Abstract
Recently, the twist angle between adjacent sheets of stacked van der Waals materials emerged as a new knob to engineer correlated states of matter in twodimensional heterostructures in a controlled manner, giving rise to emergent phenomena such as superconductivity or correlated insulating states. Here, we use an ab initio based approach to characterize the electronic properties of twisted bilayer MoS_{2}. We report that, in marked contrast to twisted bilayer graphene, slightly holedoped MoS_{2} realizes a strongly asymmetric p_{x}p_{y} Hubbard model on the honeycomb lattice, with two almost entirely dispersionless bands emerging due to destructive interference. The origin of these dispersionless bands, is similar to that of the flat bands in the prototypical Lieb or Kagome lattices and coexists with the general band flattening at small twist angle due to the moiré interference. We study the collective behavior of twisted bilayer MoS_{2} in the presence of interactions, and characterize an array of different magnetic and orbitallyordered correlated phases, which may be susceptible to quantum fluctuations giving rise to exotic, purely quantum, states of matter.
Introduction
Twodimensional van der Waals materials constitute a versatile platform to realize quantum states by design, as they can be synthesized in many different stacking conditions^{1}, offer a wide variety of chemical compositions, and are easily manipulated by back gates, strain and the like. Stacking two sheets of van der Waals materials atop each other at a relative twist has recently emerged as a vibrant research direction to enhance the role of electronic interactions, with first reports on twisted bilayer graphene^{2,3,4,5,6} and another van der Waals materials stacked atop each other at a twist^{7,8,9,10,11,12,13,14,15,16,17} displaying features of correlated physics that afford a high level of control. In particular, bi, tri, and quadruplelayer graphene^{18} as well as twisted fewlayer transition metal dichalcogenides (TMDs)^{19,20} are currently under intense experimental scrutiny^{13,21,22,23,24,25,26,27,28,29}. By forming a moiré supercell at small twist angles, a large unit cell in real space emerges for twisted systems, which due to quantum interference effects leads to a quasitwodimensional system with strongly quenched kinetic energy scales. This reduction in kinetic energy scale, signaled by the emergence of flat electron bands, in turn enhances the role of electronic interactions in these systems. Therefore, twisted systems enable the realization of new correlated condensed matter models, establishing a solidstate quantum simulator platform^{30}.
Whereas the flatting of band dispersions in twodimensional moiré superlattices results mainly from the localization of charge density distributions by the moiré potential, a wellknown alternate pathway to flat bands can occur in certain lattices such as the Lieb and the Kagome lattices. Here, purely geometric considerations lead to the formation of perfectly localized electronic states that have weight only on single plaquettes or hexagons, respectively, and that are eigenstates of the kinetic Hamiltonian due to destructive interference between lattice hopping matrix elements^{31}. To put it differently, linear combinations of the macroscopically degenerate extended Bloch states in these systems allows to form localized Wannierlike eigenstates (living on single plaquettes or hexagons in the examples above) with no dispersion (for a review on the subject see, e.g,^{32}). Such flat band systems can give rise to many interesting phenomena, such as the formation of nontrivial topology when timereversal symmetry is broken, or other exotic quantum phases of matter due to their susceptibility to quantum fluctuations and electronic correlations^{32}.
Here, we demonstrate that both flat band mechanisms can be engineered to coexist in twisted bilayers of MoS_{2} (tbMoS_{2}): a TMD of direct experimental relevance that has been extensively studied from synthesis to applications^{33,34}. We confirm that families of flat bands emerge when two sheets of MoS_{2} in the 2H structure are stacked at a twist^{12,35} due to moiré potentials. Our largescale ab initio based simulations show that while the first set of engineered flat bands closest to the edge of the bandgap with twist angles close to Θ ≈ 0^{∘} can be used to effectively engineer a nondegenerate electronic flat band in analogy to a single layer of graphene at meV energy scales, more intriguingly, the next set of flat bands instead realizes a strongly asymmetric flat band p_{x}–p_{y} honeycomb lattice^{36,37}. Both of these families of bands should be accessible experimentally via gating. The strongly asymmetric nature of this p_{x}–p_{y} honeycomb lattice is in marked contrast to the muchdiscussed case of twisted bilayer graphene, where an approximately symmetric version of such a Hamiltonian is now believed to describe the lowenergy flat band structures found at small twist angle^{38,39,40,41,42}. The strongly asymmetric p_{x}–p_{y} honeycomb model itself features two almost entirely dispersionless flat bands that touch the top and the bottom of graphenelike Dirac bands at the Gamma point, respectively. These flat bands in this model originate from destructive interference, in analogy to flat bands in the Lieb and the Kagome lattices^{31} discussed above, and will be referred to as ultraflat bands in the following discussion. On top of that, the total bandwidth of the strongly asymmetric p_{x}–p_{y} honeycomb effective model realized here (all four bands) can be further flattened by decreasing the twist angle. In addition, these ultraflat bands can be topologically nontrivial in the presence of spinorbital coupling (SOC)^{43}. Although all the flat bands discussed here originate from the Γpoint states of MoS_{2} and are not affected by intrinsic SOC (see Supplementary Fig. 3), we expect that substrate engineering^{44} can be used to introduce SOC coupling into these bands and invoke topologically nontrivial behavior of the ultraflat band states. Previously, the p_{x}p_{y} model was studied in the context of cold gases where exotic correlated phases were predicted^{36,45,46}, as well as in semiconductor microcavities^{47} and certain 2D systems such as organometallic frameworks^{48,49} and Bismuth deposited on SiC^{50} with a focus on their nontrivial topology properties. Our findings elevate tbMoS_{2} to an interesting platform where effects of ultraflat bands can be studied systematically in a strongly correlated solidstate setting.
