Spatially Indirect Exciton Condensation in Two-Dimensional Strongly Correlated Semimetals Preprint: APS/123-QED Yao Zeng, Shi-Cong Mo, Wéi Wú Abstract Identifying materials hosting an excitonic insulator ground state has been one of the major pursuits in condensed matter physics in recent years. Promising candidates in transition metal chalcogenide compounds (TMC), including 1 ​ T − TiSe 2 1T-\mathrm{TiSe_{2}} , Ta 2 ​ Pd 3 ​ Te 5 \mathrm{Ta_{2}Pd_{3}Te_{5}} , and Ta 2 ​ NiSe 5 \mathrm{Ta_{2}NiSe_{5}} , share a crucial common characteristic: their low-energy physics is governed by electrons in d − d- orbitals subject to strong on-site Coulomb interactions. In this work, we investigate spatially indirect exciton condensation in two-dimensional semimetals on triangular lattice. Using a combination of dynamical mean-field theory and the determinant quantum Monte Carlo method, we study two- and three-orbital Hubbard models incorporating strong on-site ( U U ) and inter-orbital interactions ( V V ). Our results demonstrate that on-site Hubbard U U can strongly suppress the condensation temperature T c T_{c} , an effect that is particularly pronounced at higher electron-hole pair densities. This behavior contrasts sharply with the case without on-site U U , where T c T_{c} grows with pair density at fixed V V . Moreover, we uncover competition among multiple electron-hole pairing channels in the three-orbital model, which also acts to suppress T c T_{c} of exciton condensation. An orbital-selective electron-hole pairing state is identified. These findings may help explain the large discrepancy between strong binding-energy and relative low transition temperature for indirect excitons in TMCs materials, offering important insights for understanding and engineering exciton condensation in materials with strongly correlated d − d- shell electrons. Introduction- The excitonic insulator (EI) constitutes a paradigmatic correlated state arising from a bosonic condensation of Coulomb-bound electron-hole pairs in narrow-gap semiconductors or semimetals 39 ; 5 ; 24 ; 15 . This fundamental phase of matter has been investigated extensively in both experiment 8 ; 25 ; 35 ; 2 ; 66 ; 23 ; 22 ; 17 ; 56 and theory 5 ; 24 ; 34 ; 60 ; 26 ; 25 ; 21 ; 50 ; 62 ; 65 ; 61 . The spatially separated electron–hole bilayer structures 15 provide particularly favorable platforms for achieving this state, as spatial separation suppresses electron-hole recombination, hence prolonging exciton lifetimes and promotes the condensation of indirect excitons. The exploration of many exotic emergent phenomena related to EI, such as the BCS-BEC crossover 52 ; 20 ; 70 ; 71 ; 33 ; 42 ; 38 ; 18 ; 58 , supersolidity 12 ; 67 ; 13 , finite momentum pairing 9 ; 3 ; 64 ; 14 , excitonic topological phases 57 ; 22 , and perfect Coulomb Drag 41 ; 45 , can thus be facilitated in the indirect exciton systems. Quantum wells in \mathrm{InAs}/\mathrm{GaSb} heterostructures have served as pioneering platforms in the pursuit of these exotic phases 72 ; 57 . These semiconductor systems revealed signatures of excitonic topological order and superfluidity 57 ; 51 ; 15 . More recently, transition metal chalcogenide (TMC) platforms, such as \mathrm{MoSe}_{2}/\mathrm{hBN}/\mathrm{WSe}_{2} heterostructures, have been pivotal systems for hosting high-temperature EI 41 ; 45 . By exploiting Type-II band alignment to spatially separate carriers, these devices host long-lived excitons that exhibit macroscopic coherence, evidenced by nonlinear electroluminescence up to 100~\text{K} 58 . Crucially, transport measurements revealed perfect Coulomb drag at zero magnetic field 41 ; 45 , establishing a key foundation for enabling frictionless flow of charge-neutral excitons 55 ; 47 ; 40 ; 48 in transport experiments. In bulk EI materials, on the other hand, the layered TMC compounds 1T-\mathrm{TiSe_{2}} 29 ; 8 ; 43 , \mathrm{Ta_{2}Pd_{3}Te_{5}} 23 ; 66 , and \mathrm{Ta_{2}NiSe_{5}} 62 ; 63 ; 23 ; 66 ; 1 have sparked intense debate regarding the existence of exiction condensation. This is because the gap opening in these materials, a hallmark of the excitonic ordered state, can also be attributed to non-excitonic structure phase transitions 37 ; 32 ; 2 ; 59 ; 10 . Compared to \mathrm{Ta_{2}NiSe_{5}} , which may host significant structural distortion 28 ; 59 , \mathrm{Ta_{2}Pd_{3}Te_{5}} offers a cleaner paradigm, as in which the lattice distortion is negligible in the insulating phase 66 ; 23 . This structural stability strongly supports an exciton instability driven by Coulomb interactions in \mathrm{Ta_{2}Pd_{3}Te_{5}} . While experimental advances in the study of EI have progressed rapidly, a comprehensive theoretical understanding 30 ; 6 ; 16 ; 60 ; 11 ; 52 ; 13 ; 27 of this phase in real materials remains incomplete. We notice that in theory, large exciton binding energies being order of \sim\mathrm{1eV} have been predicted in semi-metals or narrow gap semiconductors 62 ; 54 ; 68 ; 53 , a high temperature EI in experiments, however, is still elusive. Conventional theoretical frameworks are often established on the basis of mean-field or perturbative approaches. In EI candidate materials based on TMC, the strong on-site Hubbard interaction U in d- orbitals is usually larger than the electron bandwidth W 46 , and it can significantly exceed the inter-orbital (inter-layer) electron-hole attraction V 37 . This strong-correlation regime calls into question the adequacy of weak-coupling theories that consider V while neglecting the substantial on-site interaction U . In this work, we investigate excitonic condensation in two-dimensional (2D) semimetals using a combination of Cellular dynamical Mean-Field theory (CDMFT) 19 and determinant quantum Monte Carlo (DQMC) 4 . The non-perturbative CDMFT approach enables us to capture the quantum fluctuation of strongly interacting d- electrons beyond weak-coupling treatments. We show that the exciton condensation temperature T_{c} can be significantly suppressed by on-site Coulomb repulsion U , although it in general remains non-zero for finite V . We also show that a finite U fundamentally modifies the density dependence of T_{c} . Specifically, T_{c} drops with increasing carrier concentration n , which contrasts with the U=0 case, where T_{c} grows with pair density n at fixed V . Using a three-orbital model, we further uncover the competition between different pairing channels, which provides an additional mechanism suppressing T_{c} . An orbital-selective EI state is also revealed in this model. Finally, we discuss the implications of these findings for experimental observations. Figure 1: (a) Schematic illustration of the two-orbital (two-layer) Hubbard model on 2D triangular lattice. U represents the on-site intra-orbital(intra-layer) Coulomb interaction, while V denotes the inter-orbital (inter-layer) Coulomb interaction in a unit cell. The orbital/layer a (upper) is designated as electron-active, while orbital/layer b (lower) acts as hole-active. Dotted line indicates a possible electron–hole pairing. (b) Schematic illustration of the non-interacting band structure we use which is a semimetal at U=0,V=0 . Red (black) curves indicate the conduction (valence) bands. The pairing of holes at the CB bottom (red hollow circles) and electrons at the VB top (black filled circles) leads to the formation of indirect excitons. Dashed line indicates Fermi level. Model and Methods - Our two-orbital Hubbard model we use is given by, \displaystyle H \displaystyle=-\sum_{\langle ij\rangle\sigma m}t_{m}c^{\dagger}_{i\sigma m}c_{j\sigma m}+\text{H.c.