betaFeedback

WorksPhysics and Astronomy

Emergent flat band and topological Kondo semimetal driven by orbital-selective correlations

Abstract

Flat electronic bands are expected to show proportionally enhanced electron correlations, which may generate a plethora of novel quantum phases and unusual low-energy excitations. They are increasingly being pursued in dd-electron-based systems with crystalline lattices that feature destructive electronic interference, where they are often topological. Such flat bands, though, are generically located far away from the Fermi energy, which limits their capacity to partake in the low-energy physics. Here we show that electron correlations produce emergent flat bands that are pinned to the Fermi energy. We demonstrate this effect within a Hubbard model, in the regime described by Wannier orbitals where an effective Kondo description arises through orbital-selective Mott correlations. Moreover, the correlation effect cooperates with symmetry constraints to produce a topological Kondo semimetal. Our results motivate a novel design principle for Weyl Kondo semimetals in a new setting, viz. dd-electron-based materials on suitable crystal lattices, and uncover interconnections among seemingly disparate systems that may inspire fresh understandings and realizations of correlated topological effects in quantum materials and beyond.

Introduction

Certain crystalline lattices feature flat bands, via frustration caused by destructive interference in electron motion[1], which are increasingly being explored in dd-electron-based systems[2, 3]. The reduced bandwidth correspondingly enhances the effect of electron correlations. In addition, such flat bands are often topologically nontrivial. As such, these systems represent a new platform to uncover novel physics for both correlation and topology as well as their interplay[4]. For example, kagome metals may host flat bands and have been the subject of considerable recent interest for realizing unusual forms of charge-density-wave order[5, 6, 7, 8, 9, 10]. They have also been implicated to exhibit a type of strange metal behavior[11, 12, 13, 14] that resembles what has been extensively studied in quantum critical heavy fermion metals[4, 15, 16].

In order to strongly influence the low-energy physics, the flat bands need to be placed near the Fermi energy. However, this typically is not the case at the level of bare (noninteracting) electron bandstructure. There have been considerable recent experimental efforts to tune the bare flat bands to the vicinity of the Fermi energy. With the rare exception coming from materials search[13], the tuning study has met with only a limited success[17, 18]. Because the flat bands are associated with dd-electrons, the energy scales that determine the flat-band placement (with respect to the Fermi energy) are relatively large and, as a result, it is challenging to achieve the required tuning. We are thus motivated to ask the following important questions: Can electron correlations generate emergent flat bands at the Fermi energy in dd-electron-based systems? And, if so, to what extent do the resulting phases display nontrivial correlation and/or topological physics?

We address both issues in a Hubbard model in which the noninteracting limit features a topologically nontrivial flat band that is far away from the Fermi energy. Due to a well-separated hierarchy in the widths of the flat band and wide bands that it is coupled to, and through the formation of compact molecular orbitals[19], orbital-selective Mott correlations develop[11]. We show that such orbital-selective correlations lead to emergent flat bands that are pinned to the Fermi energy. Moreover, using symmetry constraints in interacting settings[20], which are based on Green’s function eigenstates (as opposed to Bloch states[21, 22, 23, 24, 25]), we demonstrate that the emergent flat bands lead to a topological Kondo semimetal. The latter is in the same family as Weyl Kondo semimetals that appear in topological Kondo-lattice models[26, 27] and materials[28, 29] of both existing and designed[20, 30] heavy fermion systems. The qualitative physics is illustrated in Figure 1a-Figure 1c. Importantly, our approach is based on an exact construction of molecular orbitals and an exact mapping to a heavy fermion description; this is in addition to the exact constraints that symmetry places, which will also come into our analysis. Our results motivate a design principle for the Weyl Kondo semimetals in the new setting of dd-electron-based systems, and point to the realization of fractional Chern insulators[31] in transition-metal compounds.

Figure 1. Illustration of the bare and emergent flat bands and lattice geometry. a, In the noninteracting case, a flat band (red solid line) appears far away from the Fermi energy. b, In the presence of orbital-selective correlations, an interaction-driven flat band emerges at the Fermi energy (red solid line), while leaving incoherent excitations far away from the Fermi energy (red dashed lines). c, The emergent flat band crosses a dispersive band, leading to a topological Kondo semimetal with symmetry-protected Dirac/Weyl nodes that are pinned close to the Fermi energy, within an effective Kondo energy scale. d, Geometry of the clover lattice with 5 sublattices per unit cell. The lattice does not have inversion symmetry. This can be seen from the mismatch between the (dark) blue sublattices and their inversion counterparts (dots in light blue). e, The Wannier orbitals are near the geometric centers (shaded blue circles) of the unit cells, which form a triangular lattice.

Results

One-orbital Hubbard model on the clover lattice

For a proof-of-principle demonstration, we consider a variant of the kagome lattice, the two-dimensional (2D) clover lattice. As shown in Figure 1d, it contains five sublattices per unit cell. Leaving the details of the model to be given in the Methods and in Supplementary Note 1, we note that this lattice features a flat band (Supplementary Note 2). As a case study, the model is simplified while preserving the topological nature of the flat band; we do so by removing the C3C_3 symmetry of the clover lattice (Methods), leaving only a mirror symmetry MxM_x. There is one orbital per site. The Hubbard model takes the form H=H0+H1\mathcal{H} = \mathcal{H}_0 + \mathcal{H}_1, where H0\mathcal{H}_0 is the kinetic term and H1\mathcal{H}_1 represents the onsite Hubbard interaction. We consider the generic setting that has not been analyzed before, namely with the flat band of the noninteracting Hamiltonian being far away from the Fermi energy, as illustrated in Figure 1a and shown in Figure 2a.

Figure 2. Bare and emergent flat bands. a, The DOS of the dd and cc electrons for the noninteracting case. b, The corresponding result at u=1.6u = 1.6; here the red solid (dashed) lines denote the coherent (incoherent) part of the dd-electron excitations. The purple backgrounds mark the DOS of the conduction cc electrons.

The lattice can be divided into two groups of sublattices (denoted by blue and yellow dots in Figure 1d), which contain different numbers of sites per unit cell. The flat-band formation can be seen by considering only the nearest-neighbor hopping between the blue and yellow sites, reflecting a destructive interference of the electronic wavefunction on the lattice[32] (see Supplementary Note 2). The flat band overlaps with the wide bands.

Molecular orbitals, effective extended Hubbard model and the solution method

A flat band that is topologically nontrivial cannot by itself be represented by exponentially localized symmetry-preserving (Kramers-doublet) Wannier orbitals[25]. Such a Wannierization only becomes possible when other bands are considered along with the flat band. In addition, a flat band coming from destructive interference comprises states from multiple (inequivalent) atomic sites. If one Wannier orbital is to primarily capture this flat band, this Wannier orbital (and, by extension, the others accompanying it) must involve multiple atomic orbitals. In other words, in this case, the Wannier orbitals are necessarily molecular orbitals. We again stress that the mapping we use is exact.

In our case, we can restrict to three bands (see Supplementary Note 3). We find the centers of the three localized Kramers-doublet Wannier orbitals to be located near the geometric center of the unit cell, which forms a triangular lattice (see Figure 1e). Importantly, one Wannier orbital primarily captures the flat band[11]; it is the most localized and is denoted as the dd orbital. The other two Wannier orbitals are dominated by the wide bands; they decay much more slowly and are marked as cc orbitals. The large difference in the width of the flat band (DflatD_{\mathrm{flat}}) and the wide bands (DwideD_{\mathrm{wide}}) opens up a range of interactions that are in between. In this range, the electron correlations are strongly orbital-selective and the system affords a Kondo/Anderson-lattice model description[11].

We project the Hubbard model of the original lattice to the Wannier basis. This leads to the effective model expressed in terms of the dd and cc Wannier orbitals with Heff=H0+HintH_{\mathrm{eff}} = H_0 + H_{\mathrm{int}}. The kinetic term is specified in the Methods. For the interaction terms, we keep the most dominant interactions on the Wannier basis. They include the onsite Hubbard interaction among the dd electrons and the density-density interactions between the dd and cc electrons:

Hint=Hd+HF=∑iu2(ni↑d+ni↓d−1)2+∑i,αFαnidnic,α(1)\begin{aligned} H_{\mathrm{int}} &= H_d + H_F \\ &= \sum_i \frac{u}{2}\left(n_{i\uparrow}^{d}+n_{i\downarrow}^{d}-1\right)^2+\sum_{i,\alpha}F_\alpha n_i^d n_i^{c,\alpha} \tag*{(1)} \end{aligned}

where ni,σa=aiσ†aiσn_{i,\sigma}^{a}=a_{i\sigma}^{\dagger}a_{i\sigma}, with a=d,cαa=d,c_\alpha, α=1,2\alpha=1,2, and nia=∑σni,σan_i^a=\sum_\sigma n_{i,\sigma}^a. The onsite Hubbard interaction on the dd-orbital, uu, is the most dominant one, given the much more localized nature of this Wanner orbital. The density-density interactions between the dd and cc electrons, FαF_\alpha, are weaker but also sizable: F1/u≈0.3F_1/u \approx0.3 and F2/u≈0.25F_2/u \approx0.25. These effective interaction parameters are determined by those of the original Hubbard model (see Supplementary Note 3). The interactions among the cc electrons are relatively small compared to their bandwidths and, accordingly, will be unimportant.

To take into account the effect of the interactions, we use the U(1) slave spin (SS) method[33]. Given that only the onsite interaction of the dd orbital is important, we need to introduce a SS representation for the dd orbital only: diσ†=oiσ†fiσ†d_{i\sigma}^{\dagger}=o_{i\sigma}^{\dagger}f_{i\sigma}^{\dagger}, where the auxiliary bosonic and fermionic operators, o†o^\dagger and f†f^\dagger, carry the charge and spin degrees of freedom, respectively. We treat the SS formulation at the saddle-point level and self-consistently solve the corresponding Hamiltonians for the SS and the auxiliary fermion parts. The SS method is also used to obtain the contributions to the single-electron excitations from the (interaction-driven) incoherent part of the spectrum. The details are found in the Methods and Supplementary Note 6.