Notably, in the strongcoupling regime, the p_{x}–p_{y} model amended by Hubbard and Hund’s interactions gives rise to a spinorbital honeycomb model which – depending on the specific parameters and symmetries of the model – hosts magnetic, orbital as well as valencebond orderings, or even more exotic quantum spinorbital liquid phases^{51,52,53}. With this, our work adds an interesting type of lattice model – the highly asymmetric p_{x}–p_{y} Hubbard model – to the growing list of systems that can effectively be engineered using the twist angle between multiple layers. This is particularly intriguing as we maintain the full advantages that come with twodimensional van der Waals materials, such as relative simplicity of the chemical composition and controllability of the material properties; e.g. of the filling (by a back gate), electric tunability (by displacement fields) or the bandwidth of the model (by the twist angle).
Results
Ab initio characterization of twisted MoS_{2}
We first characterize the lowenergy electronic properties of twisted bilayer MoS_{2} using density functional theory (DFT) calculations (see Methods). DFT in particular has established itself as a reliable tool to provide theoretical guidance and to predict the band structure of many twisted bi and multilayer materials^{8,13,15}. However, such a firstprinciples characterization becomes numerically very demanding as the twist angle Θ approaches small values and the unit cell becomes very large entailing many atoms (of the order of a few thousands and more). Nevertheless, it is that limit in which strong bandnarrowing effects and as a consequence prominent effects of correlations are expected. The results of such a characterization are summarized in Fig. 1. Note that atomic relaxation has been shown to affect the electronic properties of twisted 2D materials^{12,35,54}. While for twisted bilayer graphene this effect is only significant at twist angles smaller than 1 degree^{54}, it noticeably alters the lowenergy band dispersions and charges density localization for twisted transition metal dichalcogenides bilayer (such as MoS_{2}) even with relatively large twist angles above 1 degree^{12,35}. Therefore, we relax all the systems in our DFT calculations. Panel (a) shows the relaxed atomic structure of two sheets of MoS_{2} in real space, twisted with respect to each other. A moiré interference pattern forms at a small twist angle yielding a large unit cell, within which we identify different local patterns of stacking of the two sheets of MoS_{2}, indicated via areas framed by cyan, magenta or purple dashed lines. The local stacking arrangements of the respective areas are given in the right subpanels of the panel (a). Note that the B^{Mo/S} and the B^{S/Mo} regions are equivalent as they are related by C2 symmetry. These equivalent B^{Mo/S}/B^{Mo/S} regions form a hexagonal network. In panel (b) we show the ab initio band structure of the twisted material after relaxation, where we find two families of bands that will become increasingly flat and start to detach from all other bands, as the twist angle is lowered. We mark these bands by blue and red color in panel (b), which shows results for decreasing angles from Θ = 3.16^{∘}–2.28^{∘}. The bandwidth of these two energetically separated groups of bands is summarized in panel (c) of Fig. 1. We find that the bandwidth of these two bands shrinks drastically as the angle is decreased, yielding bandwidths of the order of 10 meV as the angle approaches Θ ≈ 2°. Similar features are also shown in the work of Naik et al.^{35}. The bandwidth and the shape of the flat bands (in particular for the second set) in our calculations are slightly quantitatively different from the previous work probably because we relax the structure directly with DFT while the authors of ref. ^{35} use a forcefield approach. Note that these flat bands near the top of the valence bands originate from the states around the Γ point in the Brillouin zone of the primitive unit cell of untwisted MoS_{2}, with both S p_{z} and Mo d\({}_{{{{{{{{\rm{{z}}}}}}}^{2}}}}\) characters (see Supplementary Fig. 2 for a DFT characterization of the orbital contribution to the different bands). This is different to the case of twisted WSe_{2}, where the top valence flat bands originate from the states around the K point in the Brillouin zone of the primitive unit cell (dominated by W d\({}_{{x}^{2}{y}^{2}}\) and d_{xy} orbitals), which experience different interlayer moiré potentials compared with those of the Γpoint flat bands discussed here leading to an effective triangular lattice Hubbard model^{13}. Since also in other TMDs, such as MoSe_{2} and WS_{2}, the top of the valence band in the untwisted bilayer is also located at the Γ point in the Brillouin zone^{55,56}, the physics we discussed here transfers also to those materials being twisted.
The upper bands in Fig. 1 (marked in blue) show a Dirac cone at the K point and behave very similar to the bands found for monolayer graphene (with the exception of a reduced bandwidth). They are spin degenerate in nature, but feature no additional degeneracy except at certain high symmetry points. Instead, the next set of bands (marked in red) is essential to our work. They too feature a Dirac cone at the K point, but also feature two additional ultraflat bands at the top and bottom in addition to a band structure similar to graphene. The ratio between the width of the ultraflat and the flat bands decreases as the angle is decreased, but saturates in our calculations as a twist angle of Θ ≈ 2.28° is approached. We attribute this saturation to lattice relaxation effects; note however that the overall bandwidth keeps decreasing. To access this second set of bands we need to empty the bands marked in blue first. The effects of this doping are of minor quantitative nature (see Supplementary Fig. 5).
