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

Learn More →

Gate-tunable flat bands in van der Waals patterned dielectric superlattices

Gate-tunable flat bands in van der Waals patterned dielectric superlattices potential, gapped vdW dielectric Gate-tunable at bands in van der Waals patterned dielectric superlattices 1 2 1;2 Li-kun Shi , Jing Ma , and Justin C.W. Song Institute of High Performance Computing, Agency for Science, Technology, & Research, Singapore 138632 and Division of Physics and Applied Physics, Nanyang Technological University, Singapore 637371 Superlattice engineering provides the means to reshape the fabric felt by quasiparticles moving in a material. Here we argue that bandstructure engineering with superlattices can be pushed to the extreme limit by stacking gapped van der Waals (vdW) materials on patterned dielectric substrates. Speci cally, we nd that high quality vdW patterned dielectric superlattices (PDS) realize a series of robust at bands that can be directly switched on and o by gate voltage in situ. In contrast to existing superlattice platforms, these at bands are realized without the need for ne tuning. Instead, the bands become at as the gate voltage increases in magnitude. The characteristics of PDS atbands are highly tunable: the type of atband (single non-degenerate or dirac-cone-like), localization length, and interaction energy are sensitive to the applied gate voltage. As a result, electron-electron interactions in the PDS atbands can become stronger than both the bandwidth and disorder broadening, providing a setting for correlated behavior such as atband ferromagnetism. We expect PDS atbands can be experimentally realized in a range of readily available gapped vdW materials such as monolayer transition metal dichalcogenides, e.g. WSe . (a) (b) van der Waals (vdW) heterostructures have become -5 5 0.3 a powerful platform to tailor the electronic properties of materials [1{3]. One case in point is moir e superlat- tices, formed when two vdW materials are stacked and 0.2 twisted. In such moir e materials, electrons and other quasiparticles experience slowly varying (emergent) ef- fective periodic potentials. Even when the potentials are relatively weak (as compared with the kinetic en- ergy of electrons in each layer), a wide variety of phe- nomena can be realized that include for e.g., emergent -10 electronic bandgaps [4], moir e excitons [5{9], Hofstadter spectra [4, 10, 11], as well as topological bands [12, 13], to name a few. However, when top and bottom layers cou- -5 5 ple strongly, extreme bandstructure reconstruction takes e ect, allowing nearly at electronic bands to form at FIG. 1: (a) Schematic of a triangular patterned dielectric magic [14, 15] or low-twist angles [16, 17]. These provide superlattice (gray) and the spatial potential pro le (purple) a vdW venue to realize correlated behavior, with intense sustained for the target gapped van der Waals material (red). interest sparked by reports of correlated insulating be- Here the top gate (yellow) and back gate (blue) sustain a havior and superconductivity in moir e materials under potential drop in the out-of-plane direction. The spacing be- such conditions [18{20]. tween two neighboring holes is a. (b) Electrostatic potential for a single hole inside the dielectric substrate (gray) in be- However, achieving good twist angle control over moir e tween top and bottom gates, see SI for full details. The top panel shows the electric potential across the red line (plane of supelattices can often be experimentally dicult; this the vdW material) displayed in the bottom panel. Thin white can be further complicated by lattice reconstruction that lines indicate equi-potential contours. Here we have used po- arise at low twist angle [21]. Good twist/registration tential di erence between the two (top and bottom) gates of control becomes critical given that extreme bandstruc- 5 eV for illustration giving a superlattice potential amplitude ture engineering typically occurs only at speci c \magic" (V see Eq. 1) of about 10 meV; we note these correspond to twist angles or in a small range of low-twist angles [14{ electric elds smaller than the dielectric breakdown voltage 17]. Recently, an inverted electrostatic strategy | pat- of hBN. terned dielectric superlattices (PDS) | wherein a dielec- tric material is patterned into a superlattice and placed (where the patterned dielectric did not signi cantly de- on top of a gate electrode (Fig. 1a) has been experimen- grade the device performance [22]), as well as high gate tally demonstrated in graphene [22], producing high qual- electrode tunability that could turn electrically on and ity superlattices. In these, dielectric contrast between o the superlattice potential. patterned hole and dielectric substrate material enables the gate electrode to sustain a spatially modulated su- Here we argue that PDS can be pushed into the strong perlattice potential (Fig. 1b). Perhaps most remarkable coupling regime, where the superlattice potential (V ) is that the PDS devices maintained an ultrahigh mobility can exceed the kinetic energy of the electrons in a gapped top gate bottom gate arXiv:1904.07877v1 [cond-mat.mes-hall] 16 Apr 2019 2 vdW material. As we describe below, in this regime, not a ect the physics we discuss below. Further, we note at electronic minibands can be achieved and are highly that the particularly large Ising spin splitting ( 100s gate-tunable. Strikingly, these atbands do not require of meV) in the valence bands of many TMD materials, ne tuning of twist angle. Indeed, the bands become e ectively lock the valley and spin in the valence bands at as gate voltage is switched on, achieving small band- making the model in Eq. (1) a good descriptor of the widths even for modest superlattice potentials (Fig. 2). low-energy mini-bandstructure. This stands in contrast to strategies using moir e super- As shown in Fig. 1b, dielectric inhomogeneity in the lattices that only feature at bands at magic or low twist substrate material can enable electric control of a spa- angles [14{17]. tially inhomgeneous potential with variations on the or- PDS atbands can be achieved in gapped vdW materi- der of the hole width (e.g., several tens of nanometers) als such as the transition metal dichalcogenides. The key [22]. Here gray indicates a dielectric (shown SiO with ingredients are a large bandgap as well as a large e ective  = 4) and empty indicates air ( = 1). As shown, when mass. The former ensures that conduction and valence a gate potential is applied, the dielectric contrast yields bands do not cross when V is applied, and the latter a spatially modulated potential even in the vdW mate- gives a small initial kinetic energy of the electrons. These rial layer (red) with potential amplitudes V that can be enable the superlattice potential to con ne the electrons switched on and o electrically [22]. Fig. 1b was plotted e ectively and form at minibands. A particularly good using a numerical solution for a generalized Poisson equa- candidate for PDS atbands is WSe , wherein high mo- tion in a spatially inhomogeneous dielectric environment 4 2 bility (& 310 cm =Vs) samples have been isolated [23]; (see Supplementary Information, SI). these have mean free paths of several hundred nanome- To illustrate the ( at) band structure engineering in ters [24]. For WSe , we anticipate at electronic mini- the PDS of vdW materials, we rst use a triangular su- bands that are well separated can be achieved for mod- perlattice (see other lattices in SI) so that est V and superlattice periods. Interestingly, PDS yields f (r) = 2cos(G  r +  ); (2) i i bunches of at bands that proliferate throughout the pa- i=1;2;3 rameter space. For example, for V > 0, we nd that in addition to the top most valence miniband which is a sin- where G are the triangular superlattice wave vector ori- gle non-degenerate atband (per valley), the other lower ented 60 relative to each other, jG j = 2=a, and  is i i (valence) minibands also bunch up into atband bundles. a relative phase. For simplicity, we will set  = 0 in Similar atband bundles can be found for V < 0, as well s the following. It does not qualitatively a ect the physics as in the conduction band. we discuss below. As expected, the superlattice folds the Perhaps the most exciting aspect of PDS atbands is original TMD bands into a series of superlattice mini- the possibility of direct gate access to correlated behav- bands (band folding is illustrated for V = 0 as the light ior. As an illustration, we consider ferromagnetism in a dashed lines in Fig. 2a). Here the bandstructure is plot- PDS atband; this is in close analogy with quantum-hall ted in a superlattice de ned mini Brilluoin zone (MBZ), ferromagnetism. In such a case, we nd a ferromagnetic where the tilde symbol indicates the MBZ. When V is state, where the spins in either valley K=K are favored, switched on, Bragg scattering mixes the bands and pro- can be achieved for readily available samples and super- duces a mini-band structure (colored lines and gray lines lattice gate voltages. with V = 0:5 meV) in both the conduction band (pos- Gate tunable at minibands { We begin with a discus- itive energies) and the valence band (negative energies) sion of the PDS scheme using monolayer transition metal as shown in Fig. 2a. We have colored only the rst three dichalogenides (TMD). The PDS that we consider here conduction and valence minibands so as to focus our dis- (and that can be experimentally fabricated) are of fairly cussion on them; in solving for the mini-bandstructure long-wavelength a & 20 nm. The resulting superlattice a set of 162 mini-bands are employed to ensure conver- bandstructure can be captured by an e ective kp model gence of the lower bands (we only show 12 minibands) expanded close to the K and K points as that we focus on in the main text. Physically, the inclu- sion of many minibands in our numerical calculation is H = v~( k +  k ) +  + V  f (r); (1) x x y y z s 0 to to capture the physics of the extreme bandstructure where v is the velocity,  are the Pauli matrices, 2 reconstruction regime in which the superlattice strength x;y;z is the bandgap of the TMD,  = 1 for K and K valleys, V is larger than that of a kinetic energy of the electrons. V is the amplitude of the superlattice potential applied Here we have used material parameters corresponding to by the PDS scheme in Fig. 1, and f (r) is the spatial WSe with v~ = 3:94 eVA,  = 0:8 eV [26]. pattern of the superlattice potential. Here we have sup- Strikingly, a clear gap between the top most (va- pressed spin indices since the bandstructure physics we lence) miniband (blue, Fig. 2a) and the rest of the mini- discuss does not di erentiate between spin/valley species bands opens up. This is distinct from that expected (i.e. the superlattice potential is a scalar potential); spin from graphene PDS [22] where a  berry phase pre- can be included in a straightforward fashion that does vents backscattering and gap opening; in the case of top valence bandwidth (a) (b) -800 -760 0 2.5 5 -780 -805 -800 20 40 -10 -5 0 5 10 FIG. 2: (a) Superlattice miniband dispersion (left) and the corresponding density of states (right) for vdW PDS with monolayer WSe . Here the electronic structure is described by Eq. (1) with v~ = 3:94 eVA,  = 0:8 eV; the substrate is a triangular patterned dielectric superlattice with a = 20 nm [see Fig. (1a)]. Solid (dashed) lines are for V = 0:5 meV (V = 0) [see Eq. (1)], s s and the inset shows the mini Brilluoin zone de ned by the superlattice. (b) Miniband minima and maxima that indicate its bandwidth are plotted as a function of V for 12 mini-bands close to the intrinsic band gap. Here solid colored lines indicate miniband maxima and minima, whereas shaded region indicate region in between maxima/minima in the minibands. All other parameters (aside from varying V ) are the same with those in (a). The inset shows the top valence bandwidth as a function of  [with v~ and a xed to the values indicated in (a)]. graphene PDS secondary Dirac cones form at the MBZ the other minibands in the valence band (that can reach corners [4, 10, 11, 22, 25]. In contrast, the wavefunction tens of meV); the minigaps increase with applied gate (AB-sublattice pseudospinor) in TMD close to the band potential [24]. edge has weight mostly on a single sublattice (aligned to One unusual feature is that the application of superlat- a pole in the Bloch sphere), allowing maximal backscat- tice PDS also renormalizes the e ective bandgap between tering and mini-gap formation. the conduction and valence band states, with changes in e ective bandgap of up to 60 meV for the largest su- Due to the large  in TMDs, minigap opening squeezes perlattice potential amplitude shown in the gure (see the top most valence miniband creating a con ned en- Fig. 2b). This arises from the signi cant miniband re- ergy window for it to exist with a sharp density of states construction. Indeed locally in real space, the electrons (shaded blue in right panel of Fig. 2a). Indeed, as V feel large variations in superalattice potential that range is increased, the top most (valence) mini-band is fur- from +6V at the peak to 6V in the