Emergent flat band at the Fermi energy

We are now in position to discuss the effect of interactions on the single-electron excitations. Consider first the density of states (DOS). In the noninteracting case, as shown in Figure 2a, the dd electron DOS (the red curve) has a sharp peak compared with the background (purple color) cc electron component. This reflects the dd electrons as primarily describing the flat band. As can be seen, the peak is located far away from the Fermi level, which also reflects its origin from the noninteracting flat band (see Supplementary Note 1).

Importantly, under the influence of electron correlations, a new flat band emerges. This is demonstrated in Figure 2b with u=1.6u = 1.6. The emergent flat band is pinned to the Fermi energy, as captured by the coherent peak (the red solid lines) in the DOS. The background DOS associated with the conduction cc electrons is largely unchanged from its noninteracting counterpart. Varying the interaction strongly influences the spectral weight of the emergent flat band (see Supplementary Note 7): This part of the spectral weight is reduced as the interaction increases (comparing Figure 2b and Supplementary Fig. S5a); the reduced spectral weight is transferred to the incoherent part (the red dashed lines). This form persists until the weight of the coherent peak is completely lost and the system goes through an orbital-selective Mott transition. When that happens, the incoherent parts of the single-electron excitations develop into the full-fledged lower and upper Hubbard bands (see Supplementary Note 7 and Figure S5).

Figure S5. Evolution of the DOS across the orbital-selective Mott transition. a, The DOS of the dd and cc electrons for u=2.2u = 2.2, which is just below the critical value uc=2.4u_{\mathrm{c}} = 2.4. The coherent part of the dd-electron spectrum corresponds to the central peak (the red solid line), and its incoherent part appears as side peaks (the red dashed lines). The cc-electron spectral function is shown in purple. b, The case of u=2.5u = 2.5, which is larger than ucu_{\mathrm{c}} and, thus, the system lies inside the OSMP. The dd-electrons are in a Mott insulating state, and their spectrum contains only the lower and upper Hubbard bands.

Orbital-selective Mott correlations

To expound the origin of the emergent flat band, we further analyze the orbital-selective Mott correlations in the regime of interactions of our interest, viz. Dflat<u<DwideD_{\mathrm{flat}} < u < D_{\mathrm{wide}}. As shown in Figure 3, the metal-to-insulator transition of the dd electrons occurs at uc=2.4u_{\mathrm{c}} = 2.4. We reiterate that the range of the interactions being considered here is weaker than the width of the wide bands associated with the cc electrons (see Supplementary Fig. S3). Thus, the cc electrons are fully itinerant. This justifies the neglect of the interactions among the cc electrons, as we have done, so that the quasiparticle weight of the cc-electrons remains to be 1 as seen in Figure 3. The existence of an orbital-selective Mott transition is further illustrated in the nature of the Fermi surface. As shown in Figure 3 (insets), the Fermi surface undergoes a dramatic change across the transition. This change parallels the electron localization-delocalization (Kondo destruction) physics of heavy fermion systems[34, 35, 36, 37, 38]. The phase with the dd-electrons being itinerant corresponds to the Kondo-screened phase, in which the local moment is converted into (fragile, or heavy) electronic excitations that hybridize with the conduction electrons to form the quasiparticles. By contrast, the orbital-selective Mott phase (OSMP) is analogous to the Kondo-destroyed phase of the heavy fermion systems, in which the Fermi surface is formed entirely from the conduction electrons. The fact that our noninteracting flat band is, to begin with, far away from the Fermi energy makes the parallel with the heavy fermion systems especially clear. Our analysis of this flat-band system provides a realization of the Kondo physics in a one-band Hubbard model that physically describes a dd-electron-based system.

Figure 3. Orbital-selective correlations. The quasiparticle weights of the dd and cc electrons, ZdZ_d and ZcZ_c, as a function of the effective interaction uu. The Left (right) inset plots the Fermi surface for the values of interactions right below (above) uc=2.4u_c = 2.4.

More specifically, the value of the interaction illustrated in Figure 2b is marked by an arrow in Figure 3. The differentiation between the quasiparticle weights of the dd and cc electrons at this interaction characterizes the orbital-selective nature of the electron correlations. This is reflected in the dd-electron spectral weight: as seen in Figure 2b, the incoherent peaks (the red dashed lines) are well formed, which corresponds to the precursor of the lower and upper Hubbard bands of the OSMP (see Supplementary Fig. S5b). The coherent spectral weight, i.e. the central peak (the red solid line of Figure 2b) is thus described in terms of the Kondo resonance of a Kondo lattice model, in which the local moments correspond to the effective spin degrees of freedom associated with the lower and upper Hubbard bands. This description makes precise the notion that the flat band at the Fermi energy is emergent, driven by the orbital-selective Mott correlations.

Topological Kondo semimetal

The energy dispersion of the electronic states is shown in Figure 4. From the dispersion of the interacting (u=1.6u = 1.6) case, we again see that a Fermi-energy-bound flat band emerges in the interacting case.

Figure 4. Topological Kondo semimetal. a, The noninteracting band structure. b, The dispersion of the single-electron excitations at u=1.6u = 1.6. The red solid curve denotes the emergent flat band close to the Fermi energy. The grey lines mark the incoherent single-electron excitations. c, The zoomed-in view of the emergent flat band. d, The band structure at u=1.6u = 1.6 with a Zeeman splitting mz=0.03m_z = 0.03.

We are then in position to analyze the symmetry constraints[21, 22, 23, 24, 25]. In the noninteracting limit, the three Wannier orbitals have different MxM_x eigenvalues. The flat dd orbital has the MxM_x eigenvalue +1+1, while the two cc orbitals have the MxM_x eigenvalues of −1-1 and +1+1 respectively[11]. In the presence of time-reversal symmetry, along the Γ−K\Gamma- K line, the flat band from the dd orbital has a symmetry-protected Dirac crossing with the cc orbital of the opposite mirror eigenvalue. This same symmetry constraint also applies to the Kondo-driven flat band. The Dirac node for the emergent flat band is shown in Figure 4b. A node from the flat band close to the Fermi energy allows a high tunability. As shown in Figure 4d, a relatively small Zeeman coupling (illustrated here with m=0.03m = 0.03, which is small compared to the width of the emergent flat band of ∼0.1\sim0.1) causes a substantial separation of the nodes. These nodes now have two-fold degeneracy.

The orbital-selective Mott correlations are caused by local correlations. While we have provided a case study of how such correlations give rise to emergent flat bands in a particular 2D model, a similar conclusion is expected in general cases, including for models in three dimensions (3D). There is an important distinction though. In 3D, topological nodes develop in the presence of SOC under symmetry constraints[25, 39]. For noncentrosymmetric systems, or for centrosymmetric systems with the breaking of time-reversal symmetry, we can then expect the emergent flat bands to feature Weyl nodes leading to a Weyl Kondo semimetal.

To further expound on the generality of our theoretical results, we note that the metallic regime with strong orbital-selective correlations can be viewed through the Kondo analogy. From this perspective, the emergent flat band describes low energy coherent electronic excitations associated with the Kondo-driven composite fermions. Because low energy electronic excitations are always Fermi-energy bound, and also based on the well-established understanding that Kondo-driven composite fermions occur in the immediate vicinity of the Fermi energy, the emergent flat bands that develop through our proposed mechanism must be pinned near the Fermi energy. This represents a general principle. To explicate on this generality, we have mapped out a phase diagram to show that the proposed mechanism operates over an extended region in the uu-ϵd0\epsilon_d^{0} parameter space (region “II” of the phase diagram given in Supplementary Note 8 and Figure S6). Furthermore, we have carried out related calculations in a more general setting and find a similar development of an emergent flat band when the noninteracting flat band is located substantially away from the Fermi energy; the details of this analysis appear in Supplementary Note 9 and Figure S7.

Figure S6. Phase diagram for the regime of emergent flat bands. a, The quasiparticle weight of the more strongly correlated species of electrons (dd) versus the interaction strength and for various ϵd0\epsilon_d^0. b, The corresponding colormap plot. In the parameter setting between the two black lines (region “II”), an emergent flat band develops in the immediate vicinity of the Fermi energy. c-e, The dispersion of the coherent single-electron excitations are shown for parameters corresponding to the purple points marked in b. f-h. The counterparts for the red points marked in b.

Figure S7. Topological Kondo effect in a general setting. a, The noninteracting band structure. b, The dispersion of the coherent single-electron excitations at u=2.1u = 2.1. The red solid curve denotes the emergent flat band close to the Fermi energy. The grey lines mark the incoherent single-electron excitations. c, The zoomed-in view of the emergent flat band. d, The band structure at u=2.1u = 2.1 with a Zeeman splitting mz=0.05m_z = 0.05.

Design principle for Weyl Kondo semimetals in physical dd-electron systems

Weyl Kondo semimetals have so far been explored in ff-electron-based materials, Ce₃Bi₄Pd₃[28, 29] and several newly proposed Ce-, Pr- and U-based compounds[20, 30]. The present work leads us to propose a design principle for realizing Weyl Kondo semimetals in a new setting. Importantly, our theoretical results are expected to be robust against the effect of the residual interactions among the quasiparticles. This is so because the origin of the topological nodes lies in the symmetry constraints, which have recently been shown to operate on the eigenvectors of the matrix associated with the exact single-electron Green’s function of an interacting system[20].

Accordingly, our proof-of-principle demonstration enables us to advance a new materials design procedure for Weyl Kondo semimetals in the setting of dd-electron-based systems. The procedure would start from 3D lattices that can host flat bands from quantum interference. Examples include the pyrochlore lattice[40], the perovskite lattices[41], and other 3D versions of the bipartite crystalline lattices[42]. We seek materials with dd-elements, and utilize orbital-selective correlations to drive interacting flat bands that are Fermi-energy-bound. Symmetry constraints can then lead to either Dirac or (with the breaking of inversion or time-reversal symmetry) Weyl nodes in these emergent flat bands. The latter case corresponds to a Weyl Kondo semimetal.

The procedure for this materials identification approach goes beyond that for Weyl Kondo semimetals in ff-electron-based systems[30]. In addition to the requirement for both correlations and crystalline symmetry constraints, it also involves the crystal lattice conditions for the formation of flat bands in the bare dispersion. We reiterate that the noninteracting flat bands are not required to be near the Fermi energy. This is an important feature in the proposed materials design principle, given that the noninteracting flat bands in relevant materials are generically away from their Fermi energy.