Remarkably, this second family of flat bands is welldescribed by an effective p_{x}–p_{y} tightbinding model on a honeycomb lattice, depicted schematically in Fig. 2a, and conveniently described by the following Hamiltonian:
where \({{{{{{{{\bf{c}}}}}}}}}_{i,s}={({c}_{i,{{{{{{{\rm{x}}}}}}}},s},{c}_{i,{{{{{{{\rm{y}}}}}}}},s})}^{T}\) with c_{i,x(y),s} annihilating an electron with p_{x(y)}orbital at site i and with spin s = ↑, ↓. \(\left\langle i,j\right\rangle\) (\(\left\langle \left\langle i,j\right\rangle \right\rangle\)) denotes (next) nearest neighbors. For each sum in Eq. (1), the first term describes the σ hopping (head to tail) between the porbitals and the second term denotes the π hopping (shoulder to shoulder). Furthermore, \({{{{{{{{\bf{n}}}}}}}}}_{ij}^{\parallel }=({{{{{{{{\bf{r}}}}}}}}}_{i}{{{{{{{{\bf{r}}}}}}}}}_{j})/ {{{{{{{{\bf{r}}}}}}}}}_{i}{{{{{{{{\bf{r}}}}}}}}}_{j}\), with r_{i} being the position of site i and \({{{{{{{{\bf{n}}}}}}}}}_{ij}^{\perp }=U{{{{{{{{\bf{n}}}}}}}}}_{ij}^{\parallel }\) with U being the twodimensional 90 degree rotation matrix \(U=\left(\begin{array}{ll}0&1\\ 1&0\end{array}\right)\). Finally, t_{σ} and t_{π} (\({t}_{\sigma }^{N}\) and \({t}_{\pi }^{N}\)) are the nearest neighbor (nextnearest neighbor) hopping amplitudes for the σbonding term and πbonding term, respectively. Figure 2b, c depict the corresponding dispersions, density of states, and wave functions in comparison to model predictions, illustrating that the four moiré bands at low energies are well captured by Eq. (1) upon the choice of hopping parameters t_{π} = 0.25t_{σ}, \({t}_{\sigma }^{N}=0.07{t}_{\sigma }\) and \({t}_{\pi }^{N}=0.04{t}_{\sigma }\). The density of states exhibits a characteristic four van Hove singularities structure, with two originating from the Dirac bands and two stemming from the additional two ultraflat bands. The small ratio between the nearest neighbor hopping amplitudes t_{π}/t_{σ} determines the residual small dispersion in the ultraflat bands we report. This ratio is controllable by the twist angle, which is summarized in the inset of Fig. 1c. All these parameters are related to the interlayer moiré potential and are thus expected to be also affected and controllable by the uniaxial pressure perpendicular to the layers as demonstrated for twisted bilayer graphene^{4}.
The flat band wavefunctions consist of atomic wavefunctions from the p_{z} orbital on S atoms and the d\({}_{{{{{{{{\rm{{z}}}}}}}^{2}}}}\) orbital on Mo atoms. Modulated by the moiré potential, the weighting of the atomic wavefunctions and their modulus square (i.e., charge density) vary at different atomic sites across the whole supercell, showing distinct patterns for different flat band states at the K point in the supercell Brillouin zone as shown in Panel (c) of Fig. 2. These patterns of the charge density as well as the real and the imaginary part of the total wavefunctions obtained from DFT show features consistent with those of the p_{x}–p_{y} Hamiltonian of Eq. (1). Note, that we call this the p_{x}p_{y} Hamiltonian to connect to established literature on the subject; whereas the actual moiré wave functions are composed of p_{z} and d_{z} orbitals, they transform like p_{x}, p_{y} orbitals according to the irreps of the reduced symmetry group of the moiré supercell. Interestingly, the charge density distribution of the top ultraflat band state displays a Kagome lattice structure. We have thus unambiguously established twisted MoS_{2} to be a candidate system to realize a p_{x}–p_{y} model on the honeycomb lattice with strongly asymmetric hoppings t_{σ} and t_{π}, giving rise to a new set of ultraflat bands.
Correlations and magnetic properties
We now study the role of electronic interactions. As the highlyanisotropic p_{x}p_{y} orbital structure constitutes the essential novelty of twisted bilayer MoS_{2}, we focus on quarter filling (one electron per sublattice in the Moié unit cell) where orbital fluctuations can be expected to be crucial. This filling fraction is straightforwardly accessible in the experiment via back gating, and we defer a discussion of the halffilled case to Supplementary Note 1. To proceed, we assume purely local electronic interactions, which can be generically parameterized in terms of the HubbardKanamori Hamiltonian:
for two orbitals with rotational symmetry. More realistic modelling should include longrange interactions. However, for our choice of commensurate quarter filling, any longerranged component of the Coulomb interaction at strongcoupling will serve merely to renormalize the effective spinorbital interactions of the resulting KugelKhomskii model and we therefore concentrate on purely local interactions for simplicity. Furthermore, our DFT calculations suggest t_{π} ≈ 0.25t_{σ} and only weak nextnearest neighbor hopping at small twist angles; we therefore neglect nextnearest neighbor hopping in the analysis below (see Supplementary Fig. 4 for a comparison of the band structures with and without nextnearest neighbor hopping). An ab initio based characterization of the values of U and J requires numerically expansive Wannierzation of the wave functions and is unfortunately beyond the scope of this work. However, by substrate engineering^{22} it is likely that a whole range of values can be accessed and therefore it is useful to vary these parameters to explore all possible phases accessible in experiments to make concrete predictions. Vice versa given a future experimental observation our results can be used to estimate the strength of correlations.
Figure 3c depicts the local density of states as a function of Hubbard U and Hund’s exchange J interactions, calculated via an exact diagonalization study of Eqs. (1) and (2) for a cluster depicted schematically in (a). Clear evidence of a charge gap beyond U/t_{σ} ~ 4 at small J signifies the onset of a correlated insulator which could be directly observed via transport and scanning tunnelling microscopy. The behavior of the gap is depicted in Fig. 3b as a function of U, J and signifies that charge fluctuations are strongly suppressed for large U. Establishing the existence of a charge gap motivates to set up a strongcoupling Hamiltonian routinely employed for the types of systems under scrutiny here.
