betaFeedback

WorksPhysics and Astronomy

Exciton collective modes in a bilayer of axion insulator MnBi2Te4\text{MnBi}_2 \text{Te}_4

Abstract

We investigate the emergence of an exciton condensate and associated collective modes in a bilayer configuration of MnBi2Te4\text{MnBi}_2\text{Te}_4, an antiferromagnetic topological insulator and van der Waals material, recognized for hosting axion physics. Utilizing a minimal low-energy Hamiltonian for the two layer system which is gapped by the intrinsic Néel order, we first employ mean-field theory to establish the conditions for exciton condensation. Our analysis identifies a nonzero, spin-singlet exciton order parameter which is tuned by external displacement field, temperature, and Coulomb attraction. Beyond the mean-field, we explore collective mode fluctuations in the uncondensed phase via many-body perturbation theory and the random phase approximation. From this, we derive the exciton spectral function which allows for a direct comparison between theoretical prediction and experimental observation. We detail how the softening of the collective mode peak is a function of the competition between interlayer detuning and thermal fluctuations. This work elucidates how the unique topological and magnetic environment of MnBi2Te4\text{MnBi}_2\text{Te}_4 offers a tunable platform for the realization and manipulation of exciton condensates and the corresponding collective excitations. Our findings contribute to understanding the interplay of topology and bosonic condensates, which could inspire application in optically accessing topological properties, dissipationless transport, and gate-tunable optoelectronics.

Introduction

Excitons, defined as bosonic composite particles composed of bound electron-hole pairs, are crucial for semiconductor optics and hold significant promise for new functionalities in optoelectronics, driving substantial recent research interest [3, 4, 5] [1, 2]. The long-standing quest to achieve Bose-Einstein Condensation (BEC) of excitons opens a pathway to exploring coherent bosonic states and emergent quantum phenomena. A number of systems have realized exciton condensates, such as: quantum Hall bilayers [6, 7], cavity exciton-polariton condensates out of equilibrium [8], and more recently some moiré materials [9] in the broader class of doped transition metal dichalcogenides [10].

Within this exciting landscape, MnBi2_{2}Te4_{4} stands out as a unique platform due to its classification as an intrinsic antiferromagnetic (AFM) topological insulator and its predicted capacity to host axion physics, where the axion field is denoted by θ\theta. The static axion field takes the quantized value of θ=π\theta= \pi [11, 12, 13], and is symmetry-protected by inversion P\mathcal{P} and magnetic-crystalline symmetry S=T1/2\mathcal{S} = \mathcal{T}_{1/2} where T\mathcal{T} is the time-reversal symmetry (TRS) operator and T1/2\mathcal{T}_{1/2} is the half-translation operator [14, 15]. However, long-range AFM order being established in the system corresponds to a staggered Zeeman effect and resultant broken TRS. Fluctuations in this state promotes θ\theta to a dynamical quantity and allows the axion field to take continuous values between 00 and π\pi. Here the dynamical axion collective mode is understood as the AFM spin wave [15].

MnBi2_{2}Te4_{4} is expected to host rich physics, with even-layer MnBi2_{2}Te4_{4} understood as an axion insulator, odd-layer MnBi2_{2}Te4_{4} as a Chern insulator, and bulk MnBi2_{2}Te4_{4} realizing type-I or -II Weyl semimetal phases in the presence of an external magnetic field, dependent on the orientation of the field [13]. Its unique layered, van der Waals (vdW) structure, combined with predictions for a tunable dynamical magnetoelectric effect, further underscores its significance for advanced materials research.

Here, we show that MnBi2_{2}Te4_{4} may harbor even more exotic physics, by demonstrating how bilayer excitons can form in a two-layer configuration of MnBi2_{2}Te4_{4}. Moreover, under certain conditions, we show how these excitons may condense to form a BEC of excitons. In general, bilayer excitons also have the advantage of more control over excitonic properties and new physics due to the tunability of the distance between the layers, applied tunable bias across the layers, and in some circumstances a tunable twist angle as in twisted bilayer graphene [16, 17]. All this motivates our selection of the bilayer MnBi2_{2}Te4_{4} to study the interplay between topology and exciton order, and how the exciton spectrum may be used to detect nontrivial topology.

Although the physics of excitons and their condensation have garnered significant attention, particularly in systems with extended lifetimes [12, 18], and axion insulators such as MnBi2_{2}Te4_{4} are recognized for their robust topological properties [13], the interplay between exciton condensation and the unique topological-magnetic environment of materials such as MnBi2_{2}Te4_{4} remains largely unexplored. Existing research has predicted spin-dependent exciton transport and vortex formation in exciton condensates due to Berry phase effects, and provided compelling evidence for exciton condensation in bilayer electron systems, even under conditions of electron-hole imbalance leading to TRS-breaking excitonic topological orders [6, 19]. Separately, the concept of a dynamical axion field, where magnetic fluctuations are coupled with electromagnetic fields, has been proposed in topological magnetic insulators, offering new avenues for detection [13, 20, 21, 22]. However, a detailed theoretical investigation into the formation and collective dynamics of excitons within the complex intrinsic antiferromagnetic and axion-insulator phases of MnBi2_2Te4_4, especially in a bilayer configuration, is currently missing. Understanding how the Néel order of the material and an external displacement field modulate exciton properties could unveil new functionalities.

Here, we investigate the emergence of an exciton condensate and its associated collective modes in bilayer MnBi2_2Te4_4, establishing it as a novel and tunable platform for studying the competition between magnetic versus excitonic order. Our study develops a minimal low-energy Hamiltonian and employs mean-field theory to delineate the conditions necessary for exciton condensation in this system. Due to the simplicity of the model it is physically motivated to consider only the exciton order parameter with S=0S = 0 no spin-polarization. However, the framework is sufficiently generic to allow for spin-polarization in future work. The order parameter is critically dependent on factors such as an external displacement field and the intrinsic Néel order of MnBi2_2Te4_4. Furthermore, by extending our analysis beyond mean-field theory to include many-body perturbation theory and the random phase approximation, we characterize the spectral function of these collective modes, demonstrating their emergence from layer-coupling and the gapped Dirac cones induced by the Néel order. These findings highlight how MnBi2_2Te4_4’s unique topological and magnetic properties

provide an unprecedented tunable platform for realizing and manipulating exciton condensates and their collective excitations against a backdrop of topology. By linking exciton condensation to the magnetic and topological degrees of freedom of MnBi2_2Te4_4, these results open avenues for magnetoelectric control of excitonic order and for leveraging axionic excitons in topological sensors and field-tunable quantum materials. This framework provides a route toward electrically tunable exciton and polariton devices based on magnetic topological insulators, suggesting opportunities for dissipationless interlayer transport, exciton-based logic, and hybrid light–matter platforms for terahertz optoelectronics [23, 24, 25, 26].

Low energy model for bilayer MnBi2_2Te4_4

We consider for simplicity a bilayer configuration of MnBi2_2Te4_4: an antiferromagnetic topological insulator whose manganese atoms host the spin moment, as well as being a vdW material, see Figure 1(a) for the layer-septuple crystal structure. Indeed, it has already been experimentally demonstrated to achieve atomically thin flakes of MnBi2_2Te4_4 [27] [28]. Each layer additionally contains spin S=1/2S = 1/2 electrons, with spin σ\sigma that move in the plane. The two layers are connected by an externally applied displacement field, denoted DD, which allows for tunable charge separation between the upper and lower layers, see Figure 1(b) for a schematic of the proposed set-up.