Implications for fractional Chern insulators

Fractional Chern insulators, with a fractional quantum Hall effect and the associated fractional charge in a lattice setting, have been proposed in correlated models with an appropriate (such as 1/3) filling ratio of a flat band when the latter crosses the Fermi energy[43, 44, 45]. Experimental evidence has recently been identified in twisted bilayer graphene[31], in which the moiré bands are located near the Fermi energy, in a small external magnetic field. Our results on the Fermi-energy-bound emergent flat bands raise the possibility of another potential platform to realize the fractional Chern insulators, namely in dd-electron-based 2D systems. Indeed, when a spin-orbit coupling is included in the 2D model, the Dirac node is gapped leading to flat Z2Z_{2} topological bands (see Supplementary Note 4). The residual interactions (which develop beyond the saddle-point analysis in the slave-spin approach that we have carried out) could be ferromagnetic (see Supplementary Note 5). In that case, the flat band can develop a nonzero Chern number and can be analyzed for a lattice realization of fractional quantum Hall effect[46]. Indeed, the combination of the flatness of the associated bands (see Supplementary Fig. S4) and the aforementioned residual interactions among the heavy quasiparticles represents a condition that is similar to what happens in the moiré systems[31]; however, the Z2Z_{2} nature of the flat bands makes them distinct and rare[47]. Accordingly, with appropriate fillings, our results suggest that the corresponding dd-electron-based 2D materials provide a new setting for realizing a fractional Chern insulator. The naturalness of the emergent flat band crossing the Fermi energy makes our proposal robust. Thus, this represents a promising new direction for a systematic examination.

Discussion

Our work opens a new bridge between topological flat bands and correlation physics. The interaction effect tends to localize the molecular orbital that has the most overlap with the flat band. As a result, these molecular orbitals play the role of local moments, by analogy with the local spins of Kondo systems. Our work provides a rare non-perturbative way to address the interplay between correlations and topology effects in such flat band systems and a variety of correlated materials[48]. As such, it promises to elucidate the physics of correlated kagome transition-metal compounds[2, 3, 13, 14] and related materials. We also expect that our analysis will inspire new understandings of the correlation effects in moiré structures, which are increasingly being viewed from a Kondo perspective[49, 50, 51, 52, 53], as well as in other flat band systems[54]. Our work has also allowed us to advance a new materials design principle to identify Weyl Kondo semimetals in the new setting of dd-electron-based systems. We expect the interconnections that our work reveals among seemingly disparate systems to inspire new realizations and understandings of correlated topological effects in a wide variety of quantum materials and structures. Finally, we note that our theoretical result for the emergent flat band is now supported by experiment: In a frustrated-lattice material, an emergent flat band has been observed by angle resolved photoemission spectroscopy at the Fermi energy, even though the ab initio noninteracting band structure predicts a flat band that is considerably away from the Fermi energy[12].

Methods

Hubbard model on the clover lattice

The clover lattice has been discovered in real materials such as the van der Waals system Fe5_5GeTe2_2[32, 55]. As shown in Figure 1d, it contains five sites in each unit cell, which are reclassified into two groups as marked by the yellow and blue colors. For an illustrative purpose, we restrict our model to have only a dz2d_{z^2} orbital on each site. The case with other dd-orbitals, such as dxz/dyzd_{xz}/d_{yz}, have a similar realization of the geometry-induced flat bands[32]. We consider the Hubbard model written as H=H0+H1\mathcal{H}=\mathcal{H}_0+\mathcal{H}_1, where H0\mathcal{H}_0 is the kinetic term that connects the two different groups of sublattices and H1\mathcal{H}_1 represents the onsite one-orbital Hubbard interaction. We label the orbitals based on the group of sublattices to be A/BA/B and C/D/EC/D/E respectively. For each site, we consider the onsite Hubbard interaction,

H1=U∑i,αni,↑ηαni,↓ηα,(2)\mathcal{H}_1=U\sum_{i,\alpha}n_{i,\uparrow}^{\eta_\alpha}n_{i,\downarrow}^{\eta_\alpha}, \tag*{(2)}

where ηα\eta_\alpha (α=1∼5\alpha=1\sim5) goes through all the five orbitals in each unit cell. The kinetic Hamiltonian is written as

H0=∑ij,αβ,σtηiασ†ηjβσ−μ0∑iασηiασ†ηiασ+∑i,σ,α∈{C,D,E}mηiασ†ηiασ+∑iσα∈{D,E}γηiασ†ηiασ.(3)\mathcal{H}_0=\sum_{ij,\alpha\beta,\sigma}t\eta_{i\alpha\sigma}^{\dagger}\eta_{j\beta\sigma}-\mu_0\sum_{i\alpha\sigma}\eta_{i\alpha\sigma}^{\dagger}\eta_{i\alpha\sigma}+\sum_{i,\sigma,\alpha\in\{C,D,E\}}m\eta_{i\alpha\sigma}^{\dagger}\eta_{i\alpha\sigma}+\sum_{i\sigma\alpha\in\{D,E\}}\gamma\eta_{i\alpha\sigma}^{\dagger}\eta_{i\alpha\sigma}. \tag*{(3)}

Here tt denotes the nearest-neighbor hopping between the two sites that are connected by the solid lines shown in Figure 1d, and μ0\mu_0 is the chemical potential. In addition, mm denotes the energy splitting between the two groups of sublattices, which have a different local environment and thus generically have different energy levels. Finally, γ\gamma represents an additional energy splitting between CC and D/ED/E, which breaks the C3C_3 rotational symmetry. As mentioned earlier, we work with the case that breaks C3C_3 symmetry to simplify the symmetry characterization and, thus, the Wannier construction. It is possible that this C3C_3 symmetry breaking spontaneously appears as a result of interactions that drive a nematic order, although, for our illustrative purpose, we do not pursue this route specifically. A detailed analysis of the dispersion is shown in Supplementary Note 1.

Effective extended Hubbard model

We project the Hubbard model of the original lattice to the Wannier basis. This leads to the effective model expressed in terms of the dd and cc Wannier orbitals with Heff=H0+HintH_{\mathrm{eff}} = H_0 + H_{\mathrm{int}}. The kinetic term takes the following form:

H0=Hd+Hc+HV=∑ij,σtijd(diσ†djσ+h.c.)−∑i,σμdiσ†diσ+∑ij,αβ,σtijαβ(ciασ†cjβσ+h.c.)+∑iασ(Δα−μ)ciασ†ciασ+∑ij,ασVijα(diσ†cjασ+h.c.).(4)\begin{aligned} H_0 &= H_d + H_c + H_V \\ &= \sum_{ij,\sigma} t_{ij}^{d}\left(d_{i\sigma}^{\dagger}d_{j\sigma} + h.c.\right) - \sum_{i,\sigma}\mu d_{i\sigma}^{\dagger}d_{i\sigma} \\ &\quad+ \sum_{ij,\alpha\beta,\sigma} t_{ij}^{\alpha\beta}\left(c_{i\alpha\sigma}^{\dagger}c_{j\beta\sigma} + h.c.\right) + \sum_{i\alpha\sigma}(\Delta_{\alpha}-\mu)c_{i\alpha\sigma}^{\dagger}c_{i\alpha\sigma} \\ &\quad+ \sum_{ij,\alpha\sigma} V_{ij}^{\alpha}\left(d_{i\sigma}^{\dagger}c_{j\alpha\sigma} + h.c.\right). \tag*{(4)} \end{aligned}

Here, diσ†d_{i\sigma}^{\dagger} (ciασ†c_{i\alpha\sigma}^{\dagger}) creates a heavy (light) electron at the position ii with spin σ\sigma (and orbital α\alpha). In addition, tijdt_{ij}^{d} (tijαβt_{ij}^{\alpha\beta}) denotes the hopping parameter between the dd (cc) electrons at the positions ii and jj (orbitals α\alpha and β\beta). Moreover, VαV^{\alpha} represents the hybridization between the light orbital α\alpha and the heavy orbital. Finally, Δα\Delta_{\alpha} describes the difference in the energy levels between the dd orbital and cc orbitals, and μ\mu specifies the chemical potential. We will focus on the case with the dd-electron level being deep below the Fermi energy and will show that the interaction effect creates a heavy band near the Fermi energy. The noninteracting dispersion is shown in Figure 4a, where there is a Dirac crossing between the flat band and lower wide band located deep below the Fermi energy; the crossing is protected by the MxM_x lattice symmetry[25].

Slave spin method and self-consistent equations

We describe the U(1) slave spin approach[33]. Because the bandwidth of the heavy orbital is much smaller than those of the light orbitals, the interaction effect is most pronounced on the dd orbital. We therefore only introduce the SS representation on the dd orbital: diσ†=oiσ†fiσ†d_{i\sigma}^{\dagger} = o_{i\sigma}^{\dagger}f_{i\sigma}^{\dagger}. The auxiliary bosonic field oσ†=P+S+P−o_{\sigma}^{\dagger} = P^{+}S^{+}P^{-} is represented by the spin operator accompanied with the projection operators Piσ±=11/2±SiσzP_{i\sigma}^{\pm} = \frac{1}{\sqrt{1/2 \pm S_{i\sigma}^{z}}} suitable for a system that is away from half filling. We treat the SS formulation at the saddle-point level by fully decoupling the SS and auxiliary fermion operators.

