betaFeedback

WorksPhysics and Astronomy

Charge and spin orders of the generalized Wigner crystal phases at fractional fillings in twisted MoTe2_2: unrestricted Hartree-Fock and variational Monte Carlo studies

Abstract

Recent experiments have revealed interaction-driven insulating states at various fractional fillings of twisted bilayer MoTe2_2. Many of these insulating states are believed to be generalized Wigner crystals (GWC) condensed in the background of triangular moireˊ\acute{e} lattice. In this work, we focus on revealing the charge and spin order of the insulating states at 4/3 and 3/2 hole fillings over a range of twist angles and applied displacement fields using unrestricted Hartree-Fock (HF) and variational Monte Carlo (VMC) methods. We find that, at 4/3 filling for low twist angles the charge ordered ground state is valley-polarized (VP), characterized by weak charge delocalization approaching the classical limit. Application of the displacement field induces layer-polarization and suppresses the charge fluctuations even further. In contrast, at 3/2 filling, the VP phase exhibits signatures of a two-component GWC phase, which is particularly robust at finite displacement field. Additionally, by increasing the twist angle one pushes the system into an intervalley-coherent (IVC) phase, which is found to be fluctuation-dominated. Our VMC-based analysis shows that the correlation energy is negligible in the VP phase, which is consistent with its classical nature. However, the IVC phase shows significant gain in correlation energy, which is dominated by the nonlocal correlations captured by the Jastrow factors, while the local Gutzwiller factors play a minor role.

Introduction

In 1934, Eugene Wigner predicted that electrons moving in an inert neutralizing background would spontaneously form a crystalline phase at sufficiently low density where the Coulomb interaction energy dominates over the kinetic energy [1]. The first experimental realization of WC phase came more than 40 years after its prediction, in 1979 in the electrons on the liquid He surface [2]. Later in GaAs/GaAlAs 2DEG, the WC was realized by quenching the kinetic energy in the presence of an external magnetic field [3]. Very recently, a direct observation of the WC phase has been achieved in Bernal-stacked bilayer graphene under external magnetic field [4]. Without the external magnetic field, the WC phase has been observed in monolayer MoSe2MoSe_2 [5] and a heterostructure of MoSe2MoSe_2, where two monolayers are separated by hBN [6]. Additionally, the advent of moiré materials has allowed to realize such a phase even without the need of an external magnetic field. In the moiré flat bands, carriers experience a large Coulomb interaction, which can lead to the formation of the WC phase at certain carrier densities, as experimentally shown in bilayer WSe2WSe_2-WS2WS_2 at fractional fillings n=1/3n = 1/3, 1/21/2 and 2/32/3 [7, 8, 9], as well as in twisted WS2WS_2 [10]. From a theoretical perspective, multiple studies have confirmed or predicted the WC phase in various moiré systems [11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28]. It should be noted that the mentioned realizations of the WC phase are not exactly the phase Wigner predicted. Namely, in contrast to the original concept, the moiré materials possess an underlying lattice, which constrains the real-space motion of the electrons and consequently influences the order of the resultant Wigner crystal. In the literature, such lattice-modified WC phases are referred to as generalized Wigner Crystals (GWCs) [29, 30]. Similar to GWC, the charge density wave (CDW) phenomenon also exhibits spontaneous charge modulation in real space. However, the CDW phase can even appear in the weak-coupling regime as a result of the Fermi surface instability due to the nesting effect [31, 32], while the WC/GWC state is due to substantial interactions leading to charge localization. Additionally, the charge density in GWC shows strong real-space inhomogeneity, which is commensurate with the underlying lattice and appears at specific band fillings. On the other hand, in CDW a much smoother modulation around the average value is seen and both commensurate and incommensurate periodicities can be realized within a wider range of band fillings also coexisting with a metallic state.

Due to a significant Coulomb repulsion and large degree of tunability, transition metal dichalcogenide (TMD) moiré systems can be considered as potential hosts of generalized Wigner-crystal phases with distinct charge and spin configurations. Apart from the structures previously studied in the context of possible GWC realization, experiments on twisted bilayer WSe2WSe_2 (t-WSe2WSe_2) and bilayer MoTe2MoTe_2 (t-MoTe2MoTe_2) have revealed numerous insulating states beyond fractional Chern insulator and anomalous quantum Hall phase, which are induced by tuning the experimentally accessible parameters such as the carrier density, displacement field, and magnetic field [33, 34, 35]. Importantly, those insulating states appear at specific band fillings which together with significant strength of inter-site Coulomb repulsion points to a possible interaction-driven charge localization and formation of the GWC. In this work, we investigate the charge and spin orThe remainder of the paper is organized as follows. Section II, introduces the model Hamiltonian of the system and discusses the HF and VMC methodologies. In section III, the phase diagrams along with the details of the charge and spin textures of the phases are presented. Section IV discusses the VMC results and the effects of correlation beyond HF. Finally, section V summarizes the main findings of the paper.

Model and Method

Model Hamiltonian

The strong Ising spin-orbit coupling in monolayer MoTe2_2 leads to spin-valley locking. As a result, the low-energy valence states near the KK and K′K' valleys of the monolayer Brillouin zone are primarily spin-↑\uparrow and spin-↓\downarrow states, respectively. The twisted bilayer system inherits this spin-valley locking property [36]. Consequently, in t-MoTe2_2, the low-energy moiré states originating from the KK and K′K' valleys of the monolayer Brillouin zone can be identified with the spin-↑\uparrow and spin-↓\downarrow sectors, respectively. Therefore, within the low-energy subspace of t-MoTe2_2, spin and valley degrees of freedom are effectively locked to each other. Hence, throughout this work, we use the terms spin and valley interchangeably. The schematic in Figure 1(a) shows the Brillouin zones of the top and bottom monolayers, denoted by blue and red hexagons, respectively, and the moiré Brillouin zone (mBZ) is denoted by the black hexagon. In this work, we retain the two highest moiré valence bands from each valley as active bands and model the system using the two-band tight-binding model introduced in [37]. Within such an approach the two bands result from two Wannier orbitals residing in MX and XM stacking configurations of the original moiré lattice, where MM and XX correspond to metal (Mo) and chalcogen (Te) ions, respectively. This leads to two interpenetrating triangular sublattices denoted here by AA and BB, which together form a honeycomb structure. The resulting single-particle Hamiltonian is provided below.

Figure 1. (a) Schematic representation of BZs. The blue and red hexagons denote the BZs of the top and bottom monolayers, respectively. The black hexagon denotes the mBZ of t-MoTe2_2 system. The uparrow and downarrow in the KK and K′(−K)K'(-K) valley represent the spin-valley locking property. (b),(c) The bandstructures of K(+K)K(+K) and K′(−K)K'(-K) valleys for θ=1.7∘\theta= 1.7^\circ, and 2.8∘2.8^\circ, respectively at Δ=0\Delta= 0. (d) The bandstructure of θ=2.8∘\theta= 2.8^\circ at Δ=10\Delta= 10 meV.

H0=∑n,m,α,σ∑r={1,3}trc^nασ†c^mα~σ+∑n,m,α,β,σ,τ∈{2,4,5,6}∣tr∣eiσ~νnα,mβτc^nασ†c^mβσ(1)H_0 = \sum_{n,m,\alpha,\sigma}\sum_{r=\{1,3\}} t_r \hat{c}^{\dagger}_{n\alpha\sigma}\hat{c}_{m\tilde{\alpha}\sigma} + \sum_{n,m,\alpha,\beta,\sigma,\tau\in\{2,4,5,6\}} |t_r|e^{i\tilde{\sigma}\nu_{n\alpha,m\beta}^{\tau}} \hat{c}^{\dagger}_{n\alpha\sigma}\hat{c}_{m\beta\sigma} \tag*{(1)}

where c^nασ†\hat{c}^{\dagger}_{n\alpha\sigma} (c^nασ\hat{c}_{n\alpha\sigma}) creates (annihilates) an electron with spin σ\sigma in the sublattice α\alpha in unit cell nn of the honeycomb lattice. The spin index σ\sigma represents the two spin components ↑\uparrow and ↓\downarrow. Additionally, σ~=1\tilde{\sigma}=1 (σ~=−1\tilde{\sigma}=-1) for σ=↑\sigma=\uparrow (σ=↓\sigma=\downarrow) and α~=B\tilde{\alpha}=B (α~=A\tilde{\alpha}=A) for α=A\alpha=A (α=B\alpha=B). The hopping terms are taken up to the 6th6^{\mathrm{th}}} nearest neighbor with amplitudes trt_r corresponding to the rr-th nearest neighbor among which only t1t_1 and t3t_3 are purely real. The remaining ones are complex, with the complex phase depending both on spin and hopping direction. Namely, νnα,mβ\nu_{n\alpha,m\beta} is +1+1 or −1-1 depending on the relative orientation of (n,α)(n,\alpha) and (m,β)(m,\beta), i.e., the two sites involved in the hopping process. For the sign convention of νnα,mβ\nu_{n\alpha,m\beta} refer to the supplementary material [38]. The noninteracting bands of t-MoTe2_2 obtained by diagonalizing Hamiltonian of Eq. (1) and are shown in Figure 1(b) and (Figure 1c) for selected values of the twist angles (θ\theta) 1.7∘1.7^\circ and 2.8∘2.8^\circ, respectively. The bands originating from the two valleys KK and K′K' exhibit degeneracy due to the time-reversal symmetry and the emergent inversion symmetry of the two-band tight-binding model. As can be seen, the bandwidth of the two bands increases with the twist angle.

It should be noted that the Wannier orbitals of the moiré system centered on the AA and BB sublattices are predominantly localized in two different monolayers [37]. Therefore, within the present model, the layer and sublattice degrees of freedom are effectively locked, and we use the terms layer and sublattice interchangeably throughout this work. Due to this correspondence, the effect of an out-of-plane displacement field is incorporated through an on-site energy difference between the AA and BB sublattices.

HD=−Δ2∑n,α∈A,σc^nασ†c^nασ+Δ2∑n,α∈B,σc^nασ†c^nασ(2)H_D = -\frac{\Delta}{2}\sum_{n,\alpha\in A,\sigma}\hat{c}^{\dagger}_{n\alpha\sigma}\hat{c}_{n\alpha\sigma} + \frac{\Delta}{2}\sum_{n,\alpha\in B,\sigma}\hat{c}^{\dagger}_{n\alpha\sigma}\hat{c}_{n\alpha\sigma} \tag*{(2)}

where Δ=Ddϵ\Delta= \frac{Dd}{\epsilon}, dd is the distance between layers, DD is the displacement field, and ϵ\epsilon is the dielectric constant. The effect of the displacement field on the band structure is shown in Figure 1 (d) for a selected value of the twist angle θ=2.8∘\theta= 2.8^\circ. As one can see the displacement field destroys the emergent inversion symmetry of the two-band tight-binding model and lifts the valley degeneracy. It also increases the bandwidth of the individual bands.