In this regime, a natural followup questions concerns possible orderings of the orbital and magnetic degrees of freedom. The corresponding strongcoupling KugelKhomskii Hamiltonian^{57,58,59} for the p_{x}p_{y} model at quarter filling is given in refs. ^{51,52,53} and reads:
Here, \({\xi }_{ij}^{1}=3/4+{{{{{{{{\bf{S}}}}}}}}}_{i}{{{{{{{{\bf{S}}}}}}}}}_{j}\) denotes the projector onto triplet states, whereas \({\xi }_{ij}^{0}=1/4{{{{{{{{\bf{S}}}}}}}}}_{i}{{{{{{{{\bf{S}}}}}}}}}_{j}\) selects the singlet spin states instead. Note that the orbital operators, for example Q_{ij}, are bond dependent, giving rise to a strong spatial anisotropy of the resulting spinorbit model. To be more precise following ref. ^{52}, the operators Q_{ij} and \({\bar{Q}}_{ij}\) describe processes where orbital occupations of sites i and j are reversed, that is they are defined as \({Q}_{ij}=({\tau }_{i}^{+}{\tau }_{j}^{+}+{\tau }_{i}^{}{\tau }_{j}^{})/2\) and \({\bar{Q}}_{ij}=({\tau }_{i}^{+}{\tau }_{j}^{}+{\tau }_{i}^{}{\tau }_{j}^{+})/2\), with \({\tau }_{i}^{\pm }={{{{{{{{\bf{n}}}}}}}}}_{ij}^{\perp }{{{{{{{{\boldsymbol{\tau }}}}}}}}}_{i}\pm i{\tau }_{i}^{y}\) where \({{{{{{{{\boldsymbol{\tau }}}}}}}}}_{i}={({\tau }_{i}^{z},{\tau }_{i}^{x},{\tau }_{i}^{y})}^{T}\). The orbital projection operators can then be expressed as \({P}_{ij}^{xx}=(1+{{{{{{{{\bf{n}}}}}}}}}_{ij}^{\parallel }{{{{{{{{\boldsymbol{\tau }}}}}}}}}_{i})(1+{{{{{{{{\bf{n}}}}}}}}}_{ij}^{\parallel }{{{{{{{{\boldsymbol{\tau }}}}}}}}}_{j})/4\), \({P}_{ij}^{yy}=(1{{{{{{{{\bf{n}}}}}}}}}_{ij}^{\parallel }{{{{{{{{\boldsymbol{\tau }}}}}}}}}_{i})(1{{{{{{{{\bf{n}}}}}}}}}_{ij}^{\parallel }{{{{{{{{\boldsymbol{\tau }}}}}}}}}_{j})/4\), \({P}_{ij}^{xy}=(1+{{{{{{{{\bf{n}}}}}}}}}_{ij}^{\parallel }{{{{{{{{\boldsymbol{\tau }}}}}}}}}_{i})(1{{{{{{{{\bf{n}}}}}}}}}_{ij}^{\parallel }{{{{{{{{\boldsymbol{\tau }}}}}}}}}_{j})/4\) and \({P}_{ij}^{yx}=(1{{{{{{{{\bf{n}}}}}}}}}_{ij}^{\parallel }{{{{{{{{\boldsymbol{\tau }}}}}}}}}_{i})(1+{{{{{{{{\bf{n}}}}}}}}}_{ij}^{\parallel }{{{{{{{{\boldsymbol{\tau }}}}}}}}}_{j})/4\), where e.g. \({P}_{ij}^{xx}\) selects states where the superposition \(({p}_{x}{{{{{{{{\bf{e}}}}}}}}}_{x}+{p}_{y}{{{{{{{{\bf{e}}}}}}}}}_{y}){{{{{{{{\bf{n}}}}}}}}}_{ij}^{\parallel }\) is occupied on nearest neighbor sites i and j connected by the bond \({{{{{{{{\bf{n}}}}}}}}}_{ij}^{\parallel }\).
To study its ground state phase diagram using the ab initio parameters found in the previous section, we employ a meanfield analysis of competing for orbital orderings with ferromagnetic and antiferromagnetic spin order. Note, that the simplifying assumption of vanishing temperature – a standard one in condensed matter research – still allows to draw conclusions for the lowtemperature physics accessible in experiments as fingerprints of the phases we discuss extend into this regime as well. To this end, we note that on the bipartite honeycomb lattice the SU(2) invariant spin sector would, on its own, order either ferro or antiferromagnetically, depending on the sign of the exchange couplings. As an Ansatz, we therefore assume that one of the respective states is stabilized and decouple the spin from the orbital degrees of freedom by replacing S_{i}S_{j} with its expectation value 〈S_{i}S_{j}〉 = ±1/4 such that \({\xi }_{ij}^{1}=1,{\xi }_{ij}^{0}=0\) for ferromagnetic spin order and \({\xi }_{ij}^{1}={\xi }_{ij}^{0}=1/2\) for Neél order.