Image not extracted

Figure 1. Crystal structure of MnBi2_2Te4_4 with space group Rˉ3m:H\bar{R}3m:H with (a) showing the isotropic view. A single septuple layer is highlighted, with the AFM magnetic moments of Mn atoms denoted with vectors. (b) Schematic of the configuration of the two layers of MnBi2_2Te4_4, where each sheet indicates a single septuple layer of material. The oppositely spin-polarized Dirac cones on each layer illustrate that this material is a topological insulator. The two layers are connected by an external displacement field, DD. The AFM ordering on each layer is indicated by Néel vector, NzN_z, which points along the zz-direction. The layer operators are denoted as τ±\tau^{\pm} which correspond to taking an electron from the lower layer to the upper layer, and vice versa. The electrostatic Coulomb interaction, VV, arises due to chemical potential imbalance between the upper and lower layers, which is driven by the displacement field.

We use an effective k⋅p\mathbf{k}\cdot\mathbf{p} approach, where k\mathbf{k} is the crystal momentum and p\mathbf{p} is the quantum mechanical momentum operator, such that the Hamiltonian describing the single-particle physics is

H^0=∑kΨk†[H0(k)−μ]Ψk,(1)\hat{H}_0=\sum_{\mathbf{k}}\Psi_{\mathbf{k}}^\dagger[H_0(\mathbf{k})-\mu]\Psi_{\mathbf{k}}, \tag*{(1)}

where Ψk\Psi_{\mathbf{k}} is the second-quantized electronic spinor in spin/layer space, and μ\mu is the chemical potential, which we set to zero without loss of generality. We take a Bloch Hamiltonian of

H0(k)=[vF(k×σ)⋅e^z+D+N⋅σ]τ3,(2)H_0(\mathbf{k})=[v_F(\mathbf{k}\times\boldsymbol{\sigma})\cdot\hat{\mathbf{e}}_z+D+\mathbf{N}\cdot\boldsymbol{\sigma}]\tau_3, \tag*{(2)}

for the bilayer MnBi2_2Te4_4 system, which captures the relevant physics [29, 30]. Each term in this Hamiltonian represents a crucial physical aspect of the system. These terms are, respectively, a Rashba-type Dirac surface state on each layer with chirality fixed to the layer (implying the material has a broken mirror plane symmetry but preserves overall inversion); DD represents an external bias or displacement field applied between the two layers, which

serves as a critical tuning parameter for driving a chemical potential imbalance between the upper and lower layers; and N⋅σ\mathbf{N}\cdot\boldsymbol{\sigma} incorporates the intrinsic N'eel order of the antiferromagnet. This last term reflects the observed magnetic ground state where spins are ferromagnetically aligned within each layer and antiferromagnetically between layers. The layers are indexed by the pseudospin operator τ3=±1\tau_3=\pm1 for the upper and lower layers, respectively.

The displacement field DD is the control knob for charge imbalance: it drives a finite population of electrons on the upper layer, and a finite density of holes on the lower layer. There will be a resultant Coulomb attraction due to this charge imbalance. In the case of the spatially separated bilayer, we distinguish between two different channels for the Coulomb interaction: interlayer and intralayer. We introduce the layer-resolved charge densities, which in real-space read as

δρU(r)=Ψ†(r)Π^UΨ(r),(3)\delta\rho_U(\boldsymbol{r}) = \Psi^\dagger(\boldsymbol{r})\hat{\Pi}_U\Psi(\boldsymbol{r}), \tag*{(3)}
δρL(r)=Ψ†(r)Π^LΨ(r).(4)\delta\rho_L(\boldsymbol{r}) = \Psi^\dagger(\boldsymbol{r})\hat{\Pi}_L\Psi(\boldsymbol{r}). \tag*{(4)}

Here the projection operators Π^U/L\hat{\Pi}_{U/L} of the electronic spinor Ψ(r)\Psi(\boldsymbol{r}) act to localize the charge onto a layer, so they count the number of electrons on each layer. To be explicit, this matrix is written as Ψ(r)=(ψU(r)ψL(r))\Psi(\boldsymbol{r}) = \begin{pmatrix} \psi_U(\boldsymbol{r}) \\ \psi_L(\boldsymbol{r}) \end{pmatrix}, where each element is a spinor. In this treatment we will consider only the interlayer Coulomb interactions, VV, which can lead to very stable excitons due to the large degree of spatial separation between the electron and hole.

The interacting Hamiltonian due to Coulomb interaction is thus

H^int=V∫d2r δρU(r)δρL(r),(5)\hat{H}_{int} = V \int d^2r \, \delta\rho_U(\boldsymbol{r})\delta\rho_L(\boldsymbol{r}), \tag*{(5)}

where the Coulombic interaction VV is taken to be short-ranged and constant. This interacting Hamiltonian is quartic in the fermionic creation and annihilation operators. The full Hamiltonian is the sum of the interacting and non-interacting terms, H=H0+HintH = H_0 + H_{int}. While the single-particle Hamiltonian H0H_0 is linear and diagonalizable, the interacting term requires further attention, as we show in the next section.

Exciton condensate within mean field theory

We employ mean field theory to handle the many-body interaction term HintH_{int} which couples the bilinear density operators from the upper and lower layers, producing a quartic fermionic term. Within mean field, this quartic interaction term gets replaced by an effective averaged field, which we denote Δ^\hat{\Delta}, by employing the Hubbard–Stratonovich transformation in the particle–hole sector, which serves to decouple the interaction channel and reduce it to an effective single particle problem. Detailed derivations are provided in Appendix § 1.

We begin in the imaginary-time path integral formalism, which allows us to consider finite temperature effects, where the fermionic action reads

S[ψˉ,ψ]=∫0βd3x(ψˉ[∂∂τ+H0(k)]ψ)+Sint,(6)S[\bar{\psi},\psi] = \int_0^\beta d^3x \left(\bar{\psi}\left[\frac{\partial}{\partial\tau} + H_0(k)\right]\psi\right) + S_{int}, \tag*{(6)}

with the notation ∫d3x=∫0βdτ∫d2r\int d^3x = \int_0^\beta d\tau\int d^2r. Here τ∈(0,β)\tau\in(0,\beta) is imaginary time, for β=1/T\beta= 1/T inverse temperature. Note we use natural units in this work, where c=ℏ=kB=1c = \hbar= k_B = 1. The interlayer density-density interaction is simply Sint=∫dτHintS_{int} = \int d\tau H_{int}. One can show (see Appendix § 1) that the interacting term becomes decoupled using the Hubbard–Stratonovich transformation with the introduction of the new auxiliary complex matrix field, Δ^(x)\hat{\Delta}(x), as

S^int=∫d3xTr⁡[Δ^†Δ^V]+ψˉ(x)[Δ^τ++Δ^†τ−]ψ(x).(7)\hat{S}_{int} = \int d^3x \operatorname{Tr}\left[\frac{\hat{\Delta}^\dagger\hat{\Delta}}{V}\right] + \bar{\psi}(x)\left[\hat{\Delta}\tau^+ + \hat{\Delta}^\dagger\tau^-\right]\psi(x). \tag*{(7)}

Note that τ±=(τ1±iτ2)/2\tau^\pm= (\tau_1 \pm i\tau_2)/2 act as raising/lowering operators between layers. This new field can be physically interpreted as the bosonic exciton order parameter, given by the correlation function Δ^=V⟨ψU†ψL⟩\hat{\Delta} = V\langle\psi_U^\dagger\psi_L\rangle; or in the form of a complex matrix field Δ^=Δ0+Δ⋅σ\hat{\Delta} = \Delta_0 + \boldsymbol{\Delta}\cdot\boldsymbol{\sigma}. Here Δ0\Delta_0 denotes the un-spin-polarized order parameter and Δ⋅σ\boldsymbol{\Delta}\cdot\boldsymbol{\sigma} allows for spin-polarized excitons.