}-\sum_{i\sigma m}(\mu-\epsilon_{m})\hat{n}_{i\sigma m} \displaystyle\quad+\sum_{im}U_{m}\hat{n}_{i\uparrow m}\hat{n}_{i\downarrow m}+V\sum_{i\sigma\sigma^{\prime},m\neq m^{\prime}}\hat{n}_{i\sigma m}\hat{n}_{i\sigma^{\prime}m^{\prime}} (1) where the m=a,b labels the orbital (or layer) degrees of freedom, corresponding to orbitals in bulk systems and layers in 2D materials, respectively. c_{i\sigma m} ( c^{\dagger}_{i\sigma m} ) denotes the electron annihilation (creation) operator at unit cell i , orbital (layer) m , with spin \sigma ; n_{i\sigma m} is the particle number operator. The nearest-neighbor hopping integral t_{a} is set to t_{a}=t=1 as energy unit throughout the paper. U_{m} is the on-site repulsion between electrons at orbital (layer) - m , V is the local inter-orbital (inter-layer) interaction. In this work, we have omitted long-range Coulomb interactions to simplify the numerical study. We adjust chemical potential \mu and orbital(layer) energy \epsilon_{m} to tune the orbital (layer) densities. Here we adopt a pure electron picture to study the problem, i.e., \langle\hat{n_{a}}\rangle=n_{e}\equiv n denote the carrier density in the electron orbital (orbital- a ), implying a hole orbital (layer) of n_{h}=1-\langle\hat{n_{b}}\rangle=n_{e}=n in the orbital- b . We assume equal mass for both carriers, i.e., , t_{a}=-t_{b} . The schematic of our model and the simplified semimental band structure are illustrated in Fig. 1 . Orbitals (layers) a (upper) and b (lower) are designated as electron-active and hole-active, respectively. Our theoretical investigation of the bilayer Hubbard model on a triangular lattice is carried out by the Cellular Dynamical Mean-Field Theory, a cluster extension of DMFT 19 designed to incorporate non-local spatial correlations. In this framework, the lattice problem is mapped onto an effective quantum impurity model for a small cluster, which is then solved self-consistently within a dynamical bath that represents the rest of the lattice degrees of freedom. Here we typically use a two-site effective cluster incorporating a,b orbitals within a unit cell. The effective cluster problem is solved using the continuous-time quantum Monte Carlo method (CTQMC) 49 ; 44 . Our simulations typically accumulate 10^{9} Monte Carlo sweeps in each DMFT loop. Figure 2: The inverse susceptibility 1/\chi as a function of temperature T . When 1/\chi approaches zero, the system enters the excitonic condensate phase, and the corresponding temperature is defined as the critical transition temperature T_{c} . Here, U_{a}=U_{b}=1,V=4 , density, \langle n_{a}\rangle=0.2,\langle n_{b}\rangle=0.8 (thus n_{e}=n_{h}=0.2 ). An excitonic condensation T_{c} is found about T_{c}\sim 0.07 . The DQMC calculations were performed on a 4\times 4\times 2 lattice, while DMFT calculations were performed on 1\times 2 - and 12\times 2 -site clusters. Result - The onset of the excitonic condensation phase can be identified by the divergence of the uniform excitonic pairing susceptibility \chi , which can be defined as, \displaystyle\chi=\frac{1}{N}\sum_{i,j}\int_{0}^{\beta}d\tau\langle\mathcal{T}_{\tau}p_{j}(\tau)p_{i}(0)\rangle (2) where p_{i} is the uniform spin-singlet electron-hole pair operator 20 , p_{i}=\frac{1}{\sqrt{2}}\sum_{\sigma}\langle c^{\dagger}_{ia\sigma}c_{ib\sigma}\rangle , and N is the number of unit cells in the system. Thus the vanishing inverse pairing susceptibility 1/\chi\to 0 can be used to determine the critical transition temperature T_{c} for excitonic condensation ( i.e. , T_{c} for EI). As shown in Fig. 2 , we plot the CDMFT result of 1/\chi as a function of