To take into account interaction within the effective two-band model, the Coulomb term is projected on the top two noninteracting valence bands of each valley, which are considered as active bands in the study. The projection generates many terms, including onsite Coulomb, intersite Coulomb, ferromagnetic exchange, and a three-site correlated hopping [37]. The intersite Coulomb and the ferromagnetic exchange are considered up to the 3rd3^{rd} and 2nd2^{nd} nearest neighbors, respectively. Taking into account all these terms, the effective interaction Hamiltonian takes the following form.

Hint=U∑in^i↑n^i↓+∑i,j3rdNN∑σ,σ′Vijn^iσn^jσ′+∑i,j2ndNN[∑σ,σ′(−12Jijn^iσn^jσ′)−2JijS^izS^jz]+∑⟨⟨i,j⟩⟩,i′∑σTij′n^i′c^iσ†c^jσ,(3)\begin{aligned} H_{int} ={}& U \sum_i \hat{n}_{i\uparrow}\hat{n}_{i\downarrow} \\ &+ \sum_{i,j}^{3^{rd}NN}\sum_{\sigma,\sigma'} V_{ij}\hat{n}_{i\sigma}\hat{n}_{j\sigma'} \\ &+ \sum_{i,j}^{2^{nd}NN}\left[\sum_{\sigma,\sigma'}\left(-\frac{1}{2}J_{ij}\hat{n}_{i\sigma}\hat{n}_{j\sigma'}\right)-2J_{ij}\hat{S}^{z}_{i}\hat{S}^{z}_{j}\right] \\ &+ \sum_{\langle\langle i,j\rangle\rangle,i'}\sum_{\sigma}T'_{ij}\hat{n}_{i'}\hat{c}^{\dagger}_{i\sigma}\hat{c}_{j\sigma}, \tag*{(3)} \end{aligned}

where, i,ji,j denotes the combined unit cell (n/mn/m) and sublattice indices (α/β\alpha/\beta), while UU and VV represent the onsite and intersite density-density interaction, respectively. Here, JijJ_{ij} is the direct ferromagnetic spin exchange and the zz-component of the spin operator is S^iz=12[n^i↑−n^i,↓]\hat{S}^{z}_{i}=\frac{1}{2}[\hat{n}_{i\uparrow}-\hat{n}_{i,\downarrow}]. The last term corresponds to correlated hopping with amplitude Tij′T'_{ij}, where the hopping strength between two 2nd2^{nd} nearest neighbor sites ii and jj depends on the particle density at the site i′i', which is the common nearest neighbor of ii and jj. The values of all the parameters mentioned in Eqs. (1), (2), and (3) for various θ\theta values can be found in the supplementary material of Ref. [37].

We note that, although the filling fractions are referred to in terms of holes throughout this work, all calculations are performed in the electron picture.

Unrestricted Hartree-Fock mean-field theory

We have used Hartree-Fock mean-field theory to study the total Hamiltonian, H=H0+HD+HintH = H_0 + H_D + H_{int} given by Eqs. (1), (2), and (3). In this method, the two-body (quartic) interaction terms are reduced to one-body (quadratic) terms by considering fluctuations up to the first order [39]. Using this approximation we write c^iσ1†c^jσ2c^kσ3†c^lσ4≈⟨c^iσ1†c^jσ2⟩c^kσ3†c^lσ4−⟨c^iσ1†c^lσ4⟩c^kσ3†c^jσ2\hat{c}^{\dagger}_{i\sigma_1}\hat{c}_{j\sigma_2}\hat{c}^{\dagger}_{k\sigma_3}\hat{c}_{l\sigma_4} \approx\langle\hat{c}^{\dagger}_{i\sigma_1}\hat{c}_{j\sigma_2}\rangle\hat{c}^{\dagger}_{k\sigma_3}\hat{c}_{l\sigma_4}-\langle\hat{c}^{\dagger}_{i\sigma_1}\hat{c}_{l\sigma_4}\rangle\hat{c}^{\dagger}_{k\sigma_3}\hat{c}_{j\sigma_2}. With these simplifications, Eq. (3) takes the following mean-field form

HintMF=U∑i[⟨c^i↑†c^i↑⟩c^i↓†c^i↓−⟨c^i↑†c^i↓⟩c^i↓†c^i↑]+∑i,j3rdNN∑σ,σ′Vij[⟨c^iσ†c^iσ⟩c^jσ′†c^jσ′−⟨c^iσ†c^jσ′⟩c^jσ′†c^iσ]−∑i,j2ndNN∑σJij[⟨c^iσ†c^iσ⟩c^jσ†c^jσ−⟨c^iσ†c^jσ⟩c^jσ†c^iσ]+∑⟨⟨i,j⟩⟩,i′∑σ,σ′Tij′[⟨c^i′σ′†c^i′σ′⟩c^iσ†c^jσ−⟨c^i′σ′†c^jσ⟩c^iσ†c^i′σ′].(4)\begin{aligned} H^{MF}_{int} ={}& U\sum_i\left[\langle\hat{c}^{\dagger}_{i\uparrow}\hat{c}_{i\uparrow}\rangle\hat{c}^{\dagger}_{i\downarrow}\hat{c}_{i\downarrow}-\langle\hat{c}^{\dagger}_{i\uparrow}\hat{c}_{i\downarrow}\rangle\hat{c}^{\dagger}_{i\downarrow}\hat{c}_{i\uparrow}\right] \\ &+\sum_{i,j}^{3^{rd}NN}\sum_{\sigma,\sigma'}V_{ij}\left[\langle\hat{c}^{\dagger}_{i\sigma}\hat{c}_{i\sigma}\rangle\hat{c}^{\dagger}_{j\sigma'}\hat{c}_{j\sigma'}-\langle\hat{c}^{\dagger}_{i\sigma}\hat{c}_{j\sigma'}\rangle\hat{c}^{\dagger}_{j\sigma'}\hat{c}_{i\sigma}\right] \\ &-\sum_{i,j}^{2^{nd}NN}\sum_{\sigma}J_{ij}\left[\langle\hat{c}^{\dagger}_{i\sigma}\hat{c}_{i\sigma}\rangle\hat{c}^{\dagger}_{j\sigma}\hat{c}_{j\sigma}-\langle\hat{c}^{\dagger}_{i\sigma}\hat{c}_{j\sigma}\rangle\hat{c}^{\dagger}_{j\sigma}\hat{c}_{i\sigma}\right] \\ &+\sum_{\langle\langle i,j\rangle\rangle,i'}\sum_{\sigma,\sigma'}T'_{ij}\left[\langle\hat{c}^{\dagger}_{i'\sigma'}\hat{c}_{i'\sigma'}\rangle\hat{c}^{\dagger}_{i\sigma}\hat{c}_{j\sigma}-\langle\hat{c}^{\dagger}_{i'\sigma'}\hat{c}_{j\sigma}\rangle\hat{c}^{\dagger}_{i\sigma}\hat{c}_{i'\sigma'}\right]. \tag*{(4)} \end{aligned}

The total mean-field Hamiltonian, HMF=H0+HD+HintMFH^{MF}=H_0+H_D+H^{MF}_{int}, is solved self-consistently to determine the ground state of the system. We have performed the self-consistent calculation in an unrestricted manner, i.e., without imposing any symmetry-related constraints on the mean fields, allowing for spontaneous symmetry breaking without assuming any predefined supercell structure. The unrestricted HF program of the mVMC package [40] has been used for this purpose.

Variational Monte Carlo

In addition to using the HF mean-field method, we have employed the variational Monte Carlo method to determine the approximate ground state of the model Hamiltonian and to study the effects of the correlations beyond the mean-field approximation. The VMC method is one of the available tools for treating correlation physics beyond mean-field theory, where a suitable many-body variational wavefunction with many variational parameters is optimized to represent the ground state of the system [41, 42, 43, 44, 45, 46]. The variational ansatz is given by

∣ΨT⟩=P^L^∣Ψ^0⟩,(5)|\Psi_T\rangle= \hat{P}\hat{\mathcal{L}}|\hat{\Psi}_0\rangle, \tag*{(5)}

where ∣Ψ0⟩|\Psi_0\rangle is the pair-product or Pfaffian wavefunction [40, 47], which has the following form

∣Ψ0⟩=(∑i,j,σ,σ′fiσ,jσ′c^iσ†c^jσ′†)Ne/2∣0⟩,(6)|\Psi_0\rangle= \left(\sum_{i,j,\sigma,\sigma'} f_{i\sigma,j\sigma'}\hat{c}_{i\sigma}^{\dagger}\hat{c}_{j\sigma'}^{\dagger}\right)^{N_e/2}|0\rangle, \tag*{(6)}

where (i,j)(i,j) are the combined unit cell and sublattice indices, (σ,σ′)(\sigma,\sigma') are spin indices as introduced earlier. NeN_e is the total number of electrons, which should be even, and fiσ,jσ′f_{i\sigma,j\sigma'}’s are the variational parameters to be determined. The projector P^\hat{P} is used to describe correlation beyond the mean-field approximation, which contains factors like Gutzwiller and Jastrow. The projector L^\hat{\mathcal{L}} denotes quantum-number projections like total-spin and momentum. The method of VMC is established on the variational principle, which states that

EG≤ET=⟨ΨT∣H∣ΨT⟩⟨ΨT∣ΨT⟩,(7)E_G \leq E_T = \frac{\langle\Psi_T|H|\Psi_T\rangle}{\langle\Psi_T|\Psi_T\rangle}, \tag*{(7)}

where EGE_G is the energy of the ground state. Any expectation value including the energy is computed by expanding ∣ΨT⟩|\Psi_T\rangle using the basis {∣x⟩}\{|x\rangle\}, which is given by

∣x⟩=∏i=1Nec^riσi†∣0⟩,(8)|x\rangle= \prod_{i=1}^{N_e}\hat{c}_{r_i\sigma_i}^{\dagger}|0\rangle, \tag*{(8)}

where rir_i and σi\sigma_i are the position and the spin of the ithi^{th} electron. The configurations are sampled according to the probability density function ρ(x)∝∣⟨x∣ΨT⟩∣2\rho(x) \propto|\langle x|\Psi_T\rangle|^2 and ETE_T is estimated as

ET≈1M∑l=1Mρ(xl)⟨ΨT∣H∣xl⟩⟨ΨT∣xl⟩,(9)E_T \approx\frac{1}{M}\sum_{l=1}^{M}\rho(x_l)\frac{\langle\Psi_T|H|x_l\rangle}{\langle\Psi_T|x_l\rangle}, \tag*{(9)}

i.e., as an average over the MM samples. The expectation value of any operator AA is estimated similarly.

For this work, in the projector P^\hat{P}, we have considered only the Gutzwiller (P^G\hat{P}_G) and the Jastrow (P^J\hat{P}_J) factors, i.e. P^=P^GP^J\hat{P}=\hat{P}_G\hat{P}_J. These factors have the following forms.