The exciton order parameter describes the coherent binding of an electron in the upper layer to a hole in the lower layer. To determine the conditions for the formation of a bilayer excitonic condensate, one may self-consistently solve for where Δ^\hat{\Delta} is nonzero via its gap equation. This is obtained using the saddle point of the action with respect to Δ^\hat{\Delta}

δSeffδΔ^=0,(8)\frac{\delta S_{\mathrm{eff}}}{\delta\hat{\Delta}} = 0, \tag*{(8)}

which yields the finite-temperature, self-consistent exciton gap equation as:

1VΔμ=−Tr⁡1β∑iεm∫d2p(2π)2G[Δ]τ−σμ,(9)\frac{1}{V}\Delta^\mu= -\operatorname{Tr}\frac{1}{\beta}\sum_{i\varepsilon_m}\int\frac{d^2p}{(2\pi)^2}\mathcal{G}[\Delta]\tau^-\sigma^\mu, \tag*{(9)}

where μ=0,1,2,3\mu= 0,1,2,3 labels the elements of the order parameter. The Matsubara Green’s function is

G=[iεm−[vF(k×σ)⋅e^z+D+Nzσz]τ3⋯ \mathcal{G} = \left[i\varepsilon_m - [v_F(\boldsymbol{k}\times\boldsymbol{\sigma})\cdot\hat{\boldsymbol{e}}_z + D + N_z\sigma_z]\tau^3 \cdots \right.
⋯−Δμστ+−Δˉμστ−]−1,(10)\left. \cdots- \Delta^\mu\sigma\tau^+ - \bar{\Delta}^\mu\sigma\tau^-\right]^{-1}, \tag*{(10)}

and εm\varepsilon_m denotes the imaginary time, Matsubara frequencies.

The mean-field phase diagram shown in Figure 2(a)-(Figure 2c) illustrates the conditions for exciton condensation obtained by solving the self-consistent gap equations (8)–(10). We assume the spin-unpolarized excitons will be the dominant species due to the simplicity of the model, and so we restrict our attention to this flavor going forwards. At low temperatures and moderate displacement field DD, the exciton order parameter with no spin polarization, Δ0\Delta_0, acquires a finite value, marking the onset of a condensate. As temperature increases, thermal fluctuations suppress coherence, and the order parameter vanishes above a critical temperature threshold. Similarly, at small displacement fields the condensate is disfavored: as DD goes to zero, the charge imbalance between layers is insufficient to stabilize bound electron–hole pairs. This leads to a dome-like region of stability in the (D,T)(D,T) plane. Notably, the overall scale of Δ0\Delta_0 and the width of the condensate dome are set by the Néel order, which controls the underlying Dirac gap and thus the available phase space for exciton formation. These results highlight the strong interplay between displacement field, temperature, and magnetic order in tuning exciton condensation in bilayer MnBi2_2Te4_4.

Mean-field phase diagrams and exciton spectral functions for a bilayer of axion insulator MnBi$_2$Te$_4$

Figure 2. Upper panel (a)-(c): Mean-field phase diagram of interlayer exciton order parameter Δ0\Delta_0: color encodes ∣Δ0∣/Nz|\Delta_0|/N_z obtained by solving the self-consistent gap equations derived from the Hubbard–Stratonovich decoupling of the interlayer Coulomb term from Eqs. (8)-(10). Axes are the dimensionless displacement field D/NzD/N_z and temperature T/NzT/N_z; panels (a)-(c) compare the interaction strengths VNz/vF2=0.285VN_z/v_F^2=0.285, 0.2930.293, and 0.3010.301. At moderate values of VNz/vF2VN_z/v_F^2, a finite Δ0\Delta_0 appears only once DD exceeds a threshold Dc(T)D_c(T) which creates a population imbalance of electrons (upper layer) and holes (lower layer). The critical line Tc(D)T_c(D) rises with both DD and VV as the interlayer particle-hole susceptibility grows, while the overall energy scale of Δ0\Delta_0 and TcT_c is set by the N'eel order NzN_z, which fixes the Dirac gap and available phase space for pairing. All energies are normalized by NzN_z, and vFv_F is the Fermi velocity. Lower panel (d)-(f): exciton spectral function A00(Ω,q=0)A_{00}(\Omega,q=0), in the uncondensed phase (T>TcT>T_c). Each subfigure shows two values of displacement field DD (solid vs. dashed curves) at fixed interaction strengths (a) VNz/vF2=0.285VN_z/v_F^2=0.285, (b) 0.2930.293, (c) 0.3000.300, and compares two temperatures (red vs. blue curves). Note that as we consider higher DD values, we must also increase the temperature so that we stay outside of the condensate dome shown in the phase diagram above each subfigure. The displacement field and temperature for each subfigure are indicated on the phase diagram plot immediately above it.

Obtaining the spectral function

To compare theoretical modeling to experimental findings requires derivation of the exciton spectral function. In order to characterize the exciton spectrum in the uncondensed phase (T>TcT > T_c for TcT_c temperature of exciton condensation), we expand the Matsubara action beyond mean field using the random-phase approximation (RPA). Integrating out the fermionic degrees of freedom yields the quadratic fluctuation effective action

S=∑qLμν(iΩm,q)ΔqμΔˉqν,(11)S = \sum_q \mathcal{L}_{\mu\nu}(i\Omega_m,q)\Delta_q^\mu\bar{\Delta}_q^\nu, \tag*{(11)}

where Lμν(iΩm,q)\mathcal{L}_{\mu\nu}(i\Omega_m,q) is the inverse exciton propagator that encodes the collective-mode correlations of the auxiliary Hubbard-Stratonovich field Δ^\hat{\Delta}. This is given explicitly as

Lμν(iΩm,q)=1Vδμν+12∫pTr⁡(G^0(p−q/2)σμτ−G^0(p+q/2)σντ++G^0(p−q/2)σμτ+G^0(p+q/2)σντ−),(12)\mathcal{L}_{\mu\nu}(i\Omega_m,q) = \frac{1}{V}\delta_{\mu\nu} + \frac{1}{2}\int_p \operatorname{Tr}\left(\hat{G}_0(p-q/2)\sigma^{\mu\tau_-}\hat{G}_0(p+q/2)\sigma^{\nu\tau_+} + \hat{G}_0(p-q/2)\sigma^{\mu\tau_+}\hat{G}_0(p+q/2)\sigma^{\nu\tau_-}\right), \tag*{(12)}

with G^0(p)\hat{G}_0(p) the single-particle Green’s function evaluated in the uncondensed state ⟨Δ^⟩=0\langle\hat{\Delta}\rangle= 0 (see Appendix B for detailed derivation). Note p=(p,iεm)p = (p,i\varepsilon_m) label the internal fermionic degrees of freedom, where pp are the fermion momentum and εm=2πT(m+1/2)\varepsilon_m = 2\pi T(m+1/2) is the fermionic Matsubara frequencies. Meanwhile q=(q,iΩn)q = (q,i\Omega_n) label the external bosonic collective modes, where qq is the momentum carried by the exciton collective mode, and Ωn\Omega_n are the bosonic Matsubara frequencies, which enter as arguments of the bosonic propagator. For the un-spin-polarized channel (μ=ν=0\mu= \nu= 0) and q=0q = 0, equation (12) becomes

L00(iΩn,0)=1V+∫pd2p(2π)2∑iεm(iεm−D)(iεm+D+iΩn)−Ep2[(iεm−D)2−Ep2][(iεm+D+iΩn)2−Ep2]+{D→−D},(13)\mathcal{L}_{00}(i\Omega_n,0) = \frac{1}{V} + \int_p \frac{d^2p}{(2\pi)^2}\sum_{i\varepsilon_m} \frac{(i\varepsilon_m-D)(i\varepsilon_m+D+i\Omega_n)-E_p^2}{[(i\varepsilon_m-D)^2-E_p^2][(i\varepsilon_m+D+i\Omega_n)^2-E_p^2]} + \{D\to-D\}, \tag*{(13)}

where Ep=Nz2+vF2p2E_p = \sqrt{N_z^2+v_F^2p^2} is the gapped Dirac dispersion. After performing the Matsubara-frequency summation using the residue theorem, analytic continuation iΩn→Ω+i0+i\Omega_n \to\Omega+i0^+ gives the retarded Green’s function response. The excitonic spectral function for q=0q=0 follows as