troughs. This large s s ther attened (Fig. 2b) with very narrow band-widths peak-to-trough di erence enables electrons to be con ned ( 1 meV) achievable with modest applied gate voltage locally to produce a atband structure (Fig. 2a). creating a nearly at band. To see this, we have plot- ted the maximum and minimum energy in each mini- The blue band is not the only at band that occurs in band (solid lines Fig. 2b, this indicates the bandwidth) PDS. For example, at suciently high superlattice po- with the colors used corresponding to the colored 3 con- tential (V > 0), both green and purple minibands in duction and colored 3 valence minibands of Fig. 2a; the the valence band atten out. Unlike the blue band they shaded region between miniband maxima and minima tend to stick together. In fact, the PDS scheme yields are shaded in corresponding colors. We note that the sets of at minibands in the valence band (some of which at top most (valence) miniband is well separated from are nearly degenerate with each other forming bunched other bands with the large   eV to the minibands bundles of bands) with severely compressed bandwidths. in the conduction band, and gate tunable minigaps to Each of these bundles of at bands are well separated energy, energy, 4 10 7 (a) (b) from each other with large electrically tunable minigaps (in Fig. 2b, we show three at miniband bundles in the valence band at V > 0). We note that the PDS scheme, when applied to gapped Dirac materials such as WSe , naturally breaks particle- 3 1 2 FM phase hole symmetry with (positive and negative energy) mini- bands exhibiting contrasting behavior (see Fig. 2a). For 0.1 1 5 10 15 20 1 5 10 15 example, when V > 0 the top (valence) miniband (blue) gets squeezed into a single non-degenerate at miniband and is well separated from the other minibands, while the FIG. 3: (a) Interaction energy U as a function of the extent bottom (conduction) miniband (yellow) adheres closely of electronic wavefunction in the at band, a , and e ective to the next miniband (red); this displays a particle-hole screening radius, r . Dashed contour lines indicate lines of asymmetric behavior. Symmetry in the miniband struc- constant U and are in units of meV. (b) Dimensionless Z ( = " )U , see Eq. (5) indicating propensity for the ferromagnetic ture, however, is restored when both V ! V and 0 s s instability with Z ( = " ) = [ 2 ] and taken at a " ! " are interchanged (see detailed discussion in 0 0 k k xed r = 5 nm for illustration. For Z ( )U > 1 (boundary s 0 SI). denoted by dashed line), the system enters a ferromagnetic This unconventional feature allows the type of at instability. Here lg denotes log . mini-bandstructure to be tuned by gate voltage. When V > 0 the top (valence) miniband (blue) is a single non- degenerate at band (per valley). However, when V < 0 between layers are required (in the case of twisted bilayer this same blue miniband, while attening out, adheres graphene) or low twist angles required in other moir e su- closely to the green band (Fig. 2b); when large enough perlattice strategies. The key to achieving at bands in gate voltage is applied, they form a close pair of at- PDS is the large  as well as low velocities. Indeed, bands that bunch together. This atband bundle has a as shown in the inset of Fig. 2b, when we x v~ and width  1 meV. Even though both green and blue mini- a (same parameters as panel a), the larger the , the bands stay together, nevertheless, we nd that they are smaller the bandwidth at a given superlattice potential separated by a very small energy gap. Indeed, close to 0 V . For WSe , we nd  = 800 meV allowing very small ~ ~ s 2 K; K points, the blue and green miniband structure (for bandwidths to be achieved even for small V applied, e.g., V < 0) bands resemble Dirac cones (with an extremely the bandwidth is eV when V = 5 meV. We note, par- small gap; we estimate the gaps to be of order several enthetically, that when  is small so that V applied is eV, see SI). We note that away from the small mini- on order , the minibands in the conduction and valence gaps, the minibands spectra mimic the attened Dirac band can start to mix, vastly complicating the miniband bands in twisted bilayer graphene [14]; interestingly, the 0 structure and making the conditions for at bands in ~ ~ chirality of electrons at K; K points are the same [24] PDS hard to achieve. mirroring the behavior found in twisted bilayer graphene. We anticipate that the PDS scheme can be applied As a result, PDS enables to achieve multiple types of to other van der Waals materials with large bandgaps. atbands via in situ gate voltage tuning (e.g., from sin- For example, we have computed the miniband structure gle non-degenerate at-miniband for positive V to Dirac- for a range of TMD materials and have found that well cone-like for negative V ). This unusual asymmetric be- separated sets of at minibands generically occur { see havior (for both conduction/valence bands and opposite SI for full band structure. Further, other superlattices signs of V ) arises from the gapped Dirac pseudo-spinor can also be easily implemented and produce qualitatively form of their wavefunctions (such asymmetry does not the same results: we have computed the minibands for arise for a simple massive two-dimensional electron gas). square as well as hexagonal superlattices as well and nd Indeed, the role of the TMD pseudospinor texture is fur- similar well separated sets of at minibands [24]. This ther evidenced by how PDS induced Bragg scattering also versatility with superlattice structure provides the abil- changes the winding of the gapped Dirac pseudo-spinor ity to study atbands and its concomittant interaction wavefunctions. This dramatically reconstructs the Berry e ects in other types of lattices which can have a di er- curvature distribution in each of the minibands (for dis- ent type of symmetry (that may be di erent from the cussion of miniband Berry curvature, see SI). underlying lattice). One of the most attractive feature of PDS at bands, is that they do not occur at ne-tuned gate voltages Flatband ferromagnetism | Perhaps the most striking or superlattice wavelength. The larger the superlattice consequence of PDS at bands is the ability to enhance wavelength, the smaller the applied superlattice poten- correlation e ects. This is because the extremely small tial needed to atten the PDS minibands. This strat- bandwidth of single-particle mini-bandstructure of PDS egy stands in stark contrast to the at bands found in ( 1 meV) quenches the kinetic energy of the electrons. moir e superlattices where speci c \magic" twist angles As a result, other energy scales such as that arising from 5 electron-electron interactions can determine the behavior a ect the qualitative conclusions we discuss below. Here of the electronic system. For clarity, in the following, we A = [ 2 ] is a normalization constant. will focus on the top-most valence miniband for V > 0 For small U , the two spin species are degenerate with which is well separated from the other superlattice mini-  =  =  is the spin chemical potential. However, for " # 0 bands. Further, we note that due to the large Ising split- large enough U this spin symmetry can become broken. ting in TMDs, the spin and valley degree of freedom are To see this, we rst write n = n  n where 2n = s 0 0 locked; at the non-interacting level, this band has only a n + n . Expanding Eq. (4) to O(n ) we obtain the " # two-fold degeneracy with spin up and spin down occur- energy density for the atband as ring (original) valley at K and K . 2 4 E = F + n [1=Z ( ) U ] +O(n ); (5) MF 0 0 To proceed, we rst estimate the strength of electron- 0 2 electron interactions in the superlattice by modeling the where F = 2 d" "Z (") Un is the energy in the average interaction energy of electrons con ned in the symmetry unbroken phase, and the n term describes troughs of the superlattice potential as [27, 28] the energy cost to imbalance the spin species. Cru- cially, an instability in the spin up/down population is 2e 2 2 induced when the coecient of the second term is nega- U = exp(q a ); (3) (q + q ) s tive Z ( )U > 1. The conditions for symmetry breaking depend on a competition between the broadening and the exchange where q = r is the e ective inverse screening length, energy. In our model above, this is parameterized by and a is the extent of the PDS atband electronic wave- r , a , as well as the broadening energy/width . To function con ned in the troughs of the superlattice. Here s W illustrate this, we plot Z ( = " )U in Fig. 3b; this is we have chosen a simple Gaussian ansatz for the extent 0 0 the maximal value that is obtained at half- lling. From of the wavefunction; other models do not qualitatively a ect the results we discuss below. In Fig. 3a, we plot Fig. 3b, we can see the competition clearly: for larger (smaller) , we require more (less) localized electron the strength of U as a function of both the screening states, or larger (smaller) exchange energy to enter the length r as well as a taking  = 4. This yields large s W broken symmetry phase. Taking r = 5 nm as a demon- interaction energies of order several tens of meV for a stration, we nd large Z ( = " )U for a wide swathes of wide range of r and a . For WSe , we estimate e ec- 0 0 s W 2 the -a parameter space (Fig. 3b). We estimate that tive r of a few nanometers arising either from intrinsic a  5 10 nm can be achieved by modest superlat- screening of the electron gas or from proximal gates, see tice potentials in the PDS scheme. At low temperatures, discussion in SI. To estimate the extent of the electronic broadening is typically dominated by disorder [29{31]. wavefunction a , we employ a variational approach on Taking values for high quality WSe [23], we estimate the Gaussian wavefunction ansatz. This yields typical disorder-induced broadening  of order several meV a that decreases for increasing superlattice potential, dis are available in high quality present day samples, see SI. reaching a fairly con ned state of  5 10 nm even for As a result, we expect that the conditions for realizing a modest applied superlattice potential, see SI. ferromagnetic state using PDS atbands can be attained Fig. 3a indicates that U can dominate over the ex- in WSe . Interestingly, this analysis is not con ned to tremely small non-interacting PDS atband bandwidths 2 half- lling  =  . The ferromagnetic instability can (see e.g., Fig. 2). As a result, we anticipate interaction ef- 0 0 occur at a variety of chemical potentials so long as the fects can become signi cant. For example, large exchange criterion is satis ed Z ( )U > 1, see below. energy can drive the spin-degenerate atband system into 0 Discussion | There are numerous probes of the ex- a ferromagnetic state. To describe this, we use a simple treme mini-bandstructure reconstruction we discuss here. mean- eld model for the energy density E of a PDS MF For example, we anticipate a lightly hole doped WSe at miniband with spins indexed s =";# as 2 PDS to exhibit a dramatic change in its low-frequency h  i 2 (THz) optical absorption characterisitcs as V > 0 is E = d" "Z (") n ; (4) MF switched on. While at V = 0 such a sample will exhibit a Drude peak around ! = 0, when V > 0 is switched on, where  is the chemical potential of spin s, and " the Drude peak will diminish as the topmost (valence) is the band energy, and U is the strength of the ex- miniband attens, and exhibit additional sharp THz ab- change energy and can be estimated by Eq. 3 [for full sorption peaks corresponding to transitions between the description, see SI]. Here we have used a simple Gaus- sets of at minibands (in the valence band). 