P^G=e∑igin^i↑n^i↓\hat{P}_G=e^{\sum_i g_i\hat{n}_{i\uparrow}\hat{n}_{i\downarrow}}
P^J=e12∑i≠jvij(n^i−1)(n^j−1)(10)\hat{P}_J=e^{\frac{1}{2}\sum_{i\ne j}v_{ij}(\hat{n}_i-1)(\hat{n}_j-1)} \tag*{(10)}

where the value of the variational parameter gig_i controls the local correlations through the double occupancy at the site ii and the variational parameters vijv_{ij} introduce long-range correlation. To obtain the minimum energy state, we have not imposed any symmetry-related constraints on the variational parameters fiσ,jσ′f_{i\sigma,j\sigma'}, gig_i, and vijv_{ij}. In this work, we have not considered the spin number projection. We also ignore the possibility of the backflow correlation.

The VMC calculations have been performed using the mVMC package [40], which utilizes the stochastic reconfiguration method to optimize the variational parameters while minimizing the trial ground state energy ETE_T.

We used a 6×66\times6 lattice system for both the unrestricted HF and VMC calculations. Although this size of cluster may appear relatively small, every unit cell carries two orbital and two spin degrees of freedom, which results in a substantially large Hilbert space. Moreover, the VMC simulations are performed without imposing any symmetry constraints on the variational parameters; any larger systems than considered here are computationally very expensive. Also this choice of system size is sufficiently commensurate to accommodate a broad range of symmetry-breaking patterns, including 2×12\times1, 2×22\times2, 2×32\times3, 3×3\sqrt{3}\times\sqrt{3}, and 3×33\times3 reconstructions. Moreover, all the calculations correspond to zero temperature.

Hartree-Fock Phase diagram

We study various possible reconstructions of the charge and spin densities of the considered system in an unbiased manner using the unrestricted HF method. In our study, we primarily obtain two types of symmetry-broken phases, which we analyze in detail in the following. In the first case, there is a net imbalance in the spin-↑\uparrow and spin-↓\downarrow population, which results in an out-of-plane magnetization of the ground state leading to breaking of the time-reversal symmetry. Since the spin and valley degrees of freedom are locked, this situation corresponds to a valley polarized (VP) state. In the second case, the spin-↑\uparrow and spin-↓\downarrow sectors are in a coherent superposition, or equivalently, the valleys KK and K′K' mix with each other, resulting in an intervalley coherent (IVC) phase with broken valley U(1)U(1) symmetry of the model.

In the following sections, we present the phase diagrams and details of the charge and spin order of the respective phases in two selected fractional fillings (n=4/3n = 4/3 and n=3/2n = 3/2). The minimum energy state at a given (θ,Δ)(\theta, \Delta) has been confirmed through multiple simulations, each starting with random initial mean fields.

Hole-filling 4/3: phase diagram, charge, and spin order of the phases

The phase diagram of the ground state as a function of θ\theta and Δ\Delta at hole-filling 4/3 is presented in Figure 2(a). We observe that while the VP phase is stabilized at small twist angles, the IVC phase is preferred at large twist angles. The transition from VP to IVC phase occurs around θ≃2.9∘\theta\simeq2.9^\circ.

Figure 2. Phase diagram of hole-filling 4/34/3 case. (a) The phase of the ground state as a function of θ\theta and Δ\Delta. The open black circles denote the phase boundary determined from the HF calculations as a function of the displacement field at selected angles. The phases VP1_1, VP2_2, IVC1_1, and IVC2_2 are characterized in (b), (c), (d), and (e), respectively, for Δ>0\Delta> 0. (b) The 6×66 \times6 lattice system is shown by red (AA sublattice) and green (BB sublattice) circles. The colorbar shows the charge density. The arrows show spin. And the dashed lines depict the reconstructed unitcell. At the bottom, the left inset and the right inset plot the amplitudes of the FT of mean-subtracted charge and spin density, respectively. The details of (c), (d), and (e) are same as (b). (c) Note that the arrows represent the zz-component of spin in (b) & (c) and the in-plane component of spin in (d) & (e).

First, we focus on the range of lower angles and analyze the valley-polarized states which appear in two slightly different forms labeled VP1_1 and VP2_2. The former corresponds to a layer-hybridized situation, while in the latter a non-zero layer-polarization emerges. Figure 2(b) and (Figure 2c) show the electron density and spin texture in the VP1_1 and VP2_2 phases, respectively. The color bar denotes the total charge density and the arrows show the spin direction. Note that the plotted spin component is ⟨S^iz⟩\langle\hat{S}_i^z\rangle, as in the VP phase the in-plane magnetization is homogeneously zero. We find that in both VP1_1 and VP2_2 states each site is occupied by one spin-↑\uparrow electron while the remaining spin-↓\downarrow electrons are distributed throughout the system. This leads to the maximal possible spin polarization, with the spin-↑\uparrow sector forming an inert background due to strong ferromagnetic exchange in the system [Eq. (3)]. In the layer-hybridized VP1_1 phase, these spin-↓\downarrow electrons are equally distributed between AA and BB sublattices, leading to an equal number of doubly occupied sites on both layers [Figure 2(b)]. However, with increasing displacement field due to the appearance of staggered onsite potential [Eq. (2)] a charge transfer from one sublattice to the other is generated and the layer polarized state (VP2_2) becomes stable above some critical value of Δ\Delta. As a result in the layer-polarized VP2_2 phase, all the spin-↓\downarrow electrons occupy either the A(Δ>0)A(\Delta> 0) or B(Δ<0)B(\Delta< 0) sublattice depending on the displacement-field direction. Consequently, in VP2_2 phase, all the doubly occupied sites reside in one layer [Figure 2(c)] only.

In both VP1_1 and VP2_2 states, a 3×3\sqrt{3} \times\sqrt{3} reconstruction is realized, as shown by the unit cells depicted by the dashed line in Figure 2(b) and (Figure 2c). For the sake of completeness, the insets to the figures show amplitudes of the Fourier transforms (FT) of (⟨n^i⟩−nˉ)(\langle\hat{n}_i\rangle-\bar{n}) (left inset) and (⟨S^iz⟩−Sˉz)(\langle\hat{S}_i^z\rangle-\bar{S}_z) (right inset) inside the mBZ. The mean values are subtracted to remove the large peak at k=0k = 0, which does not influence other peaks. The occurrences of the peaks at the mBZ corners (κ,κ′)(\kappa,\kappa') confirm the 3×3\sqrt{3} \times\sqrt{3} reconstruction. It should be noted that the maximal spin polarization of the VP phases takes ⟨S^iz⟩≃12(1−n^i)\langle\hat{S}_i^z\rangle\simeq\frac{1}{2}(1-\hat{n}_i) at every lattice site, hence, the spin/charge modulation is determined by the active spin-↓\downarrow sector. As a result, the spin and the charge degrees of freedom are tied to each other and the subtle interplay between the ferromagnetic term, the intersite Coulomb repulsion term, and the staggered potential tuned by the displacement field determine the resulting charge and spin modulation. While the ferromagnetic term maximizes the spin/valley polarization, the Coulomb repulsion terms have a significant influence on the resultant charge ordering while the displacement field additionally initiates polarization between the two layers.

To further characterize the VP phases, we focus on the spatial distribution of the spin-↓\downarrow electrons only, as the spin-↑\uparrow electrons form an essentially inert background. Figure 3(a) and (Figure 3b) show the spin-↓\downarrow electron distributions in VP1_1 and VP2_2 phases, respectively. Clearly, the occupation of the sites is near 0 or 1 in both phases. Hence the local variance of spin-↓\downarrow electrons ⟨n^i2⟩/(1−⟨n^i⟩)\langle\hat{n}_i^2\rangle/(1-\langle\hat{n}_i\rangle) is strongly suppressed. In the following we examine the average local variance of the spin-↓\downarrow electrons

Figure 3. Classical nature of VP1_1 and VP2_2 phase. (a) Distribution of spin-↓\downarrow electrons at θ=1.4∘\theta= 1.4^\circ and Δ=0\Delta= 0 (VP1_1 phase). (b) Distribution of spin-↓\downarrow electrons at θ=1.4∘\theta= 1.4^\circ and Δ=12\Delta= 12 meV (VP2_2 phase). (c) Average local variance of spin-↓\downarrow electrons as function of θ\theta at Δ=0\Delta= 0 (squares) and Δ=12\Delta= 12 meV (circles), i.e. in VP1_1 and VP2_2 phase, respectively. (d) Plots the log of the ratios of the Coulomb interaction potentials and the bandwidth (WW) of the top moiré valence band as a function of θ\theta. UU is the onsite interaction. V1V_1, V2V_2, and V3V_3 denote the interaction between 1st1^{\mathrm{st}}, 2nd2^{\mathrm{nd}}, and 3rd3^{\mathrm{rd}} nearest neighbours. XX is used as a common notation for onsite and intersite Coulomb potentials.

F↓loc=1Nsite∑i(⟨n^i↓2⟩−⟨n^i↓⟩2).(11)F_{\downarrow}^{\mathrm{loc}} = \frac{1}{N_{\mathrm{site}}}\sum_i\left(\langle\hat{n}_{i\downarrow}^{2}\rangle-\langle\hat{n}_{i\downarrow}\rangle^{2}\right). \tag*{(11)}

Note that sites with occupation 0 or 1 does not contribute to the variance, whereas fractionally occupied sites produce a finite contribution. For the considered 6×66 \times6 system, we have 24 spin-↓\downarrow electrons and the maximum value of F↓locF_{\downarrow}^{\mathrm{loc}} is ≃0.22\simeq0.22, which is obtained when all these electrons are equally distributed over all the 72 sites. Figure 3(c) plots F↓locF_{\downarrow}^{\mathrm{loc}} as a function of θ\theta in the VP1_1 phase (Δ=0\Delta= 0) denoted by black squares. We note that F↓locF_{\downarrow}^{\mathrm{loc}} remains small in magnitude over the entire twist range. This demonstrates weak delocalization induced by significant electron-electron interactions. Figure 3(d) plots the log of the interaction-to-bandwidth ratios as a function of twist. Here, the bandwidth (WW) of the topmost noninteracting moiré valence band is considered, within which the Fermi level of the noninteracting system lies for the hole fillings considered in the study. The large values of the ratios at low twist regime support the dominance of the interaction over the kinetic energy. We further observe that the variance is more suppressed in the VP2_2 phase as shown by the black circles in Figure 3(c). This occurs because, due to the finite layer polarization in VP2_2 phase, the spin-↓\downarrow electrons are largely restricted to the AA-sublattice (for Δ>0\Delta> 0) leading to the suppression of the relevant kinetic-energy scale. This consequently enhances the effective interaction-to-bandwidth ratios, resulting in even smaller local charge fluctuations. With an increase of θ\theta, the variance increases, signaling a gradual departure from the classical limit. This behavior is consistent with the decreasing relative importance of interactions at larger twist angles [Fig. 3(d)], which allows stronger charge delocalization and hence larger fluctuations in the local occupations. We therefore conclude that the electrons in the VP phase exhibit weak charge delocalization and the corresponding density distribution resembles classical real-space configurations at zero temperature, where the site occupation is restricted to 0 or 1. However, with increasing twist, the system progressively

becomes more quantum in nature.