A00(Ω,0)=−1πIm⁡L00−1(Ω,0).(14)A_{00}(\Omega,0) = -\frac{1}{\pi}\operatorname{Im}\mathcal{L}_{00}^{-1}(\Omega,0). \tag*{(14)}

The peaks in the spectral function below the gap at Ω≃2Ep(k=0)=2Nz\Omega\simeq2E_p(k=0)=2N_z indicate an interlayer exciton collective mode: the position of the peak Ω(q)/Nz\Omega(q)/N_z give the mode energy, while the width of the peak encodes the damping, or lifetime, of the mode. As can be seen from the spectral function plots in Figure 2(d)-(f), the three variables under consideration are VV, DD, and TT. Lower values of VV allow for collective mode formation and result in strong peaks alongside sufficiently low temperature and strong DD, see Figure 2(d). Meanwhile higher VV leads to collective mode softening and an instability towards condensation, Figure 2(f). The displacement field sets the interlayer detuning, i.e. the relative distance between conduction and valence bands between the two coupled layers, see Figure 3. Before the condensate forms and DD is increased, this brings the valence and conduction bands closer together, see Figure 3(a) and (b), making it easier for excitons to form. Once a critical threshold bias field is reached, the excitons condense and the bands invert, see Figure 3(c). A lower DD means it takes more energy to couple electrons to holes, resulting in shorter lifetimes and exciton peaks nearest the gap. A larger DD field softens the modes as it takes increasingly less energy to form bound pairs. For higher DD, thermal fluctuations play less of a role as the strong chemical potential imbalance drives electrons and holes to become correlated. As the DD field is lowered, thermal fluctuations play a more significant role to block exciton formation and drive the exciton peak towards the particle-hole continuum, see Figure 2(f).

Many-body band structure for three displacement fields

Figure 3. Many-body band structure for constant T/Nz=0.15T/N_z = 0.15 and VNz/vF2=0.285VN_z/v_F^2 = 0.285. Band evolution as DD is scanned across the mean field phase diagram from the uncondensed into the condensed phase. At small displacement field D/Nz=0.3D/N_z = 0.3, the bilayer MnBi2_2Te4_4 spectrum consists of two gapped Dirac cones separated by the Néel order NzN_z. As DD increases through the condensate region of Figure 2(a)-(c), hybridization mediated by the exciton order parameter Δ0\Delta_0 drives a band inversion, signaling a possible topological phase transition. In this regime the condensate may be identified as an exciton insulator, where the excitonic gap supplants the single-particle Dirac mass.

The exciton spectral function is the bridge between theoretical prediction and experimental observation. Experimentally, several probes are used to directly measure this quantity. Momentum-resolved spectroscopies such as resonant inelastic X-ray scattering (RIXS) [31] or electron energy-loss spectroscopy (EELS) [32] measure intensities proportional to Aμν(iΩm,q)A_{\mu\nu}(i\Omega_m,q). Optical absorption or reflectivity experiments [33] probe the q=0q=0 component of the same spectral weight, while photoluminescence measurements [34] detect radiative recombination of excitons, again reflecting features in the spectral function.

Thus, the theoretical calculations of Aμν(iΩm,q)A_{\mu\nu}(i\Omega_m,q) provide a direct link to observables: sharp excitonic resonances predicted below the particle–hole continuum correspond to well-defined peaks in RIXS or optical spectra. By comparing the dispersion and linewidth of these peaks across theory and experiment, one can establish signatures of exciton formation and assess the stability of the condensate in bilayer MnBi2_2Te4_4.

Conclusion and Discussion

Our results demonstrate that a bilayer configuration of the antiferromagnetic topological insulator MnBi2_2Te4_4 can host a tunable exciton condensate whose properties are strongly governed by the interplay between displacement field, Néel order, and Coulomb interaction (related to critical temperature). Within mean-field theory, we identify a dome-shaped region in the (D,T)(D,T) plane where the non-spin-polarized order parameter Δ0\Delta_0 becomes finite, indicating a stable condensate driven by interlayer Coulomb attraction. Scanning across this phase diagram modifies the underlying band structure: at small DD, the system remains a trivial gapped Dirac insulator, while at intermediate fields exciton-mediated hybridization produces a band inversion as shown in Figure 3(c). This signals a transition into an exciton-insulator phase, in which the excitonic gap replaces the single-particle Dirac mass. Beyond mean-field, our RPA analysis of the spectral function A00(Ωm,q)A_{00}(\Omega_m,q) reveals the emergence of sharp excitonic resonances in the uncondensed phase. As DD approaches the condensation threshold, these modes soften, indicating the instability toward spontaneous coherence, while at higher DD the exciton resonance hardens and broadens as the bound state dissolves into the particle–hole continuum. The spectral peaks predicted here correspond directly to experimental observables accessible through RIXS, EELS, and optical spectroscopy, providing measurable fingerprints of exciton formation and collective-mode dynamics in MnBi2_2Te4_4 bilayers.

More broadly, these findings place the bilayer configuration of MnBi2_2Te4_4 in a unique class of correlated magnetic topological materials, where collective exciton order coexists and interacts with the system’s intrinsic axion electrodynamics. Because the axion angle θ\theta in MnBi2_2Te4_4 can vary dynamically in response to E⋅B\mathbf{E}\cdot\mathbf{B} drive, one could investigate in future work the coupling of the exciton and axion degrees of freedom. This indicates a regime where exciton condensation and axion electrodynamics become intertwined, such that the excitonic gap both reflects and renormalizes the material’s topological magnetoelectric response.

Our theoretical framework establishes the microscopic foundations for these phenomena and delineates the conditions under which they may emerge. By demonstrating how excitonic order can be tuned through an external bias, this work deepens the understanding of correlated bosonic topological states in magnetically ordered topological insulators, offering new routes toward field-tunable topological functionality. In the long term, these insights may inspire novel functionalities in gate-tunable optoelectronics [9] [2], where exciton coherence and axion-mediated magnetoelectric effects could be harnessed to engineer non-reciprocal photonic responses [23, 24] [1] or coherent interlayer tunneling devices [3, 6, 18]. Moreover, exploiting these low-energy collective modes for novel solid-state dark matter detectors which rely on the formation of axionic exciton-polaritons could offer a pathway to ultralight dark matter detection [22, 35, 36].

Acknowledgments

The authors would like to acknowledge fruitful discussions with Dr. Ioannis Petrides, Dr. Zexun Lin, and Jonathan Sanchez-Lopez. This work was supported by the Quantum Science Center (QSC), a National Quantum Information Science Center of the U.S. Department of Energy (DOE). We also acknowledge Grant Numbers GBMF8048 and GBMF12976 from the Gordon and Betty Moore Foundation.