After such a meanfield decoupling corresponding to the ground state in the spin sector, we analyze the ground states of the resulting Hamiltonian for the orbital degrees of freedom, which we approximate as classical vectors. We use an iterative energy minimization combined with simulated annealing techniques (see Methods) to converge the meanfield equations and find the phase diagram summarized in Fig. 4. Panel (a) shows the energy of ferromagnetic and antiferromagnetic spin configurations from which the magnetic phase diagram can be read off. This is given in the upper part of the plot and we find antiferromagnetic ordering with an intermittent ferromagnetic phase at intermediate ratios of 0.1 < J/U < 1/3. In the lower part of the plot, we show the corresponding subsidiary orbital order. From our simulations, we identify three different configurations of orbital vectors τ, which can be classified according to their projection on a single definite plane in space, shown in the lower left of the plots: (1) ferroorbital (FO) nematic order^{5,6,60,61,62}, where the vectors on all lattice sites align in parallel to the xzplane. Quantum mechanically, finite values of \(\langle {\tau }_{i}^{x/z}\rangle\) indicate an imbalance of the occupation of p_{x} and p_{y} orbitals, breaking rotation symmetry and thereby motivating the notion of a nematic state. (2) AFO nematic order; each vector is aligned antiparallel with its nearest neighbors corresponding to \(\langle {\tau }_{i}^{x/z}\rangle \,\ne\, 0\) on each sublattice, but without finite projections \({\tau }_{i}^{y}\) on individual sites. (3) FO magnetic order; all vectors order along the yaxis, such that \(\langle {\tau }_{i}^{y}\rangle \,\ne\, 0\), which, in the quantum mechanical system, would indicate timereversal symmetry breaking. The inclusion of quantum fluctuations can change this picture and more exotic ground states may emerge. For example, for our ab initio band structure parameters, a noncollinear spin dimer phase is predicted in a certain range of interaction couplings and even a quantum spinorbital liquid is found in its proximity^{53}. Since these exotic phases primarily occur for weak Hund’s coupling and strong orbital anisotropies, the assumptions made for our calculations can therefore be justified for sizable J_{H} and modest distances to the isotropic t_{σ} = t_{π} point.
Discussion
We have established that twisted bilayer MoS_{2} is a promising platform to realize the orbital anisotropic p_{x}–p_{y} Hubbard model by employing largescale ab initio calculations. We find that families of flat bands emerge where the first family of flat bands shows sorbital character and the second family is an intriguing realization of a strongly asymmetric p_{x}–p_{y} Hubbard model both on a honeycomb lattice, adding a lattice with nontrivial almost perfectlyflat bands due to destructive interference to the growing list of systems that can be engineered in twisted heterostructures. The symmetry of these flat bands is inherited from the hexagonal lattice formed by the equivalent B^{Mo/S} and B^{S/Mo} regions. At an even smaller angle, the sequence in the family of flat bands found with respect to their orbital character continues. Our analysis shows that the lowenergy DFT band structures in this system can be well captured by a free electron gas model modulated by a simple harmonic potential that has hexagonal (D_{6}) symmetry, which is consistent with a recent study^{63}. This simple model further shows that the next family would exhibit a dorbital character on the honeycomb lattice. Such a lattice would effectively realize a multiorbital generalization of a Kagome lattice – a prototypical model for quantum spin liquids. However, at such small angles strong relaxation is likely to become dominant, prohibiting access to this regime and potentially spoiling its experimental realization. Currently, the ab initio characterization of such small angles is numerically too exhaustive and this work sparks a direct need for novel computational methods to tackle this question.
Furthermore, our combined exact diagonalization and strongcoupling expansion approaches classify the magnetic and orbital phase diagrams, however, the inclusion of quantum fluctuations stipulates an intriguing avenue of future theoretical research.
Indeed, previous theoretical works provide some evidence for a quantum spin liquid in the SU(4)symmetric KugelKhomskii model on the honeycomb lattice^{64}, the square lattice as a related system without frustration^{65} and studied the role of perturbations that break SU(4) symmetry and isotropy^{53} In twisted MoS_{2}, this regime would in fact map to larger twist angles, where the anisotropy of the p_{x}p_{y} model is less pronounced, as well as to a regime of vanishing Hund’s coupling, placing such a putative quantum spin liquid at the transition between FO nematic and AFO nematic phases.
In addition, by proximity or variations in the chemical composition of the twisted bilayer, it might be possible to induce spinorbit coupling splitting of the ultraflat bands at the top and bottom of the asymmetric p_{x}–p_{y} dispersion. Such a bandgap opening would induce interesting topological properties^{66} in a highly tunable materials setting.
Methods
Details on ab initio calculations
We calculate the electronic properties of twisted bilayer MoS_{2} with ab initio methods based on density functional theory (DFT) as implemented in the Vienna ab initio Simulation Package (VASP)^{67}. We employ planewave basis sets with an energy cutoff of 550 eV and pseudopotentials as constructed with the projector augmented wave (PAW) method^{68}. The exchangecorrelation functionals are treated at the generalized gradient approximations (GGA) level^{69}. The supercell lattice constants are chosen such that they correspond to 3.161 Å for the 1 × 1 primitive cell of MoS_{2}. Vacuum spacing larger than 15 Å is introduced to avoid artificial interaction between the periodic images along the zdirection. Because of the large supercells, a 1 × 1 × 1 kgrid is employed for the ground state and the relaxation calculations. For all the calculations, all the atoms are relaxed until the force on each atom is less than 0.01 eV/Å. Van der Waals corrections are considered with the method of Tkatchenko and Scheffler^{70}. We extract the real and the imaginary parts of the DFT wavefunctions with the VASPKIT code^{71}.
Details on exact diagonalization
Exact diagonalization calculations were performed for the electronic tightbinding model in Eq. (1) with HubbardKanamori interactions defined in Eq. (2). All calculations were performed for a twoorbital eightsite cluster with periodic boundary conditions at quarter filling, corresponding to eight spin1/2 particles in sixteen orbitals. Rotationally symmetric Kanamori interactions are adopted, with \(U^{\prime} =U2J\). As the magnitudes of the Hubbard U and Hund’s exchange J interactions cannot be reliably predicted for a Moié supercell from first principles, all presented results are shown as a function of U, J. Calculations of the singleparticle Green’s functions and local density of states are performed starting from the ground state in the total momentum K_{tot} = 0 and total spin S_{z} = 0 sectors, using the Lanczos method and continuedfraction representation, and a spectral broadening (imaginary part of the selfenergy) of η = 0.1 is imposed.