2 2 sian Z (") = A exp[("" ) =2 ] to model a broadened Similarly, we expect dual-gate control to enable con- spectral weight of the atband, and the density of each trol of both superlattice potential as well as the lling of spin species is n = d"Z ("). The broadening, , can the minibands. Such lling control can enable to probe be induced by a number of di erent processes for e.g., both the metallic as well as the insulating states induced via disorder. Other forms of the spectral function do not by the superlattice potential. When pushed into the 6 regime where atbands exhibit ferromagnetism, the spins [11] C. R. Dean et al., Nature 497, 598 (2013). [12] J. C. W. Song, P. Samutpraphoot, L. S. Levitov, Proc. (locked to the valley) split; at half- lling of a atband this Natl. Acad. Sci. U.S.A 112, 10879 (2015). would exhibit a ferromagnetic insulating state. Inter- [13] Q. Tong et al., Nat. Phys. 13, 356 (2017). estingly, ferromagnetism can also occur away from half- [14] R. Bistritzer, A. H. MacDonald, Proc. Natl Acad. Sci. lling so long as Z ( )U > 1. As a result, the ferromag- USA 108, 12233 (2011). netic system can be metallic. Crucially, we note that the [15] D. Santos, J. M. Lopes, N. M. R. Peres, A. H. Castro minibands exhibit non-zero Berry curvature distribution Neto, Phys. Rev. Lett. 99, 256802 (2007). (see details in SI). As a result, when the spins/valleys [16] F. Wu, T. Lovorn, E. Tutuc, A. H. MacDonald, Phys. Rev. Lett. 121, 026402 (2018). are split (and  is away from half- lling) we anticipate [17] F. Wu, T. Lovorn, E. Tutuc, I. Martin, A. H. MacDonald, that a (charge) anomalous Hall e ect can ensue. Phys. Rev. Lett. 122, 086402 (2019). In summary, PDS in gapped vdW materials provide [18] Y. Cao et al., Nature 556, 80 (2018). a venue to realize atbands without the need for sensi- [19] Y. Cao et al., Nature 556, 43 (2018). tive twist angle alignment or stacking arrangement. In- [20] G. Chen et al., Nat. Phys. 15, 237 (2019). stead, extreme bandstructure engineering arise directly [21] J. S. Alden, et al. Proc. Natl. Acad. Sci. U.S.A 110, 11256 from gate-controlled superlattice potentials yielding at- (2013). [22] C. Forsythe, X. D. Zhou, K. Watanabe, T. Taniguchi, A. bands when gate voltage is applied; when gate bias van- Pasupathy, P. Moon, M. Koshino, P. Kim, C. R. Dean, ishes, the PDS system remains as an ordinary TMD sys- Nat. Nano. 13, 566 (2018). tem. As a result, PDS atbands a ord considerable elec- [23] L. Jauregui, K. Pistunova, A. Y. Joe, D. Rhodes, B. Kim, trical control over the character of their electronic exci- J. Hone, P. Kim \Superconductivity in twisted transition tations (e.g., miniband-width, miniband-gap, interaction metal dichalcogenide homobilayers" APS March Meeting energy), as well as an on/o switch for atbands. While Talk K15.00003 (2019) we have focussed on ferromagnetic ordering, highly lo- [24] See Supplementary Information for a detailed discus- sion on the superlattice band structure, Poisson's equa- calized electrons can also exhibit other types of magnetic tion and electric potential generated using PDS scheme, ordering including antiferromagnetic order as well spin particle-hole asymmetry, Berry curvature in the recon- liquids (for a recent discussion see e.g. Ref [16]); hav- structed minibands, PDS induced minigaps, other types ing two extremely at bands separated by a small gap of superlattices and the PDS scheme in other TMD ma- may provide a possible venue for unusual excitonic cor- terials, estimate for disorder broadening, mean free path, relations [32]. Given the large variety of available su- extent of the electronic wavefunction, and derivation of the e ective model for atband ferromagnetism. perlattice structures that can be patterned using PDS, [25] M. Yankowitz, et al. Nat. Phys. 8, 382?386 (2012). the characteristic length scales for PDS (several tens of [26] D. Xiao, G.-B. Liu, W. X. Feng, X. D. Xu, Wang Yao, nanometers), as well as the exposed nature of the sur- Phys. Rev. Lett. 108, 196802 (2012). face states, such heterostructures may provide a one-stop [27] A. H. MacDonald, H. C. A. Oji, K. L. Liu, Phys. Rev. B platform to realize and spatially probe exotic ordering. 34, 2681 (1986). Acknowledgements | We are grateful for useful con- [28] K. Nomura, A. H. MacDonald, Phys. Rev. Lett. 96, versations with Valla Fatemi, Ata c Imamo glu, Brian 256602 (2006). Skinner, Javier Sanchez-Yamigishi, and Noah Yuan. This [29] S. Adam, E. H. Hwang, V. M. Galitski, S. Das Sarma Proc. Natl. Acad. Sci. U.S.A 104, 18392 (2007). work was supported by the Singapore National Re- [30] B. Skinner, B. I. Shklovskii Phys. Rev. B 87, 075454 search Foundation (NRF) under NRF fellowship award (2013). NRF-NRFF2016-05, a Nanyang Technological University [31] B. Skinner, Phys. Rev. B 90, 060202 (2014). start-up grant (NTU-SUG), and Singapore MOE Aca- [32] D. J erome, T. M. Rice, W. Kohn. Phys. Rev. 158, 462 demic Research Fund Tier 3 Grant MOE2018-T3-1-002. (1967). Electronic address: justinsong@ntu.edu.sg [1] A. K. Geim, I. V. Grigorieva, Nature 499, 419 (2013). [2] K. S. Novoselov, A. Mishchenko, A. Carvalho, A. H. Cas- tro Neto, Science 353, aac9439 (2016). [3] J. C. W. Song, N. M. Gabor, Nat. Nano. 13, 986 (2018). [4] B. Hunt et al., Science 340, 1427 (2013). [5] H. Yu, G.-B. Liu, J. Tang, X. D. Xu, W. Yao, Sci. Adv. 3, e1701696 (2017). [6] Jin, C. et al., Nature 567, 76 (2019). [7] Tran, K. et al. Nature 567, 71 (2019). [8] Seyler, K. L. et al., Nature 567, 66 (2019). [9] Alexeev, E. M. et al., Nature 567, 81 (2019). [10] L. A. Ponomarenko,et al. Nature 497, 594 (2013). 7 top gate Supplementary Information for \Gate-tunable at bands in van der Waals patterned dielectric TMD monolayer superlattices" Superlattice band structure For a 2D trianglular superlattice, we model the Hamil- tonian as H = H + V (r), with bottom gate H = v~( k +  k ) +  ; (S-1) 0 x x y y z FIG. S-1: The inhomogeneous dielectric environment simu- and lated in Fig. 1b. Gray regions are SiO substrate and the white region denotes an air hole. A TMD monolayer lies in 4 j j between the substrate and the top gate. Parameters used V (r) = V  2cos(G  r); G = cos ; sin : s 0 j j 3a 3 3 in the simulation: the distance (exaggerated here for illustra- j=1 tion) between the top/bottom gate and the top/bottom of the (S-2) SiO substrate is 3 nm; the thickness of the SiO substrate in 2 2 the z-direction is 40 nm; the radius of the air hole within the Here G are the triangular superlattice reciprocal vectors SiO substrate is 5 nm;   = 5 eV. 2 b t oriented 60 relative to each other. The band structure is obtained by diagonalizing the Bloch Hamiltonian, whose (pq; mn)-th 2 2 block reads expressing (r) = (; z), (r) = (; z). Using rotational symmetry, we eliminate the azimuthal angle to obtain the H (k) = hk jH jk i pq 0 mn pq;mn 2D equation: + V hk j2 cos(G  r) jk i; s pq i 0 mn 1 @ @(; z) @ @(; z) (; z) + (; z) = 0: j=1;2;3 @ @ @z @z (S-3) (S-6) To proceed we specify (; z) in the whole region ( = 4 where k = k + pG + qG with k within the rst Bril- pq 0 1 for gray regions and  = 1 for white regions), and the louin zone, p; q 2 Z, and jk i is the plane wave basis. pq boundary conditions at the bottom and top gates for The same applies to k and jk i For any (pq; mn) mn mn (; z) (see Fig. S-1, electric potentials for top and bot- block, it is coupled to the other 6 blocks: (pq  1; mn), tom gates are set to  and  , respectively). We can t b (pq; mn 1), (pq 1; mn + 1) and (pq + 1; mn 1). numerically solve the (; z) in the whole region. In so Numerically, we have to truncate the in nitely large doing, we used Mathematica NDSolve to solve the di er- matrix H (k) into a nite sized H (k) in which p; q 2 2 ential Eq. S-6. [N; N ], and H (k) has a nite dimension 2(2N + 1) . In our simulation, we obtain V  8:3 meV when we set Throughout the paper, we nd N = 4 (162 mini-bands in = 5 eV(see Fig. 1b). This correspond to an electric b t total) is sucient to ensure the dispersion for the mini- eld strength E  0:11 V nm which is smaller than bands close to the intrinsic gap converge. the breakdown electric eld strength ( 0:5 V nm ) for hBN. Electric potential from generalized Poisson equation Particle-hole asymmetry For a non-uniform dielectric environment, we have the generalized Poisson equation The bare TMD Hamiltonian H = v~( k + k ) + 0 x x y y 2 2 2 2 has eigen energy " =  v ~ jkj +  and eigen r D(r) = r [(r)r(r)] = 4 (r): (S-4) free z states With the knowledge of (r), and the values of (r) at the i =2 i =2 k k cos( =2)e sin( =2)e k k top and bottom gates, we can solve (r) using Eq. (S-4). u = ; u = ; k i =2 k i =2 k k sin( =2)e cos( =2)e k k We now investigate the electric potential in the con g- (S-7) uration illustrated in the Fig. S-1. In this con guration, 2 2 2 2 where cos  = = v ~ jkj +  and tan  = k =k . k k y x there is a single hole in the SiO dielectric substrate, The particle-hole operation P = i K (K is the complex which is in between a top gate and bottom gate. Since conjugation) transforms Pu = u , i.e., each eigenstate we have rotational symmetry in the x-y plane, we choose k k + + + u at energy " has a copyPu at energy " . Meanwhile cylindrical coordinates (; '; z): k k k k it also satis es PH P = H . Therefore H has a 0 0 0 x =  cos '; y =  sin '; z = z; (S-5) particle-hole symmetry. 8 (a) (b) (a) (b) -799.234 c= -1 c=0 -800 c= +1 -800 -799.236 0.7 -805 -805 -799.238 -0.3 -10 FIG. S-2: (a) Berry curvature distribution for the top most FIG. S-3: Similar to Fig. S-2, but with V = 1:5 meV with valence miniband. Dashed hexagons denote the mini Brillouin Berry curvature distribution shown for the blue band. Here zone. (b) Dispersion for the mini-bands close to the intrinsic Berry curvature distribution is sharply peaked close to the ~ ~ band gap, and the valley Chern number for the top most corners of the MBZ at K; K points [clipped white spots de- 5 2 valence miniband. Parameters are the same with those in note j (k)j > 10 A ]. Berry curvature values close to the ~ ~ Fig. 2a in the main text, except that V = 1:5 meV. s K; K points are the same sign indicating the pseudospin ~ ~ winding of the same chirality. Berry curvature about K; K points both have a negative sign. The panel on the right hand side in (b) shows the tiny gap between two nearly degenerate However, for the superlattice potential V (r) we have 1 minibands. PV (r)P = V (r). This breaks the particle-hole sym- metry for the overall Hamiltonian, i.e., PHP 6= H. Instead, the hamiltonian H possesses a di erent type V < 0. Similar to the case discussed above, the pos- of symmetry. To see this, we de ne a composite particle- ~ itive sign of the Berry curvature from the bare gapped hole and voltage ip operation P = PM . Here P is the Dirac cone in the TMD is again re ected in the center particle-hole operation (as de ned above), and the mirror of the MBZ. However, Bragg scattering at MBZ corners operation M ips V (r) ! V (r) by interchanging the in the top valence miniband for V < 0 is signi cant and top and bottom gate potential keeping H intact. Us- dramatically changes the winding of the pseudo-spinor ~ ~ ing this composite operation we recover PHP = H. wavefunction; indeed the Berry curvature distribution is More explicitly, when V ! V then " ! " are s s k k concentrated at the MBZ corners displaying large and interchanged. negative value (see small white dots). Note that the ~ ~ white dots appear at both K and K points; their com- mon sign indicates that chirality of the pseudospinors at Mini-band Berry curvature distribution, valley these points is the same. As a result, within this (blue) Chern number, and miniband gapped-Dirac-cone band (Fig. S-3b), we obtain a non-vanishing net Berry ux of C = 1 in a single (original) valley (see Fig. S-3). The PDS superlattice potential can also modify the The large Berry ux in the the top valence blue mini- winding of the pseudo-spins as well as the miniband band (for V < 0) arises from a very small minigap with quantum geometry. To display this, we plot the Berry the adjacent miniband just below it (green). Indeed, curvature distribution in the top most valence mini-band when zoomed-in a small minigap between the two top va- for V > 0 (see Fig. S-2a) and V < 0 (see Fig. S-3a). In s s lence minibands of order   eV appears at the K point what follows, we concentrate on the Berry curvature for (see Fig. S-3b); a similar gap also appears at K point. We minibands from the (original) K valley with  = 1 (see note parenthetically that this small gap is larger than the Eq. 1). We rst note that the Berry curvature around a numerical mesh resolution from our numerical diagonal- single bare gapped Dirac cone in the valence band of the ization routine, allowing us to extract the minigap value TMD (when V = 0) has a positive sign for K valley, shown. Indeed, the lower miniband (green) also displays and monotonically decreases in magnitude away from the the same behavior with a sharp distribution of Berry cur- Dirac point. ~ ~ vature peaked at the K; K points; these have opposite We now focus on the topmost at valence miniband sign to the blue miniband above it with a concomitant when V > 0. Now the (positive) sign of the Berry cur- C = 1 in a single valley. vature from the bare gapped Dirac cone in the TMD is re- ected in the center of the MBZ (see Fig. S-2a). However, While the nite valley Chern numbers indicate a non- PDS induced Bragg scattering close to the MBZ bound- trivial topology, the minigap is so small as to make it aries produces a di erent winding of the pseudospin and very dicult to meaningfully resolve any real experimen- a di erent sign of Berry curvature close to MBZ bound- tal signatures of the nite valley Chern number. Indeed, aries. This reconstructs the Berry curvature distribution. for practical purposes, when energy resolution is larger As a result, the net Berry ux through the MBZ for the than the small minigap   eV, the minigap cannot be top most valence miniband is C = 0. resolved. As a result, the blue and green minibands close ~ ~ Next, we turn to the topmost at miniband when to K; K points resemble Dirac cones with the same chi- energy, energy, 9 rality. This closely mirrors the behavior of the at mini- Figs. S-5 and S-6, respectively. These display that at bands in twisted bilayer graphene close to magic angle, minibands are also obtained for these superlattices under- ~ ~ where electrons close to K; K in the moir e Brilliouin scoring the generic nature of the PDS atband scheme. zone are severely slowed, and possess the same chirality. The main di erence between the green and blue bands in Other types of TMDs PDS scheme using TMDs is a reduced degeneracy (only two for spin). Here we discuss the PDS scheme (using triangular lat- tice as illustration) as applied to other types of TMDs, PDS induced minigaps speci cally, MoS , MoSe , and WS . The results are dis- 2 2 2 played in Fig. S-7. Here we only show the valence band widths, since the miniband structure possess a symmetry In this section we discuss how the minigaps between when V ! V and " ! " are interchanged (see the s s k k the bundles of atbands evolve as a function of applied above section, \Particle-hole asymmetry"). As expected, superlattice potential V . Here we show the rst minigap at minibands can be readily achieved. between minibands in the valence band (i.e., the gap be- tween blue and green bands when V > 0, and the gap between green and purple band when V < 0, see Fig. 2 Estimate for disorder broadening in the main text) as a function of V . This indicates that the minigaps increase as larger superlattice potential is In this section, we estimate the disorder broadening. applied. We begin by observing that charged impurity scatter- ing is responsible for most of the observed bulk di u- sive transport behaviors at low temperature (the con- tribution of phonons is frozen out). They reside either inside the substrate, or can be desposited near the in- terface between the substrate and the 2D system during the processing/sample preparation. Here we rst extract the impurity concentration n from transport measure- imp -5 ment results, and then use n to estimate the disorder imp -10 -5 0 5 10 broadening. We will concentrate on modeling bare WSe , where FIG. S-4: PDS induced minigaps between blue and green recent low temperature transport measurements have re- 4 2 1 1 minibands in the valence band (for V > 0) as well as between s vealed large mobilities  = 3 10 cm V s at T = 12 2 green and purple minibands (for V < 0). The gray shaded 1:4 K for carrier density n = 2 10 cm [23]. To pro- area indicates the region where the minibands are not well ceed, we rst note that close to the band edge of a TMD, separated. 