4 0.5 the ratio of the characteristic energy scales along the 0.2 0.2 zigzag and armchair directions, respectively. Through- 0 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 out the paper, we use c = 1 and the spin degeneracy g = 2 for the calculation. We will discuss the e ect of higher-order terms omitted in Eq. (1) in Sec. V. 6 5 7 0 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 FIG. 2. (a)-(c) Calculated DOS and (d)-(e) the carrier den- sity as a function of Fermi energy for the (a), (c) insulator phase, (b), (e) semi-Dirac transition point, and (c), (f ) Dirac semimetal phase. Here,   is the band gap tuning pa- 2" rameter, and g = 2 and c = 1 are used for calculation. FIG. 1. (a)-(c) Energy dispersions and (d)-(f ) the corre- sponding Fermi surfaces of few-layer BP for the (a), (c) insu- lator phase, (b), (e) semi-Dirac transition point, and (c), (f ) a function of Fermi energy for each phase. At the Dirac semimetal phase. semi-Dirac transition point, the DOS is simply given by 1=2 D(")  " [Fig. 2(a)], and the carrier density (which Figure 1 shows the energy dispersion and the corre- is an energy integral of the DOS up to " ) is given by sponding Fermi surface of few-layer BP in each phase. 3=2 n  " [Fig. 2(d)]. (See Appendix A for the detailed Initially, few-layer BP without band gap tuning is in the derivations of the DOS and the carrier density.) In the gapped insulator phase, as shown in Fig. 1(a). As the gapped insulator phase, both DOS at " and carrier den- band gap " decreases (for example, upon applying a sity vanish for " < " =2, whereas for " > " =2, they F g F g perpendicular electric eld), eventually it vanishes and follow those of the semi-Dirac transition point as " in- the system is described by the semi-Dirac Hamiltonian creases [Figs. 2(b) and 2(e)]. In the Dirac semimetal in Eq. (1) with " = 0, as shown in Fig. 1(b). If the phase, when " is very small, the system resembles band gap decreases even further and becomes negative a typical 2D Dirac semimetal such as graphene; thus, (" < 0), band inversion occurs, which has been achieved g 1 D(")  " . As " increases and approaches " =2 near the F g experimentally using surface doping [17, 18] and external top of the inverted band, the band dispersion e ectively pressure [19] becomes hyperbolic paraboloid with a di erent sign in In the gapped insulator phase, the inherent anisotropy each direction in momentum space. Subsequently, a van of the system is less evident and the system at low Hove singularity occurs in the DOS, diverging logarith- densities resembles typical semiconductors with a dif- mically with D(")  log(jj") [37]. If " increases ferent e ective mass in each direction. At the semi- further, the DOS and the carrier density follow those of Dirac transition point, the energy dispersion becomes the semi-Dirac transition point with a discontinuous en- linear (quadratic) along the armchair (zigzag) direction, ergy derivative in the DOS at the van Hove singularity as shown in Fig. 1(e). At the Dirac semimetal phase, [Figs. 2(c) and 2(f )]. the anisotropy in the energy dispersion becomes more pronounced and the Fermi surface vastly changes its shape depending on the value of the Fermi energy " . Notably, as the energy dispersion and the Fermi surface For " < " =2, the Fermi surface becomes two distinct are anisotropic, and the Fermi energy can cross multiple F g lines, as shown in Fig. 1(f ), whereas for " > " =2, the bands, we cannot naively use the conventional Boltzmann F g two Fermi surfaces become joined completely, forming a transport theory assuming an isotropic single-band sys- closed line. At " = " =2, a van Hove singularity occurs tem. Thus, the anisotropic multiband Boltzmann trans- F g in the density of states (DOS), as explained below. port theory is necessary to calculate the dc conductivity Figure 2 shows the DOS and the carrier density as of such systems, as explained in Sec. II B. 3 B. Boltzmann transport theory in anisotropic obtain the following integral equation for the relaxation multiband systems time: (i) d 0 d k v (i) 0 0 (i) We use semiclassical Boltzmann transport theory to 1 = W   : (7) 0 0 0 kk k k (i) (2) calculate the density and temperature dependence of the 0 v dc conductivity of few-layer BP in each phase in the pres- This is a coupled integral equation relating the relax- ence of impurities, assuming elastic scattering (see Sec. V ation times at di erent angles in di erent bands, which for the limitation of the current approach). In the Boltz- correctly considers the anisotropy and multiple bands of mann transport theory, electron states are described by the system. Note that, for an isotropic single-band sys- the non-equilibrium distribution function f = f (r;k; t). (i) Its time rate of change is balanced out by the collision tem [ =  (") for a given energy " = " ], Eq. (7) is term, which represents the total scattering probability reduced to the well-known expression for the relaxation df df time given by [43] per unit time, i.e., = . dt dt coll d 0 We assume a spatially homogeneous system without 1 d k 0 0 = W (1 cos  ): (8) explicit time dependence in the distribution function, i.e., kk kk (2) f = f . Thus, the time derivative of the distribution df @f _ k function is given by = k  , whereas the collision The current density J induced by an electric eld E is dt @k term is given by thus given by d 0 X X df d k d k (i) (i) (j) 0 0 = W (f f ); (2) kk k k J = g (e)v f   E ; (9) k ij d k dt (2) (2) coll 2 2 where W 0 = n jV 0j (" " 0 ) is the transition kk imp kk k k ~ where  is the conductivity tensor given by ij rate from k to k for an elastic scattering with the impu- 0 d rity potential V and the impurity density n . In the X kk imp d k (i) (j) (j) 2 (0) _  = ge S (")v v  : (10) ij presence of a uniform electric eld E, ~k = (e)E, and k k k (2) to the leading order in E, We nd that the Hall conductivity (i 6= j) vanishes, thus (0) (0) df @f @f k k we consider only the diagonal part of the dc conductivity (e)E  = (e)E  v ; (3) dt ~@k @" k (i = j). (0) 1 @" k (0) (" ) where v = and f = f (" ) = e + 1 k k ~ @k is the Fermi{Dirac distribution function at equilibrium III. DENSITY DEPENDENCE OF DC with = and the chemical potential . Assume CONDUCTIVITY k T that, to the leading order in E, the non-equilibrium dis- (0) tribution function f is given by f  f (") + f at k k k Using the anisotropic multiband Boltzmann transport df df theory developed in Sec. II B, we calculate the dc conduc- energy " = " . Thus, from = , we obtain dt dt coll tivity of few-layer BP as a function of the carrier density d 0 or Fermi energy at zero temperature for each phase: the d k (0) (e)E  v S (") = W 0 (f f 0 ); (4) semi-Dirac transition point ( = 0), gapped insulator k kk k k (2) phase ( > 0), and Dirac semimetal phase ( < 0), all (0) of which can be expressed by Eq. (1). @f (") (0) where S (") = . If the Fermi energy crosses @" As for the impurity potential, we consider two types multiple energy bands, Eq. (4) is generalized to [38, 39] of impurity scattering: short-range impurities and long-range Coulomb impurities (or charged impurities). d 0 0 0 d k (0) (e)E  v S (") = W f f ; Short-range impurities originate from lattice defects, va- 0 0 kk k k (2) cancies, dislocations, etc., and their potential form is (5) 0 given by a constant in momentum space, V = V , kk short where and are band indices. as they are approximately represented by the delta func- We parameterize f in the following form [40{42]: tion in real space. For charged impurities distributed randomly in the background, the impurity potential is d 2 X 2e given by V 0 = in 2D, where (q) is the dielec- (i) (i) (i) (0) kk (q)jqj f = (e) E v  S ("); (6) k k tric function for q = k k . Within the Thomas{Fermi i=1 approximation, (q) can be approximated as (q) (i) (i) (i) (1 + q =jqj), where  is the background dielectric con- TF where E , v , and  are the electric eld, velocity, k k 2e and relaxation time, respectively, along the ith direction stant, q = D(" ) is the Thomas{Fermi wave vec- TF F (i) for each band. After matching each coecient in E , we tor, and D(" ) is the total DOS at the Fermi energy " F F 4 1 2 (including all the contributions from the bands crossing (g  1), V  q  D (" )  " ; thus, we obtain 0 F F TF F " and the spin degeneracy). The interaction strength for charged impurities can be characterized by an e ective 2 2 e 3 "  n ; (12a) xx ne structure constant = . Note that q / g . F 0 TF 0 ~v Thus, the screening strength for Coulomb impurities is "  n : (12b) yy also characterized by . At general screening strength, the power-law behavior is determined by the competition between the screening A. Semi-Dirac transition point wave vector and the momentum transfer. We present the numerically calculated power-law behavior for the semi- Dirac transition point and for the other phases in Fig. 6. 