Details on minimization procedure for classical Hamiltonian
Metropolis Monte Carlo simulations are a prime tool for the investigation of classical spin models, since they allow for offdiagonal, spatially anisotropic spin couplings to be included, even when onespin terms, such as magnetic fields, are involved. Here we employ a special variant of the algorithm to the meanfield version of (3), keeping in mind that the ‘spins’ used in the simulation are approximations to orbital operators τ. First, a lattice site i is randomly chosen, and its respective gradient field h_{i} = ∇_{i}H is computed for the current spin configuration {τ_{i}}. Second, a random orientation \({{{{{{{{\boldsymbol{\tau }}}}}}}}}_{i}^{\prime}\) for the vector at site i is proposed and the weight
is computed for an effective inverse temperature β. Performing several Metropolis updates with increasing values of β we are able to efficiently lower the energy of a random initial configuration, minimizing the odds to converge to a local minimum by only allowing optimal updates (i.e. τ_{i} = −h_{i}) right from the start. After N_{a} sweeps over the full lattice, the soobtained configuration is ameliorated by N_{o} optimization sweeps, where the randomly selected spin is rotated antiparallel to the local gradient field such that the energy is deterministically lowered in every step and we converge as close to the global energy minimum as possible. Hence, this algorithm is reminiscent of Monte Carlo simulations with simulated annealing, but at zero temperature where thermal fluctuations are frozen out.
To benchmark our implementation we have carried out the minimization procedure in the isotropic limit t_{σ} = t_{π} for N_{a} = N_{o} = 10^{5}, where the optimization sweeps are terminated when the energy change after one sweep, ϵ, becomes small (usually ϵ ≤ 10^{−10}). Mapping out the phase diagram for both the FM, 〈S_{i}S_{j}〉 = 1/4, as well as the AFM, 〈S_{i}S_{j}〉 = −1/4, spin sector on a lattice with N = 1250 spins subject to periodic boundary conditions we find the result in Fig. 5, which is consistent with the one presented in ref. ^{52}. For J < 0 the AFM spin sector has lower energy, with the orbitals forming a ferroorbital (FO) nematic state where \(\langle {\tau }_{i}^{x/z}\rangle \,\ne\, 0\) and \(\langle {\tau }_{i}^{y}\rangle =0\). For J > 0 one finds the FM spin sector (for which the orbital degrees of freedom restore their rotation invariance) to dominate as long as J < 1/3, where the AFM sector takes over again and establishes a FO magnetic state, i.e. \(\langle {\tau }_{i}^{x/z}\rangle =0\) and \(\langle {\tau }_{i}^{y}\rangle \,\ne\, 0\).
Data availability
The raw data sets used for the presented analysis within the current study are available from the corresponding authors on reasonable request.
Code availability
The tailored developed codes used in this work can be provided from the corresponding author on reasonable request. Ab initio calculations are done with the code VASP (version 5.4.4).
References
Liu, F. et al. Disassembling 2d van der waals crystals into macroscopic monolayers and reassembling into artificial lattices. Science 367, 903–906 (2020).
Cao, Y. et al. Unconventional superconductivity in magicangle graphene superlattices. Nature 556, 43–50 (2018).
Cao, Y. et al. Correlated insulator behaviour at halffilling in magicangle graphene superlattices. Nature 556, 80–84 (2018).
Yankowitz, M. et al. Tuning superconductivity in twisted bilayer graphene. Science 363, 1059–1064 (2019).
Kerelsky, A. et al. Maximized electron interactions at the magic angle in twisted bilayer graphene. Nature 572, 95–100 (2019).
Choi, Y. et al. Electronic correlations in twisted bilayer graphene near the magic angle. Nat. Phys. 15, 1174–1180 (2019).
Jin, C. et al. Observation of moiré excitons in wse_{2}/ws_{2} heterostructure superlattices. Nature 567, 76–80 (2019).
Tran, K. et al. Evidence for moiré excitons in van der waals heterostructures. Nature 567, 71–75 (2019).
Alexeev, E. M. et al. Resonantly hybridized excitons in moiré superlattices in van der waals heterostructures. Nature 567, 81–86 (2019).
Seyler, K. L. et al. Signatures of moirétrapped valley excitons in mose_{2}/wse_{2} heterobilayers. Nature 567, 66–70 (2019).
Xian, L., Kennes, D. M., TancogneDejean, N., Altarelli, M. & Rubio, A. Multiflat bands and strong correlations in twisted bilayer boron nitride: Dopinginduced correlated insulator and superconductor. Nano Lett. 19, 4934–4940 (2019).
Naik, M. H. & Jain, M. Ultraflatbands and shear solitons in moiré patterns of twisted bilayer transition metal dichalcogenides. Phys. Rev. Lett. 121, 266401 (2018).
Wang, L. et al. Correlated electronic phases in twisted bilayer transition metal dichalcogenides. Nat. Mater. 19, 861–866 (2020).
Scuri, G. et al. Electrically tunable valley dynamics in twisted wse_{2}/wse_{2} bilayers. Phys. Rev. Lett. 124, 217403 (2020).
Kerelsky, A. et al. Moiréless correlations in abca graphene. Proc. Natl. Acad. Sci. U.S.A. 118, e2017366118 (2021).
Andersen, T. I. et al. Excitons in a reconstructed moiré potential in twisted wse_{2}/wse_{2} homobilayers. Nat. Mater. 20, 480–487 (2021).
Kennes, D. M., Xian, L., Claassen, M. & Rubio, A. Onedimensional flat bands in twisted bilayer germanium selenide. Nat. Commun. 11, 1124 (2020).