2 2 2 2 the dispersion can be written as " = v ~ q + with v~ = 3:94 eV A and  = 0:8 eV [26]. For small q, we adopt an e ective mass approximation "   + 2 2 2 v ~ q =2. This produces Other types of superlattices m = =v = 0:39 m ; (S-10) Here we show that square and hexagonal superlattices which is consistent with values obtained in the literature can also lead to at bands. For the square superlattice, [S1,S2]. Also, within the e ective mass approximation, we follow Ref. [22] and use the superlattice potential its Fermi wave vector is q = n = 0:25 nm ; (S-11) V (r) = 2V  [cos(b r)+cos(b r)+cos(b r) cos(b r)]; f squ s 0 1 2 1 2 (S-8) 12 2 where we used n = 2 10 cm [23]. where b = (2=a)(1; 0) and b = (2=a)(0; 1) with a the 1 2 In order to extract an estimate for the impurity density superlattice spacing. For hexagonal superlattice, we use from the mobility, we write the conductivity as the superlattice potential = en = e n=m ; (S-12) V (r) = V  cos(G  r) + cos(G  r ) ; (S-9) hex s 0 i i where n is the carrier density (regardless of spin or val- j=1;2;3 ley),  is the mobility, and  the scattering time. At low 0 temperature, the impurity scattering time reads as where r = r + (a; 0), and G are the same reciprocal vectors de ned in Eq. S-2 which also apply to hexago- = n dzhw( )(1 cos  )i ; (S-13) imp i i z nal lattices. The reconstructed minibands are shown in first valence gap, 10 (a) (b) -800 -835 -820 -845 -840 20 40 -10 -5 0 5 10 FIG. S-5: Dispersion and DOS at V = 0:5 meV (a) and bandwidth (b) for a square superlattice with lattice spacing of 20 nm. Other parameters are the same with those in Fig. 2a in the main text. (a) (b) -810 -830 -820 -835 -830 20 40 -4 -2 0 2 4 FIG. S-6: Dispersion and DOS at V = 0:25 meV (a) and bandwidth (b) for a hexagonal superlattice with inter-hole spacing of 20 nm. Other parameters are the same with those in Fig. 2a in the main text. energy, energy, energy, energy, 11 its magnitude equals to Fermi wave vector q , and q is f s the e ective screening wave vector. Eqs. S-14 above enables to estimate n from a known imp mobility. Using this impurity concentration, we can es- -800 timate the disorder potential width. To do so, we note that randomly distributed positive and negative charged -820 impurities create uctuations of the disorder potential. Taking uncorrelated impurity positions, the amplitude of the uctuations [29{31] is given by -10 -5 0 5 10 Z Z dq 2 2 2 = h[V (r)] i = n dz V (q; z): (S-17) imp (2) A numerical estimate for n and both require the imp -700 value for the screening wave vector q = r . There are a number of processes that can contribute to screen- ing. For example, at long-wavelengths, Thomas-Fermi -720 screening [S3] yields a screening wavevector as q = 2 2 2 2 1 (2e =) = (4e =)(=v ~ ) = 7:4 nm , where  is 0 0 the density of state at the Fermi energy, and  = 4 for -10 -5 0 5 10 SiO . Screening can also arise from a proximal metal- lic top gate displaced several nanometers away from the vdW layer by 2d [16], in this case q  d (assum- ing a perfect metallic gate). At low densities, non-linear screening can even take e ect. The exact value of the dis- -860 order potential depends on details of how the Coulomb potential is screened. Instead of identifying the source screening which is beyond the scope of this work, here -880 we take an e ective screening wavevector approach. To obtain a rough estimate of the range of values can take on, we plot for a wide range of e ective q values -900 s (Fig. S-8). We nd that for ultra-clean WSe samples -10 -5 0 5 10 several meV (see Fig. S-8). We note the above yields an estimate of disorder broad- FIG. S-7: Valence miniband widths for triangular PDS for ening for bare WSe . The precise disorder broadening MoS (v~ = 3:51 eVA,  = 0:88 eV), MoSe (v~ = 3:11 eVA, 2 2 when the PDS scheme is applied will depend on the de- = 0:73 eV), and WS (v~ = 4:38 eVA,  = 0:89 eV). The tails of screening, as well as the actual impurity concen- lattice spacing a = 20 nm. tration of the PDS sample. The detailed form as well as the values for these are, at present, unavailable. As a re- sult, the disorder broadening obtained above represents where we assumed an uncorrelated charged impurities an estimate for the possible disorder broadening in WSe n (units cm ) in the substrate, and w is the scatter- imp PDS atbands. ing rate, see below. Eqs. S-12 and S-13 lead to h i n = (e=m ) dzhw( )(1 cos  )i : (S-14) imp i i z Estimate for mean free path ` Within the e ective mass approximation, Here we estimate the mean free path ` for the sample 12 2 1 dq which has n = 2  10 cm , q = n  0:25 nm 2 f hw( )(1cos  )i = jV (k; z)j (1cos )(" " ); i i z q q i 4 2 1 1 4~ and  = 3 10 cm V s [23]: (S-15) ~q m 2 ` = v  = = 0:49 m: (S-18) 2e m e V (q; z) = exp(qz); (S-16) (q + q ) Since `  a, the electronic quasiparticles are able to where k = jq q j,  =   , tan  = q =q , q is experience Bragg scattering induced by the PDS super- i q q q y x i the initial electron wave vector at the Fermi surface with lattice potentials in this sample. energy, energy, energy, 12 0.1 4 8 0 5 10 FIG. S-8: Estimated disorder broadening (red) and im- FIG. S-9: Estimate of wavefunction spatial extent a (as a purity concentration n (blue) for a wide range of e ec- function of applied V ) when con ned in a trough of a super- imp s 12 2 tive q , for a WSe sample having n = 2  10 cm and s 2 lattice. This was obtained using a variational approach and 4 2 1 1 = 3  10 cm V s at T = 1:4 K recently reported for with an e ective mass approximation. Blue, yellow, green, WSe from the Philip Kim group [23]. 2 and red lines correspond to superlattice wavelengths of a = 20, 30, 40, 50 nm, respectively. For very small V , a rapidly s w increases and gets cut o ; below this critical V the electrons are not well con ned in the superlattice troughs. Variational estimate for wavefunction extent a Here we use a variational approach to estimate a . As a result, only a large enough V  V is able to trap We rst adopt an e ective mass approximation for an an electron at its local minimum, and to create a nearly electron in conduction band: at band. Thankfully for the TMDs we consider in this 2 2 ~ r work, this critical V is small due to their large e ective H = + V (r); (S-19) 2m masses (see Fig. S-9). To illustrate the typical extent of the trapped wavefunction we plot the behavior of a where V (r) = V 2cos(G  r) is the triangular s j j=1;2;3 versus V in Fig. S-9. Here we have used m = 0:39 m s e superlattice potential used in the main text. The minus with a = 20; 30; 40; 50 nm corresponding to blue, yellow, sign in the potential is to create a local potential mini- green and red lines, respectively. As displayed, even fairly mum for an electron at r = 0. modest values of V yield fairly localized a . s W To proceed in the variational approach, we use a Gaus- sian trial wave function 1 r (r) = p exp ; dr[ (r)] = 1; 4a Flat band ferromagnetism 2a (S-20) to describe an electron trapped at r = 0. The energy for After discussing how nearly at bands can be created this trial wave function is in TMDs using the PDS scheme, we now demonstrate a 2 possible spontaneous symmetry breaking that arises due 2 2 2 ~ a 8 a hHi = 6V exp ; (S-21) the quenching of kinetic energy. 4m 9 a We start from the electronic Hamiltonian and the equation for its extremum reads as X H =  c c + H ; (S-24) k int 2 2 2 2 2 @hHi 32 a 8 a ~ a W W W = V exp = 0: 2 2 @a 3 a 9 a 2m where  describes the dispersion, and (S-22) Eq. S-22 has a nite and positive solution for a that W X y y H = c c c c ; (S-25) decreases with increasing V (e is Euler's number): 0 0 0 0 int k  k s k+q k q 2V 0 0 k kq; 2 2 2 2 3a e  ~ a 1=2 a = W e V =V ; V = : W 0  s is the electron-electron interaction. Here we only retain 2 54 2m (S-23) the on-site repulsive interaction U and neglect inter-site Here W (z) is the the Lambert W function, and W (z) 0 0 interactions, which are strongly suppressed in nearly at only has a real, negative solution when e  z  0. bands [16]. Generally, spin polarization in the z-direction Speci cally, W (e ) = 1 and W (0) = 0. Therefore, 0 0 is favored due to repulsive interactions [S4] thus we will V  V is required for consistency of our initial bound s  focus on the possibility of the spin polarization in the state trial wave function/ansatz, i.e., a nite and positive z-direction, and use the mean eld parameterization as a . When V < V , a becomes a complex number, in- W s  W dicating the bound state trial wave function/ansatz fails. hc c i =  0 0n  : (S-26) 0 0 kk  k k k  13 Using this mean eld ansatz, we obtain the mean eld From this we see the normal spin-degenerate state is un- Hamiltonian stable when the second term is negative, i.e., Z ( )U > H = ( + Un  ) c c UV n  n  ; (S-27) MF k   " # where   is opposite to , and n  = V hc c i is MF k k k the spin  density with respect to the mean eld ground Electronic address: justinsong@ntu.edu.sg state. For a nearly at band, we can neglect the k- S1. A. Korm anyos, G. Burkard, M. Gmitra, J. Fabian, V. dependence of    , and its energy density is k 0 Z olyomi, N. D. Drummond, and V. Fa ko, 2D Mater. 2, 022001 (2015). hH i = ( + Un  ) n  + ( + Un  ) n  U n  n  ; MF 0 # " 0 " # " # S2. B. Fallahazad, H. C. P. Movva, K. Kim, S. Larentis, T. (S-28) Taniguchi, K. Watanabe, S. K. Banerjee, E. Tutuc, Phys. where hH i = hH i=V . In anticipation of the (spin- MF MF Rev. Lett. 116, 086601 (2016). split) broken symmetry state, we can re-write hH i to MF S3. S. Das Sarma, S. Adam, E. H. Hwang, E. Rossi, Rev. emphasize that the exchange interaction favors aligned Mod. Phys. 83, 407 (2011). spins: S4. Y.-H. Zhang, D. Mao, Y. Cao, P. Jarillo-Herrero, T. Senthil, Phys. Rev. B 99, 075127 (2019). hH i = " n  n  ; (S-29) MF 0 where " =  + U (n  + n  ). 0 0 " # When there is a nite bandwidth [for e.g., arising from disorder broadening (see main text and above)], " no longer resides at a single at energy, but instead uctuates. To capture this, the [single-particle] energy density can be described by a broadened spectral func- tion. Here we have used a simple spectral function 2 2 Z (") = A exp[(" " ) =2 ] to model this level broad- ening, where A = [ 2 ] . Other choices of spectral function do not modify the qualitative conclusions we de- scribe below and in the main text. As a result, the energy density can be described in close analogy with that used for the energy density in Landau levels [28]. Using the spectral function above, we nd the energy density (including disorder broadening) can be modeled as hH i = d" "Z (") n  ; (S-30) MF where  is the Fermi level for spinor-component  that satis es n  = d" Z ("): (S-31) As we can see, Eq. S-30 mirrors Eq. S-29 except that the single-particle energy density is now broadened by the spectral function (due to disorder broadening). In the normal state the two degenerate at bands are equally occupied. To see whether it is a stable state, we perturb n  = n  + n  with n  = (n  + n  )=2. By using 0  0 " # n  = 0 and expanding  to the second order, the mean eld energy density reads as hH i = d" "Z (") n MF n  1 + U + : (S-32) 2 Z ( ) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Condensed Matter arXiv (Cornell University)

Gate-tunable flat bands in van der Waals patterned dielectric superlattices

Condensed Matter , Volume 2019 (1904) – Apr 16, 2019

Loading next page...