This leads to the decoupled Hamiltonian:

Hf=∑k,σ⟨Oσ⟩2ϵkffkσ†fkσ+∑kσ(−μ−λσ+λσ0)fkσ†fkσ+∑k,ασ(Vkα⟨Oσ†⟩fkσ†ckασ+h.c.)+HcHS=∑σ[ϵ~f(⟨Oσ⟩Oσ†+h.c.)+∑α(V~αOσ†+h.c.)]+∑σλσ(Sσz+12)+HintS,(5)\begin{aligned} H^f &= \sum_{k,\sigma} \langle O_\sigma\rangle^2 \epsilon_k^f f^\dagger_{k\sigma}f_{k\sigma} + \sum_{k\sigma}(-\mu-\lambda_\sigma+\lambda_\sigma^0)f^\dagger_{k\sigma}f_{k\sigma} \\ &\quad+ \sum_{k,\alpha\sigma}\left(V_k^\alpha\langle O_\sigma^\dagger\rangle f^\dagger_{k\sigma}c_{k\alpha\sigma}+h.c.\right)+H_c \\ H^S &= \sum_\sigma\left[\widetilde{\epsilon}_f\left(\langle O_\sigma\rangle O_\sigma^\dagger+h.c.\right)+\sum_\alpha\left(\widetilde{V}_\alpha O_\sigma^\dagger+h.c.\right)\right] \\ &\quad+ \sum_\sigma\lambda_\sigma\left(S_\sigma^z+\frac{1}{2}\right)+H_{\mathrm{int}}^S, \tag*{(5)} \end{aligned}

where

HintS=U2(∑σSσz)2+∑αFα⟨nαc⟩∑σSσz,(6)H_{\mathrm{int}}^S=\frac{U}{2}\left(\sum_\sigma S_\sigma^z\right)^2+\sum_\alpha F_\alpha\langle n_\alpha^c\rangle\sum_\sigma S_\sigma^z, \tag*{(6)}

and ϵf(k)\epsilon_f(\mathbf{k}) and VkαV_k^\alpha are the Fourier transforms of tijt_{ij} and VijαV_{ij}^\alpha, respectively, ϵ~f=∑kϵf(k)⟨fk†fk⟩\widetilde{\epsilon}_f=\sum_k\epsilon_f(\mathbf{k})\langle f_k^\dagger f_k\rangle, and V~α=∑kVkα⟨fk†ck⟩\widetilde{V}_\alpha=\sum_kV_k^\alpha\langle f_k^\dagger c_k\rangle. In addition, Oσ=⟨P+⟩Sσ+⟨P−⟩O_\sigma=\langle P^+\rangle S_\sigma^+\langle P^-\rangle, λσ0=2(ϵ‾f+∑αV‾α)ησ\lambda_\sigma^0=2\left(\overline{\epsilon}_f+\sum_\alpha\overline{V}_\alpha\right)\eta_\sigma with ϵ‾f=ϵ~f⟨Oσ⟩⟨Oσ†⟩+c.c.\overline{\epsilon}_f=\widetilde{\epsilon}_f\langle O_\sigma\rangle\langle O_\sigma^\dagger\rangle+c.c., V‾=V~⟨Oσ†⟩+c.c.\overline{V}=\widetilde{V}\langle O_\sigma^\dagger\rangle+c.c. and ησ=12nσf−1/2(1−nσf)nσf\eta_\sigma=\frac{1}{2}\frac{n_\sigma^f-1/2}{(1-n_\sigma^f)n_\sigma^f}. Finally, we introduce the Lagrangian multiplier λσ\lambda_\sigma to remove the unphysical Hilbert space (see Supplementary Note 6). The pseudo-spin carries the U(1) charge degree of freedom; the quasiparticle weight associated with the coherent part near the Fermi level is described by Z=⟨Oσ⟩⟨Oσ†⟩Z=\langle O_\sigma\rangle\langle O_\sigma^\dagger\rangle. An (orbital-selective) Mott localization transition happens when some quasiparticle weight ZZ goes to zero.

In addition to the coherent quasiparticle peak, the SS method also calculates the contributions from the incoherent excitations. The Green’s function of the dd electron GdG_d is obtained by the convolution of the SS and the auxiliary fermions, with G(k,iωn)=∑iΩmGS(iΩm)Gf(k,iωn−iΩm)G(\mathbf{k},i\omega_n)=\sum_{i\Omega_m}G_S(i\Omega_m)G_f(\mathbf{k},i\omega_n-i\Omega_m), where GS(τ)=⟨Tτoσ(τ)oσ†(0)⟩G_S(\tau)=\langle T_\tau o_\sigma(\tau)o_\sigma^\dagger(0)\rangle and Gf(τ)=−⟨Tτfσ(τ)fσ†(0)⟩G_f(\tau)=-\langle T_\tau f_\sigma(\tau)f_\sigma^\dagger(0)\rangle. The spectral function is then obtained from A(k,ω)=−1πℑGdR(k,ω)A(\mathbf{k},\omega)=-\frac{1}{\pi}\Im G_d^R(\mathbf{k},\omega).

Data availability

The data that support the findings of this study are either presented in the manuscript or available at https://doi.org/10.5281/zenodo.11247849.

Code availability

The computer codes that were used to generate the data that support the findings of this study are available from the corresponding authors upon request.

References

Acknowledgment

We thank Gabriel Aeppli, Joseph Checkelsky, Han Wu, Yonglong Xie and Ming Yi for useful discussions. Work at Rice has primarily been supported by the U.S. DOE, BES, under Award No. DE-SC0018197 (model construction, L.C.,F.X.,S.S.), by the Air Force Office of Scientific Research under Grant No. FA9550-21-1-0356 (orbital-selective Mott transition, L.C.,F.X.,S.S.,H.H.,Q.S.), by the Robert A. Welch Foundation Grant No. C-1411 (model calculation, L.C.), and by the Vannevar Bush Faculty Fellowship ONR-VB N00014-23-1-2870 (conceptualization, Q.S.).The majority of the computational calculations have been performed on the Shared University Grid at Rice funded by NSF under Grant EIA-0216467, a partnership between Rice University, Sun Microsystems, and Sigma Solutions, Inc., the Big-Data Private-Cloud Research Cyberinfrastructure MRI-award funded by NSF under Grant No. CNS-1338099, and the Extreme Science and Engineering Discovery Environment (XSEDE) by NSF under Grant No. DMR170109. H.H. acknowledges the partial support of the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program (Grant Agreement No. 101020833). Work in Vienna was supported by the Austrian Science Fund (project I 5868-N - FOR 5249 - QUAST) and the ERC (Advanced Grant CorMeTop, No. 101055088). J.C. acknowledges the support of the National Science Foundation under Grant No. DMR-1942447, support from the Alfred P. Sloan Foundation through a Sloan Research Fellowship and the support of the Flatiron Institute, a division of the Simons Foundation. J.C. and Q.S. acknowledge the hospitality of the Aspen Center for Physics, which is supported by NSF grant No. PHY-2210452.

Author contributions

Q.S. conceived the research. L.C., F.X., S.S., H.H., J.C. and Q.S. carried out model studies. L.C., S.P., J.C. and Q.S. contributed to the development of the design principle for correlated topological phases in dd-electron systems. L.C. and Q.S. wrote the manuscript, with inputs from all authors.

Competing interests

The authors declare no competing interests.

Additional information

Correspondence and requests for materials should be addressed to Q.S. ([email protected])

Supplementary information for: Emergent flat band and topological Kondo semimetal driven by orbital-selective correlations

Supplementary Note 1: Band structure of the original Hamiltonian

In this section, we show the dispersion in the original lattice. We fix the reference energy scale t=2t = 2. The noninteracting band structure is shown in Figure S1, in which the bands are clearly separated into the lower two bands and upper three bands. The spectral function is provided in Figure S1, where the sharp peak denotes the position of the flat band. We focus on the upper three bands for our Wannier construction.

Figure S1. The noninteracting electronic structure of the original Hubbard model. a, The band structure in the original clover lattice, with μ=2.2\mu= 2.2, m=2m = 2 and γ=−0.2\gamma= -0.2. b, The corresponding density of states.

Supplementary Note 2: Destructive interference and the origin of the flat band

In this section, we discuss the origin of the flat band in the clover lattice. Consider the nearest-neighbor hoppings between the blue and yellow sublattices. The kinetic part of the Hamiltonian takes the form as

H0(k)=(02×2HkHk03×3)(7)\mathcal{H}_0(\mathbf{k}) = \begin{pmatrix} 0_{2\times2} & \mathcal{H}_{\mathbf{k}} \\ \mathcal{H}_{\mathbf{k}} & 0_{3\times3} \end{pmatrix} \tag*{(7)}

where Hk\mathcal{H}_{\mathbf{k}} is the hopping matrix between the two groups of sublattices. The localized wavefunction takes the following form[32]:

ψ(kx,ky)=(0 0 eky3−e−2ky3 e−kx23−ky6−ekx3+ky3 ekx23−ky6−e−kx3+ky3).(8)\psi(k_x,k_y)=\left(0\ 0\ e^{\frac{k_y}{3}}-e^{-\frac{2k_y}{3}}\ e^{-\frac{k_x}{2\sqrt{3}}-\frac{k_y}{6}}-e^{\frac{k_x}{\sqrt{3}}+\frac{k_y}{3}}\ e^{\frac{k_x}{2\sqrt{3}}-\frac{k_y}{6}}-e^{-\frac{k_x}{\sqrt{3}}+\frac{k_y}{3}}\right). \tag*{(8)}

Here, the amplitude is identically zero on the yellow sites because of the destructive interference effect. The real space wavefunction is shown in Figure S2.

Figure S2. Wavefunction for the electronic states of the flat band. Illustrated here is the real space representation of the electronic wavefunction for the flat band in the clover lattice.

Effective Multi-orbital Hubbard Model {#sec:Supplementary Note 3}

For the kinematic hoppings of the effective molecular orbitals, we refer to Table II in the SM of Ref. 11. However, we focus on the case such that, in the noninteracting case, the flat band lies considerably below the Fermi energy. In Figure S3, we present the DOS for the dd and cc electrons in the absence of any hybridization between the two orbitals. For the interaction part, the ratio between the Hubbard interaction of the effective dd electron and the one on the original lattice is u/U=0.15u/U = 0.15[11]. For the same reason that the onsite interaction among the cc-orbitals is unimportant, the Hund’s coupling does not play an important role here[11].

Figure S3. The noninteracting density of states projected to the Wannier orbitals. a,b, The DOS of the dd and cc electrons, respectively, in the absence of any hybridization between the two orbitals.

Analysis of the Effective Multi-orbital Hubbard Model with a Spin-Orbit Coupling

In this section, we discuss the influence of the spin-orbit coupling (SOC). For our purpose, it is adequate to consider a SOC that is small compared to the other energy scales. We can then analyze the effect of the SOC perturbatively, by using the same Wannier orbitals described above. We thus project the SOC of the original lattice onto the Wannier orbital basis. We consider the SOC along the zz direction of the original clover lattice, which takes the form as

Hsoc=∑i,σ=±itsocσ(ηiσC†ηiσD+ηiσD†ηiσE+ηiσE†ηiσC)+h.c.(9)\mathcal{H}_{\mathrm{soc}} = \sum_{i,\sigma=\pm} i t_{\mathrm{soc}}\sigma\left(\eta_{i\sigma C}^{\dagger}\eta_{i\sigma D} + \eta_{i\sigma D}^{\dagger}\eta_{i\sigma E} + \eta_{i\sigma E}^{\dagger}\eta_{i\sigma C}\right) + \mathrm{h.c.} \tag*{(9)}

Here, tsoct_{\mathrm{soc}} is the strength of the SOC, and σ=±\sigma= \pm represents spin up and spin down, respectively. The hopping parameters on the basis of the three Wannier orbitals are shown in Supplementary Table. I and Supplementary Table. II.