7 8 20 6 15 B. Insulator phase 4 10 2 5 6 6 14 0 0 0 5 5 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 4 4 3 3 3 3 8 2 2 2.5 2.5 6 1 1 0 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 2 1.5 4 1.5 2 1.4 1.4 4 0.5 1.2 1.2 1 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 1 1 0.8 0.8 0.6 0.6 FIG. 3. Calculated dc conductivities (a)-(c)  and (d)-(f ) xx 0.4 0.4 as a function of Fermi energy at the semi-Dirac transi- yy 0.2 0.2 tion point ( = 0) for (a), (d) short-range impurities, (b), 0 0 0 (d) charged impurities with = 1000, and (c), (f ) charged 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 2 2 2 ge k c impurities with = 1. Here,  = . 0 0 2~n imp FIG. 4. Calculated dc conductivities (a)-(c)  and (d)-(f ) xx as a function of Fermi energy in the insulator phase with yy First, let us consider the semi-Dirac transition point = 1 for (a), (d) short-range impurities, (b), (d) charged ( = 0). Figure 3 shows the Fermi energy depen- impurities with = 1000, and (c), (f ) charged impurities dence of dc conductivity at the semi-Dirac transition with = 1. point. The characteristic density or Fermi energy depen- dence of the dc conductivity can be understood as fol- (i) Figure 4 shows the Fermi energy dependence of the 1 2 lows. From Eq. (10) with   D (" )=V , we expect dc conductivity in the insulator phase ( > 0). In the (i) (i) (i) (i) (i) 2 2 2 D(" )[v ]   [v ] =V , where  and v ii F F F F F F F insulator phase, the power-law dependence of the dc con- are the relaxation time and velocity, respectively, at the ductivity at low densities becomes similar to that of 2D Fermi energy along the ith direction, and V is the angle- electron gas (2DEG) with a di erent e ective mass in averaged squared impurity potential at the Fermi energy. each direction. (See Appendix C 1 for detailed deriva- 1=2 At the semi-Dirac transition point, D(" )  " , and the tions.) (x) 1=2 Fermi velocity in each direction is given by v  " For short-range impurities, the power-law dependence F F (y) of the dc conductivity at low densities is given by and v  " , from which we can deduce the power-law F F behavior of the dc conductivity. (See Appendix A for the " ; (13a) xx detailed derivations of the power-law dependences.) " : (13b) yy For short-range impurities, V is a constant indepen- F dent of density; in this case, we obtain For charged impurities, in the strong screening limit, at 3 low densities, we obtain "  n ; (11a) xx 0 0 "  n : (11b) yy " ; (14a) xx For charged impurities, in the strong screening limit   " : (14b) yy F 5 Note that, as the Fermi energy or the carrier density For charged impurities, due to the dominant contribution increases, the power-law dependence becomes similar to from the diverging Thomas{Fermi wave vector q / TF that of the semi-Dirac transition point. D(" ), the conductivity is largely given by the square of the DOS as follows: [log (jj " )] ; (18a) xx F C. Dirac semimetal phase [log (jj " )] : (18b) yy F As the Fermi energy or the carrier density increases 8 10 25 7 further, the power-law dependence of the dc conductiv- 8 20 ity becomes similar to that of the semi-Dirac transition 6 15 point, as in the insulator phase. 4 10 2.08 1.08 2.1 2 5 2.06 1.06 1 0 0 1.9 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 2.04 1.04 1.8 2.02 1.02 1.7 5 5 14 1.6 2 1 4 4 1.5 0 2 4 0 2 4 0 2 4 10 10 10 10 10 10 10 10 10 3 3 2 2 1.2 1.08 2.05 1 1 2 2 1.06 1.15 1.95 0 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 1.04 1.1 1.9 1.02 1.85 1.05 1.8 FIG. 5. Calculated dc conductivities (a)-(c)  and (d)- xx (f )  as a function of Fermi energy in the Dirac semimetal yy 1.75 0 2 4 0 2 4 0 2 4 10 10 10 10 10 10 10 10 10 phase with  = 1 for (a), (d) short-range impurities, (b), (d) charged impurities with = 1000, and (c), (f ) charged impurities with = 1. FIG. 6. (a)-(c) d log  =d log " and (d)-(f ) xx F d log  =d log " as a function of for charged impurities yy F 0 Figure 5 shows the Fermi energy dependence of the dc in each phase. The red dashed lines represent the Fermi en- ergy exponents obtained in the strong screening limit. Here, conductivity in the Dirac semimetal phase ( < 0). In " = " for the semi-Dirac transition point, " = 1:01" for F 0 F 0 the Dirac semimetal phase, the power-law dependence of the gapped insulator phase, and " = 0:01" for the Dirac F 0 the dc conductivity at low densities becomes similar to phase are used for the calculation. that of graphene but with a di erent Fermi velocity in each direction. (See Appendix C 2 for detailed deriva- Figure 6 shows the evolution of the Fermi-energy power tions.) law of the dc conductivity as a function of the screen- For short-range impurities, the power-law dependence ing strength for each phase in the low carrier density of the dc conductivity at low densities is given by 0 limit. For the insulator phase and the semi-Dirac transi- " ; (15a) tion point, the Fermi-energy exponent decreases, whereas xx for the Dirac semimetal phase, it shows a non-monotonic " : (15b) yy behavior with a dip structure, which originates from the For charged impurities, in the strong screening limit, at interband-like scattering between two distinct Fermi sur- low densities, we obtain faces shown in Fig. 1(f ). As the screening strength in- creases, all the Fermi-energy exponents approach the cor- " ; (16a) xx F responding power law estimated in the strong screening limit. " : (16b) yy Near the van Hove singularity, "  " =2, the DOS F g diverges logarithmically [37] and it dominates the overall IV. TEMPERATURE DEPENDENCE OF DC CONDUCTIVITY power-law behavior of conductivity [39]. Therefore, for short-range impurities, the conductivity becomes We can apply the anisotropic multiband Boltzmann [ log (jj " )] ; (17a) xx F transport theory developed in Sec. II B to the dc conduc- [ log (jj " )] : (17b) tivity at nite temperature. In Eq. (10), the nite tem- yy F 6 3 15 25 perature a ects the conductivity through the Fermi dis- 2.5 tribution and the temperature-dependent screening for the charged impurity potential. At nite temperatures, the chemical potential of the system also deviates from 1.5 the Fermi energy " due to the broadening of the Fermi distribution function. From the invariance of carrier den- sity n with respect to temperature T , we obtain the 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 temperature dependence of the chemical potential (T ). For charged impurities, the nite temperature Thomas{ 2 1.1 2.5 3 2e @n Fermi screening wave vector is given by q (T ) = TF @ 1 2 2.5 for 2D systems. (See Appendix D for the detailed deriva- 0.9 1.5 2 tion of the temperature dependence of the chemical po- 0.8 tential and Thomas{Fermi wave vector). In this section, 0.7 1 1.5 0.6 we calculate the dc conductivity of few-layer BP as a 0.5 1 0.5 function of the temperature for each phase. The detailed 0.4 0 0.5 derivation of the