 
/lp/arxiv-cornell-university/gate-tunable-flat-bands-in-van-der-waals-patterned-dielectric-V3Tomm4dxM
ISSN
2053-1583
eISSN
ARCH-3331
DOI
10.1088/2053-1583/ab59a8
Publisher site
See Article on Publisher Site

Abstract

potential, gapped vdW dielectric Gate-tunable at bands in van der Waals patterned dielectric superlattices 1 2 1;2 Li-kun Shi , Jing Ma , and Justin C.W. Song Institute of High Performance Computing, Agency for Science, Technology, & Research, Singapore 138632 and Division of Physics and Applied Physics, Nanyang Technological University, Singapore 637371 Superlattice engineering provides the means to reshape the fabric felt by quasiparticles moving in a material. Here we argue that bandstructure engineering with superlattices can be pushed to the extreme limit by stacking gapped van der Waals (vdW) materials on patterned dielectric substrates. Speci cally, we nd that high quality vdW patterned dielectric superlattices (PDS) realize a series of robust at bands that can be directly switched on and o by gate voltage in situ. In contrast to existing superlattice platforms, these at bands are realized without the need for ne tuning. Instead, the bands become at as the gate voltage increases in magnitude. The characteristics of PDS atbands are highly tunable: the type of atband (single non-degenerate or dirac-cone-like), localization length, and interaction energy are sensitive to the applied gate voltage. As a result, electron-electron interactions in the PDS atbands can become stronger than both the bandwidth and disorder broadening, providing a setting for correlated behavior such as atband ferromagnetism. We expect PDS atbands can be experimentally realized in a range of readily available gapped vdW materials such as monolayer transition metal dichalcogenides, e.g. WSe . (a) (b) van der Waals (vdW) heterostructures have become -5 5 0.3 a powerful platform to tailor the electronic properties of materials [1{3]. One case in point is moir e superlat- tices, formed when two vdW materials are stacked and 0.2 twisted. In such moir e materials, electrons and other quasiparticles experience slowly varying (emergent) ef- fective periodic potentials. Even when the potentials are relatively weak (as compared with the kinetic en- ergy of electrons in each layer), a wide variety of phe- nomena can be realized that include for e.g., emergent -10 electronic bandgaps [4], moir e excitons [5{9], Hofstadter spectra [4, 10, 11], as well as topological bands [12, 13], to name a few. However, when top and bottom layers cou- -5 5 ple strongly, extreme bandstructure reconstruction takes e ect, allowing nearly at electronic bands to form at FIG. 1: (a) Schematic of a triangular patterned dielectric magic [14, 15] or low-twist angles [16, 17]. These provide superlattice (gray) and the spatial potential pro le (purple) a vdW venue to realize correlated behavior, with intense sustained for the target gapped van der Waals material (red). interest sparked by reports of correlated insulating be- Here the top gate (yellow) and back gate (blue) sustain a havior and superconductivity in moir e materials under potential drop in the out-of-plane direction. The spacing be- such conditions [18{20]. tween two neighboring holes is a. (b) Electrostatic potential for a single hole inside the dielectric substrate (gray) in be- However, achieving good twist angle control over moir e tween top and bottom gates, see SI for full details. The top panel shows the electric potential across the red line (plane of supelattices can often be experimentally dicult; this the vdW material) displayed in the bottom panel. Thin white can be further complicated by lattice reconstruction that lines indicate equi-potential contours. Here we have used po- arise at low twist angle [21]. Good twist/registration tential di erence between the two (top and bottom) gates of control becomes critical given that extreme bandstruc- 5 eV for illustration giving a superlattice potential amplitude ture engineering typically occurs only at speci c \magic" (V see Eq. 1) of about 10 meV; we note these correspond to twist angles or in a small range of low-twist angles [14{ electric elds smaller than the dielectric breakdown voltage 17]. Recently, an inverted electrostatic strategy | pat- of hBN. terned dielectric superlattices (PDS) | wherein a dielec- tric material is patterned into a superlattice and placed (where the patterned dielectric did not signi cantly de- on top of a gate electrode (Fig. 1a) has been experimen- grade the device performance [22]), as well as high gate tally demonstrated in graphene [22], producing high qual- electrode tunability that could turn electrically on and ity superlattices. In these, dielectric contrast between o the superlattice potential. patterned hole and dielectric substrate material enables the gate electrode to sustain a spatially modulated su- Here we argue that PDS can be pushed into the strong perlattice potential (Fig. 1b). Perhaps most remarkable coupling regime, where the superlattice potential (V ) is that the PDS devices maintained an ultrahigh mobility can exceed the kinetic energy of the electrons in a gapped top gate bottom gate arXiv:1904.07877v1 [cond-mat.mes-hall] 16 Apr 2019 2 vdW material. As we describe below, in this regime, not a ect the physics we discuss below. Further, we note at electronic minibands can be achieved and are highly that the particularly large Ising spin splitting ( 100s gate-tunable. Strikingly, these atbands do not require of meV) in the valence bands of many TMD materials, ne tuning of twist angle. Indeed, the bands become e ectively lock the valley and spin in the valence bands at as gate voltage is switched on, achieving small band- making the model in Eq. (1) a good descriptor of the widths even for modest superlattice potentials (Fig. 2). low-energy mini-bandstructure. This stands in contrast to strategies using moir e super- As shown in Fig. 1b, dielectric inhomogeneity in the lattices that only feature at bands at magic or low twist substrate material can enable electric control of a spa- angles [14{17]. tially inhomgeneous potential with variations on the or- PDS atbands can be achieved in gapped vdW materi- der of the hole width (e.g., several tens of nanometers) als such as the transition metal dichalcogenides. The key [22]. Here gray indicates a dielectric (shown SiO with ingredients are a large bandgap as well as a large e ective  = 4) and empty indicates air ( = 1). As shown, when mass. The former ensures that conduction and valence a gate potential is applied, the dielectric contrast yields bands do not cross when V is applied, and the latter a spatially modulated potential even in the vdW mate- gives a small initial kinetic energy of the electrons. These rial layer (red) with potential amplitudes V that can be enable the superlattice potential to con ne the electrons switched on and o electrically [22]. Fig. 1b was plotted e ectively and form at minibands. A particularly good using a numerical solution for a generalized Poisson equa- candidate for PDS atbands is WSe , wherein high mo- tion in a spatially inhomogeneous dielectric environment 4 2 bility (& 310 cm =Vs) samples have been isolated [23]; (see Supplementary Information, SI). these have mean free paths of several hundred nanome- To illustrate the ( at) band structure engineering in ters [24]. For WSe , we anticipate at electronic mini- the PDS of vdW materials, we rst use a triangular su- bands that are well separated can be achieved for mod- perlattice (see other lattices in SI) so that est V and superlattice periods. Interestingly, PDS yields f (r) = 2cos(G  r +  ); (2) i i bunches of at bands that proliferate throughout the pa- i=1;2;3 rameter space. For example, for V > 0, we nd that in addition to the top most valence miniband which is a sin- where G are the triangular superlattice wave vector ori- gle non-degenerate atband (per valley), the other lower ented 60 relative to each other, jG j = 2=a, and  is i i (valence) minibands also bunch up into atband bundles. a relative phase. For simplicity, we will set  = 0 in Similar atband bundles can be found for V < 0, as well s the following. It does not qualitatively a ect the physics as in the conduction band. we discuss below. As expected, the superlattice folds the Perhaps the most exciting aspect of PDS atbands is original TMD bands into a series of superlattice mini- the possibility of direct gate access to correlated behav- bands (band folding is illustrated for V = 0 as the light ior. As an illustration, we consider ferromagnetism in a dashed lines in Fig. 2a). Here the bandstructure is plot- PDS atband; this is in close analogy with quantum-hall ted in a superlattice de ned mini Brilluoin zone (MBZ), ferromagnetism. In such a case, we nd a ferromagnetic where the tilde symbol indicates the MBZ. When V is state, where the spins in either valley K=K are favored, switched on, Bragg scattering mixes the bands and pro- can be achieved for readily available samples and super- duces a mini-band structure (colored lines and gray lines lattice gate voltages. with V = 0:5 meV) in both the conduction band (pos- Gate tunable at minibands { We begin with a discus- itive energies) and the valence band (negative energies) sion of the PDS scheme using monolayer transition metal as shown in Fig. 2a. We have colored only the rst three dichalogenides (TMD). The PDS that we consider here conduction and valence minibands so as to focus our dis- (and that can be experimentally fabricated) are of fairly cussion on them; in solving for the mini-bandstructure long-wavelength a & 20 nm. The resulting superlattice a set of 162 mini-bands are employed to ensure conver- bandstructure can be captured by an e ective kp model gence of the lower bands (we only show 12 minibands) expanded close to the K and K points as that we focus on in the main text. Physically, the inclu- sion of many minibands in our numerical calculation is H = v~( k +  k ) +  + V  f (r); (1) x x y y z s 0 to to capture the physics of the extreme bandstructure where v is the velocity,  are the Pauli matrices, 2 reconstruction regime in which the superlattice strength x;y;z is the bandgap of the TMD,  = 1 for K and K valleys, V is larger than that of a kinetic energy of the electrons. V is the amplitude of the superlattice potential applied Here we have used material parameters corresponding to by the PDS scheme in Fig. 1, and f (r) is the spatial WSe with v~ = 3:94 eVA,  = 0:8 eV [26]. pattern of the superlattice potential. Here we have sup- Strikingly, a clear gap between the top most (va- pressed spin indices since the bandstructure physics we lence) miniband (blue, Fig. 2a) and the rest of the mini- discuss does not di erentiate between spin/valley species bands opens up. This is distinct from that expected (i.e. the superlattice potential is a scalar potential); spin from graphene PDS [22] where a  berry phase pre- can be included in a straightforward fashion that does vents backscattering and gap opening; in the case of top valence bandwidth (a) (b) -800 -760 0 2.5 5 -780 -805 -800 20 40 -10 -5 0 5 10 FIG. 2: (a) Superlattice miniband dispersion (left) and the corresponding density of states (right) for vdW PDS with monolayer WSe . Here the electronic structure is described by Eq. (1) with v~ = 3:94 eVA,  = 0:8 eV; the substrate is a triangular patterned dielectric superlattice with a = 20 nm [see Fig. (1a)]. Solid (dashed) lines are for V = 0:5 meV (V = 0) [see Eq. (1)], s s and the inset shows the mini Brilluoin zone de ned by the superlattice. (b) Miniband minima and maxima that indicate its bandwidth are plotted as a function of V for 12 mini-bands close to the intrinsic band gap. Here solid colored lines indicate miniband maxima and minima, whereas shaded region indicate region in between maxima/minima in the minibands. All other parameters (aside from varying V ) are the same with those in (a). The inset shows the top valence bandwidth as a function of  [with v~ and a xed to the values indicated in (a)]. graphene PDS secondary Dirac cones form at the MBZ the other minibands in the valence band (that can reach corners [4, 10, 11, 22, 25]. In contrast, the wavefunction tens of meV); the minigaps increase with applied gate (AB-sublattice pseudospinor) in TMD close to the band potential [24]. edge has weight mostly on a single sublattice (aligned to One unusual feature is that the application of superlat- a pole in the Bloch sphere), allowing maximal backscat- tice PDS also renormalizes the e ective bandgap between tering and mini-gap formation. the conduction and valence band states, with changes in e ective bandgap of up to 60 meV for the largest su- Due to the large  in TMDs, minigap opening squeezes perlattice potential amplitude shown in the gure (see the top most valence miniband creating a con ned en- Fig. 2b). This arises from the signi cant miniband re- ergy window for it to exist with a sharp density of states construction. Indeed locally in real space, the electrons (shaded blue in right panel of Fig. 2a). Indeed, as V feel large variations in superalattice potential that range is increased, the top most (valence) mini-band is fur- from +6V at the peak to 6V in the troughs. This large s s ther attened (Fig. 2b) with very narrow band-widths peak-to-trough di erence enables electrons to be con ned ( 1 meV) achievable with modest applied gate voltage locally to produce a atband structure (Fig. 2a). creating a nearly at band. To see this, we have plot- ted the maximum and minimum energy in each mini- The blue band is not the only at band that occurs in band (solid lines Fig. 2b, this indicates the bandwidth) PDS. For example, at suciently high superlattice po- with the colors used corresponding to the colored 3 con- tential (V > 0), both green and purple minibands in duction and colored 3 valence minibands of Fig. 2a; the the valence band atten out. Unlike the blue band they shaded region between miniband maxima and minima tend to stick together. In fact, the PDS scheme yields are shaded in corresponding colors. We note that the sets of at minibands in the valence band (some of which at top most (valence) miniband is well separated from are nearly degenerate with each other forming bunched other bands with the large   eV to the minibands bundles of bands) with severely compressed bandwidths. in the conduction band, and gate tunable minigaps to Each of these bundles of at bands are well separated energy, energy, 4 10 7 (a) (b) from each other with large electrically tunable minigaps (in Fig. 2b, we show three at miniband bundles in the valence band at V > 0). We note that the PDS scheme, when applied to gapped Dirac materials such as WSe , naturally breaks particle- 3 1 2 FM phase hole symmetry with (positive and negative energy) mini- bands exhibiting contrasting behavior (see Fig. 2a). For 0.1 1 5 10 15 20 1 5 10 15 example, when V > 0 the top (valence) miniband (blue) gets squeezed into a single non-degenerate at miniband and is well separated from the other minibands, while the FIG. 3: (a) Interaction energy U as a function of the extent bottom (conduction) miniband (yellow) adheres closely of electronic wavefunction in the at band, a , and e ective to the next miniband (red); this displays a particle-hole screening radius, r . Dashed contour lines indicate lines of asymmetric behavior. Symmetry in the miniband struc- constant U and are in units of meV. (b) Dimensionless Z ( = " )U , see Eq. (5) indicating propensity for the ferromagnetic ture, however, is restored when both V ! V and 0 s s instability with Z ( = " ) = [ 2 ] and taken at a " ! " are interchanged (see detailed discussion in 0 0 k k xed r = 5 nm for illustration. For Z ( )U > 1 (boundary s 0 SI). denoted by dashed line), the system enters a ferromagnetic This unconventional feature allows the type of at instability. Here lg denotes log . mini-bandstructure to be tuned by gate voltage. When V > 0 the top (valence) miniband (blue) is a single non- degenerate at band (per valley). However, when V < 0 between layers are required (in the case of twisted bilayer this same blue miniband, while attening out, adheres graphene) or low twist angles required in other moir e su- closely to the green band (Fig. 2b); when large enough perlattice strategies. The key to achieving at bands in gate voltage is applied, they form a close pair of at- PDS is the large  as well as low velocities. Indeed, bands that bunch together. This atband bundle has a as shown in the inset of Fig. 2b, when we x v~ and width  1 meV. Even though both green and blue mini- a (same parameters as panel a), the larger the , the bands stay together, nevertheless, we nd that they are smaller the bandwidth at a given superlattice potential separated by a very small energy gap. Indeed, close to 0 V . For WSe , we nd  = 800 meV allowing very small ~ ~ s 2 K; K points, the blue and green miniband structure (for bandwidths to be achieved even for small V applied, e.g., V < 0) bands resemble Dirac cones (with an extremely the bandwidth is eV when V = 5 meV. We note, par- small gap; we estimate the gaps to be of order several enthetically, that when  is small so that V applied is eV, see SI). We note that away from the small mini- on order , the minibands in the conduction and valence gaps, the minibands spectra mimic the attened Dirac band can start to mix, vastly complicating the miniband bands in twisted bilayer graphene [14]; interestingly, the 0 structure and making the conditions for at bands in ~ ~ chirality of electrons at K; K points are the same [24] PDS hard to achieve. mirroring the behavior found in twisted bilayer graphene. We anticipate that the PDS scheme can be applied As a result, PDS enables to achieve multiple types of to other van der Waals materials with large bandgaps. atbands via in situ gate voltage tuning (e.g., from sin- For example, we have computed the miniband structure gle non-degenerate at-miniband for positive V to Dirac- for a range of TMD materials and have found that well cone-like for negative V ). This unusual asymmetric be- separated sets of at minibands generically occur { see havior (for both conduction/valence bands and opposite SI for full band structure. Further, other superlattices signs of V ) arises from the gapped Dirac pseudo-spinor can also be easily implemented and produce qualitatively form of their wavefunctions (such asymmetry does not the same results: we have computed the minibands for arise for a simple massive two-dimensional electron gas). square as well as hexagonal superlattices as well and nd Indeed, the role of the TMD pseudospinor texture is fur- similar well separated sets of at minibands [24]. This ther evidenced by how PDS induced Bragg scattering also versatility with superlattice structure provides the abil- changes the winding of the gapped Dirac pseudo-spinor ity to study atbands and its concomittant interaction wavefunctions. This dramatically reconstructs the Berry e ects in other types of lattices which can have a di er- curvature distribution in each of the minibands (for dis- ent type of symmetry (that may be di erent from the cussion of miniband Berry curvature, see SI). underlying lattice). One of the most attractive feature of PDS at bands, is that they do not occur at ne-tuned gate voltages Flatband ferromagnetism | Perhaps the most striking or superlattice wavelength. The larger the superlattice consequence of PDS at bands is the ability to enhance wavelength, the smaller the applied superlattice poten- correlation e ects. This is because the extremely small tial needed to atten the PDS minibands. This strat- bandwidth of single-particle mini-bandstructure of PDS egy stands in stark contrast to the at bands found in ( 1 meV) quenches the kinetic energy of the electrons. moir e superlattices where speci c \magic" twist angles As a result, other energy scales such as that arising from 5 electron-electron interactions can determine the behavior a ect the qualitative conclusions we discuss below. Here of the electronic system. For clarity, in the following, we A = [ 2 ] is a normalization constant. will focus on the top-most valence miniband for V > 0 For small U , the two spin species are degenerate with which is well separated from the other superlattice mini-  =  =  is the spin chemical potential. However, for " # 0 bands. Further, we note that due to the large Ising split- large enough U this spin symmetry can become broken. ting in TMDs, the spin and valley degree of freedom are To see this, we rst write n = n  n where 2n = s 0 0 locked; at the non-interacting level, this band has only a n + n . Expanding Eq. (4) to O(n ) we obtain the " # two-fold degeneracy with spin up and spin down occur- energy density for the atband as ring (original) valley at K and K . 2 4 E = F + n [1=Z ( ) U ] +O(n ); (5) MF 0 0 To proceed, we rst estimate the strength of electron- 0 2 electron interactions in the superlattice by modeling the where F = 2 d" "Z (") Un is the energy in the average interaction energy of electrons con ned in the symmetry unbroken phase, and the n term describes troughs of the superlattice potential as [27, 28] the energy cost to imbalance the spin species. Cru- cially, an instability in the spin up/down population is 2e 2 2 induced when the coecient of the second term is nega- U = exp(q a ); (3) (q + q ) s tive Z ( )U > 1. The conditions for symmetry breaking depend on a competition between the broadening and the exchange where q = r is the e ective inverse screening length, energy. In our model above, this is parameterized by and a is the extent of the PDS atband electronic wave- r , a , as well as the broadening energy/width . To function con ned in the troughs of the superlattice. Here s W illustrate this, we plot Z ( = " )U in Fig. 3b; this is we have chosen a simple Gaussian ansatz for the extent 0 0 the maximal value that is obtained at half- lling. From of the wavefunction; other models do not qualitatively a ect the results we discuss below. In Fig. 3a, we plot Fig. 3b, we can see the competition clearly: for larger (smaller) , we require more (less) localized electron the strength of U as a function of both the screening states, or larger (smaller) exchange energy to enter the length r as well as a taking  = 4. This yields large s W broken symmetry phase. Taking r = 5 nm as a demon- interaction energies of order several tens of meV for a stration, we nd large Z ( = " )U for a wide swathes of wide range of r and a . For WSe , we estimate e ec- 0 0 s W 2 the -a parameter space (Fig. 3b). We estimate that tive r of a few nanometers arising either from intrinsic a  5 10 nm can be achieved by modest superlat- screening of the electron gas or from proximal gates, see tice potentials in the PDS scheme. At low temperatures, discussion in SI. To estimate the extent of the electronic broadening is typically dominated by disorder [29{31]. wavefunction a , we employ a variational approach on Taking values for high quality WSe [23], we estimate the Gaussian wavefunction ansatz. This yields typical disorder-induced broadening  of order several meV a that decreases for increasing superlattice potential, dis are available in high quality present day samples, see SI. reaching a fairly con ned state of  5 10 nm even for As a result, we expect that the conditions for realizing a modest applied superlattice potential, see SI. ferromagnetic state using PDS atbands can be attained Fig. 3a indicates that U can dominate over the ex- in WSe . Interestingly, this analysis is not con ned to tremely small non-interacting PDS atband bandwidths 2 half- lling  =  . The ferromagnetic instability can (see e.g., Fig. 2). As a result, we anticipate interaction ef- 0 0 occur at a variety of chemical potentials so long as the fects can become signi cant. For example, large exchange criterion is satis ed Z ( )U > 1, see below. energy can drive the spin-degenerate atband system into 0 Discussion | There are numerous probes of the ex- a ferromagnetic state. To describe this, we use a simple treme mini-bandstructure reconstruction we discuss here. mean- eld model for the energy density E of a PDS MF For example, we anticipate a lightly hole doped WSe at miniband with spins indexed s =";# as 2 PDS to exhibit a dramatic change in its low-frequency h  i 2 (THz) optical absorption characterisitcs as V > 0 is E = d" "Z (") n ; (4) MF switched on. While at V = 0 such a sample will exhibit a Drude peak around ! = 0, when V > 0 is switched on, where  is the chemical potential of spin s, and " the Drude peak will diminish as the topmost (valence) is the band energy, and U is the strength of the ex- miniband attens, and exhibit additional sharp THz ab- change energy and can be estimated by Eq. 3 [for full sorption peaks corresponding to transitions between the description, see SI]. Here we have used a simple Gaus- sets of at minibands (in the valence band). 