Because the value of the SOC is small, we use the same saddle-point parameters as we obtained in the case without the SOC. The renormalization factors ZZ for the different orbitals will then renormalize the SOC accordingly.

As shown in Figure S4, the SOC generically gaps out the Dirac nodes (located along Γ−K\Gamma- K) for our 2D model, both for the cases with and without interactions. This leads to flat Z2Z_2 topological bands. We stress that, the flatness of the emergent bands close to the Fermi energy is robust against the SOC.

Figure S4. The electronic dispersion in the presence of the SOC. Shown here are the electronic dispersions for u=0u=0 (a) and for u=1.6u=1.6 (b), with the SOC in the original lattice to have a strength of tsoc=0.06t_{\mathrm{soc}}=0.06.

(i,j)(i,j)(0,1)(0,1)(−1,0)(-1,0)(1,0)(1,0)(0,−1)(0,-1)
tia1+ja2t_{ia_1+ja_2}−0.04566+i0.01088-0.04566 + i0.01088−0.04566−i0.01088-0.04566 - i0.01088−0.04566+i0.01088-0.04566 + i0.01088−0.04566−i0.01088-0.04566 - i0.01088
±(2,−2)\pm(2,-2)(0,−2)(0,-2)(0,2)(0,2)(2,0)(2,0)
0.049910.049910.03701−i0.00310.03701 - i0.00310.03701+i0.00310.03701 + i0.00310.03701+i0.00310.03701 + i0.0031
(−2,0)(-2,0)(−1,2)(-1,2)(1,2)(1,2)(−2,1)(-2,1)
0.03701−i0.00310.03701 - i0.0031−0.03314−i0.00061-0.03314 - i0.00061−0.03314+i0.00061-0.03314 + i0.00061−0.03314+i0.00061-0.03314 + i0.00061
(2,1)(2,1)±(1,−1)\pm(1,-1)
−0.03314−i0.00061-0.03314 - i0.000610.020320.02032
(i,j)(i,j)(1,0)(1,0)(0,−1)(0,-1)(−1,2)(-1,2)(−2,1)(-2,1)
Via1+ja21V^1_{ia_1+ja_2}0.13301+i0.018930.13301 + i0.01893−0.13301+i0.01893-0.13301 + i0.018930.09264−i0.002990.09264 - i0.00299−0.09264−i0.00299-0.09264 - i0.00299
(1,−2)(1,-2)(2,−1)(2,-1)(0,−2)(0,-2)(2,0)(2,0)
0.07903−i0.007490.07903 - i0.00749−0.07903−i0.00749-0.07903 - i0.00749−0.06361+i0.00909-0.06361 + i0.009090.06361+i0.009090.06361 + i0.00909
(−2,3)(-2,3)(−3,2)(-3,2)(1,1)(1,1)(−1,−1)(-1,-1)
0.05494−i0.004310.05494 - i0.00431−0.05494−i0.00431-0.05494 - i0.00431−0.05082−i0.007-0.05082 - i0.0070.05082−i0.0070.05082 - i0.007
(i,j)(i,j)(1,−1)(1,-1)(−1,1)(-1,1)(0,1)(0,1)(−1,0)(-1,0)
Via1+ja22V^2_{ia_1+ja_2}−0.15053-0.15053−0.14577-0.145770.096−i0.008960.096 - i0.008960.096+i0.008960.096 + i0.00896
(0,−1)(0,-1)(1,0)(1,0)(0,0)(0,0)(2,−2)(2,-2)
−0.09143+i0.01766-0.09143 + i0.01766−0.09143−i0.01766-0.09143 - i0.01766−0.07918-0.079180.075630.07563
(1,1)(1,1)(−1,−1)(-1,-1)(−1,2)(-1,2)(−2,1)(-2,1)
−0.07542+i0.00449-0.07542 + i0.00449−0.07542−i0.00449-0.07542 - i0.004490.07311+i0.005090.07311 + i0.005090.07311−i0.005090.07311 - i0.00509

Table I. Parameters for the noninteracting Hamiltonian in the Wannier basis – part I. Shown here are the hopping parameters for the dd orbitals and the hybridization between the dd and cc orbitals for spin up when tsoc=0.06t_{\mathrm{soc}} = 0.06. The parameters in the spin-down case are the Hermitian conjugate of their spin-up counterparts.

Supplementary Note 5: Residual interactions and their competition

We have so far considered the dominant terms of the interactions that are projected into the Wannier basis: the on-site Coulomb interaction among the most localized molecular (Wannier) orbitals (dd) and the other on-site interactions between the dd orbitals and those of the more extended molecular orbitals (c1c_1 and c2c_2), as described by (1) in the main text. Taking into account these interactions at the saddle-point level of the effective Hamiltonian in the slave-spin approach leads to the emergent flat band, in which the (renormalized) electronic quasiparticle states are primarily associated with the effective dd orbitals.

(i,j)(i,j)(0,−1)(0,-1)(0,1)(0,1)(1,0)(1,0)(−1,0)(-1,0)
tia1+ja211t^{11}_{ia_1+ja_2}−0.35388+i0.01913-0.35388+i0.01913−0.35388−i0.01913-0.35388-i0.01913−0.35388−i0.01913-0.35388-i0.01913−0.35388+i0.01913-0.35388+i0.01913
±(1,−1)\pm(1,-1)±(2,−2)\pm(2,-2)(−1,−1)(-1,-1)(1,1)(1,1)
−0.15797-0.15797−0.08055-0.08055−0.07438+i0.00105-0.07438+i0.00105−0.07438−i0.00105-0.07438-i0.00105
(i,j)(i,j)(0,−1)(0,-1)(0,1)(0,1)(1,0)(1,0)(−1,0)(-1,0)
tia1+ja222t^{22}_{ia_1+ja_2}0.27789−i0.008260.27789-i0.008260.27789+i0.008260.27789+i0.008260.27789+i0.008260.27789+i0.008260.27789−i0.008260.27789-i0.00826
(0,−2)(0,-2)(0,2)(0,2)(−2,0)(-2,0)(2,0)(2,0)
−0.07658+i0.0008-0.07658+i0.0008−0.07658−i0.0008-0.07658-i0.0008−0.07658+i0.0008-0.07658+i0.0008−0.07658−i0.0008-0.07658-i0.0008
(−1,−1)(-1,-1)(1,1)(1,1)
0.05848+i0.00140.05848+i0.00140.05848−i0.00140.05848-i0.0014
(i,j)(i,j)(0,−1)(0,-1)(1,0)(1,0)(1,−2)(1,-2)(2,−1)(2,-1)
tia1+ja212t^{12}_{ia_1+ja_2}0.43550−i0.003950.43550-i0.00395−0.43550−i0.00395-0.43550-i0.003950.1959−i0.000790.1959-i0.00079−0.1959−i0.00079-0.1959-i0.00079
(0,1)(0,1)(−1,0)(-1,0)(2,−3)(2,-3)(3,−2)(3,-2)
0.15083−i0.001380.15083-i0.00138−0.15083−i0.00138-0.15083-i0.001380.11537−i0.000170.11537-i0.00017−0.11537−i0.000179-0.11537-i0.000179
(0,2)(0,2)(−2,0)(-2,0)(−1,2)(-1,2)(−2,1)(-2,1)
0.09257+i0.001330.09257+i0.00133−0.09257+i0.00133-0.09257+i0.001330.09107+i0.005160.09107+i0.00516−0.09107+i0.00516-0.09107+i0.00516
μ\muΔ1\Delta_1Δ2\Delta_2
0.150.151.811.812.7082.708

Table 2. Parameters for the noninteracting Hamiltonian in the Wannier basis – part II. Shown here are the hopping parameters between the cc electrons for spin up when tsoc=0.06t_{\mathrm{soc}}=0.06, the chemical potential μ\mu and the effective crystal field differences Δα\Delta_\alpha. The parameters for the spin-down case are the Hermitian conjugate of their spin-up counterparts.

The residual interactions for the emergent flat bands are mainly of three types, all of which are primarily pertinent to the dd electrons (as opposed to the c1c_1 and c2c_2 electrons). The first type is the Ruderman–Kittel–Kasuya–Yosida (RKKY) interaction between the dd electrons, mediated by the conduction cc electrons. For the generic case with typical fillings of the conduction cc electrons, the RKKY interaction tends to be antiferromagnetic. However, when the cc electrons are dilute, the RKKY interaction would be ferromagnetic.

The second type is the superexchange interaction between the dd electrons. This arises because, in the effective (projected) model, the hybridization between the dd orbitals in different unit cells is non-zero and so is the hybridization between the dd and cc orbitals. These kinetic terms induce a coupling between the low-energy renormalized (i.e., coherent) part of the dd-electron spectral weight near the Fermi energy and the incoherent high-energy part of the dd-electron spectral weight (which is controlled by both the uu-interaction and bare dd-level); when the latter are integrated out, the superexchange interaction ensues. The superexchange interaction is typically antiferromagnetic and is maximized near the (orbital-selective) Mott transition point. Deep in a Mott insulating phase, the value of the superexchange interaction can be derived using a direct perturbation theory in terms of the dd-dd hopping and dd-cc hybridization. In the bad metal region, i.e. in the regime of our interest–where the coherent part of the dd-electron spectral weight is nonzero but small, such that the system is in proximity to the orbital-selective Mott transition – the superexchange interaction can be constructed by considering the fluctuations beyond the saddle-point level such that the incoherent part of the slave-spin spectral weight is taken into account: This method is in parallel to the derivation for the superexchange interaction in a closely-related slave-particle representation, viz. the slave-rotor approach (see Fig. 2b of Ref. 56).