Next, we move on to the discussion of the IVC phases found at larger θ\theta in the phase diagram of Figure 2(a). Figure 2(d) and (Figure 2e) show the electron density and spin texture of the layer-hybridized IVC1_1 and layer-polarized IVC2_2, respectively. Since in both IVC states the charge is equally distributed between the two valleys (⟨n^i↑⟩=⟨n^i↓⟩\langle\hat{n}_{i\uparrow}\rangle= \langle\hat{n}_{i\downarrow}\rangle) we have ⟨S^iz⟩=0\langle\hat{S}^{z}_{i}\rangle= 0. Therefore, in Figure 2(d) and (Figure 2e) we plot the in-plane spin component ⟨S^i∥⟩=(⟨S^ix⟩,⟨S^iy⟩)\langle\hat{\mathbf{S}}_{i}^{\parallel}\rangle= (\langle\hat{S}^{x}_{i}\rangle,\langle\hat{S}^{y}_{i}\rangle) depicted by the black arrows. In IVC1_1, the in-plane spins on AA and BB sublattices are oppositely oriented, thus cancelling each other and giving rise to a sublattice-compensated IVC order, resulting in zero in-plane magnetization. Because of the layer-hybridized nature, in IVC1_1, the doubly occupied sites are equally distributed on both AA and BB sublattices. In contrast, in the IVC2_2 phase, all the doubly occupied sites move to a single sublattice (AA for Δ>0\Delta> 0) which is due to the staggered onsite potential induced by the displacement field. As a result, all sites belonging to the sublattice BB have a single electron, resulting in a uniform spin in the BB-sublattice network, while the sublattice AA features fewer sites with nonzero spin moment. Therefore, cancellation of in-plane spins between AA and BB sublattices is incomplete, and a net in-plane spin moment develops. We call this IVC order as ferrimagnetic-IVC. Similar to the VP phases, the IVC phases also show 3×3\sqrt{3}\times\sqrt{3} reconstruction in both the charge and spin order. In Figure 2(d) and (Figure 2e), the reconstructed unit cell is shown by the dashed lines. The left-inset and the right-inset show the amplitudes of the mean-subtracted-FTs of charge and in-plane spin densities, respectively. For in-plane spin at site ii we use Si+=Six+iSiyS_i^{+}=S_i^{x}+iS_i^{y}. The peaks in the FT amplitudes occur at the mBZ corners (κ,κ′)(\kappa,\kappa'), confirming the 3×3\sqrt{3}\times\sqrt{3} order.

We now focus on the phase transitions among various VP and IVC phases presented in the phase diagram of Figure 2(a). Figure 4(a) characterizes the VP-to-IVC phase transition as a function of θ\theta at Δ=0\Delta= 0. In the following, we define the VP and the IVC order parameters.

Figure 4. Phase transitions in the 4/3 hole-filling case. (a) VP to IVC phase transition as a function of θ\theta at Δ=0\Delta= 0. The left axis shows the VP (black circles) and the IVC (red squares) order parameters (see text). The right axis shows EgE_g (blue triangles). (b) Shows PlayerP_{\mathrm{layer}} (black circles) on the left axis and the EgE_g (blue triangles) on the right axis for θ=1.4∘\theta= 1.4^\circ. (c) Shows PlayerP_{\mathrm{layer}} (black circles) on the left axis and the magnetization per site (blue triangles) on the right axis for θ=3.0∘\theta= 3.0^\circ. The transitions are VP1_1-VP2_2 and IVC1_1-IVC2_2 for θ=1.4∘\theta= 1.4^\circ and 3.0∘3.0^\circ, respectively. For the phases refer to Figure 2. (d) The ratio of the amplitudes of the FT peaks of the mean-subtracted charge density at κ\kappa and κ′\kappa' in the mBZ. This plot holds for any θ\theta; only the location of the jump follows the transition in PlayerP_{\mathrm{layer}}. The amplitudes of the FTs are plotted in the insets of Figure 2.

ΦVP=n↑−n↓n↑+n↓,ΦIVC=2Ne∑i∣Si+∣.(12)\begin{aligned} \Phi_{\mathrm{VP}} &= \frac{n_{\uparrow}-n_{\downarrow}}{n_{\uparrow}+n_{\downarrow}},\\ \Phi_{\mathrm{IVC}} &= \frac{2}{N_e}\sum_i \left|S_i^{+}\right|. \tag*{(12)} \end{aligned}

Here, n↑n_{\uparrow}, n↓n_{\downarrow}, and Ne(=n↑+n↓)N_e(= n_{\uparrow}+n_{\downarrow}) are the numbers of spin-↑\uparrow, spin-↓\downarrow, and total electrons, respectively. Across the transition point at θ≃2.9∘\theta\simeq2.9^\circ, the VP order parameter (black circles) shows sudden collapse with concomitant emergence of the IVC order parameter (red squares). The discontinuous jumps in the order parameters reveals the first-order nature of the transition. The single-particle bandgap (EgE_g) at the Fermi level (blue triangles) also shows a sudden jump across the transition. Moving from VP to IVC, the HF Hamiltonian gains valley-off-diagonal terms. These can open additional hybridization gaps resulting in abrupt increase of the bandgap at the transition point. It should be noted that in the considered situation of fractional band filling, the single-particle Fermi surfaces do not correspond to nesting vectors which would lead to the observed 3×3\sqrt{3}\times\sqrt{3} reconstruction. Also, there is no significant change of the single particle Fermi surface across the transition between the VP and IVC states. Hence, the obtained ordering should not be considered as resulting from the momentum-space instability but rather a real-space phenomenon reflecting the competition between different types of interaction terms and the kinetic energy. In the VP state which is close to the classical limit the interaction terms dominate. However, with increasing twist angle the kinetic energy becomes gradually more important [Figure 3(d)] leading to increased role of fluctuations. At some critical value of θ\theta a phase transition takes place to the IVC state which is more quantum in nature.

Next, we discuss the influence of the displacement field, which introduces the staggered onsite potential Δ\Delta and leads to layer polarization defined in the following manner

Player=∣nA−nB∣(nA+nB),(13)P_{\mathrm{layer}}=\frac{\left|n_A-n_B\right|}{\left(n_A+n_B\right)}, \tag*{(13)}

where nA(nB)n_A(n_B) is the total number of electrons on the A(B)A(B) sublattice. Figure 4(b) characterizes the transition from VP1_1 to VP2_2 at θ=1.4∘\theta= 1.4^\circ. Across the transition at Δ≃3\Delta\simeq3 meV, PlayerP_{\mathrm{layer}} (left axis) discontinuously jumps to a nonzero value, indicating the first-order nature of the transition. The EgE_g (right axis) also shows a rapid change near the transition. Additionally, Figure 4(c) presents the transition from IVC1_1 to IVC2_2 state at θ=3.0∘\theta= 3.0^\circ. The first-order nature of the transition is evident from the abrupt changes in PlayerP_{\mathrm{layer}} (left axis). On the right axis we show magnetization per site denoted by M/Nsite(=m)M/N_{\mathrm{site}}(=m). The jump from zero to finite value of mm, confirms the emergence of net in-plane magnetization in the IVC2_2 phase. All the results taken together, we conclude that both VP1_1-VP2_2 and IVC1_1-IVC2_2 transitions fall in the category of first-order transition.

Although all the obtained phases in the phase diagram of Figure 2(a) show 3×3\sqrt{3}\times\sqrt{3} reconstruction, the layer-polarization transition influences the relative amplitudes of the FT peaks at κ\kappa and κ′\kappa'. In the layer-hybridized regime (VP1_1 and IVC1_1) both sublattices (AA and BB) individually contain the 3×3\sqrt{3}\times\sqrt{3} order. But in the layer-polarized regime (VP2_2 and IVC2_2), one of the sublattices (BB for Δ>0\Delta> 0) becomes uniform in terms of the charge and spin densities, and the 3×3\sqrt{3}\times\sqrt{3} order is entirely contained in the other layer (AA for Δ>0\Delta> 0). As a result, the FTs in layer-hybridized regime show interference between the AA and BB sublattices, producing unequal amplitudes at κ\kappa and κ′\kappa', which can be seen in the insets of Figure 2(b) and (Figure 2d). In contrast, in the layer-polarized regime, the absence of such interference results in equal amplitudes at κ\kappa and κ′\kappa', as is clear from the insets of Figure 2(c) and (Figure 2e). Figure 4(d) plots the ratio of the amplitudes for the charge density at κ\kappa and κ′\kappa' as a function of Δ\Delta. The jump in the ratio from 0.5 to 1 coincides with layer-polarization transition. Such a jump is true for any θ\theta; only the location of the jump follows the transition in PlayerP_{\mathrm{layer}}.

Hole-filling 3/2: phase diagram, charge, and spin order of the phases

We now discuss the 3/2 hole-filling case. Figure 5(a) presents the ground state phase diagram as a function of θ\theta and Δ\Delta. This phase diagram shares some similarities with the previous one corresponding to filling 4/3 in Figure 2(a). Like the 4/3 filling case, the VP phases possess maximal valley polarization and the VP-IVC transition occurs at θ≃2.9∘\theta\simeq2.9^\circ. Despite these similarities, the phase diagram of 3/2 case has some new features. A few new orders, 2×12\times1 and 2×22\times2, appear alongside the ubiquitous 3×3\sqrt{3}\times\sqrt{3} order as we discuss further on.

Figure 5. Phase diagram of hole-filling 3/23/2 case. (a) The phase of the ground state as a function of θ\theta and Δ\Delta. The open black circles denote the phase boundary determined from the HF calculations as a function of the displacement field at selected angles. The phases VP~1\widetilde{\mathrm{VP}}_1, VP3\mathrm{VP}_3, VP4\mathrm{VP}_4, and VP5\mathrm{VP}_5 are characterized in (b), (c), (d), and (e), respectively, for Δ>0\Delta> 0. (b) The 6×66 \times6 lattice system is shown by red (AA sublattice) and green (BB sublattice) circles. The colorbar shows the charge density. The arrows show spin. And the dashed lines depict the reconstructed unitcell. At the bottom, the left inset and the right inset plot the amplitudes of the FT of mean-subtracted charge and spin density, respectively. The details of (c), (d), and (e) are the same as those of (b). Note that the arrows represent the zz-component of spin.