2 2 sian Z (") = A exp[("" ) =2 ] to model a broadened Similarly, we expect dual-gate control to enable con- spectral weight of the atband, and the density of each trol of both superlattice potential as well as the lling of spin species is n = d"Z ("). The broadening, , can the minibands. Such lling control can enable to probe be induced by a number of di erent processes for e.g., both the metallic as well as the insulating states induced via disorder. Other forms of the spectral function do not by the superlattice potential. When pushed into the 6 regime where atbands exhibit ferromagnetism, the spins [11] C. R. Dean et al., Nature 497, 598 (2013). [12] J. C. W. Song, P. Samutpraphoot, L. S. Levitov, Proc. (locked to the valley) split; at half- lling of a atband this Natl. Acad. Sci. U.S.A 112, 10879 (2015). would exhibit a ferromagnetic insulating state. Inter- [13] Q. Tong et al., Nat. Phys. 13, 356 (2017). estingly, ferromagnetism can also occur away from half- [14] R. Bistritzer, A. H. MacDonald, Proc. Natl Acad. Sci. lling so long as Z ( )U > 1. As a result, the ferromag- USA 108, 12233 (2011). netic system can be metallic. Crucially, we note that the [15] D. Santos, J. M. Lopes, N. M. R. Peres, A. H. Castro minibands exhibit non-zero Berry curvature distribution Neto, Phys. Rev. Lett. 99, 256802 (2007). (see details in SI). As a result, when the spins/valleys [16] F. Wu, T. Lovorn, E. Tutuc, A. H. MacDonald, Phys. Rev. Lett. 121, 026402 (2018). are split (and  is away from half- lling) we anticipate [17] F. Wu, T. Lovorn, E. Tutuc, I. Martin, A. H. MacDonald, that a (charge) anomalous Hall e ect can ensue. Phys. Rev. Lett. 122, 086402 (2019). In summary, PDS in gapped vdW materials provide [18] Y. Cao et al., Nature 556, 80 (2018). a venue to realize atbands without the need for sensi- [19] Y. Cao et al., Nature 556, 43 (2018). tive twist angle alignment or stacking arrangement. In- [20] G. Chen et al., Nat. Phys. 15, 237 (2019). stead, extreme bandstructure engineering arise directly [21] J. S. Alden, et al. Proc. Natl. Acad. Sci. U.S.A 110, 11256 from gate-controlled superlattice potentials yielding at- (2013). [22] C. Forsythe, X. D. Zhou, K. Watanabe, T. Taniguchi, A. bands when gate voltage is applied; when gate bias van- Pasupathy, P. Moon, M. Koshino, P. Kim, C. R. Dean, ishes, the PDS system remains as an ordinary TMD sys- Nat. Nano. 13, 566 (2018). tem. As a result, PDS atbands a ord considerable elec- [23] L. Jauregui, K. Pistunova, A. Y. Joe, D. Rhodes, B. Kim, trical control over the character of their electronic exci- J. Hone, P. Kim \Superconductivity in twisted transition tations (e.g., miniband-width, miniband-gap, interaction metal dichalcogenide homobilayers" APS March Meeting energy), as well as an on/o switch for atbands. While Talk K15.00003 (2019) we have focussed on ferromagnetic ordering, highly lo- [24] See Supplementary Information for a detailed discus- sion on the superlattice band structure, Poisson's equa- calized electrons can also exhibit other types of magnetic tion and electric potential generated using PDS scheme, ordering including antiferromagnetic order as well spin particle-hole asymmetry, Berry curvature in the recon- liquids (for a recent discussion see e.g. Ref [16]); hav- structed minibands, PDS induced minigaps, other types ing two extremely at bands separated by a small gap of superlattices and the PDS scheme in other TMD ma- may provide a possible venue for unusual excitonic cor- terials, estimate for disorder broadening, mean free path, relations [32]. Given the large variety of available su- extent of the electronic wavefunction, and derivation of the e ective model for atband ferromagnetism. perlattice structures that can be patterned using PDS, [25] M. Yankowitz, et al. Nat. Phys. 8, 382?386 (2012). the characteristic length scales for PDS (several tens of [26] D. Xiao, G.-B. Liu, W. X. Feng, X. D. Xu, Wang Yao, nanometers), as well as the exposed nature of the sur- Phys. Rev. Lett. 108, 196802 (2012). face states, such heterostructures may provide a one-stop [27] A. H. MacDonald, H. C. A. Oji, K. L. Liu, Phys. Rev. B platform to realize and spatially probe exotic ordering. 34, 2681 (1986). Acknowledgements | We are grateful for useful con- [28] K. Nomura, A. H. MacDonald, Phys. Rev. Lett. 96, versations with Valla Fatemi, Ata c Imamo glu, Brian 256602 (2006). Skinner, Javier Sanchez-Yamigishi, and Noah Yuan. This [29] S. Adam, E. H. Hwang, V. M. Galitski, S. Das Sarma Proc. Natl. Acad. Sci. U.S.A 104, 18392 (2007). work was supported by the Singapore National Re- [30] B. Skinner, B. I. Shklovskii Phys. Rev. B 87, 075454 search Foundation (NRF) under NRF fellowship award (2013). NRF-NRFF2016-05, a Nanyang Technological University [31] B. Skinner, Phys. Rev. B 90, 060202 (2014). start-up grant (NTU-SUG), and Singapore MOE Aca- [32] D. J erome, T. M. Rice, W. Kohn. Phys. Rev. 158, 462 demic Research Fund Tier 3 Grant MOE2018-T3-1-002. (1967). Electronic address: justinsong@ntu.edu.sg [1] A. K. Geim, I. V. Grigorieva, Nature 499, 419 (2013). [2] K. S. Novoselov, A. Mishchenko, A. Carvalho, A. H. Cas- tro Neto, Science 353, aac9439 (2016). [3] J. C. W. Song, N. M. Gabor, Nat. Nano. 13, 986 (2018). [4] B. Hunt et al., Science 340, 1427 (2013). [5] H. Yu, G.-B. Liu, J. Tang, X. D. Xu, W. Yao, Sci. Adv. 3, e1701696 (2017). [6] Jin, C. et al., Nature 567, 76 (2019). [7] Tran, K. et al. Nature 567, 71 (2019). [8] Seyler, K. L. et al., Nature 567, 66 (2019). [9] Alexeev, E. M. et al., Nature 567, 81 (2019). [10] L. A. Ponomarenko,et al. Nature 497, 594 (2013). 7 top gate Supplementary Information for \Gate-tunable at bands in van der Waals patterned dielectric TMD monolayer superlattices" Superlattice band structure For a 2D trianglular superlattice, we model the Hamil- tonian as H = H + V (r), with bottom gate H = v~( k +  k ) +  ; (S-1) 0 x x y y z FIG. S-1: The inhomogeneous dielectric environment simu- and lated in Fig. 1b. Gray regions are SiO substrate and the white region denotes an air hole. A TMD monolayer lies in 4 j j between the substrate and the top gate. Parameters used V (r) = V  2cos(G  r); G = cos ; sin : s 0 j j 3a 3 3 in the simulation: the distance (exaggerated here for illustra- j=1 tion) between the top/bottom gate and the top/bottom of the (S-2) SiO substrate is 3 nm; the thickness of the SiO substrate in 2 2 the z-direction is 40 nm; the radius of the air hole within the Here G are the triangular superlattice reciprocal vectors SiO substrate is 5 nm;   = 5 eV. 2 b t oriented 60 relative to each other. The band structure is obtained by diagonalizing the Bloch Hamiltonian, whose (pq; mn)-th 2 2 block reads expressing (r) = (; z), (r) = (; z). Using rotational symmetry, we eliminate the azimuthal angle to obtain the H (k) = hk jH jk i pq 0 mn pq;mn 2D equation: + V hk j2 cos(G  r) jk i; s pq i 0 mn 1 @ @(; z) @ @(; z) (; z) + (; z) = 0: j=1;2;3 @ @ @z @z (S-3) (S-6) To proceed we specify (; z) in the whole region ( = 4 where k = k + pG + qG with k within the rst Bril- pq 0 1 for gray regions and  = 1 for white regions), and the louin zone, p; q 2 Z, and jk i is the plane wave basis. pq boundary conditions at the bottom and top gates for The same applies to k and jk i For any (pq; mn) mn mn (; z) (see Fig. S-1, electric potentials for top and bot- block, it is coupled to the other 6 blocks: (pq  1; mn), tom gates are set to  and  , respectively). We can t b (pq; mn 1), (pq 1; mn + 1) and (pq + 1; mn 1). numerically solve the (; z) in the whole region. In so Numerically, we have to truncate the in nitely large doing, we used Mathematica NDSolve to solve the di er- matrix H (k) into a nite sized H (k) in which p; q 2 2 ential Eq. S-6. [N; N ], and H (k) has a nite dimension 2(2N + 1) . In our simulation, we obtain V  8:3 meV when we set Throughout the paper, we nd N = 4 (162 mini-bands in = 5 eV(see Fig. 1b). This correspond to an electric b t total) is sucient to ensure the dispersion for the mini- eld strength E  0:11 V nm which is smaller than bands close to the intrinsic gap converge. the breakdown electric eld strength ( 0:5 V nm ) for hBN. Electric potential from generalized Poisson equation Particle-hole asymmetry For a non-uniform dielectric environment, we have the generalized Poisson equation The bare TMD Hamiltonian H = v~( k + k ) + 0 x x y y 2 2 2 2 has eigen energy " =  v ~ jkj +  and eigen r D(r) = r [(r)r(r)] = 4 (r): (S-4) free z states With the knowledge of (r), and the values of (r) at the i =2 i =2 k k cos( =2)e sin( =2)e k k top and bottom gates, we can solve (r) using Eq. (S-4). u = ; u = ; k i =2 k i =2 k k sin( =2)e cos( =2)e k k We now investigate the electric potential in the con g- (S-7) uration illustrated in the Fig. S-1. In this con guration, 2 2 2 2 where cos  = = v ~ jkj +  and tan  = k =k . k k y x there is a single hole in the SiO dielectric substrate, The particle-hole operation P = i K (K is the complex which is in between a top gate and bottom gate. Since conjugation) transforms Pu = u , i.e., each eigenstate we have rotational symmetry in the x-y plane, we choose k k + + + u at energy " has a copyPu at energy " . Meanwhile cylindrical coordinates (; '; z): k k k k it also satis es PH P = H . Therefore H has a 0 0 0 x =  cos '; y =  sin '; z = z; (S-5) particle-hole symmetry. 8 (a) (b) (a) (b) -799.234 c= -1 c=0 -800 c= +1 -800 -799.236 0.7 -805 -805 -799.238 -0.3 -10 FIG. S-2: (a) Berry curvature distribution for the top most FIG. S-3: Similar to Fig. S-2, but with V = 1:5 meV with valence miniband. Dashed hexagons denote the mini Brillouin Berry curvature distribution shown for the blue band. Here zone. (b) Dispersion for the mini-bands close to the intrinsic Berry curvature distribution is sharply peaked close to the ~ ~ band gap, and the valley Chern number for the top most corners of the MBZ at K; K points [clipped white spots de- 5 2 valence miniband. Parameters are the same with those in note j (k)j > 10 A ]. Berry curvature values close to the ~ ~ Fig. 2a in the main text, except that V = 1:5 meV. s K; K points are the same sign indicating the pseudospin ~ ~ winding of the same chirality. Berry curvature about K; K points both have a negative sign. The panel on the right hand side in (b) shows the tiny gap between two nearly degenerate However, for the superlattice potential V (r) we have 1 minibands. PV (r)P = V (r). This breaks the particle-hole sym- metry for the overall Hamiltonian, i.e., PHP 6= H. Instead, the hamiltonian H possesses a di erent type V < 0. Similar to the case discussed above, the pos- of symmetry. To see this, we de ne a composite particle- ~ itive sign of the Berry curvature from the bare gapped hole and voltage ip operation P = PM . Here P is the Dirac cone in the TMD is again re ected in the center particle-hole operation (as de ned above), and the mirror of the MBZ. However, Bragg scattering at MBZ corners operation M ips V (r) ! V (r) by interchanging the in the top valence miniband for V < 0 is signi cant and top and bottom gate potential keeping H intact. Us- dramatically changes the winding of the pseudo-spinor ~ ~ ing this composite operation we recover PHP = H. wavefunction; indeed the Berry curvature distribution is More explicitly, when V ! V then " ! " are s s k k concentrated at the MBZ corners displaying large and interchanged. negative value (see small white dots). Note that the ~ ~ white dots appear at both K and K points; their com- mon sign indicates that chirality of the pseudospinors at Mini-band Berry curvature distribution, valley these points is the same. As a result, within this (blue) Chern number, and miniband gapped-Dirac-cone band (Fig. S-3b), we obtain a non-vanishing net Berry ux of C = 1 in a single (original) valley (see Fig. S-3). The PDS superlattice potential can also modify the The large Berry ux in the the top valence blue mini- winding of the pseudo-spins as well as the miniband band (for V < 0) arises from a very small minigap with quantum geometry. To display this, we plot the Berry the adjacent miniband just below it (green). Indeed, curvature distribution in the top most valence mini-band when zoomed-in a small minigap between the two top va- for V > 0 (see Fig. S-2a) and V < 0 (see Fig. S-3a). In s s lence minibands of order   eV appears at the K point what follows, we concentrate on the Berry curvature for (see Fig. S-3b); a similar gap also appears at K point. We minibands from the (original) K valley with  = 1 (see note parenthetically that this small gap is larger than the Eq. 1). We rst note that the Berry curvature around a numerical mesh resolution from our numerical diagonal- single bare gapped Dirac cone in the valence band of the ization routine, allowing us to extract the minigap value TMD (when V = 0) has a positive sign for K valley, shown. Indeed, the lower miniband (green) also displays and monotonically decreases in magnitude away from the the same behavior with a sharp distribution of Berry cur- Dirac point. ~ ~ vature peaked at the K; K points; these have opposite We now focus on the topmost at valence miniband sign to the blue miniband above it with a concomitant when V > 0. Now the (positive) sign of the Berry cur- C = 1 in a single valley. vature from the bare gapped Dirac cone in the TMD is re- ected in the center of the MBZ (see Fig. S-2a). However, While the nite valley Chern numbers indicate a non- PDS induced Bragg scattering close to the MBZ bound- trivial topology, the minigap is so small as to make it aries produces a di erent winding of the pseudospin and very dicult to meaningfully resolve any real experimen- a di erent sign of Berry curvature close to MBZ bound- tal signatures of the nite valley Chern number. Indeed, aries. This reconstructs the Berry curvature distribution. for practical purposes, when energy resolution is larger As a result, the net Berry ux through the MBZ for the than the small minigap   eV, the minigap cannot be top most valence miniband is C = 0. resolved. As a result, the blue and green minibands close ~ ~ Next, we turn to the topmost at miniband when to K; K points resemble Dirac cones with the same chi- energy, energy, 9 rality. This closely mirrors the behavior of the at mini- Figs. S-5 and S-6, respectively. These display that at bands in twisted bilayer graphene close to magic angle, minibands are also obtained for these superlattices under- ~ ~ where electrons close to K; K in the moir e Brilliouin scoring the generic nature of the PDS atband scheme. zone are severely slowed, and possess the same chirality. The main di erence between the green and blue bands in Other types of TMDs PDS scheme using TMDs is a reduced degeneracy (only two for spin). Here we discuss the PDS