The third type of residual interactions consists of the direct exchange among the dd-electrons. They can be constructed from the projection of the original Hubbard interaction to the Wannier basis. Although the emergent molecular orbitals are exponentially localized, even the most compact (dd) orbital still extends over several lattice sites. The shared sites between the nearby molecular orbitals contribute to this ferromagnetic direct exchange interaction. These typically ferromagnetic interactions between the neighboring sites have the form of the Hund’s-like two-body interactions.

We have so far focused on spin exchange interactions. The processes leading to the first and third types of residual interactions naturally lead to density-density interactions between neighboring sites as well.

In summary, the forms of the residual interactions are of the spin-spin and density-density interactions among neighboring sites. An appropriate study of the dominant residual interactions should include the above terms, and the competition between them could lead to different phases. For example, if the antiferromagnetic interactions prevail over the ferromagnetic terms, they promote antiferromagnetic-ordering tendencies and the associated quantum criticality. If the ferromagnetic interactions dominate, they may split the Kramers degeneracy, leading to bands with a non-zero Chern number. With the help of other short-range interactions, including the short-range density-density interactions, this regime has the potential to yield fractional Chern insulating (FCI) states.

Slave spin method

In this section, we derive the slave spin equations we used in the main text. In the slave spin method, the creation operator for the correlated orbital is re-expressed as the product of a SS operator and an auxiliary fermionic operator: diσ†=fiσ†oiσ†d^\dagger_{i\sigma}=f^\dagger_{i\sigma}o^\dagger_{i\sigma}, where oiσ†=Piσ+Siσ+Piσ−o^\dagger_{i\sigma}=P^+_{i\sigma}S^+_{i\sigma}P^-_{i\sigma} and Piσ±=11/2±SiσzP^\pm_{i\sigma}=\frac{1}{\sqrt{1/2\pm S^z_{i\sigma}}}. To remove the unphysical Hilbert space, the SS and the auxiliary fermion need to follow the constraint:

Sioz+12=niof(10)S^z_{io}+\frac{1}{2}=n^f_{io} \tag*{(10)}

The interaction term can be expressed by the SS operators as described by Eq. (6) in the main text.

The noninteracting part of the Hamiltonian now takes the following form:

H0=∑ijσtij(fiσ†oiσ†ojσfjσ+h.c.)−∑iμfiσ†fiσ+∑ij,ασVijα(fiσ†oiσ†cjασ+h.c.)+∑ij,αβ,σtijαβ(ciασ†cjβσ+h.c.)+∑iασ(Δα−μ)ciασ†ciασ.(11)H_0 = \sum_{ij\sigma} t_{ij} \left(f_{i\sigma}^{\dagger}o_{i\sigma}^{\dagger}o_{j\sigma}f_{j\sigma} + h.c.\right) - \sum_i \mu f_{i\sigma}^{\dagger}f_{i\sigma} + \sum_{ij,\alpha\sigma} V_{ij}^{\alpha} \left(f_{i\sigma}^{\dagger}o_{i\sigma}^{\dagger}c_{j\alpha\sigma} + h.c.\right) + \sum_{ij,\alpha\beta,\sigma} t_{ij}^{\alpha\beta} \left(c_{i\alpha\sigma}^{\dagger}c_{j\beta\sigma} + h.c.\right) + \sum_{i\alpha\sigma}(\Delta_{\alpha}-\mu)c_{i\alpha\sigma}^{\dagger}c_{i\alpha\sigma}. \tag*{(11)}

We decouple the ff and oo at the saddle-point level with fiσ†oiσ†ojσfjσ=fiσ†fjσ⟨oiσ†ojσ⟩+⟨fiσ†fjσ⟩oiσ†ojσ+const.f_{i\sigma}^{\dagger}o_{i\sigma}^{\dagger}o_{j\sigma}f_{j\sigma}=f_{i\sigma}^{\dagger}f_{j\sigma}\langle o_{i\sigma}^{\dagger}o_{j\sigma}\rangle+\langle f_{i\sigma}^{\dagger}f_{j\sigma}\rangle o_{i\sigma}^{\dagger}o_{j\sigma}+const. and fiσ†oiσ†cjασ=⟨fiσ†cjασ⟩oiσ†+⟨oiσ†⟩fiσ†cjασ+const.f_{i\sigma}^{\dagger}o_{i\sigma}^{\dagger}c_{j\alpha\sigma}=\langle f_{i\sigma}^{\dagger}c_{j\alpha\sigma}\rangle o_{i\sigma}^{\dagger}+\langle o_{i\sigma}^{\dagger}\rangle f_{i\sigma}^{\dagger}c_{j\alpha\sigma}+const. We take the single site decoupling ⟨oioj†⟩≈⟨oi⟩⟨oj†⟩\langle o_i o_j^{\dagger}\rangle\approx\langle o_i\rangle\langle o_j^{\dagger}\rangle, and Taylor expand oσ†o_{\sigma}^{\dagger} at the saddle point as follows:

oσ†≈⟨Pσ+⟩S+⟨Pσ−⟩+⟨Pσ+⟩⟨Sσ+⟩⟨Pσ−⟩12(Sσz−⟨Sσz⟩)(−1nσ+11−nσ)=Oσ†+⟨Oσ†⟩12nσ−1/2(1−nσ)nσ(2Sσz−(2nσ−1))=Oσ†+⟨Oσ†⟩ησ[2Sσz−(2nσ−1)],(12)\begin{aligned} o_{\sigma}^{\dagger} &\approx\langle P_{\sigma}^{+}\rangle S^{+}\langle P_{\sigma}^{-}\rangle+\langle P_{\sigma}^{+}\rangle\langle S_{\sigma}^{+}\rangle\langle P_{\sigma}^{-}\rangle\frac{1}{2}\left(S_{\sigma}^{z}-\langle S_{\sigma}^{z}\rangle\right)\left(-\frac{1}{n_{\sigma}}+\frac{1}{1-n_{\sigma}}\right) \\ &= O_{\sigma}^{\dagger}+\langle O_{\sigma}^{\dagger}\rangle\frac{1}{2}\frac{n_{\sigma}-1/2}{(1-n_{\sigma})n_{\sigma}}\left(2S_{\sigma}^{z}-(2n_{\sigma}-1)\right) \\ &= O_{\sigma}^{\dagger}+\langle O_{\sigma}^{\dagger}\rangle\eta_{\sigma}\left[2S_{\sigma}^{z}-(2n_{\sigma}-1)\right], \tag*{(12)} \end{aligned}

where

Oσ†=⟨Pσ+⟩Sσ+⟨Pσ−⟩(13)O_{\sigma}^{\dagger}=\langle P_{\sigma}^{+}\rangle S_{\sigma}^{+}\langle P_{\sigma}^{-}\rangle \tag*{(13)}
ησ=12nσ−1/2(1−nσ)nσ.\eta_{\sigma}=\frac{1}{2}\frac{n_{\sigma}-1/2}{(1-n_{\sigma})n_{\sigma}}.

We notice that ⟨oσ⟩=⟨Oσ⟩\langle o_{\sigma}\rangle=\langle O_{\sigma}\rangle. Combining the above equations and introducing the Lagrangian multiplier λσ\lambda_{\sigma} to enforce the constraint in Eq. (10), we obtain the self-consistent equations as described in Eq. (5) of the main text.

The Green’s function of the dd electron is obtained by the convolution of the SS and the auxiliary fermion propagators, with

G(k,iωn)=∑mGS(iΩm)Gf(k,iωn−iΩm).(14)G(\mathbf{k},i\omega_n)=\sum_m G_S(i\Omega_m)G_f(\mathbf{k},i\omega_n-i\Omega_m). \tag*{(14)}

After an analytical continuation, we have the spectral function

Ad(k,ω)=2πZ∑σ∑n,m∣⟨n∣oσ∣m⟩⟨m∣oσ†∣n⟩δ(ω−Enm−ϵλk)∣Λλkf∣2×[e−βEm(1−fλk)+e−βEnfλk],(15)\begin{aligned} A_d(\mathbf{k},\omega)=\frac{2\pi}{Z}\sum_{\sigma}\sum_{n,m}\lvert\langle n\lvert o_{\sigma}\rvert m\rangle\langle m\lvert o_{\sigma}^{\dagger}\rvert n\rangle\delta(\omega-E_{nm}-\epsilon_{\lambda\mathbf{k}})\rvert\Lambda_{\lambda\mathbf{k}}^{f}\rvert^2 \\ &\times\left[e^{-\beta E_m}(1-f_{\lambda\mathbf{k}})+e^{-\beta E_n}f_{\lambda\mathbf{k}}\right], \tag*{(15)} \end{aligned}

where λ\lambda labels the bands solved from HfH^f, and Λλkf\Lambda_{\lambda\mathbf{k}}^{f} represents the corresponding ff-component in the eigenvectors. In addition, n,mn,m denote the eigenvectors of HSH^S, and Enm=En−EmE_{nm}=E_n-E_m is the energy difference between the two slave spin states. Finally, fλk=1eβελk+1f_{\lambda k} = \frac{1}{e^{\beta\varepsilon_{\lambda k}+1}} is the Fermi-Dirac distribution function.

Evolution of the spectral functions across the orbital-selective Mott transition

In this section, we provide further details on the evolution of the single-electron spectral weight when the interactions further increase from the case of u=1.6u = 1.6 shown in Figure 2b of the main text, approaching and into the OSMP. The evolution is primarily seen in the dd-electron DOS (red lines). The cc-electron DOS (marked in purple) is not markedly changed.

Figure S5 shows the DOS for u=2.2u = 2.2, which is smaller than but close to the critical value (uc=2.4u_{\mathrm{c}} = 2.4) for the orbital-selective Mott transition. Compared to the case of u=1.6u = 1.6 shown in Figure 2b of the main text, there is a significant transfer of the dd-electron spectral weight from the coherent part (the central red peak near the Fermi energy) to the incoherent part (the side peaks marked by the red dashed lines).

Figure S5 displays the DOS for u=2.5u = 2.5, where the system is in the OSMP (at u>uc=2.4u > u_{\mathrm{c}} = 2.4). Here, for the dd-electron DOS, the coherent peak has vanished. What is left is the incoherent part that comprises the lower and upper Hubbard bands marked by the red dashed lines.