Compared to the 4/3 hole-filling case, the present case has 1/6 more holes (or equivalently 1/6 fewer electrons) per unit cell. Therefore, the layer-hybridized VP1_1 phase can be understood as a hole-doped version of the VP1_1 phase of Figure 2(a). In the VP1_1 phase, the spin-↓\downarrow electrons are nearly classically localized, with site occupations close to either 0 or 1. Upon increasing the hole filling to 3/2, some of the sites that were nearly fully occupied by spin-↓\downarrow electrons become approximately half occupied. The additional holes are therefore distributed over twice as many sites, and these half-occupied sites develop a 2×12\times1 reconstruction superimposed on the underlying 3×3\sqrt{3}\times\sqrt{3} charge order inherited from the VP1_1 phase, as can be seen in Figure 5(b). The dashed lines show the 2×12\times1 reconstruction of the underlying 3×3\sqrt{3}\times\sqrt{3} order. Note that the added holes are distributed equally over the AA and BB sublattices because of the zero layer polarization. The amplitudes of mean-subtracted charge (left inset) and spin density (right inset) FTs show peaks at κ\kappa, κ′\kappa', and μ\mu points. The peaks at κ\kappa and κ′\kappa' originate from 3×3\sqrt{3}\times\sqrt{3} ordering and the peak at μ\mu confirm the further 2×12\times1 reconstruction. Similarly as for the case of 4/3 hole-filling, with an increase in Δ\Delta, the layer-hybridized VP1_1 phase reorganizes into layer-polarized VP3_3 phase [Figure 5(a)], which is visualized in Figure 5(c). As can see in this phase, all the minority spin-↓\downarrow electrons are transferred to the AA sublattice (for Δ>0\Delta> 0). Therefore, while the BB sublattice becomes uniform in terms of the charge and spin density, the actual order is contained in the AA sublattice network. In the VP3_3 phase, 2/3 of the minority spins form a 3×3\sqrt{3}\times\sqrt{3} pattern, and the remaining 1/3 form a 2×12\times1 reconstruction on top of it. The peak positions of the amplitudes of the mean-subtracted FTs shown in the insets again confirm the mentioned real-space order.

In the intermediate twist regime, the VP phase reorganizes itself into a genuinely new order rather than a hole-doped version of 4/34/3 filling. In the layer-hybridized regime a 2×22 \times2 ordered phase termed as VP4\mathrm{VP}_4 appears as shown in Figure 5(d). The quadrupling of the original unit cell is indicated by the dashed lines. Similarly as before, the amplitudes of FTs of the mean-subtracted charge (left inset) and spin (right inset) density show peaks at all three μ\mu points which is consistent with the 2×22 \times2 real space order. With the increase of Δ\Delta, all the spin-↓\downarrow electrons of VP4\mathrm{VP}_4 moves to AA sublattice and gives

rise to 2×12 \times1 ordered VP5\mathrm{VP}_5 phase, Figure 5(e). The doubling of the unit cell along one direction gives rise to the peak at a single μ\mu point in the FT amplitudes in the insets. Further increase of Δ\Delta induces more layer-polarization, and the system goes into VP2\mathrm{VP}_2, Figure 5(a), which is the hole-doped version of VP2\mathrm{VP}_2 discussed in connection with Figure 2. This phase shows reconstruction of the underlying 3×3\sqrt{3} \times\sqrt{3} order (see supplementary material [38]).

Similar to the 4/34/3 hole-filling case [Figure 2(a)], in the present case, the system leaves the VP phase and enters the IVC state at θ≃2.9∘\theta\simeq2.9^\circ [Figure 5(a)]. Both the layer-hybridized IVC‾1\overline{\mathrm{IVC}}_1 and layer-polarized IVC‾2\overline{\mathrm{IVC}}_2 show reconstruction of the underlying 3×3\sqrt{3} \times\sqrt{3} order. Further discussions of IVC‾1\overline{\mathrm{IVC}}_1 and IVC‾2\overline{\mathrm{IVC}}_2 are relegated to the supplementary material [38].

Now we further elaborate on the nature of the VP‾1\overline{\mathrm{VP}}_1 and VP3\mathrm{VP}_3 phases. These phases arise upon doping holes to VP1\mathrm{VP}_1 and VP2\mathrm{VP}_2 phases shown in Figure 2(a), in the corresponding ranges of θ\theta and Δ\Delta. As discussed in Figure 3, the VP1\mathrm{VP}_1 and VP2\mathrm{VP}_2 phases exhibit a pronounced classical character, with the local electron occupations close to the integer values. However, with additional hole doping, the system departs from this nearly classical limit. Doping holes only reduces the number of spin-↓\downarrow electrons, as the spin-↑\uparrow sector remains inert due to maximal spin polarization. Figure 6(a) shows the spin-↓\downarrow occupations in the VP3\mathrm{VP}_3 phase at θ=1.2∘\theta= 1.2^\circ and Δ=18\Delta= 18 meV. Three distinct groups of sites can be identified, with occupations close to 11, 0.50.5, and 00. The sites with occupation 11 remain strongly localized with negligible coherence with the nearby sites. Whereas the sites with occupation 0.50.5, forms dimer like entities. A given half-occupied site shows coherence only with the nearest half-occupied site, ie the electron is confined within these two sites. The dashed oval in Figure 6(a) represents such a dimer. The state therefore possesses two qualitatively distinct carrier components: a set of nearly pinned carriers and a set of two-site coherent dimer like charge objects. This is reminiscent of a pinball liquid [48, 49], in which a subset of carriers remain pinned to sites, while the rest of the carriers with higher mobility form the ball component. However, the VP3\mathrm{VP}_3 phase remains gapped as found in HF calculations and is therefore insulating rather than liquid-like. Hence, we refer to this state as a two-component GWC [20]. Figure 6(b) plots the average spin-↓\downarrow occupations of the pinned, ball-like, and nearly empty sites as a function of θ\theta. Their occupations vary only weakly with twist angle, indicating that the two-component character remains robust throughout the range in which the VP3\mathrm{VP}_3 phase is stable.

Figure 6. Signature of the two-component GWC in VP‾1\overline{\mathrm{VP}}_1 and VP3\mathrm{VP}_3 phase. (a) Distribution of spin-↓\downarrow electrons at θ=1.2∘\theta= 1.2^\circ and Δ=18\Delta= 18 meV (VP3\mathrm{VP}_3 phase). (b) Average occupation of pinned, ball, and empty sites as a function of θ\theta at Δ=18\Delta= 18 meV (VP3\mathrm{VP}_3 phase). (c) Distribution of spin-↓\downarrow electrons at θ=1.2∘\theta= 1.2^\circ and Δ=0\Delta= 0 meV (VP‾1\overline{\mathrm{VP}}_1 phase). (d) Average occupation of pinned, ball, and empty sites as a function of θ\theta at Δ=0\Delta= 0 (VP‾1\overline{\mathrm{VP}}_1 phase). The dashed ovals in (a) and (c) represent the two-site coherent dimers.

Figure 6(c) shows the spin-↓\downarrow occupation pattern in the VP‾1\overline{\mathrm{VP}}_1 phase at θ=1.2∘\theta= 1.2^\circ and Δ=0\Delta= 0. In contrast to the VP3\mathrm{VP}_3 phase, the occupation of the most strongly occupied sites is already reduced from unity, while the nearly empty sites acquire a small but finite occupation. Figure 6(d) shows the average occupation of the pinned, ball-like, and nearly empty sites as a function of θ\theta. While the occupation of the ball-like sites remains close to 0.50.5, we observe a gradual charge transfer from the most highly occupied sites to the least occupied sites as θ\theta increases. This behavior indicates that the two-component character of the VP‾1\overline{\mathrm{VP}}_1 phase weakens rapidly with increasing twist angle. The observed trend further suggests that a more pronounced two-component GWC may emerge for twist angles below θ=1.2∘\theta= 1.2^\circ.

We therefore conclude that, while the parent phases VP1\mathrm{VP}_1 and VP2\mathrm{VP}_2 are close to classical real-space occupation patterns, upon hole doping the obtained phases VP‾1\overline{\mathrm{VP}}_1 and VP3\mathrm{VP}_3 display substantially stronger charge delocalization and exhibit signatures of a two-component GWC.

Next, we move on to the characterization of the transitions between the phases shown in Figure 5(a). Namely, Figure 7(a) presents the evolution of the order parameters across the VP-IVC transition induced by tuning θ\theta at Δ=0\Delta= 0. The abrupt changes of the VP (black circles) and IVC (red squares) order parameters which appear at the critical value indicate first order phase transition with increasing angle. In the figure, we also provide the single-particle bandgap at the Fermi level, EgE_g (blue triangles), which shows two jumps as a function of θ\theta. The first one corresponds to the VP‾1\overline{\mathrm{VP}}_1-VP4\mathrm{VP}_4 transition while the second marks the previously discussed VP-IVC transition. Furthermore, as can be seen from Figure 7(b) and (Figure 7c) by increasing Δ\Delta one enhances the layer polarization (black circles) in a straightforward manner, which again leads to abrupt changes in EgE_g (blue triangles) at the critical values corresponding to the first order phase transitions. It should be noted that the two jumps in EgE_g, which appear for θ=2.5∘\theta= 2.5^\circ [Figure 7(c)] correspond to VP4_4-VP5_5 and VP5_5-VP2_2 transitions. For the case of 3∘3^\circ twist angle, a rapid increase of total in-plane magnetization appears at the transition point, as we show in Figure 7(d).

Figure 7. Phase transitions in the 3/2 hole-filling case. (a) VP to IVC phase transition as a function of θ\theta at Δ=0\Delta= 0. The left axis shows the VP (black circles) and the IVC (red squares) order parameters. The right axis shows EgE_g (blue triangles). (b) shows PlayerP_{\mathrm{layer}} (black circles) on the left axis and the EgE_g (blue triangles) on the right axis for θ=1.3∘\theta= 1.3^\circ. (c) same as (b) for 2.5∘2.5^\circ. (d) shows PlayerP_{\mathrm{layer}} (black circles) on the left axis and magnetization per site (blue triangles) on the right axis for θ=3.0∘\theta= 3.0^\circ. For the phases refer to Figure 5.

Universal features of the HF phase diagrams

Here we comment on some general features of the phase diagrams of the 4/34/3 and 3/23/2 hole-filling cases. The underlying principles that determine the phases and govern the phase transitions are discussed as well.

  1. In both 4/34/3 and 3/23/2 hole-filling, while the VP phase stabilizes at low twist angles, the IVC phase is favored at high twist. The transition from VP to IVC occurs at θ≃2.9∘\theta\simeq2.9^\circ for both fillings. The VP phase is exchange-driven, originating from the term −2JijS^izS^jz-2J_{ij}\hat{S}^{z}_{i}\hat{S}^{z}_{j} in Eq. (3). This transition reflects a reorganization of the HF state in which the energy gain from intervalley coherence becomes larger than the easy-axis exchange energy of the VP state. Once the IVC phase appears, the HF Hamiltonian gains valley-off-diagonal terms. These can lead to an additional hybridization gap leading to a decrease in the total energy of the system. This explains the jumps in EgE_g across the VP to IVC transitions in Figure 4(a) and Figure 7(a).