Mean field and exciton order parameter

Here we present a more detailed derivation of the exciton order parameter, starting from the interacting Hamiltonian:

H^int=V∫d2r δρU(r)δρL(r).(15)\hat{H}_{int}=V\int d^2r\,\delta\rho_U(\mathbf{r})\delta\rho_L(\mathbf{r}). \tag*{(15)}

Which is recast in the path integral formulation of quantum mechanics as the following action

Sint=V∫d3x δρU(r)δρL(r),(16)S_{int}=V\int d^3x\,\delta\rho_U(\mathbf{r})\delta\rho_L(\mathbf{r}), \tag*{(16)}

where again the notation used is ∫d3x=∫dτ∫d2r\int d^3x=\int d\tau\int d^2r. The matrix which comprises the spin and layer resolved charge density operators is Ψ(r)=(ψU(r)ψL(r))\Psi(\mathbf{r})=\begin{pmatrix}\psi_U(\mathbf{r})\\\psi_L(\mathbf{r})\end{pmatrix}, and its conjugate Ψˉ(r)=(ψˉU(r) ψˉL(r))\bar{\Psi}(\mathbf{r})=(\bar{\psi}_U(\mathbf{r})\ \bar{\psi}_L(\mathbf{r})). Each element of the upper and lower operators denote spin. Explicitly, this is written as the annihilation operator ψU(r)=(ψU,↑(r)ψU,↓(r))\psi_U(\mathbf{r})=\begin{pmatrix}\psi_{U,\uparrow}(\mathbf{r})\\\psi_{U,\downarrow}(\mathbf{r})\end{pmatrix} for the upper layer; while the creation operator on the lower layer ψˉL=(ψˉL,↑(r) ψˉL,↓(r))\bar{\psi}_L=(\bar{\psi}_{L,\uparrow}(\mathbf{r})\ \bar{\psi}_{L,\downarrow}(\mathbf{r})).

One may rewrite the SintS_{int} in terms of the upper and lower spinors as

Sint=V∫d3x ψˉU(x)ψU(x)ψˉL(x)ψL(x)=−V∫d3x Tr⁡[ψU(x)ψˉL(x)ψL(x)ψˉU(x)].(17)S_{int}=V\int d^3x\,\bar{\psi}_U(x)\psi_U(x)\bar{\psi}_L(x)\psi_L(x)=-V\int d^3x\,\operatorname{Tr}\left[\psi_U(x)\bar{\psi}_L(x)\psi_L(x)\bar{\psi}_U(x)\right]. \tag*{(17)}

The minus sign indicates that this is an attractive force: energy decreases if you increase the Coulomb potential. Now, defining the following complex matrices

Φ(x)=ψU(x)ψˉL(x),Φ†(x)=ψL(x)ψˉU(x),(18)\Phi(x)=\psi_U(x)\bar{\psi}_L(x),\qquad\Phi^\dagger(x)=\psi_L(x)\bar{\psi}_U(x), \tag*{(18)}

gives the action

Sint=−V∫d3x Tr⁡[Φ(x)Φ†(x)].(19)S_{int}=-V\int d^3x\,\operatorname{Tr}\left[\Phi(x)\Phi^\dagger(x)\right]. \tag*{(19)}

Note, the product Φ(x)Φ†(x)\Phi(x)\Phi^\dagger(x) is always positive semi-definite, which in turn ensures the Coulomb interaction will always be attractive.

Next, we perform the Hubbard-Stratonovich transformation with the introduction of a new, complex field Δ^\hat{\Delta}, which allows us to transform the interaction action with quadratic dependence on Φ(x)\Phi(x) to a simpler form with linear dependence on Δ^\hat{\Delta}. This gives the following expression for SintS_{int}, now in terms of this new field

Sint=−∫d3x Tr⁡[Δ^†Δ^V]+ψˉ(x)(Δ^τ++Δ^†τ−)ψ(x).(20)S_{int}=-\int d^3x\,\operatorname{Tr}\left[\frac{\hat{\Delta}^\dagger\hat{\Delta}}{V}\right]+\bar{\psi}(x)\left(\hat{\Delta}\tau^++\hat{\Delta}^\dagger\tau^-\right)\psi(x). \tag*{(20)}

which in the form of complex fields may be expressed as Δ^=Δ0+Δ⋅σ\hat{\Delta}=\Delta_0+\boldsymbol{\Delta}\cdot\boldsymbol{\sigma} where σ\boldsymbol{\sigma} denotes the triplet of Pauli spin matrices, allowing for the possibility of spin-polarized excitons.

The full action is now given as

S=∫d3x ψˉ(x)(∂∂τ+H0(k))ψ(x)+∫d3x ψˉ(x)[Δ^τ++Δ^†τ−]ψ(x)+1V∫d3x Tr⁡[Δ^†Δ^]=1V∫d3x Tr⁡[Δ^†Δ^]−Tr⁡ln⁡G^−1[Δ^],(21)\begin{aligned} S &= \int d^3x\,\bar{\psi}(x)\left(\frac{\partial}{\partial\tau}+\mathcal{H}_0(k)\right)\psi(x)+\int d^3x\,\bar{\psi}(x)\left[\hat{\Delta}\tau^++\hat{\Delta}^{\dagger}\tau^-\right]\psi(x)+\frac{1}{V}\int d^3x\,\operatorname{Tr}\left[\hat{\Delta}^{\dagger}\hat{\Delta}\right]\\ &=\frac{1}{V}\int d^3x\,\operatorname{Tr}\left[\hat{\Delta}^{\dagger}\hat{\Delta}\right]-\operatorname{Tr}\ln\hat{G}^{-1}[\hat{\Delta}], \tag*{(21)} \end{aligned}

where we’ve introduced the Matsubara Green’s function G^−1=[iεm−H0(p)−Δ^τ+−Δ^†τ−]\hat{G}^{-1}=[i\varepsilon_m-\mathcal{H}_0(\mathbf{p})-\hat{\Delta}\tau^+-\hat{\Delta}^{\dagger}\tau^-]. Note εm\varepsilon_m denotes the imaginary time, fermionic Matsubara frequencies.

To see where the exciton order parameter is nonzero and obtain the mean-field phase diagram, we use the saddle point of the action with respect to Δ^\hat{\Delta},

δSeffδΔ^=0,(22)\frac{\delta S_{\mathrm{eff}}}{\delta\hat{\Delta}}=0, \tag*{(22)}

which yields

1VΔμ=−Tr⁡1β∑iεm∫dp2(2π)2G^[Δ^]τ−σμ,(23)\frac{1}{V}\Delta^\mu=-\operatorname{Tr}\frac{1}{\beta}\sum_{i\varepsilon_m}\int\frac{d p^2}{(2\pi)^2}\hat{G}[\hat{\Delta}]\tau^-\sigma^\mu, \tag*{(23)}

where μ=0,1,2,3\mu=0,1,2,3 label the elements of the order parameter, and the Matsubara Green’s function is

G^=[iεm−[vF(p×σ)⋅e^z+D+Nzσz]τ3−Δμσμτ+−Δˉμσμτ−]−1.(24)\hat{G}=\left[i\varepsilon_m-\left[v_F(\mathbf{p}\times\boldsymbol{\sigma})\cdot\hat{\mathbf{e}}_z+D+N_z\sigma_z\right]\tau^3-\Delta^\mu\sigma^\mu\tau^+-\bar{\Delta}^\mu\sigma^\mu\tau^-\right]^{-1}. \tag*{(24)}

A condensate corresponds to a nonzero order parameter. The phase diagrams shown in Figure 2(a)-(Figure 2c) plot the non-spin-polarized Δ0\Delta_0 order parameter as a function of displacement field DD and temperature TT, at various Coulomb potentials VV, in units normalized by the Néel order NzN_z.

Beyond mean field

Since the phase diagram of the exciton order parameter has large regions of the uncondensed phase, we now turn to study this phase where the order parameter is zero. To start, we rewrite the excitonic bilayer model with mean-field Hamiltonian

H^=∑pΨp†H(p)Ψp,(25)\hat{H}=\sum_{\mathbf{p}}\Psi_{\mathbf{p}}^\dagger\mathcal{H}(\mathbf{p})\Psi_{\mathbf{p}}, \tag*{(25)}
H(p)=[vF(p×σ)⋅e^z+D+N⋅σ]τ3+Δμσμτ++Δˉμσμτ−,\mathcal{H}(\mathbf{p})=\left[v_F(\mathbf{p}\times\boldsymbol{\sigma})\cdot\hat{\mathbf{e}}_z+D+\mathbf{N}\cdot\boldsymbol{\sigma}\right]\tau_3+\Delta^\mu\sigma^\mu\tau^++\bar{\Delta}^\mu\sigma^\mu\tau^-,

where Δμ\Delta^\mu are the excitonic orders derived self-consistently.