The slave spin method incorporates the nonlocal spin correlations in the system[57], which is important in capturing the orbital-selective Mott transition. The result here has some similarities with its counterpart arising in the context of the Fe-based superconductors[58, 59]. There is however one important distinction: the existence of the flat band in the noninteracting dispersion means that the difference between the bandwidths (DflatD_{\mathrm{flat}} and DwideD_{\mathrm{wide}}) is especially large here.

Phase diagram of the effective Multi-orbital Model

In this section, we discuss the overall phase diagram in terms of uu and the bare local energy level of the dd electron (ϵd0\epsilon_{\mathrm{d}}^{0}). According to the kinetic Hamiltonian as described by Eq. (4) in the Methods, ϵd0=−μ\epsilon_{\mathrm{d}}^{0} = -\mu. As ϵd0\epsilon_{\mathrm{d}}^{0} decreases, the flat band further deviates from the Fermi energy. We scan the phase diagram as a function of both ϵd0\epsilon_{\mathrm{d}}^{0} and the interaction strength uu. The curves of quasiparticle weight versus uu for the dd electron are shown in Figure S6 for different values of ϵd0\epsilon_{\mathrm{d}}^{0}, accompanied by the corresponding colormap shown in Figure S6. The phase diagram can be divided into three regimes separated by the black dashed line, as depicted in Figure S6. In the dark blue region (region “I”), the quasiparticle weight of the dd electron completely vanishes, leading to the full localization of the dd orbital. In the bright yellow region (region “III”), the quasiparticle weight of the dd electron almost equals 1, and the flat band remains distant from the Fermi energy. The intermediate region (region “II”), marked as gradually changing shades of green, corresponds to the orbital-selective correlation region, where we realize an emergent flat band pinned to the Fermi energy. This can be further elucidated by comparing the single-particle excitations with a fixed ϵd0\epsilon_{d}^{0} and varying the interaction strengths uu. The dispersions displayed in Figure S6-Figure S6 correspond to the parameter settings marked with purple dots, as shown in Figure S6, with ϵd0=−0.15\epsilon_{d}^{0} = -0.15 and u=0.1u = 0.1, 1.11.1 and 1.51.5, respectively. These correspond to the weakly interacting region, mixed-valence region, and the Kondo limit. The dispersions with a local energy further away from the Fermi energy (ϵd0=−0.5\epsilon_{d}^{0} = -0.5) is shown in Figure S6-Figure S6, displaying a similar trend.

Orbital selective correlations in a general setting {#sec:Supplementary Note 9}

In this section, we illustrate that the conclusion we have reached is applicable to other types of geometry-induced flat band systems. (For an overall discussion, see the main text.) The example presented here is a model defined on the kagome lattice, which has recently been found to display a Kondo-destruction quantum critical point[19]. Here, we consider the case when the energy dispersion in the non-interacting limit is shown in Figure S7, where the flat band is situated away from the Fermi energy. The dispersion for the interacting case (u=2.1u = 2.1) is displayed in Figure S7, where once again, we observe the emergence of a flat band close to the Fermi energy. Due to the inclusion of spin-orbital coupling, the obtained solution is a topological insulator (TI), featuring a hybridization gap as depicted in the zoomed-in plot in Figure S7. Like in the canonical Kondo systems, the system is highly tunable. As shown in Figure S7, a relatively small Zeeman field leads to a substantial change in topology, where the TI gap is closed and nodal lines appear, which are protected by the MzM_z symmetry.