  1. In the VP phase, the spin polarization is completely saturated. The majority spin occupying each site acts as an inert background. The minority spin electrons are the actual degrees of freedom, whose reorganization determines the charge order of the phase. This also explains why the spin order is tied to the charge order in the VP phase.

The phase diagrams reveal the ubiquity of the 3×3\sqrt{3} \times\sqrt{3} order. For hole-filling 4/34/3, the entire phase diagram is occupied by 3×3\sqrt{3} \times\sqrt{3} reconstructions irrespective of the VP or IVC phase. Similarly, for hole-filling 3/23/2, the same order appears everywhere except in the intermediate range of θ\theta-Δ\Delta. The reason behind this universality is twofold: commensurability and triangular lattice formation. It is well known that through triangular lattice formation, which leads to 3×3\sqrt{3} \times\sqrt{3} reconstruction, interacting particles maintain equal and maximum distances from each other, resulting in a minimum interaction energy. For hole-filling 4/34/3, commensurability further reinforces the order. In this case, the average number of electrons per unit cell is 8/38/3, which is commensurable with the 3×3\sqrt{3} \times\sqrt{3} order. So both triangular lattice formation and commensurability work in favor of 3×3\sqrt{3} \times\sqrt{3} reconstruction. Unlike hole-filling 4/34/3, the extra dimension of commensurability is lost to some extent for the hole-filling 3/23/2 case, as the average number of electrons per unit cell is 5/25/2. Equivalently, each unit cell has 1/61/6 extra holes compared to the 4/34/3 hole-filling case. These extra holes are accommodated appropriately through a 2×12 \times1 or 2×22 \times2 reconstruction of already existing 3×3\sqrt{3} \times\sqrt{3} order.

So we see that 3×3\sqrt{3} \times\sqrt{3} is favored by the energy minimization principle and is further reinforced by the commensurability condition. These underlying principles that govern the order of the phases are independent of the size of the system. This tells us that the phases and phase diagrams obtained in our work for the finite 6×66 \times6 system will also hold for larger system sizes.

Variational Monte Carlo Results

We have further employed the VMC technique to study the effects of the interactions beyond the mean-field considerations. We performed VMC simulations using both random initialization of variational parameters and initialization with the results of the converged solutions of the unrestricted HF calculations. In general, the lowest optimized energy is obtained when the converged HF results are used for initialization, which we report here. The simulations confirm the phases obtained with the HF method for both 4/34/3 and 3/23/2 hole-filling cases. In the paper, we primarily discuss the results of 4/34/3 hole-filling case.

Although the VMC simulations produced ordered phases similar to those obtained within the HF calculations, the ground-state energies obtained in the two methods differ in some parameter range. The importance of correlations beyond the mean-field level can be directly assessed by comparing the ground state energies obtained from the two methods. For this purpose, we study the correlation energy defined in the following manner,

ΔEcorr=EVMC−EHF,(14)\Delta E_{\mathrm{corr}} = E_{\mathrm{VMC}} - E_{\mathrm{HF}}, \tag*{(14)}

where EVMCE_{\mathrm{VMC}} and EHFE_{\mathrm{HF}} denote the total energies of the system obtained in the VMC and the HF simulations, respectively. A negative value of ΔEcorr\Delta E_{\mathrm{corr}} therefore indicates an energy gain induced by correlations beyond mean-field. Figure 8 (a) shows ΔEcorr\Delta E_{\mathrm{corr}} per site as a function of θ\theta at Δ=0\Delta= 0. In the range of 1.2∘−2.8∘1.2^{\circ} - 2.8^{\circ}, where the ground state phase is VP1_1, EVMCE_{\mathrm{VMC}} agrees very well with EHFE_{\mathrm{HF}}. The small positive value of ΔEcorr\Delta E_{\mathrm{corr}} arises from the statistical uncertainty of the VMC simulations. This result is consistent with the nearly classical character of the VP1_1 phase arising from high interaction-to-bandwidth ratio. The corresponding localized charge configuration is captured well by the unrestricted HF solution, leaving only a very small room for improvement by the VMC wave function. With increasing θ\theta, the bandwidth increases relative to the interaction scale, enhancing charge delocalization. At θ≃2.9∘\theta\simeq2.9^{\circ}, the system enters the IVC1_1 phase, characterized by a coherent superposition of the two spin (valley) degrees of freedom. In this regime, EVMCE_{\mathrm{VMC}} becomes lower than EHFE_{\mathrm{HF}}, indicating a gain in correlation energy as can be seen in Figure 8(a). We further note that the gain rapidly increases with θ\theta, which implies growing importance of correlations beyond HF in the IVC phase.

Correlation energy plots for the 4/3 hole-filling case

Figure 8. Correlation energy for 4/3 hole-filling case. (a) Correlation energy per site as function of θ\theta at Δ=0\Delta= 0. (b) and (c) show the correlation energies per site as function of Δ\Delta for θ=1.4∘\theta= 1.4^{\circ} and 3.0∘3.0^{\circ}, respectively.

We previously noted that a sufficiently strong displacement field induces finite layer polarization. The increased layer polarization suppresses interlayer hybridization leading to an enhanced charge localization. Consequently, the energetic importance of fluctuations beyond HF decreases in the layer-polarized regime. Figure 8(b) plots the correlation energy at θ=1.4∘\theta= 1.4^{\circ} as function of Δ\Delta. Since the VP phase is strongly classical in nature, HF and VMC energies agree and the layer-polarization induces no further effect. The small positive value is due to the statistical uncertainty of the VMC procedure. Whereas in the IVC phase at θ=3.0∘\theta= 3.0^{\circ} in Figure 8(c), the reduced fluctuation in the IVC2_2 phase at higher Δ\Delta results in lower ΔEcorr\Delta E_{\mathrm{corr}} as a result EVMCE_{\mathrm{VMC}} approaches EHFE_{\mathrm{HF}}. Therefore, we conclude that the gain in correlation energy is progressively suppressed as the system evolves toward the layer-polarized state.

Next, we study how rigid the obtained ordered phases are within both HF and VMC approaches. We note that,

since no translational-symmetry restoration is imposed during the optimization of the variational wavefunction, the final optimized VMC states can retain the broken translational symmetry. Consequently, the static order of the converged phases can be extracted from the single-particle Greens functions like local charge (nin_i) and spin (Si\mathbf{S}_i) densities. All phases obtained for the 4/3 hole-filling case show 3×3\sqrt{3} \times\sqrt{3} order, which leads to dominant peaks at the mBZ corners (κ,κ′\kappa,\kappa') in the FT of the density modulations. Therefore, we define the strength of the static charge and spin orders order as

mρ=∣n(κ)∣2+∣n(κ′)∣2,m_{\rho} = |n(\kappa)|^2 + |n(\kappa')|^2,
mSz=∣Sz(κ)∣2+∣Sz(κ′)∣2,m_{S_z} = |S_z(\kappa)|^2 + |S_z(\kappa')|^2,
mS+=∣S+(κ)∣2+∣S+(κ′)∣2.(15)m_{S_+} = |S_+(\kappa)|^2 + |S_+(\kappa')|^2. \tag*{(15)}

To quantify the fluctuation around the mean-field order, we consider the connected 2-particle correlation functions Nconn(q)N^{\mathrm{conn}}(\mathbf{q}), Szzconn(q)S_{zz}^{\mathrm{conn}}(\mathbf{q}), and S+−conn(q)S_{+-}^{\mathrm{conn}}(\mathbf{q}), which characterize the charge, out-of-plane spin, and in-plane spin fluctuations, respectively. These quantities measure the strength of the fluctuation of the respective order at wavevector q\mathbf{q} and are defined as follows

Nconn(q)=1Nsite∑i,jeiq⋅(ri−rj)(⟨n^in^j⟩−⟨n^i⟩⟨n^j⟩),N^{\mathrm{conn}}(\mathbf{q}) = \frac{1}{N_{\mathrm{site}}} \sum_{i,j} e^{i\mathbf{q}\cdot(\mathbf{r}_i-\mathbf{r}_j)} \left(\langle\hat{n}_i\hat{n}_j\rangle- \langle\hat{n}_i\rangle\langle\hat{n}_j\rangle\right),
Szzconn(q)=1Nsite∑i,jeiq⋅(ri−rj)(⟨S^izS^jz⟩−⟨S^iz⟩⟨S^jz⟩),S_{zz}^{\mathrm{conn}}(\mathbf{q}) = \frac{1}{N_{\mathrm{site}}} \sum_{i,j} e^{i\mathbf{q}\cdot(\mathbf{r}_i-\mathbf{r}_j)} \left(\langle\hat{S}_i^z\hat{S}_j^z\rangle- \langle\hat{S}_i^z\rangle\langle\hat{S}_j^z\rangle\right),
S+−conn(q)=1Nsite∑i,jeiq⋅(ri−rj)(⟨S^i+S^j−⟩−⟨S^i+⟩⟨S^j−⟩).(16)S_{+-}^{\mathrm{conn}}(\mathbf{q}) = \frac{1}{N_{\mathrm{site}}} \sum_{i,j} e^{i\mathbf{q}\cdot(\mathbf{r}_i-\mathbf{r}_j)} \left(\langle\hat{S}_i^+\hat{S}_j^-\rangle- \langle\hat{S}_i^+\rangle\langle\hat{S}_j^-\rangle\right). \tag*{(16)}

We define the total fluctuation by summing the connected 2-particle correlation functions over all wavevector modes, which is essentially the total local variance of the corresponding quantity. Moreover, we define the order-fluctuation ratio (OFR), denoted by η\eta, as follows

ηρ=mρ1Nsite∑qNconn(q)+mρ,\eta_{\rho} = \frac{m_{\rho}}{\frac{1}{N_{\mathrm{site}}}\sum_{\mathbf{q}} N^{\mathrm{conn}}(\mathbf{q}) + m_{\rho}},
ηSz=mSz1Nsite∑qSzzconn(q)+mSz,\eta_{S_z} = \frac{m_{S_z}}{\frac{1}{N_{\mathrm{site}}}\sum_{\mathbf{q}} S_{zz}^{\mathrm{conn}}(\mathbf{q}) + m_{S_z}},
ηS+=mS+1Nsite∑qS+−conn(q)+mS+.(17)\eta_{S_+} = \frac{m_{S_+}}{\frac{1}{N_{\mathrm{site}}}\sum_{\mathbf{q}} S_{+-}^{\mathrm{conn}}(\mathbf{q}) + m_{S_+}}. \tag*{(17)}

As can be seen from the equations above, the OFRs compare the strength of the static order and the fluctuations present in a given phase. The parameter η\eta is defined to take values between 0 and 1, where η≈1\eta\approx1 represents a phase dominated by static-order, while η≈0\eta\approx0 implies a phase dominated by fluctuations.