Rozhkov, A., Sboychakov, A., Rakhmanov, A. & Nori, F. Electronic properties of graphenebased bilayer systems. Phys. Rep. 648, 1–104 (2016).
Wu, F., Lovorn, T., Tutuc, E. & MacDonald, A. H. Hubbard model physics in transition metal dichalcogenide Moiré Bands. Phys. Rev. Lett. 121, 026402 (2018).
Wu, F., Lovorn, T., Tutuc, E., Martin, I. & MacDonald, A. H. Topological insulators in twisted transition metal dichalcogenide homobilayers. Phys. Rev. Lett. 122, 086402 (2019).
Lu, X. et al. Superconductors, orbital magnets and correlated states in magicangle bilayer graphene. Nature 574, 653–657 (2019).
Stepanov, P. et al. Untying the insulating and superconducting orders in magicangle graphene. Nature 583, 375–378 (2020).
Arora, H. S. et al. Superconductivity in metallic twisted bilayer graphene stabilized by wse_{2}. Nature 583, 379–384 (2020).
Chen, G. et al. Tunable correlated chern insulator and ferromagnetism in a moiré superlattice. Nature 579, 56–61 (2020).
Liu, X. et al. Tunable spinpolarized correlated states in twisted double bilayer graphene. Nature 583, 221–225 (2020).
Shen, C. et al. Correlated states in twisted double bilayer graphene. Nat. Phys. 16, 520–525 (2020).
Cao, Y. et al. Tunable correlated states and spinpolarized phases in twisted bilayer–bilayer graphene. Nature 583, 215–220 (2020).
He, M. et al. Symmetry breaking in twisted double bilayer graphene. Nat. Phys. 17, 26–30 (2020).
Burg, G. W. et al. Correlated insulating states in twisted double bilayer graphene. Phys. Rev. Lett. 123, 197702 (2019).
Kennes, D. M. et al. Moiré heterostructures as a condensedmatter quantum simulator. Nat. Phys. 17, 155–163 (2021).
Liu, Z., Liu, F. & Wu, Y.S. Exotic electronic states in the world of flat bands: from theory to material. Chin. Phys. B 23, 077308 (2014).
Leykam, D., Andreanov, A. & Flach, S. Artificial flat band systems: from lattice models to experiments. Adv. Phys.: X 3, 1473052 (2018).
Mak, K. F., Lee, C., Hone, J., Shan, J. & Heinz, T. F. Atomically thin mos_{2}: a new directgap semiconductor. Phys. Rev. Lett. 105, 136805 (2010).
Wang, H., Li, C., Fang, P., Zhang, Z. & Zhang, J. Z. Synthesis, properties, and optoelectronic applications of twodimensional mos_{2} and mos_{2}based heterostructures. Chem. Soc. Rev. 47, 6101–6127 (2018).
Naik, M. H., Kundu, S., Maity, I. & Jain, M. Origin and evolution of ultraflat bands in twisted bilayer transition metal dichalcogenides: Realization of triangular quantum dots. Phys. Rev. B 102, 075413 (2020).
Wu, C., Bergman, D., Balents, L. & Das Sarma, S. Flat bands and wigner crystallization in the honeycomb optical lattice. Phys. Rev. Lett. 99, 070401 (2007).
Wu, C. & Das Sarma, S. p_{x,y}orbital counterpart of graphene: cold atoms in the honeycomb optical lattice. Phys. Rev. B 77, 235107 (2008).
Bistritzer, R. & MacDonald, A. H. Moiré bands in twisted doublelayer graphene. Proc. Natl. Acad. Sci. U.S.A. 108, 12233–12237 (2011).
Yuan, N. F. Q. & Fu, L. Model for the metalinsulator transition in graphene superlattices and beyond. Phys. Rev. B 98, 045103 (2018).
Po, H. C., Zou, L., Senthil, T. & Vishwanath, A. Faithful tightbinding models and fragile topology of magicangle bilayer graphene. Phys. Rev. B 99, 195455 (2019).
Carr, S., Fang, S., Po, H. C., Vishwanath, A. & Kaxiras, E. Derivation of wannier orbitals and minimalbasis tightbinding hamiltonians for twisted bilayer graphene: Firstprinciples approach. Phys. Rev. Research 1, 033072 (2019).
Kennes, D. M., Lischner, J. & Karrasch, C. Strong correlations and d + id superconductivity in twisted bilayer graphene. Phys. Rev. B 98, 241407 (2018).
Zhang, G.F., Li, Y. & Wu, C. Honeycomb lattice with multiorbital structure: topological and quantum anomalous hall insulators with large gaps. Phys. Rev. B 90, 075114 (2014).
Qi, J., Li, X., Niu, Q. & Feng, J. et al. Giant and tunable valley degeneracy splitting in mote_{2}. Phys. Rev. B 92, 121403 (2015).
Lee, W.C., Wu, C. & Das Sarma, S. fwave pairing of cold atoms in optical lattices. Phys. Rev. A 82, 053611 (2010).
Wu, C. Orbital ordering and frustration of pband mott insulators. Phys. Rev. Lett. 100, 200406 (2008).
Jacqmin, T. et al. Direct observation of dirac cones and a flatband in a honeycomb lattice for polaritons. Phys. Rev. Lett. 112, 116402 (2014).
Liu, Z., Wang, Z.F., Mei, J.W., Wu, Y.S. & Liu, F. Flat chern band in a twodimensional organometallic framework. Phys. Rev. Lett. 110, 106804 (2013).
Su, N., Jiang, W., Wang, Z. & Liu, F. Prediction of large gap flat chern band in a twodimensional metalorganic framework. Appl. Phys. Lett. 112, 033301 (2018).