temperature-dependent conductivity is 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 presented in Appendices E and F. FIG. 7. Calculated dc conductivities (a)-(c)  and (d)-(f ) xx as a function of the temperature at the semi-Dirac transi- yy A. Semi-Dirac transition point tion point ( = 0) for (a), (d) short-range impurities, (b), (d) charged impurities with = 1000, and (c), (f ) charged impu- rities with = 1. Here, if the temperature is normalized by From the power-law dependence of the DOS, D(")  0 1=2 T = " =k , the result is independent of " at the semi-Dirac F F B F " at the semi-Dirac transition point [Fig. 2(a)], we transition point. The blue dashed-dotted lines and red dashed can obtain the asymptotic behaviors of (T ) and q (T ) TF lines represent tting by the corresponding asymptotic form in a relatively straightforward manner. In the low- and [Eqs. (21) and (22)] in the low- and high-temperature limits, high-temperature limits, the chemical potential at the respectively. semi-Dirac transition point is given by >  T 1 (T  T ); F For charged impurities, the asymptotic behavior is given 12 T =   (19) by " 1 T (T  T ); 1 5 F 8 2( )( ) 2 2 1 + C (T  T ); xx F (T ) xx T whereas the Thomas{Fermi wave vector is given by =   (22a) (0) > xx D (T  T ); 8 xx F < 8 1 (T  T ); q (T ) 12 T TF F < T = 1 (20) 1 + C (T  T ); (T ) yy F yy 2 F q (0) TF 1 3 T = (22b) 2 (T  T ); 2 2 T  (0) : yy D (T  T ); yy F where is the Gamma function and  is the Dirichlet where C (D ) indicates the low- (high-) temperature co- ii ii eta function [44]. In a single-band system, q (T ) typ- TF ecients. In the strong screening limit, the coecients ically decreases with the temperature at high tempera- 2 2 become C = 0, D = , C = , and D = xx xx yy yy tures, whereas at the semi-Dirac transition point, q (T ) 6 4 TF log 2. As the screening strength decreases, the high- increases with the temperature due to the thermal exci- temperature coecients D remain positive, whereas the ii tation of carriers participating in the screening. low-temperature coecients C decrease and we expect ii Figure 7 shows the temperature dependence of the dc that the initially negative or vanishing C would even- ii conductivity at the semi-Dirac transition point, normal- tually become positive in the weak screening limit. (See ized by the zero-temperature conductivity value in each Appendix E for the detailed derivations of the coecients direction. For short-range impurities, we determine that C and D .) ii ii the asymptotic behavior is given by The temperature dependence in the high-temperature limit can be easily understood by replacing " with <  T 1 (T  T ); (T ) F xx 12 T F T in the Fermi energy dependence of dc conductivity = (21a) : T (0) [Eqs. (11) and (12)]. At high temperatures,  (T ) for xx yy log 2 (T  T ); short-range impurities decreases with the temperature, T =T <1 e (T  T ); showing a metallic behavior. Otherwise, the conductiv- (T ) yy = (21b) 2 ities increase with the temperature, showing an insulat- 1 1 T (0) : + (T  T ): yy F 1 5 2 T F ing behavior. Note that the high-temperature asymp- ( ) ( ) 2 2 7 totic form for charged impurities is obtained by consid- perature dependence of the chemical potential, Thomas{ ering the e ect of the energy averaging and that of the Fermi screening wave vector, and conductivity of the temperature-dependent screening separately. It correctly gapped 2DEG system.) predicts the temperature power-law dependence but not In the high-density limit, the temperature dependence the coecients in the asymptotic form, showing a dis- of dc conductivity in the insulator phase resembles that crepancy with the numerical result, as the e ect of tem- of the semi-Dirac transition point. perature cannot be simply separated into the energy av- eraging and the temperature-dependent screening at high temperatures. C. Dirac semimetal phase B. Insulator phase 1.1 15 0.9 25 120 400 10 0.8 1 3 0.7 2 5 80 5 0.5 0.6 60 200 0 0 0.5 0 0.5 0 0.5 40 0.4 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 0 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 1.1 15 0.9 8 50 140 0.8 1 3 0.7 0.5 1 0.6 0 0 0 0.5 0 0.5 0.5 0.4 0 40 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 0 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 FIG. 9. Calculated dc conductivities (a)-(c)  and (d)- xx (f )  in the low-density limit as a function of the temper- yy ature in the Dirac semimetal phase with  = 1 for (a), FIG. 8. Calculated dc conductivities (a)-(c)  and (d)-(f ) xx (d) short-range impurities, (b), (d) charged impurities with in the low-density limit as a function of the temperature yy = 1000, and (c), (f ) charged impurities with = 1. Here, 0 0 in the insulator phase with  = 1 for (a), (d) short-range " = 0:01" is used for the calculation. The blue dashed- F 0 impurities, (b), (d) charged impurities with = 1000, and dotted lines and red dashed lines represent tting by the cor- (c), (f ) charged impurities with = 1. Here, " = 1:1" is 0 F 0 responding asymptotic form [Eqs. (23) and (24)] in the low- used for the calculation. The blue dashed-dotted lines repre- and high-temperature limits, respectively. sent the result for the gapped 2DEG system (see Appendix F), and the red dashed lines represent power-law tting by the asymptotic form of the semi-Dirac transition point [Eqs. Figure 9 shows the calculated temperature-dependent (21) and (22)] in the high-temperature limit. conductivity in the Dirac semimetal phase, with the xed Fermi energy of " = 0:01" , which corresponds to the F 0 Figure 8 shows the temperature dependence of the dc low-density limit. At low densities, the Dirac semimetal conductivity in the insulator phase with the xed Fermi phase can be e ectively considered as graphene (with energy of " = 1:1" , which corresponds to the low- anisotropic velocities); thus, we can understand its F 0 density limit. At zero temperature, the insulator phase temperature-dependent conductivity behavior using the in the low-density limit can be e ectively considered as a result of graphene. (See Appendix F for the temperature gapped 2DEG (with anisotropic e ective masses). Simi- dependence of the chemical potential, Thomas{Fermi larly, at nite temperatures, the temperature-dependent screening wave vector, and conductivity of graphene.) conductivity of the insulator phase in the low-density For graphene with short-range impurities, the asymptotic limit resembles that of the gapped 2DEG system (blue form of the temperature-dependent conductivity becomes dash-dotted lines in Fig. 8), especially in the low- T =T temperature limit. In the high-temperature limit, the < 1 e (T  T ); (T ) gp power-law behavior of the temperature-dependent con- 2 = (23) 1 1 T (0) : gp + (T  T ); ductivity for the insulator phase becomes similar to that 2 16 log 2 T of the semi-Dirac transition point [Eqs. (21) and (22)], because thermally excited carriers above the gap con- whereas for charged impurities in the strong screening tribute to the conductivity. (See Appendix F for the tem- limit, the asymptotic form of the temperature-dependent 8 conductivity becomes tial (T ) [Fig. 13(b) in the Appendix], shifting the central point of the energy averaging. T In the high-density limit, the temperature dependence 1 (T  T ); (T ) 3 T gp F of dc conductivity in the Dirac semimetal phase resembles =   (24) (0)  T gp that of the semi-Dirac transition point. (T  T ): 6 T Similar to the result of the semi-Dirac transition point, V. DISCUSSION AND CONCLUSION the high-temperature asymptotic form for charged im- purities correctly captures the temperature power-law dependence (but not the exact coecient value, as dis- When we consider both short-range and charged impu- cussed in Sec. IV A). rities, assuming that each scattering mechanism is inde- pendent, the total scattering rate is obtained by adding 1.2 1.5 2.5 their scattering rates in accordance with Matthiessen's 1.1 rule. Note that the scattering mechanism with a higher scattering rate (or equivalently a lower conductivity) 1.5 dominates the resulting conductivity. From the obtained 0.9 Fermi-energy power-law dependence of dc conductivity, 0.8 0.5 we can determine the dominant scattering mechanism. At the semi-Dirac transition point, we can observe from 0.7 0.5 0 1 2 0 1 2 0 1 2 Eq. (11) and Figs. 6(a) and (d) that, for both  and xx , the Fermi-energy power law for short-range impuri- yy 1.4 1.6 3 ties is always smaller than that of charged impurities. 