Analysis of the HF and VMC results show that fluctuations of the charge order are generally the strongest at the ordering wavevector, i.e., at q=κ,κ′\mathbf{q} = \kappa,\kappa', throughout the entire twist and displacement field range studied here. Figure 9(a) shows the Nconn(q)N^{\mathrm{conn}}(\mathbf{q}) at θ=1.4∘\theta= 1.4^{\circ} and Δ=0\Delta= 0 in the mBZ clearly revealing the pronounced peak at κ′\kappa'. The corresponding charge OFR as a function of θ\theta is shown in Figure 9(b). Within the VP1_1 phase, which is stable over 1.2∘≤θ≤2.8∘1.2^\circ\leq\theta\leq2.8^\circ, the charge order shows high values of OFR both within the HF and VMC methods, implying a static order dominated crystal. However, with increasing angle, the ordering gradually softens. Upon entering the IVC1_1 phase, θ≥2.9∘\theta\geq2.9^\circ, the two methods lead to significantly different behavior, with the VMC showing a much lower OFR or equivalently softer crystal dominated by fluctuations. This result again shows the highly classical character of the VP1_1 phase and the increased quantum nature of the IVC phase, where the importance of correlations beyond HF leads to stronger fluctuations.

Figure 9. OFR of the charge and spin orders for 4/3 hole-filling case as function of twist at Δ=0\Delta= 0. (a) Nconn(q)N^{\mathrm{conn}}(\mathbf{q}) at θ=1.4∘\theta= 1.4^\circ. (b) Charge OFR ηρ\eta_{\rho} as a function of θ\theta. (c) Szzconn(q)S_{zz}^{\mathrm{conn}}(\mathbf{q}) at θ=1.4∘\theta= 1.4^\circ. (d) OFR of the out-of-plane spin order ηSz\eta_{S^z} in the VP1_1 phase. (e) S+−conn(q)S_{+-}^{\mathrm{conn}}(\mathbf{q}) at θ=3.0∘\theta= 3.0^\circ. (f) OFR of the in-plane spin order ηS+\eta_{S^+} in the IVC1_1 phase. In (b), (d), and (f) HF and VMC results are denoted by black circles and red squares, respectively.

Similar to charge order, the dominant fluctuation of the out-of-plane spin order also occurs at the ordering wavevectors. Figure 9(c) shows Szzconn(q)S_{zz}^{\mathrm{conn}}(\mathbf{q}) at θ=1.4∘\theta= 1.4^\circ and Δ=0\Delta= 0, where the peak is clearly seen at κ′\kappa'. The q=0\mathbf{q} = 0 component is proportional to the variance of the total out-of-plane spin, Szzconn(q=0)=1Nsite(⟨(S^ztot)2⟩−⟨S^ztot⟩2)S_{zz}^{\mathrm{conn}}(\mathbf{q} = 0) = \frac{1}{N_{\mathrm{site}}}\left(\langle(\hat{S}^{\mathrm{tot}}_z)^2\rangle- \langle\hat{S}^{\mathrm{tot}}_z\rangle^2\right), where S^ztot=∑iS^zi\hat{S}^{\mathrm{tot}}_z = \sum_i \hat{S}^i_z. Its small magnitude therefore reflects suppressed fluctuations of S^ztot\hat{S}^{\mathrm{tot}}_z, indicating that the ground state is close to an eigenfunction of the S^ztot\hat{S}^{\mathrm{tot}}_z. Figure 9(d) plots the OFR of the out-of-plane spin order at Δ=0\Delta= 0 over the stability range of the VP1_1 phase. Like the charge OFR of Figure 9(b), HF (black circles) and VMC (red squares) results are in close agreement. The order is highly rigid at low twist and gradually becomes softer with θ\theta. This effect is induced by the reduced interaction-to-bandwidth ratio, which leads to higher fluctuations around the static order for higher angles.

The character of the fluctuations of the in-plane spin order is qualitatively very different. S+−conn(q)S_{+-}^{\mathrm{conn}}(\mathbf{q}) at θ=3.0∘\theta= 3.0^\circ and Δ=0\Delta= 0 is nearly momentum-independent, as

shown in Figure 9(e), indicating that the fluctuations are predominantly short-ranged in real space and lack any preferred ordering wavevector. The weak enhancement at κ\kappa only reflects the underlying 3×3\sqrt{3} \times\sqrt{3} modulation. Figure 9(f) plots the corresponding OFR in the twist range where the IVC1_1 is stable at Δ=0\Delta= 0. The small value of OFR implies that in-plane spin order is intrinsically soft, due to the stronger fluctuations present in the IVC phase.

For the sake of completeness, we have also analyzed the influence of the layer polarization on OFR. According to our study, with increasing Δ\Delta the system generally maintains high rigidity in the OFR for both charge and out-of-plane spin orders in the VP phase, barring brief dips near Δ≃3\Delta\simeq3 meV caused by heightened fluctuations near phase transitions as shown in Figure 10(a) and (b), respectively. The inset in Figure 10(b) shows Szzconn(q)S_{zz}^{\mathrm{conn}}(\mathbf{q}) at Δ=15\Delta= 15 meV. A comparison with Figure 9(c), reveals that layer-polarization reduces the fluctuation by almost an order. The vanishingly small value at q=0\mathbf{q} = 0, indicates negligible variance of S^ztot\hat{S}^{\mathrm{tot}}_z, essentially making VP2_2 an eigenstate of S^ztot\hat{S}^{\mathrm{tot}}_z. In contrast, the charge OFR in the IVC phase experiences stronger fluctuations and a sharp drop at the layer polarization transition, recovering only partially at higher displacement fields due to suppressed interlayer hybridization as shown in Figure 10(c). However, the in-plane spin order in the IVC phase remains intrinsically soft, exhibiting a very low OFR across the entire displacement field range as shown in Figure 10(d).

Figure 10. OFR of the charge and spin orders for 4/3 hole-filling case at Δ≠0\Delta\ne0. (a) Charge OFR at θ=1.4∘\theta= 1.4^\circ. (b) Out-of-plane spin OFR at θ=1.4∘\theta= 1.4^\circ. The inset shows Szzconn(q)S_{zz}^{\mathrm{conn}}(\mathbf{q}) at Δ=15\Delta= 15 meV (VP2_2 phase). (c) Charge OFR at θ=3.0∘\theta= 3.0^\circ. (d) In-plane spin OFR at θ=3.0∘\theta= 3.0^\circ. In (a), (b), (c), and (d) the black circles and the red squares denote the HF and VMC results, respectively.

We finally investigate the origin of the correlation energy gain in the IVC phase discussed in connection with Figure 8. As mentioned in the Model and Method section, the variational wavefunction, ∣ΨT⟩=P^GP^J∣Ψ0⟩|\Psi_T\rangle= \hat{P}_G\hat{P}_J|\Psi_0\rangle, used in our VMC study has several variational components and therefore the correlations beyond HF can be incorporated at different levels. First, the pair-product wavefunction, ∣Ψ0⟩|\Psi_0\rangle, itself already spans a more general variational space than the single-determinant wavefunction of HF. Consequently, even in the absence of the Gutzwiller and Jastrow projectors, optimization of pair amplitudes fiσ,jσ′f_{i\sigma,j\sigma'} alone could lead to a solution which is closer to the true ground state of the system. The Gutzwiller and Jastrow projectors can provide additional energy gain by incorporating further many-body correlations into the variational state.

To disentangle the contributions of the different components of the variational wave function, we performed a series of VMC simulations in which different subsets of the variational parameters were optimized. In addition, we performed simulations where the pair amplitudes fiσ,jσ′f_{i\sigma,j\sigma'} are constructed from the converged solutions of the HF simulations and are held fixed, denoted by fHFf_{HF}, while the Gutzwiller (gg) and/or Jastrow (vv) parameters are optimized. Comparison among these calculations allows us to determine the importance of different variational parameters in the IVC phase of the system.

Figure 11(a) and (b) plot the correlation energies obtained by optimizing various subsets of parameters for θ=3.0∘\theta= 3.0^\circ and 3.5∘3.5^\circ, respectively, for four different displacement fields. As one can see, the results of the simultaneous optimization of all variational parameters, fiσ,jσ′f_{i\sigma,j\sigma'}, gig_i, and vijv_{ij}, denoted by the black crosses, lead to the smallest energies. As expected, such a procedure yields the largest overall correlation-energy gain. Among all the other scenarios, the one that corresponds to the simultaneous optimization of the pair-product wave function (fiσ,jσ′f_{i\sigma,j\sigma'}) and Jastrow parameters (vijv_{ij}), denoted by the green plus signs, is the closest to the fully optimized state. For θ=3.0∘\theta= 3.0^\circ the former solution is slightly above the latter in energy, which demonstrates a small but finite additional contribution from the Gutzwiller sector when all variational degrees of freedom are allowed to readjust simultaneously. For the case of θ=3.5∘\theta= 3.5^\circ the two solutions give almost identical energies.

Figure 11. Correlation energy per site obtained by optimizing different subsets of the variational parameters at Δ=0\Delta= 0, 3 meV, 8 meV, and 15 meV for θ=3.0∘\theta= 3.0^\circ (a) and 3.5∘3.5^\circ (b). Only fiσ,jσ′f_{i\sigma,j\sigma'} is optimized: blue circles. Simultaneous optimization of fiσ,jσ′f_{i\sigma,j\sigma'} and gig_i: orange squares. Simultaneous optimization of fiσ,jσ′f_{i\sigma,j\sigma'} and vijv_{ij}: green pluses. Keeping fiσ,jσ′f_{i\sigma,j\sigma'} fixed to fHFf_{HF} we optimize only gig_i (red diamonds), only vijv_{ij} (purple uparrows), gig_i and vijv_{ij} together (brown downarrows). Simultaneous optimization of all parameters fiσ,jσ′f_{i\sigma,j\sigma'}, gig_i, and vijv_{ij} is denoted by black crosses.

We conclude from the results presented in Figure 11 that the gain in the correlation energy in the IVC phase is dominated by the intersite correlations captured by the Jastrow parameters. In contrast, the Gutzwiller projector, which primarily modifies local correlations associated with the double occupancy, plays a minor role in improving the energy gain. This is consistent with the fact that for the particular band filling studied here, the number of electrons exceeds the number of sites, and, as a result, the double occupancy is unavoidable.

Summary

In this work, we have investigated the charge and spin order of the twisted MoTe2_2 bilayer system at hole fillings 4/3 and 3/2, which are experimentally found to be insulating. Using an interacting two-band tight-binding model, we construct the phase diagram as a function of the twist angle and an out-of-plane displacement field within the HF approximation. We further examined the effects of correlations beyond the mean-field using the VMC method.

For both fillings, low twist angles stabilize a VP phase with fully saturated out-of-plane ferromagnetism driven by easy-axis ferromagnetic exchange. At 4/3 filling, the VP phase exhibits a strongly classical character due to a high interaction-to-bandwidth ratio, which is further enhanced by displacement-field-induced layer polarization. In contrast, the VP phase at 3/2 filling displays a more quantum nature, forming a two-component generalized Wigner crystal that is particularly robust in the layer-polarized regime.