temperature T for a typical parameter set: U_{a}=U_{b}=1 , V=4 , and electron-hole density n=0.2 . As the temperature decreases from T=1.0 to T=0.08 , 1/\chi drops monotonically from 1/\chi\approx 2.0 to 1/\chi\approx 0 , signaling a growing tendency to El condensation as T reduces. As T approaches T_{c}\approx 0.07 , 1/\chi vanishes, denoting an El instability in the system at the given parameters. We observe excellent agreement of 1\times 2 - and 12\times 2 - site CDMFT with exact high- T DQMC results on a 4\times 4\times 2 - site lattice, thereby justifying the use of CDMFT to reliably capture the exciton physics. Figure 3: Transition temperature T_{c} as a function of the inter-orbital interaction V for different values of the on-site interaction U_{a}=U_{b}\equiv U . Here, electron-hole density is fixed at n_{e}=n_{h}=0.06 , the data points represent results for U=0 (red circles) and U=4 (blue circles). The dashed lines depict fits to the BCS scaling trend, T_{c}=Ae^{-B/V} , while shaded ribbons indicate the corresponding confidence intervals. The extracted fitting parameters are A\approx 5.73 , B\approx 19.28 for the non-interacting case ( U=0 ) and A\approx 7.59 , B\approx 25.90 for the interacting case ( U=4 ). We now focus on the dependence of critical temperature T_{c} on the inter-orbital electron-hole attraction V . In Fig. 3 , results of T_{c} as a function of V are shown. The calculations are performed at a fixed electron-hole density n=0.06 , and two on-site repulsion U=0,U=4 . In the absence of no-site repulsion ( U=0 ), T_{c} increases from T_{c}\approx 0.006 to T_{c}\approx 0.12 as V is increased from V=3 to V=5 . For the finite U case ( U=4 ), T_{c} grows from T_{c}\approx 0.012 to T_{c}\approx 0.1 as V increases from V=4 to V=6 . At given V , T_{c} is consistently smaller for U=4 than for U=0 , indicating on-site U plays a detrimental role in exciton condensation. For both curves, we find a trend qualitatively consistent with a BCS-like behavior is followed, T_{c}\propto\exp(-B/V) , namely, T_{c} are very small for V<3 , then grows rapidly, exhibiting an exponential-like rise as V\geq 3 . Figure 4: Transition temperature T_{c} as a function of the on-site Coulomb interaction U at different values of inter-orbital interaction V . Orange symbols correspond to V=5 , while blue symbols correspond to V=4 . Here n_{e}=n_{h}=n=0.06 . 1\times 2- site CDMFT is used here. To investigate the effect of on-site Hubbard repulsion U on excitonic condensation in detail, we computed T_{c} across a wide range of U values. As shown in FIG. 4 , T_{c} decreases substantially as U increases from zero, demonstrating a strong suppression of excitonic condensation by U . This decreasing trend is particularly rapid at smaller U and becomes more gradual for larger U , see both curves in FIG. 4 . Specifically, at fixed V=5 , increasing U from U=1 to U=18 reduces T_{c} from T_{c}=0.1 to T_{c}=0.015 , a decrease by nearly a factor of seven. Similarly, for V=4 , raising U from U=0.5 to U=12 lowers T_{c} from T_{c}=0.03 to T_{c}=0.005 , corresponding to a sixfold reduction. These results confirm that the on-site repulsion U is a key factor limiting the stability of the excitonic condensed phase in the system. Now we investigate the interplay between the on-site repulsion U and the electron-hole density. The density dependence of T_{c} for both U=0 and i U=8 cases is shown in FIG 5 . For the U=0 case ( U=0,V=5 ), T_{c} exhibits a monotonic increase from T_{c}=0.035 to T_{c}=0.15 as the density rises from n=0.005 to n=0.08 , indicating that a larger population of charge carriers favours condensation. In stark contrast, the on-site repulsion fundamentally alters this behavior. In the low-density regime ( 0.005