1.4 1.3 This indicates that, at low densities, charged impuri- 2.5 1.2 ties are dominant over short-range impurities, whereas 1.2 at high densities, short-range impurities are dominant 1.1 0.8 over charged impurities. In the insulator phase, at low 1.5 densities, the system can be approximated as a 2DEG 0.6 and the Fermi-energy power laws for short-range and 0.9 0.4 0 1 2 0 1 2 0 1 2 charged impurities are almost comparable (except in the no-screening limit) as shown in Eq. (13) and Figs. 6(b) and (e). At high densities, the power-law dependence FIG. 10. Calculated dc conductivities (a){(c)  and (d){ xx follows that of the semi-Dirac transition point; thus, (f )  immediately below the van Hove singularity point as a yy function of the temperature in the Dirac semimetal phase with short-range impurities dominate over charged impuri- = 1 for (a), (d) short-range impurities, (b), (d) charged ties. In the Dirac semimetal phase, at low densities, the impurities with = 1000, and (c), (f ) charged impurities 0 Fermi-energy power law for short-range impurities is al- with = 1. Here, " = 0:9" is used for the calculation. 0 F 0 ways smaller than that of charged impurities as shown in Eq. (15) and Figs. 6(c) and (f ); thus, charged impurities are dominant over short-range impurities as in the case Figure 10 shows the temperature dependence of the dc of graphene. At high densities, short-range impurities conductivity in the Dirac semimetal phase immediately become dominant over charged impurities, following the below the van Hove singularity point, exhibiting a non- trend of the semi-Dirac transition point. Note that, near monotonic behavior with temperature. As explained ear- the van Hove singularities, charged impurities are highly lier, the temperature dependence of the dc conductivity is (0) screened due to the enhanced DOS, and thus, short-range determined by the energy averaging with S (") broad- impurities are dominant over charged impurities [39]. ened by temperature, and by the temperature-dependent screening for charged impurities. Thus, if the Fermi en- Our analysis is based on the semiclassical Boltzmann transport theory, which is known to be valid in the high- ergy is near the van Hove singularity, the distance be- density limit. At low densities, the e ect of potential tween the Fermi energy and the van Hove singularity sets an important energy scale for the temperature de- uctuations induced by spatially inhomogeneous impu- rities becomes important, which is not captured by our pendence, k T  jj" jj" jj. For charged impurities, B 1 F g the conductivity rst increases, showing a peak at T , approach assuming a spatially homogeneous system. At ch the semi-Dirac transition point or in the Dirac semimetal and thereafter decreases, showing a dip at T  0:5T corresponding to the minimum of q (T ), mainly follow- phase, the potential uctuation is expected to result in TF a minimum conductivity [45{47]. In the insulator phase, ing the temperature dependence of the screening wave vector q (T ) [Fig. 13(e) in the Appendix]. For short- if the band gap is suciently large, the e ect of the po- TF tential uctuation might be limited. The interplay of the range impurities, the conductivity rst decreases, show- ing a dip at T , and thereafter increases, showing a peak impurity potential uctuation, temperature, and band sh gap would be an interesting future research direction. at T  0:25T . These dips and peaks are from the temperature-dependent evolution of the chemical poten- Finally, we wish to mention the additional parabolic 9 2 2 ~ k \interband" scattering between these two surfaces. The term  omitted in Eq. (1) along the armchair (y) 2m maximum value of  is thus given by direction beyond the lowest order [48]. This term could a ect the dc conductivity, especially at high densities arccos ( 6= 0 and jj < r); cr 2m v above the crossover Fermi energy " = , where (r) = (A3) ( < 0 and jj  r); max the e ective Hamiltonian in Eq. (1) is no longer valid. (otherwise); For example, at the semi-Dirac transition point with cr "  " , the parabolic term becomes dominant over where  2 [ (r);  (r)]. This coordinate transfor- max max the linear term along the armchair direction; thus, xx mation changes the Hamiltonian into the following form: and  will follow those of (anisotropic) 2DEG. yy In summary, we calculate the dc conductivity of few- 0 e H = " r : (A4) layer BP as a function of the density and temperature us- 0 e 0 ing the anisotropic multiband Boltzmann transport the- ory, which is essential when the e ect of anisotropic en- In the transformed coordinates, the energy dispersion is ergy dispersion or interband scattering becomes impor- given by " (r) = " r and the corresponding eigenstates tant. We nd that the dc conductivities in the Boltzmann are given by limit show characteristic density and temperature depen- dence in each phase, which could be used as a signature j+i = ; (A5a) of the tunable electronic structure of BP in transport measurements. 1 1 ji = : (A5b) The Jacobian J corresponding to this transformation ACKNOWLEDGMENTS is given by @k @k This work was supported by the National Research x x k r @r @ 0 J = = p  J (r; ): (A6) Foundation of Korea (NRF) grant funded by the Ko- @k @k y y 2c r cos @r @ rea government (MSIT) (No. 2018R1A2B6007837) and Creative-Pioneering Researchers Program through Seoul (i) Note that, for the + band, the band velocity v = National University (SNU). @" 1 +;k can be expressed as ~ @k (x) v = 2 v cos  r cos  ; (A7a) Appendix A: Eigenstates and density of states (y) v = v c sin ; (A7b) In this section, we provide a detailed explanation on where v = . ~k the model Hamiltonian of few-layer black phosphorus 0 The DOS at the semi-Dirac transition point ( = 0) (BP), and its various properties including density of at the energy " > 0 can be obtained analytically as states (DOS). In the model Hamiltonian given by Eq. (1) in the main text, the exact values of m and v depend d k D(") = g (" " ) on the number of layers and the gap tuning parameter. +;k (2) We introduce the normalization constants k  a and Z Z 2 2 1 ~ k J (r; ) "  ; thus, the Hamiltonian becomes = 2g dr d (" " r) 2m (2) ~ ~ p   1 0 k ick + y 2 x 2 H = " ; (A1) 2gk 2K (1=2) " 0 0 ~ ~ k + ick +  0 = ; (A8) c" " 0 0 where k = k=k , c = ~v k =" , and   . To avoid where g is the spin degeneracy, and the factor 2 0 0 0 0 2" diculties associated with anisotropic dispersion, we con- originates from the duplicate parts of the Fermi sur- sider the following coordinate transformation with faces parameterized by = 1. Here, K (k) = 1 2 2n [(2n 1)!!=(2n)!!] k is the complete elliptic in- n=0 tegral of the rst kind with K (1=2)  1:854 [44]. Note k ! k (r cos  ) ; x 0 (A2) that the Thomas{Fermi wave vector is determined by the k ! r sin ; DOS at the Fermi energy " given by p 1 2 2 2e 4g k 2K (1=2) " 0 0 F where = 1 represents each half of the Fermi sur- q = D(" ) = ; (A9) TF F faces. This Fermi surface splitting is especially useful for  c " the  < 0 case where there are two distinct Fermi sur- faces (see Fig. 1(f ) in the main text), accounting for the where = is the e ective ne structure constant. ~v 0 10 The carrier density is thus given by energy dispersion. To consider the anisotropy of the en- ergy dispersion, we express the multiband anisotropic Boltzmann equation in Eq. (7) using the transformed co- 4g 2K (1=2) " n = d"D(") = n ; (A10) 2 ordinates in Eq. (A6) as follows: 3 c " 0 0 2 1 3 3 where n = k . Note that "  n and D(" )  n . 