scheme (using triangular lat- tice as illustration) as applied to other types of TMDs, PDS induced minigaps speci cally, MoS , MoSe , and WS . The results are dis- 2 2 2 played in Fig. S-7. Here we only show the valence band widths, since the miniband structure possess a symmetry In this section we discuss how the minigaps between when V ! V and " ! " are interchanged (see the s s k k the bundles of atbands evolve as a function of applied above section, \Particle-hole asymmetry"). As expected, superlattice potential V . Here we show the rst minigap at minibands can be readily achieved. between minibands in the valence band (i.e., the gap be- tween blue and green bands when V > 0, and the gap between green and purple band when V < 0, see Fig. 2 Estimate for disorder broadening in the main text) as a function of V . This indicates that the minigaps increase as larger superlattice potential is In this section, we estimate the disorder broadening. applied. We begin by observing that charged impurity scatter- ing is responsible for most of the observed bulk di u- sive transport behaviors at low temperature (the con- tribution of phonons is frozen out). They reside either inside the substrate, or can be desposited near the in- terface between the substrate and the 2D system during the processing/sample preparation. Here we rst extract the impurity concentration n from transport measure- imp -5 ment results, and then use n to estimate the disorder imp -10 -5 0 5 10 broadening. We will concentrate on modeling bare WSe , where FIG. S-4: PDS induced minigaps between blue and green recent low temperature transport measurements have re- 4 2 1 1 minibands in the valence band (for V > 0) as well as between s vealed large mobilities  = 3 10 cm V s at T = 12 2 green and purple minibands (for V < 0). The gray shaded 1:4 K for carrier density n = 2 10 cm [23]. To pro- area indicates the region where the minibands are not well ceed, we rst note that close to the band edge of a TMD, separated. 2 2 2 2 the dispersion can be written as " = v ~ q + with v~ = 3:94 eV A and  = 0:8 eV [26]. For small q, we adopt an e ective mass approximation "   + 2 2 2 v ~ q =2. This produces Other types of superlattices m = =v = 0:39 m ; (S-10) Here we show that square and hexagonal superlattices which is consistent with values obtained in the literature can also lead to at bands. For the square superlattice, [S1,S2]. Also, within the e ective mass approximation, we follow Ref. [22] and use the superlattice potential its Fermi wave vector is q = n = 0:25 nm ; (S-11) V (r) = 2V  [cos(b r)+cos(b r)+cos(b r) cos(b r)]; f squ s 0 1 2 1 2 (S-8) 12 2 where we used n = 2 10 cm [23]. where b = (2=a)(1; 0) and b = (2=a)(0; 1) with a the 1 2 In order to extract an estimate for the impurity density superlattice spacing. For hexagonal superlattice, we use from the mobility, we write the conductivity as the superlattice potential = en = e n=m ; (S-12) V (r) = V  cos(G  r) + cos(G  r ) ; (S-9) hex s 0 i i where n is the carrier density (regardless of spin or val- j=1;2;3 ley),  is the mobility, and  the scattering time. At low 0 temperature, the impurity scattering time reads as where r = r + (a; 0), and G are the same reciprocal vectors de ned in Eq. S-2 which also apply to hexago- = n dzhw( )(1 cos  )i ; (S-13) imp i i z nal lattices. The reconstructed minibands are shown in first valence gap, 10 (a) (b) -800 -835 -820 -845 -840 20 40 -10 -5 0 5 10 FIG. S-5: Dispersion and DOS at V = 0:5 meV (a) and bandwidth (b) for a square superlattice with lattice spacing of 20 nm. Other parameters are the same with those in Fig. 2a in the main text. (a) (b) -810 -830 -820 -835 -830 20 40 -4 -2 0 2 4 FIG. S-6: Dispersion and DOS at V = 0:25 meV (a) and bandwidth (b) for a hexagonal superlattice with inter-hole spacing of 20 nm. Other parameters are the same with those in Fig. 2a in the main text. energy, energy, energy, energy, 11 its magnitude equals to Fermi wave vector q , and q is f s the e ective screening wave vector. Eqs. S-14 above enables to estimate n from a known imp mobility. Using this impurity concentration, we can es- -800 timate the disorder potential width. To do so, we note that randomly distributed positive and negative charged -820 impurities create uctuations of the disorder potential. Taking uncorrelated impurity positions, the amplitude of the uctuations [29{31] is given by -10 -5 0 5 10 Z Z dq 2 2 2 = h[V (r)] i = n dz V (q; z): (S-17) imp (2) A numerical estimate for n and both require the imp -700 value for the screening wave vector q = r . There are a number of processes that can contribute to screen- ing. For example, at long-wavelengths, Thomas-Fermi -720 screening [S3] yields a screening wavevector as q = 2 2 2 2 1 (2e =) = (4e =)(=v ~ ) = 7:4 nm , where  is 0 0 the density of state at the Fermi energy, and  = 4 for -10 -5 0 5 10 SiO . Screening can also arise from a proximal metal- lic top gate displaced several nanometers away from the vdW layer by 2d [16], in this case q  d (assum- ing a perfect metallic gate). At low densities, non-linear screening can even take e ect. The exact value of the dis- -860 order potential depends on details of how the Coulomb potential is screened. Instead of identifying the source screening which is beyond the scope of this work, here -880 we take an e ective screening wavevector approach. To obtain a rough estimate of the range of values can take on, we plot for a wide range of e ective q values -900 s (Fig. S-8). We nd that for ultra-clean WSe samples -10 -5 0 5 10 several meV (see Fig. S-8). We note the above yields an estimate of disorder broad- FIG. S-7: Valence miniband widths for triangular PDS for ening for bare WSe . The precise disorder broadening MoS (v~ = 3:51 eVA,  = 0:88 eV), MoSe (v~ = 3:11 eVA, 2 2 when the PDS scheme is applied will depend on the de- = 0:73 eV), and WS (v~ = 4:38 eVA,  = 0:89 eV). The tails of screening, as well as the actual impurity concen- lattice spacing a = 20 nm. tration of the PDS sample. The detailed form as well as the values for these are, at present, unavailable. As a re- sult, the disorder broadening obtained above represents where we assumed an uncorrelated charged impurities an estimate for the possible disorder broadening in WSe n (units cm ) in the substrate, and w is the scatter- imp PDS atbands. ing rate, see below. Eqs. S-12 and S-13 lead to h i n = (e=m ) dzhw( )(1 cos  )i : (S-14) imp i i z Estimate for mean free path ` Within the e ective mass approximation, Here we estimate the mean free path ` for the sample 12 2 1 dq which has n = 2  10 cm , q = n  0:25 nm 2 f hw( )(1cos  )i = jV (k; z)j (1cos )(" " ); i i z q q i 4 2 1 1 4~ and  = 3 10 cm V s [23]: (S-15) ~q m 2 ` = v  = = 0:49 m: (S-18) 2e m e V (q; z) = exp(qz); (S-16) (q + q ) Since `  a, the electronic quasiparticles are able to where k = jq q j,  =   , tan  = q =q , q is experience Bragg scattering induced by the PDS super- i q q q y x i the initial electron wave vector at the Fermi surface with lattice potentials in this sample. energy, energy, energy, 12 0.1 4 8 0 5 10 FIG. S-8: Estimated disorder broadening (red) and im- FIG. S-9: Estimate of wavefunction spatial extent a (as a purity concentration n (blue) for a wide range of e ec- function of applied V ) when con ned in a trough of a super- imp s 12 2 tive q , for a WSe sample having n = 2  10 cm and s 2 lattice. This was obtained using a variational approach and 4 2 1 1 = 3  10 cm V s at T = 1:4 K recently reported for with an e ective mass approximation. Blue, yellow, green, WSe from the Philip Kim group [23]. 2 and red lines correspond to superlattice wavelengths of a = 20, 30, 40, 50 nm, respectively. For very small V , a rapidly s w increases and gets cut o ; below this critical V the electrons are not well con ned in the superlattice troughs. Variational estimate for wavefunction extent a Here we use a variational approach to estimate a . As a result, only a large enough V  V is able to trap We rst adopt an e ective mass approximation for an an electron at its local minimum, and to create a nearly electron in conduction band: at band. Thankfully for the TMDs we consider in this 2 2 ~ r work, this critical V is small due to their large e ective H = + V (r); (S-19) 2m masses (see Fig. S-9). To illustrate the typical extent of the trapped wavefunction we plot the behavior of a where V (r) = V 2cos(G  r) is the triangular s j j=1;2;3 versus V in Fig. S-9. Here we have used m = 0:39 m s e superlattice potential used in the main text. The minus with a = 20; 30; 40; 50 nm corresponding to blue, yellow, sign in the potential is to create a local potential mini- green and red lines, respectively. As displayed, even fairly mum for an electron at r = 0. modest values of V yield fairly localized a . s W To proceed in the variational approach, we use a Gaus- sian trial wave function 1 r (r) = p exp ; dr[ (r)] = 1; 4a Flat band ferromagnetism 2a (S-20) to describe an electron trapped at r = 0. The energy for After discussing how nearly at bands can be created this trial wave function is in TMDs using the PDS scheme, we now demonstrate a 2 possible spontaneous symmetry breaking that arises due 2 2 2 ~ a 8 a hHi = 6V exp ; (S-21) the quenching of kinetic energy. 4m 9 a We start from the electronic Hamiltonian and the equation for its extremum reads as X H =  c c + H ; (S-24) k int 2 2 2 2 2 @hHi 32 a 8 a ~ a W W W = V exp = 0: 2 2 @a 3 a 9 a 2m where  describes the dispersion, and (S-22) Eq. S-22 has a nite and positive solution for a that W X y y H = c c c c ; (S-25) decreases with increasing V (e is Euler's number): 0 0 0 0 int k  k s k+q k q 2V 0 0 k kq; 2 2 2 2 3a e  ~ a 1=2 a = W e V =V ; V = : W 0  s is the electron-electron interaction. Here we only retain 2 54 2m (S-23) the on-site repulsive interaction U and neglect inter-site Here W (z) is the the Lambert W function, and W (z) 0 0 interactions, which are strongly suppressed in nearly at only has a real, negative solution when e  z  0. bands [16]. Generally, spin polarization in the z-direction Speci cally, W (e ) = 1 and W (0) = 0. Therefore, 0 0 is favored due to repulsive interactions [S4] thus we will V  V is required for consistency of our initial bound s  focus on the possibility of the spin polarization in the state trial wave function/ansatz, i.e., a nite and positive z-direction, and use the mean eld parameterization as a . When V < V , a becomes a complex number, in- W s  W dicating the bound state trial wave function/ansatz fails. hc c i =  0 0n  : (S-26) 0 0 kk  k k k  13 Using this mean eld ansatz, we obtain the mean eld From this we see the normal spin-degenerate state is un- Hamiltonian stable when the second term is negative, i.e., Z ( )U > H = ( + Un  ) c c UV n  n  ; (S-27) MF k   " # where   is opposite to , and n  = V hc c i is MF k k k the spin  density with respect to the mean eld ground Electronic address: justinsong@ntu.edu.sg state. For a nearly at band, we can neglect the k- S1. A. Korm anyos, G. Burkard, M. Gmitra, J. Fabian, V. dependence of    , and its energy density is k 0 Z olyomi, N. D. Drummond, and V. Fa ko, 2D Mater. 2, 022001 (2015). hH i = ( + Un  ) n  + ( + Un  ) n  U n  n  ; MF 0 # " 0 " # " # S2. B. Fallahazad, H. C. P. Movva, K. Kim, S. Larentis, T. (S-28) Taniguchi, K. Watanabe, S. K. Banerjee, E. Tutuc, Phys. where hH i = hH i=V . In anticipation of the (spin- MF MF Rev. Lett. 116, 086601 (2016). split) broken symmetry state, we can re-write hH i to MF S3. S. Das Sarma, S. Adam, E. H. Hwang, E. Rossi, Rev. emphasize that the exchange interaction favors aligned Mod. Phys. 83, 407 (2011). spins: S4. Y.-H. Zhang, D. Mao, Y. Cao, P. Jarillo-Herrero, T. Senthil, Phys. Rev. B 99, 075127 (2019). hH i = " n  n  ; (S-29) MF 0 where " =  + U (n  + n  ). 0 0 " # When there is a nite bandwidth [for e.g., arising from disorder broadening (see main text and above)], " no longer resides at a single at energy, but instead uctuates. To capture this, the [single-particle] energy density can be described by a broadened spectral func- tion. Here we have used a simple spectral function 2 2 Z (") = A exp[(" " ) =2 ] to model this level broad- ening, where A = [ 2 ] . Other choices of spectral function do not modify the qualitative conclusions we de- scribe below and in the main text. As a result, the energy density can be described in close analogy with that used for the energy density in Landau levels [28]. Using the spectral function above, we nd the energy density (including disorder broadening) can be modeled as hH i = d" "Z (") n  ; (S-30) MF where  is the Fermi level for spinor-component  that satis es n  = d" Z ("): (S-31) As we can see, Eq. S-30 mirrors Eq. S-29 except that the single-particle energy density is now broadened by the spectral function (due to disorder broadening). In the normal state the two degenerate at bands are equally occupied. To see whether it is a stable state, we perturb n  = n  + n  with n  = (n  + n  )=2. By using 0  0 " # n  = 0 and expanding  to the second order, the mean eld energy density reads as hH i = d" "Z (") n MF n  1 + U + : (S-32) 2 Z ( )

Journal

Condensed MatterarXiv (Cornell University)

Published: Apr 16, 2019

There are no references for this article.