We want to derive the exciton spectrum for T>TcT>T_c where TcT_c is the temperature of equilibrium excitonic condensation. This is found within the random phase approximation (RPA) from the Matsubara action after integrating out the fermions:

S=∑q1V∣Δqμ∣2−Tr⁡ln⁡G^−1(p),(26)S=\sum_q\frac{1}{V}\lvert\Delta_q^\mu\rvert^2-\operatorname{Tr}\ln\hat{G}^{-1}(p), \tag*{(26)}

where G^−1(p)=iεm−H(p)\hat{G}^{-1}(p)=i\varepsilon_m-\mathcal{H}(p).

Note, p=(p,iεm)p=(\mathbf{p},i\varepsilon_m) label the internal fermionic degrees of freedom, where p\mathbf{p} is the fermion momentum and εm=2πT(m+1/2)\varepsilon_m=2\pi T(m+1/2), m∈Zm\in\mathbb{Z} are the fermionic Matsubara frequencies. Meanwhile q=(q,iΩm)q=(\mathbf{q},i\Omega_m) label the external bosonic collective modes, where q\mathbf{q} is the momentum carried by the exciton collective mode, and Ωm\Omega_m are the bosonic Matsubara frequencies, which enter as arguments of the bosonic propagator.

We may study fluctuations of the order parameter above TcT_c assuming ⟨Δ⟩=0\langle\Delta\rangle=0. We perform a Taylor series expansion around this averaged, zero value and apply many-body perturbation theory in the small fluctuation. Moreover, [H0(k),τ3]=0[\mathcal{H}_0(k),\tau_3]=0, which amounts to diagonalizing each layer separately to determine the coupling of fluctuations between layers. Substituting in the Green’s function gives

Tr⁡ln⁡G^−1=Tr⁡ln⁡[δq,0G^0−1(p)−Δ^τ+−Δ^†τ−](27)\operatorname{Tr}\ln\hat{G}^{-1}=\operatorname{Tr}\ln\left[\delta_{q,0}\hat{G}_0^{-1}(p)-\hat{\Delta}\tau^+-\hat{\Delta}^{\dagger}\tau^-\right] \tag*{(27)}
=Tr⁡ln⁡[1−G^0(p)Δ^τ+−G^0(p)Δ^†τ−],(28)=\operatorname{Tr}\ln\left[1-\hat{G}_0(p)\hat{\Delta}\tau^+-\hat{G}_0(p)\hat{\Delta}^{\dagger}\tau^-\right], \tag*{(28)}

where Δ^=Δμσμ\hat{\Delta}=\Delta^\mu\sigma^\mu and Δ^†=Δˉνσν\hat{\Delta}^{\dagger}=\bar{\Delta}^\nu\sigma^\nu. We perform a double Taylor series expansion of Eq. (28) for Δ^\hat{\Delta} and Δ^†\hat{\Delta}^{\dagger} to second order, using expansion of the logarithm ln⁡(1−x)≈−x−x22−O(x3)\ln(1-x)\approx-x-\frac{x^2}{2}-O(x^3). The lowest surviving order is the cross term

Tr⁡[12(G^0(p)Δ^τ+)(G^0(p)Δ^†τ−)].(29)\operatorname{Tr}\left[\frac{1}{2}(\hat{G}_0(p)\hat{\Delta}\tau^+)(\hat{G}_0(p)\hat{\Delta}^{\dagger}\tau^-)\right]. \tag*{(29)}

The effective action becomes

Seff=∑μ,ξ1V∣Δμ∣2+Tr⁡[12(G^0(p)Δ^τ+)(G^0(p)Δ^†τ−)],(30)S_{\mathrm{eff}}=\sum_{\mu,\xi}\frac{1}{V}\lvert\Delta^\mu\rvert^2+\operatorname{Tr}\left[\frac{1}{2}(\hat{G}_0(p)\hat{\Delta}\tau^+)(\hat{G}_0(p)\hat{\Delta}^{\dagger}\tau^-)\right], \tag*{(30)}

and we rewrite the expression above as

S=∑qLμν(q)ΔqμΔˉqν,(31)S=\sum_q\mathcal{L}_{\mu\nu}(q)\Delta^\mu_q\bar{\Delta}^\nu_q, \tag*{(31)}

with

Lμν(q)=1Vδμν+12∫pTr⁡[G^0(p−q/2)σμτ−G^0(p+q/2)σντ++G^0(p−q/2)σμτ+G^0(p+q/2)σντ−].(32)\mathcal{L}_{\mu\nu}(q)=\frac{1}{V}\delta_{\mu\nu}+\frac{1}{2}\int_p\operatorname{Tr}\left[\hat{G}_0(p-q/2)\sigma^\mu\tau^-\hat{G}_0(p+q/2)\sigma^\nu\tau^++\hat{G}_0(p-q/2)\sigma^\mu\tau^+\hat{G}_0(p+q/2)\sigma^\nu\tau^-\right]. \tag*{(32)}

So that the Lμν(q)\mathcal{L}_{\mu\nu}(q) matrix contains all the coefficients for the terms in Eq. (30). Note the Δ^\hat{\Delta} correlation functions are found as

−⟨ΔqμΔˉqν⟩≡Dμν(q)(33)-\langle\Delta^\mu_q\bar{\Delta}^\nu_q\rangle\equiv D^{\mu\nu}(q) \tag*{(33)}
→Dμν(q)=−[L^(q)]μν−1(34)\to D^{\mu\nu}(q)=-[\hat{\mathcal{L}}(q)]^{-1}_{\mu\nu} \tag*{(34)}

Therefore we can obtain the spectral function Aμν(iΩn,q)A_{\mu\nu}(i\Omega_n,q) of Δ^\hat{\Delta} by analytically continuing Lμν−1(q)\mathcal{L}^{-1}_{\mu\nu}(q).

This can be evaluated by noting that in G^0\hat{G}_0, the exciton order parameter Δ^=0\hat{\Delta}=0 so that [G^0,τ3]=0[\hat{G}_0,\tau_3]=0 such that we can diagonalize G^0\hat{G}_0 in each layer separately (in uncondensed phase G^0\hat{G}_0 is diagonal in layer). We diagonlize in each subpsace as:

G^0−1(iϵm,p)=iϵm−τ3h^pwithh^p=D+N⋅σ+vF(p×σ)z(35)\hat{G}_0^{-1}(i\epsilon_m,\mathbf{p})=i\epsilon_m-\tau_3\hat{h}_{\mathbf{p}}\quad\text{with}\quad\hat{h}_{\mathbf{p}}=D+\mathbf{N}\cdot\boldsymbol{\sigma}+v_F(\mathbf{p}\times\boldsymbol{\sigma})_z \tag*{(35)}

This gives the following expression for the Lμν\mathcal{L}_{\mu\nu} matrix elements