0 F F Figure 2 in the main text shows the calculated DOS and the carrier density for each phase. Appendix B: Density dependence of dc conductivity in black phosphorus In this section, we derive the dc conductivity at zero temperature for 2D multiband systems with anisotropic Z Z (i) 1  (r ) max 0 0 J (r ;  ) v (i) 0 0 (i) 0 0 1 = dr d W 0 0 0 kk k k (i) 0 (2) 0  (r ) 0 max v Z Z 1 0 0 2 0 0 0 d k r 2 (i) (i) (i) 0 0 2 0 = dr p n jV j F (" r " r )  d 0 0 0 0 0 imp 0 0 kk kk k kk k 0 0 (2) 2c r cos   ~ 0 r 2 0 2 k d r 0 0 (i) (i) (i) 0 2 = n p jV j F  d  ; (B1) 0 0 0 0 0 imp kk kk k kk k ~ 2c" 2 2 r cos where = 1 represents each half of the Fermi surfaces, Since Eq. (B3) holds for both = 1, we can rewrite it (i) (i) (i) as d = v =v , and F = [1 + cos(  )] is 0 0 0 0 kk k k kk the square of the wave function overlap between k and (i) (i) (i) (i) 0 0 1 = w ~ () ~ () w ~ (;  ) ~ ( ) (B5a) k states in the same conduction (or valence) band. Let 11 1 11 1 2 r " 1 2 2 0 0 us de ne  = , V = , and = n V  , 0 2 0 0 0 imp 0 2c" k  (r) ~ d 0 0 0 (i) (i) (i) (i) 0 0 + w ~ () ~ () w ~ (;  ) ~ ( ); then we have 11 1 11 1 Z 0 0 0 X d (i) (i) (i) (i) d r 0 0 1 = w ~ () ~ () w ~ (;  ) ~ ( ) (B5b) 1 = 11 1 11 1 0 2 2 2 r cos 0 (i) (i) (i) (i) 0 0 0 0 (i) (i) (i) + w ~ () ~ () w ~ (;  ) ~ ( ): jV j F  ~ d  ~ ; (B2) 11 1 11 1 0 0 0 0 0 kk kk k kk k 0 0 Now, let us discretize  to  (n = 1; 2; ; N ) with (i) (i) (i) (i) where V = V =V and  ~ =  = . Here, 0 0 (i) (i) 0 0 kk kk k k R an interval  = 2 (r)=N . Thus, for  ~ =  ~ ( ), max n n 0 0 d represents an integration over  (r) <  < max r we have (r). Thus, Eq. (B2) becomes max (i) (i) (i) n (i) nn 1 = P  ~ P  ~ n1 11 n 1 0 0 (i) (i) (i) (i) 0 0 1 = w ~ () ~ () w ~ (;  ) ~ ( ) ; 0 0 0 0 (i) (i) (i) n (i) nn 2 + P  ~ P  ~ ; (B6a) r 11 0 n1 11 n 1 (B3) (i) 0 (i) (i) n (i) nn where 1 = P  ~ P  ~ 11 0 n1 11 n 1 d r 0 0 (i) 0 2 (i) (i) (i) n (i) nn w ~ () = p jV j F ; 0 0 0 + P  ~ P  ~ ; (B6b) 11 0 kk kk 0 n1 11 n 1 2 2 r cos (B4a) (i) (i) (i)nn 0 0 r where P 0 = w ~ ( ) is an N -vector and P = (i) (i) 0 0 0 2 n w ~ (;  ) = p jV j F d : 0 0 0 0 kk kk kk (i) 2 r cos w ~ ( ;  ) is an N  N matrix, which correlates 0 n n (B4b) the di erent -dependent relaxation times for a given 11 ( , ) combination. Note that Eq. (B6) shares the basic minimum point of the conduction band, we have structure with the multiband scattering formula [38, 39] 2 2 (which accounts for the scattering between each half of k " c k x 0 y "(k) = " + the Fermi surface) and the anisotropic scattering formula k 2 k 0 0 [42] (which accounts for the scattering between di erent 2 2 2 2 ~ k ~ k 0 y and  points). Furthermore, Eq. (B6) is a 2N  2N + ; (C1) 2m 2m x y matrix equation with two independent basis indices ( , ), i.e., index for each half of the Fermi surfaces and 2 2 2 2 ~ k ~ k 0 0 where m = and m = . the -discretization index n. x y 2 2" c " 0 0 For comparison, we rst consider a 2DEG with an Thus, the dc conductivity at zero temperature is given isotropic energy dispersion given by by 2 2 X 2 ~ k d k (i) (j) (j) "(k) = : (C2) = ge (" " )v v ij k F k k k 2m (2) Z Z (i) (j) (j) 1 0 2 As the system is isotropic, we can readily calculate the k r(" r " )v v 0 F 2 k k k = ge dr d conductivity of each case using the Einstein relation 2(2) c r cos 0 r Z Z (i) (j) (j) 1 0 X  = e D(" )D; (C3) iso F r(r r )v ~ v ~  ~ k k k =  dr p ; (B7) 2 r cos 0 r 2 where D = is the di usion constant and D(") = gm is the DOS for the isotropic 2DEG. The relaxation (i) (i) 2 2 2 2~ where  = ge  v  , r = " =" and v ~ = v =v . 0 0 0 F F 0 0 0 k k time at the Fermi energy  is given by Thus, from Eq. (A7), we have 2 0 Z 1 2n d k imp 2 0 = jV 0j (" " )(1 cos  ) xx kk F 2 (x) 2 ~ (2) = 2 r cos  r cos   ~ (); F F F 0 Z F 2 2n m d imp 2 0 = jV 0j (1 cos  ) (B8a) ~ 2~ (2) d r sin yy F 2 (y) 2n m imp = c p  ~ (): (B8b) V ; (C4) i2DEG 2 2 2 r cos 0 F ~ 2~ where V 0 is the angle-dependent potential on the Fermi Note that  , v ,  , and  are the density-independent 0 0 0 0 R 2 2 0 surface and V  jV 0j (1 cos  ) is the normalization constants in units of time, velocity, DOS, i2DEG 0 2 angle-averaged square of the impurity potential. and conductivity, respectively. Therefore, the dc conductivity of the isotropic 2DEG is given by Appendix C: Low-density approximate models for 2 2 gm v ~ 2~ 2 F the insulator phase and Dirac semimetal phase = e iso 2 2 2~ 2 2n mV imp i2DEG 2 2 2 ge ~ ~ k In this section, we derive the dc conductivity of low- 2 2 density approximate models for the insulator phase and 2n V 2m imp i2DEG Dirac semimetal phase. Note that the only anisotropy ge ~" considered in these models is the anisotropy in the ef- = ; (C5) 2n mV imp i2DEG fective mass or velocity with the same power-law depen- dence in momentum. 2 2 ~ k ~k F F where v = and " = . F F 2 m 2m Subsequently, let us consider the Fermi energy depen- dence of the dc conductivity using the Einstein relation 1. Insulator phase at low densities in Eq. (C3). For short-range impurities, V is a con- i2DEG stant independent of " ; thus, we have For the insulator phase, as well as the Dirac semimetal phase discussed later, the DOS and carrier density do " : (C6) not follow the simple power-law behavior. Therefore, 2 2 we utilize approximate models to understand the asymp- Here, we used v  k  " . For charged impurities in F F 2 2 totic behavior of dc conductivity at low densities. When the strong screening limit, V  q  D (" ) is i2DEG TF j" j > j" j but the carrier density is suciently small, F g also a constant; thus, the system can be approximated as a two-dimensional electron gas (2DEG). From the series expansion at the   " : (C7) F 12 For the anisotropic 2DEG with di erent e ective the x-direction, the conductivity  becomes xx masses in each direction, we introduce the following co- p Z 2 h i ge m m ordinate transformation [(k ; k ) ! (k; )]: x y (x) x y =  v xx F 2~ 2 m ge m m x x y ~ 2~ k ! k cos ; x = 2~ 2n m m V imp x y a2DEG r (C8) m 2 ~ d 2 2 k ! k sin ; k cos m F mm 2 p ge ~" m m m x y = : (C13) which gives the Jacobian dk dk = kdkd. The x y 2 2n V m m imp a2DEG (i) 1 @" band velocity v = can be expressed as ~ @k Similarly, when the electric eld and the current density are along the y-direction, the conductivity  becomes yy ~k (x) v = p cos ; p Z mm 2 2 h i ge m m x y (y) (C9) =  v yy F ~k F (y) 2 2~ 2 v = sin : mm ge ~" m = : (C14) 2n V m m imp y a2DEG Subsequently, the energy dispersion becomes isotropic in the transformed coordinates; thus, the DOS is given Therefore, the dc conductivities for the anisotropic case by are modi ed as =  ; (C15a) g m m xx iso x y D(") = : (C10) 2~ m =  : (C15b) yy iso The relaxation time of the anisotropic 2DEG for k at the Fermi energy can be obtained by solving the cou- Thus, for short-range impurities or charged-impurities in the strong screening limit, the Fermi energy dependence pled integral equation [Eq. (7) in the main text]. For short-range impurities or charged impurities in the strong of the dc conductivities for the anisotropic 2DEG follows 0 that of the isotropic 2DEG given by Eqs. (C6) and (C7). screening limit, the scattering potential V = V is kk 0 independent of the angle, thus it can be shown that Note that, as the Fermi energy or the carrier density (i) (i) (i) increases, the insulator phase can no longer be approxi- =    . Then the coupled equation can be k k mated by a 2DEG model, and the energy dispersion fol- simpli ed as lows that of the semi-Dirac transition point. Therefore, the power-law dependence eventually follows that of the (i) 2 0 1 2n d k v imp 0 2 k semi-Dirac transition point. 