References

  1. [1]Mielke, A. Ferromagnetic ground states for the Hubbard model on line graphs. J. Phys. A 24, L73 (1991).DOI
  2. [2]Ye, L., Kang, M., Liu, J., von Cube, F., Wicker, C. R., Suzuki, T., Jozwiak, C., Bostwick, A., Rotenberg, E., Bell, D. C., Fu, L., Comin, R. & Checkelsky, J. G. Massive Dirac fermions in a ferromagnetic kagome metal. Nature 555, 638–642 (2018).DOI
  3. [3]Yao, M., Lee, H., Xu, N., Wang, Y., Ma, J., Yazyev, O., Xiong, Y., Shi, M., Aeppli, G. & Soh, Y. Switchable Weyl nodes in topological Kagome ferromagnet Fe₃Sn₂. arXiv:1810.01514 (2018).arxiv.org/abs/1810.01514
  4. [4]Paschen, S. & Si, Q. Quantum phases driven by strong correlations. Nat. Rev. Phys. 3, 9 (2021).DOI
  5. [5]Jiang, Y.-X., Yin, J.-X., Denner, M. M., Shumiya, N., Ortiz, B. R., He, J., Liu, X., Zhang, S. S., Chang, G., Belopol­s­ki, I. et al. Unconventional chiral charge order in kagome superconductor KV₃Sb₅. Nat. Mater. 20, 1353 (2021).DOI
  6. [6]Ill, C. M., Das, D., Yin, J. X., Liu, H., Gupta, R., Jiang, Y. X., Medarde, M., Wu, X., Lei, H. C., Chang, J. J., Dai, P., Si, Q., Miao, H., Thomale, R., Neupert, T., Shi, Y., Khasanov, R., Hasan, M. Z., Luetkens, H. & Guguchia, Z. Time-reversal symmetry-breaking charge order in a correlated kagome superconductor. Nature 602, 245–250 (2022).
  7. [7]Zhou, S. & Wang, Z. Chern fermi pocket, topological pair density wave, and charge-4e and charge-6e superconductivity in kagomé superconductors. Nat. Commun. 13, 7288 (2022).DOI
  8. [8]Teng, X., Chen, L., Ye, F., Rosenberg, E., Liu, Z., Yin, J.-X., Jiang, Y.-X., Oh, J. S., Hasan, Z., Neubauer, K., Gao, B., Xie, Y., Hasimoto, M., Lu, D., Jozwiak, C., Bostwick, A., Rotenberg, E., Birgeneau, R., Chu, J.-H., Yi, M. & Dai, P. Discovery of charge density wave in a correlated kagome lattice antiferromagnet. Nature 490–495 (2022).DOI
  9. [9]Yin, J.-X., Jiang, Y.-X., Teng, X., Hossain, M. S., Mardanya, S., Chang, T.-R., Ye, Z., Xu, G., Denner, M. M., Neupert, T., Lienhard, B., Deng, H.-B., Setty, C., Si, Q., Chang, G., Guguchia, Z., Gao, B., Shumiya, N., Zhang, Q., Cochran, T. A., Multer, D., Yi, M., Dai, P. & Hasan, M. Z. Discovery of charge order and corresponding edge state in kagome magnet FeGe. Phys. Rev. Lett. 129, 166401 (2022).DOI
  10. [10]Setty, C., Lane, C. A., Chen, L., Hu, H., Zhu, J.-X. & Si, Q. Electron correlations and charge density wave in the topological kagome metal FeGe. arXiv preprint arXiv:2203.01930 (2022).arxiv.org/abs/2203.01930
  11. [11]Hu, H. & Si, Q. Coupled topological flat and wide bands: Quasiparticle formation and destruction. Sci. Adv. 9, eadg0028 (2023).DOI
  12. [12]Huang, J., Chen, L., Huang, Y., Setty, C., Gao, B., Shi, Y., Liu, Z., Zhang, Y., Yilmaz, T., Vescovo, E., Hashimoto, M., Lu, D., Yakobson, B. I., Dai, P., Chu, J.-H., Si, Q. & Yi, M. Non-fermi liquid behaviour in a correlated flat-band pyrochlore lattice. Nat. Phys. 20, 603–609 (2024).DOI
  13. [13]Ye, L., Fang, S., Kang, M., Kaufmann, J., Lee, Y., John, C., Neves, P. M., Zhao, S. Y. F., Denlinger, J., Jozwiak, C., Bostwick, A., Rotenberg, E., Kaxiras, E., Bell, D. C., Janson, O., Comin, R. & Checkelsky, J. G. Hopping frustration-induced flat band and strange metallicity in a kagome metal. Nat. Phys. 20, 610–614 (2024).DOI
  14. [14]Ekahana, S. A., Soh, Y., Tamai, A., Gosálbez-Martínez, D., Yao, M., Hunter, A., Fan, W., Wang, Y., Li, J., Kleibert, A., Vaz, C. A. F., Ma, J., Lee, H., Xiong, Y., Yazyev, O. V., Baumberger, F., Shi, M. & Aeppli, G. Anomalous electrons in a metallic kagome ferromagnet. Nature 627, 67–72 (2024).DOI
  15. [15]Wirth, S. & Steglich, F. Exploring heavy fermions from macroscopic to microscopic length scales. Nat. Rev. Mater. 1, 16051 (2016).DOI
  16. [16]Kirchner, S., Paschen, S., Chen, Q., Wirth, S., Feng, D., Thompson, J. D. & Si, Q. Colloquium: Heavy-electron quantum criticality and single-particle spectroscopy. Rev. Mod. Phys. 92, 011002 (2020).DOI
  17. [17]Zhang, J., Yilmaz, T., Meier, J., Pai, J., Lapano, J., Li, H., Kaznatcheev, K., Vescovo, E., Huon, A., Brahlek, M., Ward, T. Z., Lawrie, B., Moore, R., Lee, H. N., Wang, Y. L., Miao, H. & Sales, B. Flat band induced negative magnetoresistance in multi-orbital kagome metal. arXiv preprint arXiv:2105.08888 (2021).arxiv.org/abs/2105.08888
  18. [18]Deng, Y., Yu, Y., Song, Y., Zhang, J., Wang, N. Z., Sun, Z., Yi, Y., Wu, Y. Z., Wu, S., Zhu, J., Wang, J., Chen, X. H. & Zhang, Y. Gate-tunable room-temperature ferromagnetism in two-dimensional Fe₃GeTe₂. Nature 563, 94–99 (2018).DOI
  19. [19]Chen, L., Xie, F., Sur, S., Hu, H., Paschen, S., Cano, J. & Si, Q. Metallic quantum criticality enabled by flat bands in a kagome lattice. arXiv preprint arXiv:2307.09431 (2023).arxiv.org/abs/2307.09431
  20. [20]Hu, H., Chen, L., Setty, C., Grefe, S. E., Prokofiev, A., Kirchner, S., Paschen, S., Cano, J. & Si, Q. Topological semimetal without quasiparticles. arXiv:2110.06182 (2021).arxiv.org/abs/2110.06182
  21. [21]Bradlyn, B., Elcoro, L., Cano, J., Vergniory, M. G., Wang, Z., Felser, C., Aroyo, M. I. & Bernevig, B. A. Topological quantum chemistry. Nature 547, 298–305 (2017).
  22. [22]Cano, J., Bradlyn, B., Wang, Z., Elcoro, L., Vergniory, M. G., Felser, C., Aroyo, M. I. & Bernevig, B. A. Building blocks of topological quantum chemistry: Elementary band representations. Phys. Rev. B 97, 035139 (2018).DOI
  23. [23]Po, H. C., Vishwanath, A. & Watanabe, H. Symmetry-based indicators of band topology in the 230 space groups. Nat. Commun. 8, 50 (2017).
  24. [24]Watanabe, H., Po, H. C., Zaletel, M. P. & Vishwanath, A. Filling-enforced gaplessness in band structures of the 230 space groups. Phys. Rev. Lett. 117, 096404 (2016).DOI
  25. [25]Cano, J. & Bradlyn, B. Band representations and topological quantum chemistry. Annu. Rev. Condens. Matter Phys 12, 225–246 (2021).DOI
  26. [26]Lai, H.-H., Grefe, S. E., Paschen, S. & Si, Q. Weyl-Kondo semimetal in heavy-fermion systems. Proc. Natl. Acad. Sci. U.S.A. 115, 93–97 (2018).DOI
  27. [27]Grefe, S. E., Lai, H.-H., Paschen, S. & Si, Q. Weyl-Kondo semimetals in nonsymmor­phic systems. Phys. Rev. B 101, 075138 (2020).DOI
  28. [28]Dzsaber, S., Prochaska, L., Sidorenko, A., Eguchi, G., Svagera, R., Waas, M., Prokofiev, A., Si, Q. & Paschen, S. Kondo insulator to semimetal transformation tuned by spin-orbit coupling. Phys. Rev. Lett. 118, 246601 (2017).DOI
  29. [29]Dzsaber, S., Yan, X., Taupin, M., Eguchi, G., Prokofiev, A., Shiroka, T., Blaha, P., Rubel, O., Grefe, S. E., Lai, H.-H., Si, Q. & Paschen, S. Giant spontaneous hall effect in a nonmagnetic Weyl Kondo semimetal. Proc. Natl. Acad. Sci. U.S.A. 118, e2013386118 (2021).DOI
  30. [30]Chen, L., Setty, C., Hu, H., Vergniory, M. G., Grefe, S. E., Fischer, L., Yan, X., Eguchi, G., Prokofiev, A., Paschen, S., Cano, J. & Si, Q. Topological semimetal driven by strong correlations and crystalline symmetry. Nat. Phys. 18, 134101346 (2022).DOI
  31. [31]Xie, Y., Pierce, A. T., Park, J. M., Parker, D. E., Khalaf, E., Ledwith, P., Cao, Y., Lee, S. H., Chen, S., Forrester, P. R., Watanabe, K., Taniguchi, T., Vishwanath, A., Jarillo-Herrero, P. & Yacoby, A. Fractional chern insulators in magic-angle twisted bilayer graphene. Nature 600, 439–443 (2021).DOI
  32. [32]Wu, H., Chen, L., Malinowski, P., Jang, B. G., Deng, Q., Scott, K., Huang, J., Ruff, J. P. C., He, Y., Chen, X., Hu, C., Yue, Z., Oh, J. S., Teng, X., Guo, Y., Klemm, M., Shi, C., Shi, Y., Setty, C., Werner, T., Hashimoto, M., Lu, D., Yilmaz, T., Vescovo, E., Mo, S.-K., Fedorov, A., Denlinger, J. D., Xie, Y., Gao, B., Kono, J., Dai, P., Han, Y., Xu, X., Birgeneau, R. J., Zhu, J.-X., da Silva Neto, E. H., Wu, L., Chu, J.-H., Si, Q. & Yi, M. Reversible non-volatile electronic switching in a near-room-temperature van der waals ferromagnet. Nat. Commun. 15, 2739 (2024).DOI
  33. [33]Yu, R. & Si, Q. U(1) slave-spin theory and its application to mott transition in a multiorbital model for iron pnictides. Phys. Rev. B 86, 085104 (2012).DOI
  34. [34]Si, Q., Rabello, S., Ingersent, K. & Smith, J. L. Locally critical quantum phase transitions in strongly correlated metals. Nature 413, 804–808 (2001).DOI
  35. [35]Coleman, P., Pépin, C., Si, Q. & Ramazashvili, R. How do Fermi liquids get heavy and die? J. Phys. Cond. Matt. 13, R723 (2001).
  36. [36]Senthil, T., Vojta, M. & Sachdev, S. Weak magnetism and non-fermi liquids near heavy-fermion critical points. Phys. Rev. B 69, 035111 (2004).DOI
  37. [37]Paschen, S., Lühmann, T., Wirth, S., Gegenwart, P., Trovarelli, O., Geibel, C., Steglich, F., Coleman, P. & Si, Q. Hall-effect evolution across a heavy-fermion quantum critical point. Nature 432, 881 (2004).DOI
  38. [38]Shishido, H., Settai, R., Harima, H. & Ōnuki, Y. A drastic change of the Fermi surface at a critical pressure in CeRhIn₅: dhva study under pressure. J. Phys. Soc. Jpn. 74, 1103–1106 (2005).DOI
  39. [39]Armitage, N. P., Mele, E. J. & Vishwanath, A. Weyl and Dirac semimetals in three-dimensional solids. Rev. Mod. Phys. 90, 015001 (2018).DOI
  40. [40]Bergman, D. L., Wu, C. & Balents, L. Band touching from real-space topology in frustrated hopping models. Phys. Rev. B 78, 125104 (2008).DOI
  41. [41]Weeks, C. & Franz, M. Topological insulators on the Lieb and perovskite lattices. Phys. Rev. B 82, 085310 (2010).DOI
  42. [42]Regnault, N., Xu, Y., Li, M.-R., Ma, D.-S., Jovanovic, M., Yazdani, A., Parkin, S. S. P., Felser, C., Schoop, L. M., Ong, N. P., Cava, R. J., Elcoro, L., Song, Z.-D. & Bernevig, B. A. Catalogue of flat-band stoichiometric materials. Nature 603, 824–828 (2022).
  43. [43]Sheng, D., Gu, Z.-C., Sun, K. & Sheng, L. Fractional quantum Hall effect in the absence of landau levels. Nat. Commun. 2, 1–5 (2011).DOI
  44. [44]Neupert, T., Santos, L., Chamon, C. & Mudry, C. Fractional quantum hall states at zero magnetic field. Phys. Rev. Lett. 106, 236804 (2011).pubmed.ncbi.nlm.nih.gov/21770534
  45. [45]Regnault, N. & Bernevig, B. A. Fractional chern insulator. Phys. Rev. X 1, 021014 (2011).
  46. [46]Wang, J., Cano, J., Millis, A. J., Liu, Z. & Yang, B. Exact Landau Level Description of Geometry and Interaction in a Flatband. Phys. Rev. Lett. 127, 246403 (2021).DOI
  47. [47]Devakul, T., Crépel, V., Zhang, Y. & Fu, L. Magic in twisted transition metal dichalcogenide bilayers. Nat. Commun. 12, 6730 (2021).DOI
  48. [48]Checkelsky, J. G., Bernevig, B. A., Coleman, P., Si, Q. & Paschen, S. Flat bands, strange metals and the kondo effect. Nat. Rev. Mater. (2024). URL https://doi.org/10.1038/s41578-023-00644-z.
  49. [49]Zhao, W., Shen, B., Tao, Z., Han, Z., Kang, K., Watanabe, K., Taniguchi, T., Mak, K. F. & Shan, J. Gate-tunable heavy fermions in a moiré kondo lattice. Nature 616, 61–65 (2023).DOI
  50. [50]Ramires, A. & Lado, J. L. Emulating heavy fermions in twisted trilayer graphene. Phys. Rev. Lett. 127, 026401 (2021).DOI
  51. [51]Song, Z.-D. & Bernevig, B. A. Magic-angle twisted bilayer graphene as a topological heavy fermion problem. Phys. Rev. Lett. 129, 047601 (2022).DOI
  52. [52]Guercı, D., Wang, J., Zang, J., Cano, J., Pixley, J. H. & Millis, A. Chiral Kondo lattice in doped MoTe₂/WSe₂ bilayers. Sci. Adv. 9, eade7701 (2023).
  53. [53]Chou, Y.-Z. & Das Sarma, S. Kondo lattice model in magic-angle twisted bilayer graphene. Phys. Rev. Lett. 131, 026501 (2023).DOI
  54. [54]Addison, Z. & Trivedi, N. Flat bands arising from spin-orbit assisted orbital frustration. Phys. Rev. B 106, 235144 (2022).DOI
  55. [55]May, A. F., Ovchinnikov, D., Zheng, Q., Hermann, R., Calder, S., Huang, B., Fei, Z., Liu, Y., Xu, X. & McGuire, M. A. Ferromagnetism near room temperature in the cleavable van der Waals crystal Fe₅GeTe₂. ACS Nano 13, 4436–4442 (2019).DOI
  56. [57]Hu, H., Chen, L., Zhu, J.-X., Yu, R. & Si, Q. Orbital-selective mott phase as a dehybridization fixed point. arXiv preprint arXiv:2203.06140 (2022).arxiv.org/abs/2203.06140
  57. [58]Huang, J., Yu, R., Xu, Z., Zhu, J.-X., Oh, J. S., Jiang, Q., Wang, M., Wu, H., Chen, T., Denlinger, J. D., Mo, S.-K., Hashimoto, M., Michiardi, M., Pedersen, T. M., Gorikov, S., Zhdanovich, S., Damascelli, A., Gu, G., Dai, P., Chu, J.-H., Lu, D., Si, Q., Birgeneau, R. J. & Yi, M. Correlation-driven electronic reconstruction in FeTe₁−ₓSeₓ. Commun. Phys. 5, 29 (2022).DOI
  58. [59]Yu, R., Hu, H., Nica, E. M., Zhu, J.-X. & Si, Q. Orbital selectivity in electron correlations and superconducting pairing of iron-based superconductors. Front. Phys. 9, 578347 (2021).DOI

Paper details

vlnc:W_bqzf3mfn
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

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