Lμν(iΩn,q)=1Vδμν+12∫pTr⁡[G^0+(iϵm,p)σμG^0−(iϵm+iΩn,p)σν+G^0−(iϵm,p)σμG^0+(iϵm+iΩn,p)σν](36)\mathcal{L}_{\mu\nu}(i\Omega_n,q)=\frac{1}{V}\delta_{\mu\nu}+\frac{1}{2}\int_p\operatorname{Tr}\left[\hat{G}_0^+(i\epsilon_m,p)\sigma^\mu\hat{G}_0^-(i\epsilon_m+i\Omega_n,p)\sigma^\nu+\hat{G}_0^-(i\epsilon_m,p)\sigma^\mu\hat{G}_0^+(i\epsilon_m+i\Omega_n,p)\sigma^\nu\right] \tag*{(36)}
=1Vδμν+12∫pTr⁡[([iϵm−D−N⋅σ−vF(p×σ)z]σμ[iϵm+iΩn+D+N⋅σ+vF(p×σ)z]σν(iϵm−D)2−Ep2)(iϵm+iΩn+D)2−Ep2)+([iϵm+D−N⋅σ−vF(p×σ)z]σμ[iϵm+iΩn−D+N⋅σ+vF(p×σ)z]σν(iϵm+D)2−Ep2)(iϵm+iΩn−D)2−Ep2)](37)\begin{aligned} &=\frac{1}{V}\delta_{\mu\nu}+\frac{1}{2}\int_p\operatorname{Tr}\left[\left(\frac{[i\epsilon_m-D-\mathbf{N}\cdot\boldsymbol{\sigma}-v_F(\mathbf{p}\times\boldsymbol{\sigma})_z]\sigma^\mu[i\epsilon_m+i\Omega_n+D+\mathbf{N}\cdot\boldsymbol{\sigma}+v_F(\mathbf{p}\times\boldsymbol{\sigma})_z]\sigma^\nu}{(i\epsilon_m-D)^2-E_p^2)(i\epsilon_m+i\Omega_n+D)^2-E_p^2}\right)\right.\\ &\left.\qquad+\left(\frac{[i\epsilon_m+D-\mathbf{N}\cdot\boldsymbol{\sigma}-v_F(\mathbf{p}\times\boldsymbol{\sigma})_z]\sigma^\mu[i\epsilon_m+i\Omega_n-D+\mathbf{N}\cdot\boldsymbol{\sigma}+v_F(\mathbf{p}\times\boldsymbol{\sigma})_z]\sigma^\nu}{(i\epsilon_m+D)^2-E_p^2)(i\epsilon_m+i\Omega_n-D)^2-E_p^2}\right)\right] \tag*{(37)} \end{aligned}

We will only focus on the q=0q=0, μ=ν=0\mu=\nu=0 un-spin-polarized case, which is motivated by the fact that this was the only non-zero order parameter:

L00(iΩn,0)=1V+∫pd2p(2π)2∑iϵm((iϵm−D)(iϵm+D+iΩn)−Ep2[(iϵm−D)2−Ep2][(iϵm+D+iΩn)2−Ep2]+⋯+(iϵm+D)(iϵm−D+iΩn)−Ep2[(iϵm+D)2−Ep2][(iϵm−D+iΩn)2−Ep2]).(38)\mathcal{L}_{00}(i\Omega_n,0)=\frac{1}{V}+\int_p\frac{d^2p}{(2\pi)^2}\sum_{i\epsilon_m}\left(\frac{(i\epsilon_m-D)(i\epsilon_m+D+i\Omega_n)-E_p^2}{[(i\epsilon_m-D)^2-E_p^2][(i\epsilon_m+D+i\Omega_n)^2-E_p^2]}+\cdots+\frac{(i\epsilon_m+D)(i\epsilon_m-D+i\Omega_n)-E_p^2}{[(i\epsilon_m+D)^2-E_p^2][(i\epsilon_m-D+i\Omega_n)^2-E_p^2]}\right). \tag*{(38)}

We solve for the poles of this matrix element via the residue theorem, and obtain the following expression:

L00(iΩn,0)=1V−∫d2p(2π)2[f(β(D−Ep))2D+iΩn−2Ep+f(β(D+Ep))2D+iΩn+2Ep]+[f(β(−D−Ep))−2D+iΩn−2Ep+f(β(−D+Ep))−2D+iΩn+2Ep].(39)\mathcal{L}_{00}(i\Omega_n,0)=\frac{1}{V}-\int\frac{d^2p}{(2\pi)^2}\left[\frac{f(\beta(D-E_p))}{2D+i\Omega_n-2E_p}+\frac{f(\beta(D+E_p))}{2D+i\Omega_n+2E_p}\right]+\left[\frac{f(\beta(-D-E_p))}{-2D+i\Omega_n-2E_p}+\frac{f(\beta(-D+E_p))}{-2D+i\Omega_n+2E_p}\right]. \tag*{(39)}
Dependence of the interlayer Coulomb interaction on the critical exciton condensation temperature

Figure 4. Dependence of the interlayer Coulomb interaction VV on the critical exciton condensation temperature TcT_{\mathrm{c}}. The solid black curve delineates where the self-consistent mean-field solution yields a finite exciton order parameter Δ0\Delta_{0} at zero displacement field D=0D=0. The red dashed line indicates an interaction strength VV such that no exciton condensation can take place. The purple dashed line indicates a threshold voltage, defining the minimal interaction strength required to overcome thermal and single-particle energy scales for exciton binding. Physically, larger VV enhances interlayer electron–hole attraction, raising TcT_{\mathrm{c}} and stabilizing the condensate, whereas for V<VcV<V_{\mathrm{c}} only incoherent excitonic fluctuations persist.

The notation used above defines the weighting function as f(βz)=12tanh⁡(βz2)f(\beta z)=\frac{1}{2}\tanh\left(\frac{\beta z}{2}\right). Eq. (39) is then inverted to obtain the excitonic spectral function, whose poles correspond to the exciton frequencies.

For the case for zero displacement field and in the uncondensed phase, the spectral function is a sharp quasiparticle peak such that it is a δ\delta-function in energy, with location in energy following the dispersion of the quasiparticle as Ep=Nz2+vF2p2E_{p}=\sqrt{N_{z}^{2}+v_{\mathrm{F}}^{2}p^{2}} for vFv_{\mathrm{F}} the Fermi velocity. Therefore, one only needs to integrate over momentum. The spectral function is solved by analytic continuation iΩn→Ω+i0+i\Omega_{n}\to\Omega+i0^{+}

A00(Ω,0)=−1πIm⁡[L00−1(Ω,0)].(40)\mathcal{A}_{00}(\Omega,0)=-\frac{1}{\pi}\operatorname{Im}\left[\mathcal{L}_{00}^{-1}(\Omega,0)\right]. \tag*{(40)}