0 0 = jV j (" " ) 1 kk k k (i) 2 (i) ~ (2) m m 2n x y imp = V ; (C11) a2DEG 2 2. Dirac semimetal phase at low densities ~ 2~ (i) 2 v For the Dirac semimetal phase ( < 0), the series 2 d 2 2 where V  jV j 1 = jV j is the 0 0 (i) a2DEG (2) expansion at one of the band touching points gives angle-averaged square of the impurity potential for the (i) anisotropic 2DEG. Note that  is independent of the H (k) = 2 k  + ck  (C16) x x y y direction i. ~ (v k  + v k  ) ; x x x y y y Therefore, the conductivity of the anisotropic 2DEG is given by 2 " c" 0 0 where v = and v = . x y ~k ~k 0 0 For comparison, we rst consider an isotropic 2D Dirac d k 2 (i) (j) (j) = ge (" "(k))v v  semimetal with the Hamiltonian given by ij F (2) p Z 2 2 H (k) = ~v (k  + k  ) : (C17) ge m m x y d x x y y (i) (j) =  v v ; (C12) F F 2~ 2 The DOS is thus given by where  is the relaxation time at the Fermi energy. F gk g" D(") = = : (C18) When the electric eld and the current density are along 2 2 2~v 2~ v 13 The relaxation time at the Fermi energy  is given by the relaxation time of the anisotropic Dirac semimetals Z given by 2 0 1 2n d k imp 2 0 = jV 0j F 0(" " )(1 cos  ) kk kk F 2 Z (i) ~ (2) 2 0 1 2n d k v imp 0 0 0 0 2 = jV j F (" " ) 1 0 kk kk k k (i) 2 (i) 2n k d imp F ~ (2) 2 0 0 = jV 0j F ( )(1 cos  ) ~ 2~v 2 2n gvk imp = V ; (C26) 2n k agp imp F ~ 2~v v = V ; (C19) x y igp ~ 2~v (i) R 2 0 1 0 0 2 v jV j where F ( ) = (1 + cos  ) is the square of the wave 2 d 2 0 0 0 2 where V  jV j F ( ) 1 = 0 (i) R agp (2) 4 0 0 2 v 2 2 0 0 k function overlap and V  jV 0j F ( )(1cos  ) igp 0 2 is the angle-averaged square of the impurity potential is the angle-averaged square of the impurity potential. (i) for the anisotropic graphene. Note that  for the Therefore, the dc conductivity of the isotropic Dirac anisotropic Dirac semimetal is also independent of the semimetal is given by direction i. Therefore, the conductivity of the anisotropic Dirac gk v ~ 2~v = e iso semimetal is given by 2~v 2 2n k V imp igp 2 2 d k ge ~v 2 (i) (j) (j) = ge (" "(k))v v  (C27) = : (C20) ij F (2) 4n V imp igp ge vk d (i) (j) Subsequently, let us consider the Fermi energy depen- =  v v : F F 2~v v 2 x y 0 dence of the dc conductivity using the Einstein relation in Eq. (C3). For short-range impurities, V is a constant igp When the electric eld and the current density are along independent of " ; thus, we have the x-direction, the conductivity  becomes xx " ; (C21) 2 h i ge vk d (x) =  v whereas for charged impurities in the strong screening xx F 2~v v 2 2 2 x y 2 2 0 limit, V  q  D (" )  " ; thus, igp Z TF F 2 2 ge vk ~ 2~v v d x y 2 2 2 = v cos " : (C22)  2 F 2~v v 2n vkV 2 x y imp agp 0 2 2 2 2 2 Note that, even in the weak screening limit, V  k  ge ~v v igp F = : (C28) 2 2 2 " , and in general,   " for charged impurities. 4n V imp agp F F For the anisotropic Dirac semimetals with di erent ve- Similarly, when the electric eld and the current density locities in each direction, we introduce the following co- are along the y-direction, the conductivity  becomes ordinate transformation [(k ; k ) ! (k; )]: yy x y 2 2 h i ge vk d (y) k ! k cos ; =  v yy F x F 2~v v 2 (C23) x y k ! k sin ; 2 2 2 ge ~v = : (C29) 2 2 4n V v imp agp which gives the Jacobian dk dk = kdkd. The x y v v x y Therefore, the dc conductivities for the anisotropic case (i) 1 @" band velocity v = can be expressed as k ~ @k are modi ed as (x) v = v cos ; x x =  ; (C30a) xx iso (C24) (y) v = v sin : =  : (C30b) yy iso Subsequently, the energy dispersion becomes isotropic in the transformed coordinates; thus, the DOS is given Thus, for short-range impurities or charged-impurities in by the strong screening limit, the Fermi energy dependence gvk g" of the dc conductivities for the anisotropic graphene fol- D(") = = : (C25) 2~v v 2~ v v lows that of the isotropic graphene given by Eqs. (C21) x y x y and (C22). Similarly, using the same assumptions which were used Near the van Hove singularities, where the energy dis- in Eq. (C11), for short-range impurities or charged im- 2 2 c k 2 y persion can be expanded as "(k)="  jjk + , the purities in the strong screening limit, we can calculate 0 2jj 14 DOS diverges logarithmically [37], dominating the over- Thus, we have all power-law behavior of the conductivity. Therefore, Z q (T ) TF for short-range impurities, the conductivity becomes = d"D(") (D3) q (0) 2D(" ) TF F [ log(jj " )] ; (C31a) xx F 1 1 + : [ log(jj " )] : (C31b) 1 + cosh (" ) 1 + cosh (" + ) yy F For a given T , the chemical potential is calculated us- For the charged impurities near the van Hove singulari- ing the density invariance in Eq. (D2), and subsequently, ties, the conductivity becomes q (T ) is obtained from the above relation. TF When the DOS is given by a simple power law with [log(jj " )] ; (C32a) xx F respect to energy, we can analytically obtain the temper- [log(jj " )] : (C32b) yy F ature dependence of the chemical potential and Thomas{ Fermi wave vector, and their asymptotic behaviors at low Note that, as the Fermi energy or the carrier density and high temperatures. increases, the power-law dependence of the dc conductiv- Consider a gapless electron{hole system with a DOS ity follows that of the semi-Dirac transition point, as in given by D(") = C j"j ("), where C is a constant the gapped insulator case. and (") is a step function. Using the results from the Supplemental Materials in Ref. [42], we can obtain Appendix D: Temperature dependence of chemical < 1 (T  T ); 12 T potential and Thomas{Fermi wave vector in black = (D4) " : 1 T phosphorus (T  T ); 2( 1)( +1) T where T = " =k is the Fermi temperature,  is the F F B In this section, we derive the temperature-dependent Dirichlet eta function, and is the gamma function [44]. chemical potential and Thomas{Fermi wave vector of For the temperature-dependent Thomas{Fermi wave vec- few-layer BP. When the temperature is nite, the chemi- tor q (T ), we obtain TF cal potential  deviates from the Fermi energy " due to the broadening of the Fermi distribution function (0) (") 1 ( 1) (T  T ); f ("; ) = e + 1 where = . As the q (T ) TF 6 T k T =   (D5) charge carrier density n does not vary under the temper- q (0) > T TF 2( 1)( ) (T  T ); ature change, we have For few-layer BP at the semi-Dirac transition point, (0) n = d"D(")f ("; ) 2 the DOS is given by D(") / " ; thus, = . Thus, we have 1 h i (0) (0) = d"D(") f ("; ) + f ("; ) >  T 1 (T  T ); 0 F 12 T F 1 =   (D6) " > 1 T d"D("): (D1) (T  T ); 1 5 ( ) ( ) 1 2 2 and Thus, the carrier density measured from the charge neu- tral point, n  nj nj , is given by 2 =0 1 (T  T ); q (T ) 12 T TF F 1 h i =   1 (D7) (0) (0) q (0) TF 1 3 T n = d"D(") f ("; ) f (";) 2 (T  T ); 2 2 T where q (0) = q is given by Eq. (A9). TF TF d"D("); (D2) Figure 11, Figure 12, and Figure 13 show the calcu- lated temperature dependence of the chemical potential where the rst and second lines represent the carrier den- (T ) and Thomas{Fermi wave vector q (T ) in various TF sity evaluated at the nite and zero temperatures, respec- phases of BP using Eqs. (D2) and (D3), respectively. tively. Here, we used f ("; ) = 1f (";). By solving this equality in terms of , we can calculate the chemical potential of the system for a given temperature T . See Appendix E: Temperature dependence of dc the Supplemental Material in [42] for the simpli ed cases. conductivity at the semi-Dirac transition point Subsequently, consider the temperature-dependent Thomas{Fermi wave vector q (T ). Note that, in 3D, TF Using Eq. (10) in the main text, we can general- 2 2 2e 2e @n q (0) = D(" ) and at nite T , q (T ) = . ize the conductivity tensor at zero temperature to that TF F TF @ 15 1 2.2 1 1 1 0.9 0.9 0.8 0.8 0.8 1.8 0.8 0.7 0.6 0.6 1.6 0.6 0.7 0.5 1.4 0.4 0.4 0.6 0.4 1.2 0.3 0.2 0.2 0.5 0.2 0.4 0.8 0 0 0.1 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 0.3 0.6 0 1 2 3 4 0 1 2 3 4 2.5 2.5 2.4 2.2 FIG. 11. Calculated temperature dependence of (a) chem- 2 2 ical potential and (b) Thomas{Fermi wave vector for the 1.8 