Reis, F. et al. Bismuthene on a sic substrate: a candidate for a hightemperature quantum spin hall material. Science 357, 287–290 (2017).
Smerald, A. & Mila, F. Exploring the spinorbital ground state of ba_{3}cusb_{2}o_{9}. Phys. Rev. B 90, 094422 (2014).
Venderbos, J. W. F. & Fernandes, R. M. Correlations and electronic order in a twoorbital honeycomb lattice model for twisted bilayer graphene. Phys. Rev. B 98, 245103 (2018).
Natori, W. M. H., Nutakki, R., Pereira, R. G. & Andrade, E. C. Su(4) heisenberg model on the honeycomb lattice with exchangefrustrated perturbations: Implications for twistronics and mott insulators. Phys. Rev. B 100, 205131 (2019).
Yoo, H. et al. Atomic and electronic reconstruction at the van der waals interface in twisted bilayer graphene. Nat. Mater. 18, 448 (2019).
Kuc, A., Zibouche, N. & Heine, T. Influence of quantum confinement on the electronic structure of the transition metal sulfide ts_{2}. Phys. Rev. B 83, 245213 (2011).
Zhang, Y. et al. Direct observation of the transition from indirect to direct bandgap in atomically thin epitaxial mose_{2}. Nat. Nanotechnol. 9, 111 (2014).
Tokura, Y. & Nagaosa, N. Orbital physics in transitionmetal oxides. Science 288, 462–468 (2000).
Corboz, P., Lajkó, M., Läuchli, A. M., Penc, K. & Mila, F. Spinorbital quantum liquid on the honeycomb lattice. Phys. Rev. X 2, 041013 (2012).
Khomskii, D. & Mostovoy, M. Orbital ordering and frustrations. J. Phys. A: Math. Gen. 36, 9197 (2003).
Sboychakov, A. O., Rozhkov, A. V., Rakhmanov, A. L. & Nori, F. Manybody effects in twisted bilayer graphene at low twist angles. Phys. Rev. B 100, 045111 (2019).
Sboychakov, A., Rozhkov, A., Rakhmanov, A. & Nori, F. Spin density wave and electron nematicity in magicangle twisted bilayer graphene. Phys. Rev. B 102, 155142 (2020).
RubioVerdú, C. et al. Universal moiré nematic phase in twisted graphitic systems. Preprint at https://arxiv.org/abs/2009.11645 (2020).
Angeli, M. & MacDonald, A. H. γ valley transition metal dichalcogenide moiré bands. Proc. Natl. Acad. Sci. USA 118, e2021826118 (2021).
Corboz, P., Lajkó, M., Läuchli, A. M., Penc, K. & Mila, F. Spinorbital quantum liquid on the honeycomb lattice. Phys. Rev. X 2, 041013 (2012).
Wang, F. & Vishwanath, A. z_{2} spinorbital liquid state in the square lattice kugelkhomskii model. Phys. Rev. B 80, 064413 (2009).
Zhang, G.F., Li, Y. & Wu, C. Honeycomb lattice with multiorbital structure: topological and quantum anomalous hall insulators with large gaps. Phys. Rev. B 90, 075114 (2014).
Kresse, G. & Hafner, J. Ab initio molecular dynamics for liquid metals. Phys. Rev. B 47, 558 (1993).
Blöchl, P. E. Projector augmentedwave method. Phys. Rev. B 50, 17953 (1994).
Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation made simple. Phys. Rev. Lett. 77, 3865 (1996).
Tkatchenko, A. & Scheffler, M. Accurate molecular van der waals interactions from groundstate electron density and freeatom reference data. Phys. Rev. Lett. 102, 073005 (2009).
Wang, V., Xu, N., Liu, J.C., Tang, G. & Geng, W.T. Vaspkit: a userfriendly interface facilitating highthroughput computing and analysis using vasp code. Comput. Phys. Commun. 267, 108033 (2021).
Acknowledgements
We are grateful to Abhay Pasupathy, Cory Dean and Dmitri Basov for fruitful discussions. This work is supported by the European Research Council (ERC2015AdG694097), Grupos Consolidados (IT124919), and SFB925. MC and AR are supported by the Flatiron Institute, a division of the Simons Foundation. We acknowledge funding by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy  Cluster of Excellence Matter and Light for Quantum Computing (ML4Q) EXC 2004/1  390534769 and Advanced Imaging of Matter (AIM) EXC 2056  390715994 and funding by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under RTG 1995 and RTG 2247. Support by the Max Planck Institute  New York City Center for NonEquilibrium Quantum Phenomena is acknowledged. DK, MMS, and ST acknowledge support from the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation), Projektnummer 277146847 – CRC 1238 (projects C02, C03).
Funding
Open Access funding enabled and organized by Projekt DEAL.
Author information
Authors and Affiliations
Contributions
L.X., D.M.K. and A.R. conceived the project. D.M.K., S.T. and A.R. designed and coordinated the research. L.X., M.C., D.K. and M.S. performed all the simulations. All authors discussed and analyzed the results and contributed to writing the manuscript.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Peer review information Nature Communications thanks Fengcheng Wu and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Xian, L., Claassen, M., Kiese, D. et al. Realization of nearly dispersionless bands with strong orbital anisotropy from destructive interference in twisted bilayer MoS_{2}. Nat Commun 12, 5644 (2021). https://doi.org/10.1038/s41467021259228
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41467021259228
Further reading

Unconventional superconductivity in magicangle twisted trilayer graphene
npj Quantum Materials (2022)

A tunable bilayer Hubbard model in twisted WSe2
Nature Nanotechnology (2022)

Ultrastrong spin–orbit coupling and topological moiré engineering in twisted ZrS2 bilayers
Nature Communications (2022)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.