References

  1. [3]Lujun Wang, Sotirios Papadopoulos, Fadil Iyikanat, Jian Zhang, Jing Huang, Takashi Taniguchi, Kenji Watanabe, Michel Calame, Mickael L. Perrin, F. Javier García de Abajo, and Lukas Novotny. Exciton-assisted electron tunnelling in van der Waals heterostructures. Nature Materials, 22(9), 2023.DOI
  2. [4]Wojciech Jankowski, Joshua Thompson, Bartomeu Monserrat, and Robert-Jan Slager. Excitonic topology and quantum geometry in organic semiconductors. Nature Communications, 16, 05 2025.
  3. [5]Juan Esteve-Paredes, M. García-Blázquez, Alejandro Uría-Álvarez, María Camarasa-Gómez, and Juan Palacios. Excitons in nonlinear optical responses: shift current in MoS₂ and GeS monolayers, 06 2024.
  4. [6]J. P. Eisenstein and A. H. MacDonald. Bose-Einstein condensation of excitons in bilayer electron systems. Nature (London), 432(7018):691–694, December 2004.
  5. [7]Xiaomeng Liu, Kenji Watanabe, Takashi Taniguchi, Bertrand Halperin, and Philip Kim. Quantum Hall drag of exciton condensate in graphene. Nature Physics, 13, 05 2017.
  6. [8]J. Kasprzak, Maxime Richard, Stefan Kundermann, A Baas, P Jeambrun, J Keeling, Francesca Maria Marchetti, M Szymańska, R André, J Staehli, Vincenzo Savona, Peter Littlewood, Benoit Deveaud, and Dang Le Si. Bose-Einstein condensation of exciton polaritons. Nature, 443:409–14, 10 2006.pubmed.ncbi.nlm.nih.gov/17006506
  7. [9]Richien Xiong, Jacob H. Nie, Samuel L. Brantly, Patrick Hays, Renee Sailus, Kenji Watanabe, Takashi Taniguchi, Sefaattin Tongay, and Chenhao Jin. Correlated insulator of excitons in WSe₂/WS₂ moiré superlattices. Science, 380(6647):860–864, 2023.DOI
  8. [10]Benjamin Remez and Nigel R. Cooper. Leaky exciton condensates in transition metal dichalcogenide moiré bilayers. Phys. Rev. Res., 4:L022042, May 2022.DOI
  9. [11]Yu-Hang Li and Ran Cheng. Identifying axion insulator by quantized magnetoelectric effect in antiferromagnetic MnBi₂Te₄ tunnel junction. Phys. Rev. Res., 4:L022067, Jun 2022.DOI
  10. [12]Wang Yao and Qian Niu. Berry phase effect on the exciton transport and on the exciton bose-einstein condensate. Phys. Rev. Lett., 101:106401, Sep 2008.
  11. [13]Dongqin Zhang, Minji Shi, Tongshuai Zhu, Dingyu Xing, Haijun Zhang, and Jing Wang. Topological axion states in the magnetic insulator mnbi₂te₄ with the quantized magnetoelectric effect. Phys. Rev. Lett., 122:206401, May 2019.DOI
  12. [14]Jinlong Zhang, Dinghui Wang, Minji Shi, Tongshuai Zhu, Haijun Zhang, and Jing Wang. Large dynamical axion field in topological antiferromagnetic insulator mn2bi2te5. Chinese Physics Letters, 37(7):077304, jun 2020.
  13. [15]Rundong Li, Jing Wang, Xiao-Liang Qi, and Shou-Cheng Zhang. Dynamical axion field in topological magnetic insulators. Nature Physics, 6(4):284–288, 2010.DOI
  14. [16]Hiral Patel, Lujie Huang, Cheol-Joo Kim, Jiwoong Park, and Matt W. Graham. Stacking angle-tunable photoluminescence from interlayer exciton states in twisted bilayer graphene. Phys. Rev. Res., 10, 2019.DOI
  15. [17]Dohun Kim, Seyoung Jin, Takashi Taniguchi, Kenji Watanabe, Jurgen Smet, Gil Cho, and Youngwook Kim. Observation of 1/3 fractional quantum Hall physics in balanced large angle twisted bilayer graphene, 12 2024.
  16. [18]Ruishi Qi, Qize Li, Zuocheng Zhang, Jiahui Nie, Bo Zou, Zhiyuan Cui, Haleem Kim, Collin Sanborn, Sudi Chen, Jingxu Xie, Takashi Taniguchi, Kenji Watanabe, Michael F. Crommie, Allan H. MacDonald, and Feng Wang. Competition between excitonic insulators and quantum hall states in correlated electron–hole bilayers. Nature Materials, Aug 2025.arxiv.org/abs/2501.18168
  17. [19]Rui Wang, Tigran A. Sedrakyan, Baigeng Wang, Lingjie Du, and Rui-Rui Du. Excitonic topological order in imbalanced electron–hole bilayers. Nature, 619:045115, 2023.DOI
  18. [20]Rundong Li, Jing Wang, Xiaoliang Qi, and Shou-Cheng Zhang. Dynamical Axion Field in Topological Magnetic Insulators. Nature Phys., 6:284, 2010.DOI
  19. [21]Olivia Liebman, Jon Curtis, Ioannis Petrides, and Prineha Narang. Multiphoton spectroscopy of a dynamical axion insulator. Phys. Rev. Lett., Oct 2025.
  20. [22]Jian-Xiang Qiu, Barun Ghosh, Jan Schütte-Engel, Tiema Qian, Michael Smith, Yueh-Ting Yao, Junyeong Ahn, Yu-Fei Liu, Anyuan Gao, Christian Tzschaschel, Houchen Li, Ioannis Petrides, Damien Bérubé, Thao Dinh, Tianye Huang, Olivia Liebman, Emily M. Been, Joanna M. Blawat, Kenji Watanabe, Takashi Taniguchi, Kin Chung Fong, Hsin Lin, Peter P. Orth, Prineha Narang, Claudia Felser, Tay-Rong Chang, Ross McDonald, Robert J. McQueeney, Arun Bansil, Ivar Martin, Ni Ni, Qiong Ma, David J. E. Marsh, Ashvin Vishwanath, and Su-Yang Xu. Observation of the axion quasiparticle in 2D MnBi₂Te₄. Nature, 641(8061):62–69, April 2025.DOI
  21. [23]Ming Sun, Xue Gao, Zhiyuan Li, and Zhizhen Yu. Exciton–Polariton Topological Insulator with an Array of Magnetic Dots. Physical Review Applied, 12(6):064028, 2019.DOI
  22. [24]Filippo L. Ruta, Harish K. Krishnamoorthy, Luis Zepeda-Núñez, and et al. Hyperbolic Exciton Polaritons in a van der Waals Magnet. Nature Communications, 14:1064, 2023.DOI
  23. [25]D. N. Basov, M. M. Fogler, and F. J. García de Abajo. Polaritons in van der Waals Materials. Science, 354(6309):aag1992, 2016.pubmed.ncbi.nlm.nih.gov/27738142
  24. [26]Elizabeth Mertz, Wei-Ting Hsu, and Vinod M. Menon. Materials for Excitons–Polaritons: Exploiting the Diversity of Light–Matter Engineering. MRS Bulletin, 2024.
  25. [27]Chang Liu, Yongchao Wang, Hao Li, Ouyang Chun, Yaoxin Li, Jiacheng Li, Ke He, Yong Xu, Jinsong Zhang, and Yayu Wang. Robust axion insulator and chern insulator phases in a two-dimensional antiferromagnetic topological insulator. Nature Materials, 19:1–6, 05 2020.DOI
  26. [31]Luuk J. P. Ament, Michel van Veenendaal, Thomas P. Devereaux, John P. Hill, and Jeroen van den Brink. Resonant inelastic x-ray scattering studies of elementary excitations. Rev. Mod. Phys., 83:705–767, Jun 2011.DOI
  27. [32]Colliex, Christian. From early to present and future achievements of eels in the tem. Eur. Phys. J. Appl. Phys., 97:38, 2022.DOI
  28. [33]Emilia Ridolfi, Caio H. Lewenkopf, and Vitor M. Pereira. Excitonic structure of the optical conductivity in MoS₂ monolayers. Phys. Rev. B, 97:205409, May 2018.DOI
  29. [34]Samuel Brem, August Eckmann, Dominik Christiansen, Florian Katsch, Malte Selig, Cedric Robert, Xavier Marie, Bernhard Urbaszek, Andreas Knorr, and Ermin Malic. Phonon-assisted photoluminescence from indirect excitons in monolayers of transition-metal dichalcogenides. Nano Letters, XXX, 02 2020.DOI
  30. [35]David J. E. Marsh, Kin Chung Fong, Erik W. Lentz, Libor Šmejkal, and Mazhar N. Ali. Proposal to detect dark matter using axionic topological antiferromagnets. Phys. Rev. Lett., 123:121601, Sep 2019.DOI
  31. [36]Andrea Mitridate, Tanner Trickle, Zhengkang Zhang, and Kathryn M. Zurek. Detectability of axion dark matter with phonon polaritons and magnons. Phys. Rev. D, 102:095005, Nov 2020.DOI

Paper details

vlnc:W_5gk5xug9
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