1.5 1.5 semi-Dirac transition point ( = 0). Here, the black solid 1.6 lines represent the numerical result, and the red/blue dashed 1.4 1 1 1.2 lines represent the high-/low-temperature asymptotic forms 0.5 0.5 in Eqs. (D6) and (D7). If the chemical potential and temper- 0.8 ature are normalized by " and T , respectively, the result is F F 0 0 0.6 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 independent of " at the semi-Dirac transition point. FIG. 12. Calculated temperature dependence of (a)-(c) chemical potential and (d)-(f ) Thomas{Fermi wave vector for the gapped insulator phase with  = 1 at (a), (d) " = 1:01" , F 0 (b), (e) " = 1:1" , and (c), (f ) " = 1:5" . F 0 F 0 (0) 1 " at nite temperature. For f (") = z e + 1 , (0) @f (") (0) where z = e is the fugacity, S (") = = @" 1 " z e (0) (0) f (") 1 f (") = . Thus, the conduc- 1 " 2 (z e +1) tivity at nite temperature is given by 2 (0) d k @f (" ) k (i) (j) (j) (T ) = ge v v ij k k k (2) @" Z Z 1 0 2 1 " r X 0 k r z e (i) (j) (j) 2 0 = ge dr d p v v k k k 1 " r 2 2c r cos   (z e + 1) 0 r Z Z 1 0 1 " r X 0 d r " z e 0 (i) (j) (j) =  dr p v ~ v ~  ~ : (E1) k k k 1 " r 2 2 (z e + 1) 2 r cos 0 r Thus, from Eq. (A7), we have Z Z 1 0 1 " r X 0 p " z e d 2 (x) (T ) = 2 dr r cos  r cos  ~ (); (E2a) xx 0 1 " r 2 (z e + 1) 2 0 r Z Z 1 0 2 1 " r X 0 " z e d r sin 2 (y) (T ) = c  dr p  ~ (): (E2b) yy 0 1 " r 2 (z e + 1) 2 2 r cos 0 r To derive the asymptotic behaviors of  (T ) at low and is the energy-independent correction term from the ii (i) high temperatures, assume that the relaxation time can temperature-dependent screening e ect with g (0) = 1. be decomposed into energy- and temperature-dependent (i) T For short-range impurities, g = 1. For charged (i) (i) (i) T (i) T parts as  ("; T ) =  (")g where g T T F F 1.2 1.1 1.1 0.9 0.9 For charged impurities in the strong screening limit, 0.8 0.8 0.8 (i) 0.7 A = , which is two times the low-temperature co- 0.6 0.7 0.6 0.6 ecient in Eq. (D7), and = . Thus, we obtain 0.4 0.5 12 2 0.5 0.4 0.2 0.4 0.3 0 0.2 0.3 1 (T  T ); 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 F (T ) xx = (E5a) (0) : (T  T ); xx 6 T 6 1.6 1.5 1.4 1 (T  T ); 5  (T ) yy 1.4 4 T 1.3 = (E5b) 4  (0) : yy log 2 (T  T ): 1.2 1.2 3 1.1 As the screening strength decreases, the low-temperature 0.9 coecient in Eq. (E3) increases, because the screening 0.8 0.8 (i) coecient A decreases whereas the other part remains 0 0.6 0.7 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 positive. FIG. 13. Calculated temperature dependence of (a)-(c) Appendix F: Temperature dependence of dc chemical potential and (d)-(f ) Thomas{Fermi wave vector conductivity in the low-density approximate models for the Dirac semimetal phase with  = 1 at (a), (d) for the insulator phase and Dirac semimetal phase " = 0:01" , (b), (e) " = 0:9" , and (c), (f ) " = 1:1" . F 0 F 0 F 0 In this section, we present the temperature dependence T (i) T of the chemical potential, Thomas{Fermi wave vector, Coulomb impurities, we expect g  1 A T T F F and conductivity of the low-density approximate models at low temperatures, whereas at high temperatures, the for the insulator phase and Dirac semimetal phase, which energy averaging typically dominates over the screen- are the gapped 2DEG and graphene, respectively. ing contribution and we can assume g  1. Sup- pose the following power-law dependence: D(")  " , (i)  (i) v (")  " , and  (")  " . Subsequently, by 1. Insulator phase rewriting Eq. (E1) as an energy-integral form, we obtain the asymptotic power-law behavior of the temperature- We introduce the gapped 2DEG model system with dependent conductivity at low and high temperatures as the energy dispersion given by "(k) = "  + (k=k ) 0 0 h i 2 with  > 0, to account for the thermal excitation behav- (i) 1 + ( ) A (T  T ); (T ) ii 6 T ior involving the band gap between the valence and con- (0) > duction bands, similar to the insulator phase. Note that ii ( + 1)() (T  T ); the e ects of the di erence between the e ective mass (E3) of each direction are canceled out by zero-temperature normalization. where  = 1 + 2 + . (See the Supplemental Material Figure 14 shows the calculated dc conductivities as a of Ref. [42] for the detailed derivation of the power-law function of the temperature for the gapped 2DEG system analysis of the temperature-dependent dc conductivity.) in the low-density limit with " = 1:1" along with the F 0 For the semi-Dirac transition point ( = 0), = . result of the insulator phase with the same Fermi energy For short-range impurities, g = 1 and from the (see also Fig. 8 in the main text). At low temperatures, 1 the calculated results of temperature-dependent conduc- energy dependence of the relaxation time, = . Thus, tivity in the insulator phase show a similar behavior as from Eq. (E2), the asymptotic behavior is given by that of the low-density approximate model. However, as T the temperature increases, the discrepancy between the 1 (T  T ); (T ) xx 12 T two results increases, and in the high-temperature limit, = (E4a) (0) : xx log 2 (T  T ); the conductivity becomes similar to that of the semi- Dirac transition point. T =T <1 e (T  T ); (T ) yy = (E4b) 1 1 T (0) : + (T  T ): yy 1 5 F 2 T 8 F ( ) ( ) 2 2 2. Dirac semimetal Phase Here, the extra terms in  (T )= (0) were obtained yy yy through the next-order expansion of the temperature cor- For graphene (which is an approximate model for the rections. Dirac semimetal phase in the low-density limit) from 17 30 120 400 whereas those for the Thomas{Fermi wave vector are 25 100 given by 1 3 20 80 250 2 0.5  T 1 1 (T  T ); q (T ) F TF 15 60 200 6 T 0 0 = (F2) 0 0.5 0 0.5 q (0) : TF 2 log 2 (T  T ): 10 40 F 5 20 As shown in Figs. 13(a) and 13(d), the result of 0 0 0 the low-density approximate model and the numeri- 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 cally calculated result of the Dirac semimetal phase in the low-density limit are consistent with each other. FIG. 14. Calculated dc conductivities as a function of the For short-range impurities, the asymptotic form of the temperature for the gapped 2DEG system in the low-density temperature-dependent conductivity becomes [Eq. (E4) limit with  = 1 for (a) short-range impurities, (b) charged with = 0] impurities with = 1000, and (c) charged impurities with = 1. Here, " = 1:1" is used for the calculation. The red 0 F 0 T =T < F 1 e (T  T ); (T ) dashed lines and blue dashed-dotted lines represent the con- gp = (F3) 1 1 T ductivity of the insulator phase (with the same Fermi energy) : (0) gp + (T  T ); 2 16 log 2 T ins ins F and  , respectively. xx yy whereas for charged impurities in the strong screening limit, [Eq. (E4) with = 2] Eqs. (D4) and (D5) with = 2, the low{ and high{ 8 temperature asymptotic behaviors for chemical potential < 1 (T  T ); (T ) gp 3 T are given by = (F4) (0) > gp (T  T ): 6 T 8 F 1 (T  T ); 6 T =   (F1) " > 1 T (T  T ); 4 log 2 T [1] A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. [9] Z. J. Xiang, G. J. Ye, C. Shang, B. Lei, N. Z. Wang, Novoselov and A. K. Geim, The electronic properties of K. S. Yang, D. Y. Liu, F. B. Meng, X. G. Luo, L. J. graphene, Rev. Mod. Phys. 81, 109 (2009). Zou, Z. Sun, Y. Zhang and X. H. Chen, Pressure-Induced [2] S. Das Sarma, S. Adam, E. H. Hwang and E. Rossi, Elec- Electronic Transition in Black Phosphorus, Phys. Rev. tronic transport in two-dimensional graphene, Rev. Mod. Lett. 115, 186403 (2015). Phys. 83, 407 (2011). [10] B. Deng, V. Tran, Y. Xie, H. Jiang, C. Li, Q. Guo, [3] G. Rui, S. Zdenek and P. Martin, Black Phosphorus Re- X. Wang, H. Tian, S. J. Koester, H. Wang, J. J. Cha, discovered: From Bulk Material to Monolayers, Angew. Q. Xia, L. Yang and F. Xia, Ecient electrical control of Chem. Int. Ed. 56, 8052 (2017). thin- lm black phosphorus bandgap, Nature Communi- [4] A. Chaves, W. Ji, J. Maassen, T. Dumitric a and T. Low, cations 8, 14474 (2017). Theoretical Overview of Black Phosphorus, in P. Avouris, [11] Y. Liu, Z. Qiu, A. Carvalho, Y. Bao, H. Xu, S. J. R. Tan, T. Low and T. F. Heinz, eds., 2D Materials: Properties W. Liu, A. H. Castro Neto, K. P. Loh and J. Lu, Gate- and Devices, 381{412, Cambridge University Press, Cam- Tunable Giant Stark E ect in Few-Layer Black Phospho- bridge (2017). rus, Nano Lett. 17, 1970 (2017). [5] R. Rold an and A. Castellanos-Gomez, A new bandgap [12] L. L. Li, B. Partoens and F. M. Peeters, Tuning the elec- tuning knob, Nature Photonics 11, 407 (2017). tronic properties of gated multilayer phosphorene: A self- [6] J. Yang, R. Xu, J. Pei, Y. W. Myint, F. Wang, Z. Wang, consistent tight-binding study, Phys. Rev. B 97, 155424 S. Zhang, Z. Yu and Y. Lu, Optical tuning of exciton and (2018). trion emissions in monolayer phosphorene, Light: Science [13] J. Kim, S. S. Baik, S. H. Ryu, Y. Sohn, S. Park, B.-G. & Applications 4, e312 (2015). Park, J. Denlinger, Y. Yi, H. J. Choi and K. S. Kim, [7] D. akr, C. Sevik and F. M. Peeters, Signi cant e ect of Observation of tunable band gap and anisotropic Dirac stacking on the electronic and optical properties of few- semimetal state in black phosphorus, Science 349, 723 layer black phosphorus, Phys. Rev. B 92, 165406 (2015). (2015). [8] J. Quereda, P. San-Jose, V. Parente, L. Vaquero-Garzon, [14] S. Banerjee, R. R. P. Singh, V. Pardo and W. E. Pickett, A. J. Molina-Mendoza, N. Agra t, G. Rubio-Bollinger, Tight-Binding Modeling and Low-Energy Behavior of the F. Guinea, R. Rold an and A. Castellanos-Gomez, Strong Semi-Dirac Point, Phys. Rev. Lett. 103, 016402 (2009). Modulation of Optical Properties in Black Phosphorus [15] V. Pardo and W. E. Pickett, Half-Metallic Semi-Dirac- through Strain-Engineered Rippling, Nano Lett. 16, 2931 Point Generated by Quantum Con nement in TiO =VO 2 2 (2016). Nanostructures, Phys. Rev. Lett. 102, 166803 (2009). 18 [16] V. Pardo and W. E. Pickett, Metal-insulator transition [31] D. J. P. de Sousa, L. V. de Castro, D. R. da Costa, through a semi-Dirac point in oxide nanostructures: VO J. M. Pereira and T. Low, Multilayered black phospho- (001) layers con ned within TiO , Phys. Rev. B 81, rus: From a tight-binding to a continuum description, 035111 (2010). Phys. Rev. B 96, 155427 (2017). [17] J. Kim, S. S. Baik, S. W. Jung, Y. Sohn, S. H. Ryu, [32] S. S. Baik, K. S. Kim, Y. Yi and H. J. Choi, Emergence of H. J. Choi, B.-J. Yang and K. S. Kim, Two-Dimensional Two-Dimensional Massless Dirac Fermions, Chiral Pseu- Dirac Fermions Protected by Space-Time Inversion Sym- dospins, and Berrys Phase in Potassium Doped Few- metry in Black Phosphorus, Phys. Rev. Lett. 119, 226801 Layer Black Phosphorus, Nano Lett. 15, 7788 (2015). (2017). [33] H. Doh and H. J. Choi, Dirac-semimetal phase diagram [18] N. Ehlen, A. Sanna, B. V. Senkovskiy, L. Petaccia, A. V. of two-dimensional black phosphorus, 2D Materials 4, Fedorov, G. Profeta and A. Gruneis,  Direct observation 025071 (2017). of a surface resonance state and surface band inversion [34] G. Montambaux, F. Pi echon, J.-N. Fuchs and M. O. Go- control in black phosphorus, Phys. Rev. B 97, 045143 erbig, Merging of Dirac points in a two-dimensional crys- (2018). tal, Phys. Rev. B 80, 153412 (2009). [19] P. Di Pietro, M. Mitrano, S. Caramazza, F. Capitani, [35] G. Montambaux, F. Pi echon, J.-N. Fuchs and M. O. Go- S. Lupi, P. Postorino, F. Ripanti, B. Joseph, N. Ehlen, erbig, A universal Hamiltonian for motion and merging of A. Gruneis,  A. Sanna, G. Profeta, P. Dore and A. Peruc- Dirac points in a two-dimensional crystal, The European chi, Emergent Dirac carriers across a pressure-induced Physical Journal B 72, 509 (2009). Lifshitz transition in black phosphorus, Phys. Rev. B 98, [36] R. de Gail, J.-N. Fuchs, M. Goerbig, F. Pichon 165111 (2018). and G. Montambaux, Manipulation of Dirac points in [20] F. Xia, H. Wang and Y. Jia, Rediscovering black phos- graphene-like crystals, Physica B: Condensed Matter phorus as an anisotropic layered material for optoelec- 407, 1948 , proceedings of the International Workshop tronics and electronics, Nature Communications 5, 4458 on Electronic Crystals (ECRYS-2011) (2012). (2014). [37] M. P. Marder, Condensed Matter Physics, 2nd edition, [21] A. Mishchenko, Y. Cao, G. L. Yu, C. R. Woods, R. V. John Wiley & Sons (2010). Gorbachev, K. S. Novoselov, A. K. Geim and L. S. Lev- [38] E. D. Siggia and P. C. Kwok, Properties of Electrons itov, Nonlocal Response and Anamorphosis: The Case in Semiconductor Inversion Layers with Many Occupied of Few-Layer Black Phosphorus, Nano Lett. 15, 6991 Electric Subbands. I. Screening and Impurity Scattering, (2015). Phys. Rev. B 2, 1024 (1970). [22] T. Li, Z. Zhang, X. Li, M. Huang, S. Li, S. Li and Y. Wu, [39] S. Woo, E. H. Hwang and H. Min, Large negative dif- High eld transport of high performance black phos- ferential transconductance in multilayer graphene: the phorus transistors, Applied Physics Letters 110, 163507 role of intersubband scattering, 2D Materials 4, 025090 (2017). (2017). [23] Y. Y. Illarionov, M. Waltl, G. Rzepa, J.-S. Kim, S. Kim, [40] J. Schliemann and D. Loss, Anisotropic transport in a A. Dodabalapur, D. Akinwande and T. Grasser, Long- two-dimensional electron gas in the presence of spin-orbit Term Stability and Reliability of Black Phosphorus Field- coupling, Phys. Rev. B 68, 165311 (2003). E ect Transistors, ACS Nano 10, 9543 (2016). [41] K. Vyb  orny,  A. A. Kovalev, J. Sinova and T. Jungwirth, [24] J. Qiao, X. Kong, Z.-X. Hu, F. Yang and W. Ji, High- Semiclassical framework for the calculation of transport mobility transport anisotropy and linear dichroism in anisotropies, Phys. Rev. B 79, 045427 (2009). few-layer black phosphorus, Nature Communications 5, [42] S. Park, S. Woo, E. J. Mele and H. Min, Semiclassical 4475 (2014). Boltzmann transport theory for multi-Weyl semimetals, [25] Y. Liu, T. Low and P. P. Ruden, Mobility anisotropy in Phys. Rev. B 95, 161113 (2017). monolayer black phosphorus due to scattering by charged [43] N. W. Ashcroft and N. D. Mermin, Solid State Physics, impurities, Phys. Rev. B 93, 165402 (2016). Brooks Cole (1976). [26] Y. Liu and P. P. Ruden, Temperature-dependent [44] G. B. Arfken, H. J. Weber and F. E. Harris, Mathematical anisotropic charge-carrier mobility limited by ionized im- Methods for Physicists, Academic Press, 7th ed. (2012). purity scattering in thin-layer black phosphorus, Phys. [45] S. Adam, E. H. Hwang, V. M. Galitski and S. Das Sarma, Rev. B 95, 165446 (2017). A self-consistent theory for graphene transport, Proceed- [27] F. W. Han, W. Xu, L. L. Li, C. Zhang, H. M. Dong and ings of the National Academy of Sciences 104, 18392 F. M. Peeters, Electronic and transport properties of n- (2007). type monolayer black phosphorus at low temperatures, [46] Q. Li, E. H. Hwang and S. Das Sarma, Disorder-induced Phys. Rev. B 95, 115436 (2017). temperature-dependent transport in graphene: Puddles, [28] M. Zare, B. Z. Rameshti, F. G. Ghamsari and R. As- impurities, activation, and di usion, Phys. Rev. B 84, gari, Thermoelectric transport in monolayer phospho- 115442 (2011). rene, Phys. Rev. B 95, 045422 (2017). [47] N. Ramakrishnan, M. Milletari and S. Adam, Trans- [29] P. Adroguer, D. Carpentier, G. Montambaux and port and magnetotransport in three-dimensional Weyl E. Orignac, Di usion of Dirac fermions across a topo- semimetals, Phys. Rev. B 92, 245120 (2015). logical merging transition in two dimensions, Phys. Rev. [48] J. Jang, S. Ahn and H. Min, Optical conductivity of B 93, 125113 (2016). black phosphorus with a tunable electronic structure, [30] A. N. Rudenko and M. I. Katsnelson, Quasiparticle band arXiv:1811.07529 (2018). structure and tight-binding model for single- and bilayer black phosphorus, Phys. Rev. B 89, 201408 (2014).

Journal

Condensed MatterarXiv (Cornell University)

Published: Nov 9, 2018

There are no references for this article.