Above θ≃2.9∘\theta\simeq2.9^\circ a transition from the VP to an IVC state appears for both fillings. In the layer-hybridized regime, opposing in-plane spin moments on the AA and BB sublattices cancel out, whereas layer polarization induces incomplete cancellation and a net ferrimagnetic in-plane magnetization. We found that, while 4/3 filling universally favors 3×3\sqrt{3} \times\sqrt{3} order due to commensurability and Coulomb repulsion, 3/2 filling accommodates excess holes through superimposed 2×12 \times1 and 2×22 \times2 reconstructions.

In the second part of our analysis we have applied the VMC method for the case of 4/3 filling to study the effects of correlations beyond HF. The phases obtained in the VMC simulations remain consistent with those found in the HF simulations. The ground state energies obtained in both methods agree well in the VP phase, which is consistent with its strong classical nature.

However, VMC yields noticeably lower energy in the IVC phase. By selectively optimizing different components of the variational wavefunction, we find that this gain in correlation energy is dominated by the nonlocal density correlations, which are captured by the Jastrow projector. The contribution from the local density correlations described by the Gutzwiller factor is rather small.

The rigidity analysis provides a further understanding of the obtained phases. Both charge and spin orders show high rigidity in the low-twist VP phase. In contrast, the high-twist IVC phase is fluctuation-dominated, which leads to substantially lower rigidity. In particular, the in-plane spin order in the IVC phase is very soft. The VMC also predicts a much lower charge rigidity in this regime. The pronounced softening of the charge order at higher twist angle indicates a possible melting of the insulating charge order. Whether the high fluctuations in the IVC states ultimately lead to the metallic state is left for future study.

Overall, our study reveals a transition from an interaction-dominated classical generalized Wigner crystal at low twist angles to an increasingly fluctuation-dominated quantum state at high twist angles. The results further demonstrate that correlations beyond the mean-field approximation are primarily nonlocal Jastrow-type and are operative only in the quantum phase at high twist angles.

ACKNOWLEDGMENT

This research was funded by National Science Centre, Poland (NCN) according to Decision No. 2021/42/E/ST3/00128. This work was partly supported by the AGH University of Krakow subvention. We gratefully acknowledge Poland’s high-performance Infrastructure PLGrid ACK Cyfronet AGH for providing computer facilities and support within computational grant no. PLG/2026/019274.

References

References

  1. [1]E. Wigner, Phys. Rev. 46, 1002 (1934).
  2. [2]C. C. Grimes and G. Adams, Phys. Rev. Lett. 42, 795 (1979).
  3. [3]E. Y. Andrei, G. Deville, D. C. Glattli, F. I. B. Williams, E. Paris, and B. Etienne, Phys. Rev. Lett. 60, 2765 (1988).
  4. [4]Y. C. Tsui, M. He, Y. Hu, E. Lake, T. Wang, K. Watanabe, T. Taniguchi, M. P. Zaletel, and A. Yazdani, Nature 628, 287 (2024).
  5. [5]T. Smoleński, P. E. Dolgirev, C. Kuhlenkamp, A. Popert, Y. Shimazaki, P. Back, X. Lu, M. Kroner, K. Watanabe, T. Taniguchi, I. Esterlis, E. Demler, and A. Imamoglu, Nature 595, 53–57 (2021).
  6. [6]Y. Zhou, J. Sun, E. Brutschea, I. Esterlis, Y. Wang, G. Scuri, R. J. Gelly, H. Heo, T. Taniguchi, K. Watanabe, G. Zarand, M. D. Lukin, P. Kim, E. Demler, and H. Park, Nature 595, 48–52 (2021).
  7. [7]Y. Xu, S. Liu, D. A. Rhodes, K. Watanabe, T. Taniguchi, J. Hone, V. Elser, K. F. Mak, and J. Shan, Nature 587, 214–218 (2020).
  8. [8]H. Li, S. Li, E. C. Regan, D. Wang, W. Zhao, S. Kahn, K. Yumigeta, M. Blei, T. Taniguchi, K. Watanabe, S. Tongay, A. Zettl, M. F. Crommie, and F. Wang, Nature 597, 650–654 (2021).
  9. [9]X. Huang, T. Wang, S. Miao, C. Wang, Z. Li, Z. Lian, T. Taniguchi, K. Watanabe, S. Okamoto, D. Xiao, S. F. Shi, Y. T. Cui, Nat. Phys. 17, 715–719 (2021).
  10. [10]H. Li, Z. Xiang, A. P. Reddy, T. Devakul, R. Sailus, R. Banerjee, T. Taniguchi, K. Watanabe, S. Tongay, A. Zettl, L. Fu, M. F. Crommie, and F. Wang, Science 385, 86 (2024).
  11. [11]B. Padhi, R. Chitra, and P. W. Phillips, Phys. Rev. B 103, 125146 (2021).
  12. [12]H. Pan, F. Wu, and S. Das Sarma, Phys. Rev. B 102, 201104(R) (2020).
  13. [13]M. Matty and E.-A. Kim, Nat. Commun. 13, 7098 (2022).
  14. [14]N. Morales-Durán, P. Potasz, and A. H. MacDonald, Phys. Rev. B 107, 235131 (2023).
  15. [15]S. F. Ung, J. Lee, and D. R. Reichman, Phys. Rev. B 108, 245113 (2023).
  16. [16]Y. Tan, P. K. H. Tsang, V. Dobrosavljević, and L. Rademaker, Phys. Rev. Res. 5, 043190 (2023).
  17. [17]Z. Song, U. F. P. Seifert, Z.-X. Luo, and L. Balents, Phys. Rev. B 108, 155109 (2023).
  18. [18]Y. Zhou, D. N. Sheng, and E.-A. Kim, Phys. Rev. Lett. 133, 165601 (2024).
  19. [19]D. Erkensten, S. Brem, R. Perea-Causin, and E. Malic, Phys. Rev. B 110, 155132 (2024).
  20. [20]A. Kumar, C. Lewandowski, and H. J. Changlani, npj Quantum Mater. 10, 95 (2025).
  21. [21]A. Biborski and M. Zegrodnik, Phys. Rev. B 111, 075116 (2025).
  22. [22]A. Biborski, Acta Phys. Pol. B 57, 5-A17 (2026).
  23. [23]S. Mishra, T. M. R. Wolf, and A. H. MacDonald, arXiv:2603.05714 (2026).arxiv.org/abs/2603.05714
  24. [24]Y. Zhou, I. Esterlis, T. Smoleński, arXiv:2509.21222 (2025).arxiv.org/abs/2509.21222
  25. [25]P. Saha, L. Rademaker, and M. Zegrodnik, Phys. Rev. B 112, 045147 (2025).
  26. [26]L. Peng, C. De Beule, Y. Lai, D. Li, L. Yang, E. J. Mele, S. Adam, Phys. Rev. B 113, L161119 (2026).
  27. [27]J. Chen, Q. Li, X. Wang, W. Li, Sci. Bull. 71, 1034–1042 (2026).
  28. [28]B. Wang, J. Yu, and Cheng-Cheng Liu, Phys. Rev. B 113, L201409 (2026).
  29. [29]J. Hubbard, Phys. Rev. B 17, 494 (1978).
  30. [30]Y. Noda, M. Imada, Phys. Rev. Lett. 89, 176803 (2002).
  31. [31]G. Grüner (1994). Density Waves In Solids (1st ed.). CRC Press. https://doi.org/10.1201/9780492501012.
  32. [32]G. Grüner, Rev. Mod. Phys. 60, 1129 (1988)
  33. [33]P. Knüppel, J. Zhu, Y. Xia, Z. Xia, Z. Han, Y. Zeng, K. Watanabe, T. Taniguchi, J. Shan, and K. F. Mak, Nat. Commun. 16, 1959 (2025).
  34. [34]F. Xu, Z. Sun, J. Li, C. Zheng, C. Xu, J. Gao, T. Jia, Y. Su, K. Watanabe, T. Taniguchi, B. Tong, L. Lu, J. Jia, Z. Shi, S. Jiang, J. Lin, Y. Zhang, Y. Zhang, S. Lei, X. Liu, and T. Li, arXiv:2504.06972v2.
  35. [35]F. K. Wolff, S. Jiao, H. Park, J. Cai, E. Anderson, X. Liu, C. Wang, T. Taniguchi, K. Watanabe, X. Xu, D. Xiao, S. Vaitiekėnas, arXiv:2610.02493.
  36. [36]F. Wu, T. Lovorn, E. Tutuc, I. Martin, and A. H. MacDonald, Phys. Rev. Lett. 122, 086402 (2019).
  37. [37]C. Xu, J. Li, Y. Xu, Z. Bi, and Y. Zhang, Proc. Natl. Acad. Sci. U.S.A. 121 (8) e2316749121 (2024).
  38. [38]The supplementary material contains the sign convention for the phase related to the hopping, details of the V$_{P}$, \bar{V}C$_1$, and \bar{V}C$_2$ phases.
  39. [39]H. Bruus and K. Flensberg, Many-Body Quantum Theory in Condensed Matter Physics: An Introduction, Oxford University Press, Oxford, England, 2004.
  40. [40]T. Misawa, S. Morita, K. Yoshimi, M. Kawamura, Y. Motoyama, K. Ido, T. Ohgoe, M. Imada, and T. Kato, Comput. Phys. Commun. 235, 447 (2019).
  41. [41]F. Becca and S. Sorella, Quantum Monte Carlo Approaches For Correlated Systems, Cambridge University Press, Cambridge, England, 2017.DOI
  42. [42]J. Toulouse, R. Assaraf, and C. J. Umrigar, in Electron Correlation in Molecules–ab initio Beyond Gaussian Quantum Chemistry, Adv. Quantum Chem., Vol. 73, edited by P. E. Hoggan and T. Ozdoğan (Academic, New York, 2016), pp. 285–314.
  43. [43]W. M. C. Foulkes, L. Mitas, R. J. Needs, and G. Rajagopal, Rev. Mod. Phys. 73, 33 (2001).
  44. [44]D. Ceperley, G. V. Chester, and M. H. Kalos, Phys. Rev. B 16, 3081 (1977).
  45. [45]H. Yokoyama and H. Shiba, J. Phys. Soc. Jpn. 56, 1490-1506 (1987).
  46. [46]A. Biborski, M. Zegrodnik, and J. Spałek, Phys. Rev. B 101, 214504 (2020).
  47. [47]D. Tahara and M. Imada, J. Phys. Soc. Jpn. 77, 114701 (2008).
  48. [48]C. Hotta, N. Furukawa, Phys. Rev. B 74, 193107 (2006).
  49. [49]J. Merino, A. Ralko, S. Fratini, Phys. Rev. Lett. 111, 126403 (2013).

Paper details

vlnc:W_s5ymwe5q
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 1

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents