betaFeedback

WorksPhysics and Astronomy

Photon condensation driven by scalar spin chirality

Abstract

Gauge invariance forbids equilibrium photon condensation for spatially uniform electric-dipole coupling, motivating alternative routes based on orbital magnetic response. We show that triangular molecular magnets coupled to a cavity support photon condensation driven by emergent scalar spin chirality. Starting from the half-filled Hubbard model in the Mott regime, we derive an effective Hamiltonian in which virtual ring exchange couples the cavity flux directly to scalar spin chirality. Mean-field theory, collective-fluctuation calculations, and finite-size exact diagonalization reveal a phase with spontaneous chirality ordering and a macroscopic cavity displacement, together with hybrid chirality-photon excitations. The condensation threshold can be tuned electrically through the molecular chirality splitting. Our results establish emergent orbital degrees of freedom in frustrated magnets as a route to equilibrium light-matter ordering beyond conventional dipolar cavity quantum electrodynamics.

Equilibrium superradiant or photon-condensed phases in cavity quantum electrodynamics have long been constrained by gauge-invariance arguments. For spatially uniform electric-dipole coupling, the cancelation between paramagnetic and diamagnetic terms forbids a macroscopic ground-state photon occupation—the superradiant no-go theorem [1, 2, 3, 4, 5]. This restriction motivates alternative coupling mechanisms that allow photon condensation without external driving or dissipation.

Theoretical work has established that spatially varying cavity fields can evade the uniform-field no-go theorem and support equilibrium photon condensation through the orbital magnetic response of electrons [6, 7, 8, 9]. Unlike a uniform vector potential, a cavity field carrying magnetic flux modifies the electronic spectrum and can induce orbital currents. For continuous transitions, the instability is governed by the orbital susceptibility, connecting photon condensation to the magnetostatic Condon instability [7, 10, 11]. Related transitions have been explored in two-dimensional electron gases with spin-orbit coupling [12] and extended ladders [13]. When the electronic energy gain outweighs the photon cost, a coherent cavity field develops without external driving.

At the molecular scale, Van Vleck paramagnetism provides a route to orbital photon condensation in ensembles of square plaquettes [14]. Including Hubbard interactions and varying the electronic filling modify both the threshold and the order of the transition [15]. Here, we show that frustrated triangular Mott molecules support photon condensation driven by an emergent orbital degree of freedom: scalar spin chirality. Virtual ring exchange [16, 17] couples this orbital-current degree of freedom directly to cavity flux. Invariant under spin rotations yet odd under time reversal, the scalar chirality is a key

ingredient in frustrated magnets and chiral spin liquids [18, 19].

Light-induced control of correlated magnetism has been explored through laser-generated scalar-chirality interactions [20, 21], photon-dependent exchange [22, 23, 24, 25], and cavity modification of magnetic frustration and critical fluctuations [26, 27]. Vacuum fluctuations in gyrotropic cavities can also stabilize chiral spin liquids [28]. Here, the orbital response of correlated spins drives the cavity field to condense, spontaneously breaking time-reversal symmetry and selecting a finite scalar spin chirality.

System and model Hamiltonian. We consider a triangular Hubbard ring at half filling (Nel=3N_{\mathrm{el}}=3), coupled to a single cavity mode as illustrated in Figure 1, and neglect cavity losses. The cavity magnetic flux enters the electronic hopping through Peierls phases [29, 30], so that the Hamiltonian is H=Hel+ωca†aH=H_{\mathrm{el}}+\omega_{\mathrm{c}}a^{\dagger}a (ℏ=1\hbar=1), with

Cavity-coupled molecular rings

Figure 1. Cavity-coupled molecular rings. (a) A standing-wave microwave mode with molecular triangles at the magnetic field Bcav\mathbf{B}_{\mathrm{cav}} antinode (green) and electric field Ecav\mathbf{E}_{\mathrm{cav}} node (gray) to maximize inductive coupling. (b) Individual ring detail: an out-of-plane flux Φ^cav\hat{\Phi}_{\mathrm{cav}} couples the cavity to the scalar spin chirality CzC_z with strength g∝∣Bcav∣Ag \propto\left|\mathbf{B}_{\mathrm{cav}}\right| A, with AA the area of the molecule (assumed, for simplicity, to lie in the xx–yy plane). (c) Spontaneous symmetry breaking for g>gcg > g_c, where a macroscopic photon displacement ⟨a†+a⟩\langle a^\dagger+a\rangle emerges alongside a finite scalar spin chiral order parameter ⟨Cz⟩\langle C_z\rangle.

Hel=∑jσ(tjeigj(a†+a)cjσ†cj+1σ+h.c.)+U∑jnj↑nj↓.(1)H_{\mathrm{el}}=\sum_{j\sigma}\left(t_j e^{\mathrm{i}g_j(a^{\dagger}+a)}c_{j\sigma}^{\dagger}c_{j+1\sigma}+\mathrm{h.c.}\right)+U\sum_j n_{j\uparrow}n_{j\downarrow}. \tag*{(1)}

Here j=1,2,3j=1,2,3 labels sites cyclically around the triangle, tjt_j is the real hopping amplitude on bond (j,j+1)(j,j+1), and UU is the uniform on-site Coulomb repulsion. The operator cjσ†c_{j\sigma}^{\dagger} creates a spin-σ\sigma electron on site jj, while a†a^{\dagger} creates a cavity photon of frequency ωc\omega_{\mathrm{c}}. The dimensionless coefficient gjg_j specifies the photon-induced Peierls phase on bond (j,j+1)(j,j+1) [31].

Effective spin-photon coupling. In the Mott regime (U≫tjU\gg t_j), the cavity couples to the low-energy spin degrees of freedom through virtual charge fluctuations. We project onto the singly occupied subspace P\mathcal{P}, while virtual charge excitations in Q=1−P\mathcal{Q}=1-\mathcal{P} are treated perturbatively through a Schrieffer-Wolff transformation (SWT) [32, 33]. Splitting the Hamiltonian as H=H0+VH=H_0+V, with H0=U∑ini↑ni↓+ωca†aH_0=U\sum_i n_{i\uparrow}n_{i\downarrow}+\omega_{\mathrm{c}}a^{\dagger}a and VV the photon-dressed hopping, we obtain the effective Hamiltonian to third order in tj/Ut_j/U:

Hs−ph≈∑jJjSj⋅Sj+1−Ksin⁡Φ^cav χ+ωca†a,(2)H_{\mathrm{s-ph}}\approx\sum_j J_j\mathbf{S}_j\cdot\mathbf{S}_{j+1}-K\sin\hat{\Phi}_{\mathrm{cav}}\,\chi+\omega_{\mathrm{c}}a^{\dagger}a, \tag*{(2)}

where Sj=(1/2)∑σ,σ′Pcjσ†(σ)σσ′cjσ′P\mathbf{S}_j=(1/2)\sum_{\sigma,\sigma'}\mathcal{P}c_{j\sigma}^{\dagger}(\boldsymbol{\sigma})_{\sigma\sigma'}c_{j\sigma'}\mathcal{P} are the spin 1/21/2 operators, Jj=4∣tj∣2/UJ_j=4|t_j|^2/U, K=24t1t2t3/U2K=24t_1t_2t_3/U^2, Φ^cav=g(a†+a)\hat{\Phi}_{\mathrm{cav}}=g(a^{\dagger}+a), and g=∑j=13gjg=\sum_{j=1}^3g_j [34]. The first term describes the antiferromagnetic superexchange, while the second couples the cavity field to the scalar spin chirality generated by three-site virtual hopping [16, 17, 35, 36]

χ=S1⋅(S2×S3).(3)\chi=\mathbf{S}_1\cdot(\mathbf{S}_2\times\mathbf{S}_3). \tag*{(3)}

The scalar spin chirality is the low-energy representation of the orbital current generated by virtual electron hopping around the triangular plaquette. To make this connection explicit, consider the microscopic current operator on bond (j,j+1)(j,j+1),

I^j=ie tj∑σ(cjσ†cj+1σ−cj+1σ†cjσ).(4)\hat{I}_j=\mathrm{i}e\,t_j\sum_{\sigma}\left(c_{j\sigma}^{\dagger}c_{j+1\sigma}-c_{j+1\sigma}^{\dagger}c_{j\sigma}\right). \tag*{(4)}

At zero flux, the SWT of the current operator yields PeSI^je−SP=−I0χ\mathcal{P}e^S\hat{I}_je^{-S}\mathcal{P}=-I_0\chi, where I0=24e t1t2t3/U2I_0=24e\,t_1t_2t_3/U^2 and the sign follows the bond-current convention in Eq. (4). The scalar spin chirality therefore represents the emergent orbital-current degree of freedom of the Mott triangle, providing the microscopic link between correlated electron motion and orbital magnetism [17, 37].

In deriving Eq. (2), we further assumed U≫ωcU\gg\omega_{\mathrm{c}}, allowing photon-energy shifts in the virtual intermediate states to be neglected. The Supplemental Material (SM) [33] gives the corresponding expressions with photon-energy shifts retained in the virtual intermediate states. Because the cavity couples only through the Peierls phases, the global SU(2) spin symmetry is preserved. Consequently, the low-energy spectrum separates into a spin-3/2 quartet and two spin-1/2 doublets. The spin-3/2 sector is completely decoupled from the cavity, so the light-matter dynamics is entirely governed by the spin-1/2 manifold. Projecting onto this subspace therefore yields the minimal low-energy Hamiltonian governing the coupled chirality-photon dynamics,

Hc−ph=12Δ⋅C−K′Czsin⁡Φ^cav+ωca†a,(5)H_{\mathrm{c-ph}}=\frac{1}{2}\boldsymbol{\Delta}\cdot\mathbf{C}-K'C_z\sin\hat{\Phi}_{\mathrm{cav}}+\omega_{\mathrm{c}}a^{\dagger}a, \tag*{(5)}

where C=(Cx,Cy,Cz)\mathbf{C}=(C_x,C_y,C_z), with

Cx=23(2S1⋅S2−S2⋅S3−S3⋅S1),C_x=\frac{2}{3}(2\mathbf{S}_1\cdot\mathbf{S}_2-\mathbf{S}_2\cdot\mathbf{S}_3-\mathbf{S}_3\cdot\mathbf{S}_1),
Cy=23(S3⋅S1−S2⋅S3),(6)C_y=\frac{2}{\sqrt{3}}(\mathbf{S}_3\cdot\mathbf{S}_1-\mathbf{S}_2\cdot\mathbf{S}_3), \tag*{(6)}

and Cz=4χ/3C_z=4\chi/\sqrt{3}. These operators act as Pauli matrices in the S=1/2S=1/2 subspace and vanish in the S=3/2S=3/2 sector. Here, K′=3K/4K'=\sqrt{3}K/4 and Δ=[(2J1−J2−J3)/2,3(J3−J2)/2,0]\boldsymbol{\Delta}=[(2J_1-J_2-J_3)/2,\sqrt{3}(J_3-J_2)/2,0], with Δz=0\Delta_z=0 reflecting the absence of a chirality bias at zero cavity flux. (5) is the central result of this work: in the Mott regime, the cavity couples directly to the emergent scalar spin chirality, a many-body orbital degree of freedom generated by strong electronic correlations.

Mean-field description of the photons. We now consider NN identical triangles coupled to a common cavity mode, with the per-triangle coupling scaled as g/Ng/\sqrt{N}. For the triangle ensemble, the Hamiltonian is

HN=ωca†a+∑n=1N[12Δ⋅Cn−K′Cz,nsin⁡(Φ^cavN)].(7)H_N=\omega_{\mathrm{c}}a^{\dagger}a+\sum_{n=1}^{N}\left[\frac{1}{2}\boldsymbol{\Delta}\cdot\mathbf{C}_n-K'C_{z,n}\sin\left(\frac{\hat{\Phi}_{\mathrm{cav}}}{\sqrt{N}}\right)\right]. \tag*{(7)}

At zero temperature, we approximate the ground state by ∣Ψtot⟩=∣ψc⟩⊗∣Nα⟩|\Psi_{\mathrm{tot}}\rangle=|\psi_{\mathrm{c}}\rangle\otimes|\sqrt{N}\alpha\rangle, where α=⟨a⟩/N\alpha=\langle a\rangle/\sqrt{N} is real, following the coherent-state mean-field treatment of orbital photon condensation [14, 15]. The coherent-state average of the sinusoidal coupling is e−g2/(2N)sin⁡(2gα)e^{-g^2/(2N)}\sin(2g\alpha). Its prefactor tends to unity as N→∞N\to\infty, giving the ground-state energy per triangle

EGS(α)=ωcα2−(Δ/2)2+K′2sin⁡2(2gα).(8)E_{\mathrm{GS}}(\alpha)=\omega_{\mathrm{c}}\alpha^2-\sqrt{(\Delta/2)^2+K'^2\sin^2(2g\alpha)}. \tag*{(8)}

where Δ=∣Δ∣\Delta= |\Delta| is the chirality splitting. Expanding the energy around the normal phase, EGS(α)=EGS(0)+A(g)α2+⋯E_{\mathrm{GS}}(\alpha) = E_{\mathrm{GS}}(0) + A(g)\alpha^2 + \cdots, with A(g)=ωc−4g2K′2/ΔA(g) = \omega_c - 4g^2K^{\prime2}/\Delta shows that the quadratic coefficient changes sign at gc=ωcΔ2∣K′∣g_c = \frac{\sqrt{\omega_c\Delta}}{2|K'|}, marking the onset of a superradiant transition accompanied by spontaneous scalar chirality, ⟨Cz⟩≠0\langle C_z\rangle\ne0, and a macroscopic cavity displacement. The transition is therefore driven by the collective ordering of an emergent chirality degree of freedom rather than by conventional spin polarization.

To benchmark the effective theory, Figure 2(a) compares independently determined mean-field solutions of the full three-electron Hubbard model and the effective chirality model. The Hubbard chirality signal is inferred from the circulating current using the third-order current-chirality relation. Their close agreement supports the low-energy description of the orbital response in the Mott regime. At fixed K′K', the dependence gc∝Δg_c \propto\sqrt{\Delta} enables in situ electrical control of the superradiant threshold through Stark tuning of Δ(E)\Delta(E) [35, 36, 38]. This separates the control mechanism from the light-matter interaction: an applied electric field tunes the molecular chirality splitting, while the cavity couples through orbital magnetic flux, without requiring an external magnetic bias. In molecular magnets with triangular spin units, such as Cu3_3 and V15_{15}, exchange inequivalence produces a chirality splitting that can be modified by an applied electric field [38, 39]. Recently, such electric-field tuning of the exchange interaction has been demonstrated in a Cu3_3 spin triangle [40].

Figure 2. Mean-field order parameters and collective excitation spectra of the full half-filled Hubbard model and the effective chirality model. (a) Mean-field order parameters as a function of the light-matter coupling gg, showing the photon amplitude ∣α∣|\alpha| and the chirality response for the full Hubbard model and the effective chirality model. The Hubbard chirality ⟨Cz(I)⟩\langle C_z^{(I)}\rangle is inferred from the circulating current using the third-order current–chirality relation and compared with the effective-model expectation value ⟨Cz⟩\langle C_z\rangle. In both cases, the order parameters turn on at their respective critical couplings, and the chirality model closely reproduces the Hubbard mean-field behavior. (b) Corresponding polaritonic excitations, ωp\omega_p, with p=±p = \pm, plotted in units of the cavity frequency ωc\omega_c. The soft polariton frequency decreases toward zero at each model’s transition, while the local gap and the upper polariton frequency remain finite and are likewise well reproduced by the chirality model. Parameters: t1=t2=1t_1 = t_2 = 1, t3=1.05t_3 = 1.05, U=20U = 20 (Δ=0.0205\Delta= 0.0205), and ωc=0.1\omega_c = 0.1.

Chirality polaritons

We now investigate collective light–matter excitations around the mean-field solution, neglecting spin–orbit coupling. Writing a=Nα+da = \sqrt{N}\alpha+ d and describing the chirality fluctuations through a Holstein–Primakoff representation [41, 42, 43] gives the quadratic Hamiltonian of the cavity-active bright sector [33]

Hpol=ωcd†d+Aα(d†+d)2+Δαb†b+Gα(d+d†)(b+b†)(9)H_{\mathrm{pol}} = \omega_c d^\dagger d + A_\alpha(d^\dagger+ d)^2 + \Delta_\alpha b^\dagger b + G_\alpha(d + d^\dagger)(b + b^\dagger) \tag*{(9)}

where only the fully symmetric bright mode b=1N∑nbnb = \frac{1}{\sqrt{N}}\sum_n b_n hybridizes with the photon field, where bnb_n is the annihilation operator for a local chirality excitation on triangle nn. This collective bright mode is a direct consequence of the permutation symmetry of the molecular triangles and is analogous to the Dicke model, whereas the remaining N−1N-1 orthogonal modes remain dark. The equilibrium cavity displacement determines the effective chirality gap Δα\Delta_\alpha, the light–matter coupling GαG_\alpha, and the cavity curvature coefficient AαA_\alpha:

Δα=Δ2+[2K′sin⁡(2gα)]2,Gα=−gK′Δcos⁡(2gα)Δα,Aα=g2K′2sin⁡2(2gα)Δα,(10)\begin{aligned} \Delta_\alpha&= \sqrt{\Delta^2 + [2K'\sin(2g\alpha)]^2},\\ G_\alpha&= -\frac{gK'\Delta\cos(2g\alpha)}{\Delta_\alpha},\\ A_\alpha&= \frac{g^2K^{\prime2}\sin^2(2g\alpha)}{\Delta_\alpha}, \tag*{(10)} \end{aligned}

where AαA_\alpha arises from the curvature of the sinusoidal Peierls coupling. Diagonalization yields the upper (++) and lower (−-) polariton branches:

ω±2=ω~c2+Δα22±12(ω~c2−Δα2)2+16Gα2ωcΔα(11)\omega_\pm^2 = \frac{\widetilde{\omega}_c^2 + \Delta_\alpha^2}{2} \pm\frac{1}{2}\sqrt{(\widetilde{\omega}_c^2 - \Delta_\alpha^2)^2 + 16G_\alpha^2\omega_c\Delta_\alpha} \tag*{(11)}

with ω~c=ωc(ωc+4Aα)\widetilde{\omega}_c = \sqrt{\omega_c(\omega_c + 4A_\alpha)}. As shown in Figure 2(b), the lower polariton branch softens and vanishes at the continuous superradiant transition, providing a direct spectroscopic signature of the collective instability. Remarkably, the resulting polariton spectrum is in excellent quantitative agreement with that obtained from collective fluctuations around the mean-field solution of the full microscopic Hubbard model, demonstrating that the effective chirality theory faithfully captures the low-energy collective physics of the microscopic model.

Exact diagonalization. To benchmark the theory beyond mean field, we perform exact diagonalization (ED) of the effective chirality-photon Hamiltonian and the full microscopic electron-photon model. For the effective model, we work in the permutation-symmetric sector with up to N=240N=240 triangles, supplemented by calculations through N=400N=400 at two representative parameter points. We locate the finite-size pseudocritical coupling from the maximum slope of the parity-even photon amplitude αfl=max⁡[0,(⟨X2⟩−1)/(4N)]\alpha_{\mathrm{fl}}=\sqrt{\max[0,(\langle X^2\rangle-1)/(4N)]}, where X=a+a†X=a+a^\dagger [42]. Figure 3(a) compares these couplings and their extrapolated estimates with the mean-field phase boundary. Fits based on the leading N−2/3N^{-2/3}

Figure 3. Quantum characterization of cavity-induced superradiance. (a) Critical coupling versus chirality splitting, obtained by varying t3=t2+UΔ/4t_3=\sqrt{t^2+U\Delta/4} at fixed t1=t2=t=1t_1=t_2=t=1, U=20U=20, and ωc=0.1\omega_c=0.1. The mean-field curve uses the corresponding K′(Δ)=63 t32/U2K'(\Delta)=6\sqrt{3}\,t_3^2/U^2. Quantum results are shown for N=200N=200, together with extrapolated thermodynamic estimates [33]. (b) Chirality-photon entanglement entropy Sc−pS_{c-p} for N=40N=40, 100, and 200 triangles. Its maximum sharpens and approaches the mean-field critical coupling gcg_c (vertical dashed line) as NN increases. A selecting field hN(a+a†)h\sqrt{N}(a+a^\dagger), with h=10−4h=10^{-4}, is included. (c) Photon displacement for the microscopic Hubbard model (solid lines) and effective chirality model (dashed lines), for N=1,…,4N=1,\ldots,4 triangles, demonstrating their close agreement in the Mott regime. (d) Photon displacement for N=20N=20 in a fixed, fully polarized spin sector for several values of ΔSO/Δ\Delta_{\mathrm{SO}}/\Delta, with selecting-field magnitude h=10−5h=10^{-5}. SOI acts as a longitudinal chirality bias. The curves illustrate finite-size crossovers rather than thermodynamic critical thresholds. Here X=a+a†X=a+a^\dagger, and the per-triangle cavity coupling is g/Ng/\sqrt{N}. In (c), U=30U=30 and Hbias=hNXH_{\mathrm{bias}}=h\sqrt{N}X, while the remaining microscopic parameters are as in Figure 2.

scaling [44] closely reproduce the mean-field prediction. Numerical convergence and extrapolation sensitivity are discussed in the SM [33].

As an independent diagnostic, Figure 3(b) shows the chirality-photon entanglement entropy [45, 46],

Sc−p=−Tr⁡(ρphln⁡ρph),(12)S_{c-p}=-\operatorname{Tr}(\rho_{\mathrm{ph}}\ln\rho_{\mathrm{ph}}), \tag*{(12)}

where ρph=Tr⁡χ∣Ψ0⟩⟨Ψ0∣\rho_{\mathrm{ph}}=\operatorname{Tr}_{\chi}|\Psi_0\rangle\langle\Psi_0| is the reduced photonic density matrix. The ground state ∣Ψ0⟩|\Psi_0\rangle is evaluated with a weak selecting field Hbias=hNXH_{\mathrm{bias}}=h\sqrt{N}X, with h=10−4h=10^{-4}. The entropy peak sharpens and moves toward the mean-field critical coupling with increasing system size, providing an additional quantum signature of the collective instability. Finally, Figure 3(c) compares the photon displacement obtained by ED of up to N=4N=4 half-filled Hubbard triangles (Nel=3N_{\mathrm{el}}=3 per triangle) with that of the effective chirality model. Their close agreement shows that the three-spin ring-exchange interaction captures the low-energy light-matter coupling in the Mott regime, with only small quantitative corrections from charge fluctuations.

Effects of spin-orbit interaction. We now examine the robustness of the superradiant transition in the presence of the spin-orbit interaction (SOI). The leading effect of SOI is to couple the emergent scalar chirality to the local spin polarization [35, 36], pertaining to the Hamiltonian

HSO=ΔSO∑n=1NCz,nSz,n,H_{\mathrm{SO}}=\Delta_{\mathrm{SO}}\sum_{n=1}^{N}C_{z,n}S_{z,n},

while neglecting SOI-dependent corrections to the orbital light-matter coupling. Within this projected model, each Sz,nS_{z,n} is conserved, and the Hamiltonian separates into sectors of fixed molecular spin projections. For nonzero SOI and a nonzero equilibrium cavity displacement, the unconstrained ground state selects a fully polarized spin sector. Within a fixed spin sector, the SOI acts as a longitudinal bias on chirality. However, when both spin sectors are allowed, the two displaced branches remain symmetry-related. At zero temperature, minimizing the mean-field energy over both spin sectors produces a cusp at zero cavity displacement, so that any nonzero light-matter coupling favors symmetry-related displaced minima. Figure 3(d) shows the corresponding displacement response within a fixed polarized sector. By contrast, fixing the total magnetization to zero gives equal populations of the two spin sectors, whose linear chirality biases cancel. The ground-state energy is then analytic around zero displacement, and the normal-state instability occurs at

gc(ΔSO)=gc(0)[1+(ΔSOΔ)2]3/4.(13)g_c(\Delta_{\mathrm{SO}})=g_c(0)\left[1+\left(\frac{\Delta_{\mathrm{SO}}}{\Delta}\right)^2\right]^{3/4}. \tag*{(13)}

This threshold increases with SOI because the spin-chirality coupling reduces the susceptibility to cavity-induced chirality ordering. This coincides with the transition coupling when the transition is continuous.

Estimates and candidate molecular systems. We focus on superconducting LC resonators because the proposed mechanism relies on the orbital coupling to the magnetic flux generated by the cavity mode, rather than on the electric-dipole interaction dominant in conventional cavity QED. The zero-point current produces a flux Φzpf,i\Phi_{\mathrm{zpf},i} through each molecular triangle and hence a dimensionless Peierls coupling gi=2πΦzpf,i/Φeg_i = 2\pi\Phi_{\mathrm{zpf},i}/\Phi_e, where Φe=h/e\Phi_e = h/e is the electronic flux quantum. For a conventional circular loop and an ensemble of N∼1012N \sim10^{12} molecules, we obtain the conservative estimate [33]

g=∑igi2∼10−6.(14)g = \sqrt{\sum_i g_i^2} \sim10^{-6}. \tag*{(14)}

The corresponding collective energy scale is Δcav∼∣K′∣g\Delta_{\mathrm{cav}} \sim|K'|g, giving Δcav/h∼10\Delta_{\mathrm{cav}}/h \sim10–100100 kHz for an illustrative range ∣K′∣/h∼10|K'|/h \sim10–100100 GHz, bracketed by the molecular estimates below. Increasing this scale motivates resonator geometries that concentrate the zero-point magnetic field within the molecular sample, such as narrow inductive constrictions [47].

As a quantitative benchmark, approximate Hubbard parameters reported for Cu3_3 molecular triangles [39], t≃50t \simeq50 meV and U≃9.06U \simeq9.06 eV, give ∣K′∣/h≃4|K'|/h \simeq4 GHz using ∣K′∣=63∣t∣3/U2|K'| = 6\sqrt{3}|t|^3/U^2. For V3_3 molecular triangles [39], t≃53t \simeq53 meV and U≃1U \simeq1 eV yield ∣K′∣/h≃370|K'|/h \simeq370 GHz. The enhancement primarily reflects the smaller effective Coulomb repulsion, illustrating the potential of molecular design to strengthen the orbital coupling. These values remain indicative: for V3_3, the same one-band parameters give J=4t2/U≃11J = 4t^2/U \simeq11 meV, substantially exceeding the reported DFT exchange of 1.3 meV. A quantitative prediction, therefore, requires a consistent microscopic parametrization. More broadly, the scale t1t2t3/U2t_1t_2t_3/U^2 provides a microscopic guide to enhancing the orbital chirality-photon coupling.

Conclusions and outlook. We have shown that orbital cavity coupling can drive equilibrium photon condensation and spontaneous scalar spin chirality in triangular Mott molecules. Our microscopic derivation and numerical benchmarks establish an effective chirality-photon description, with collective polariton excitations and an electrically tunable condensation threshold.

A natural extension is to evaluate the finite-frequency exchange and chirality couplings derived in the SM, following the approach developed for cavity-engineered exchange [23]. This would reveal how photon-assisted ring exchange modifies chirality and orbital currents beyond the low-frequency limit. Beyond molecular magnets, gate-defined quantum-dot triangles offer another setting for exploring chirality-photon coupling, motivated by the recent demonstration of simultaneous control of all three exchange interactions [48]. Coupling a single triangle to the electric field of a resonator could enable manipulation of its chirality degree of freedom through electrically modulated exchange. Progress in controlling two-dimensional quantum-dot arrays [49] also motivates extending this approach to several triangular units coupled to a common resonator. More broadly, our work opens a route to manipulating frustrated magnetism through cavity control of emergent orbital degrees of freedom.

Acknowledgments. We thank V. K. Kozin for very useful discussions. M. T. acknowledges support from the National Science Centre (Poland) OPUS Grant No. 2021/42/B/ST3/04475, the Foundation for Polish Science project “MagTop” (No. FENG.02.01-IP.05-0028/23) cofinanced by the European Union from the funds of Priority 2 of the European Funds for a Smart Economy Program 2021–2027 (FENG), and by the NAWA Bekker Grant No. BPN/BEK/2024/1/00310 (Poland). D. S. acknowledges funding provided by the Institute of Physics Belgrade, through a grant from the Ministry of Science, Technological Development and Innovation of the Republic of Serbia.

Data availability. The data and exact-diagonalization scripts supporting this work are available from the authors upon reasonable request.

References

Schrieffer-Wolff derivation of three-site ring exchange

We start by writing the total Hamiltonian as H=HU+Ht+HpH = H_U + H_t + H_p, with

HU=U∑jnj↑nj↓,Ht=∑j,σ[tjeigj(a†+a)cjσ†cj+1σ+H.c.],Hp=ωca†a,H_U = U \sum_j n_{j\uparrow}n_{j\downarrow}, \qquad H_t = \sum_{j,\sigma} \left[t_j e^{ig_j(a^\dagger+a)} c^\dagger_{j\sigma}c_{j+1\sigma} + \mathrm{H.c.}\right], \qquad H_p = \omega_c a^\dagger a,

where we assume tjt_j are real. To systematically account for photon-assisted virtual processes, it is convenient to work in a frame rotating at the cavity frequency. In this representation, the explicit photon Hamiltonian is removed, while the Peierls phases become explicitly time periodic. This allows us to employ a Floquet-SWT expansion [23]. Specifically, we perform a unitary transformation U(τ)=eiωca†aτU(\tau) = e^{i\omega_c a^\dagger a\tau} on the total triangle-cavity Hamiltonian, which removes HpH_p and leads to the replacement a→ae−iωcτa \to ae^{-i\omega_c\tau}. The transformed Hamiltonian becomes

Hrot(τ)−i∂τ=U(τ)[H−i∂τ]U†(τ),Hrot(τ)=HU+Ht(τ),(15)H_{\mathrm{rot}}(\tau) - i\partial_\tau= U(\tau)[H - i\partial_\tau]U^\dagger(\tau), \qquad H_{\mathrm{rot}}(\tau) = H_U + H_t(\tau), \tag*{(15)}
Ht(τ)=∑j,σ[tjeigjA(τ)cjσ†cj+1σ+H.c.],(16)H_t(\tau) = \sum_{j,\sigma} \left[t_j e^{ig_jA(\tau)} c^\dagger_{j\sigma}c_{j+1\sigma} + \mathrm{H.c.}\right], \tag*{(16)}

where A(τ)=(a†eiωcτ+ae−iωcτ)A(\tau) = (a^\dagger e^{i\omega_c\tau} + ae^{-i\omega_c\tau}). Note that this transformation renders the Hamiltonian periodic, Hrot(τ+Tc)=Hrot(τ)H_{\mathrm{rot}}(\tau+ T_c) = H_{\mathrm{rot}}(\tau), with Tc=2π/ωcT_c = 2\pi/\omega_c. This periodicity is a consequence of the rotating-frame representation of the quantized cavity mode and does not introduce an external drive. At half filling, the Hilbert space is partitioned into the singly occupied sector, with projector PP, and its complement Q=1−PQ = 1 - P. The projectors act on the electronic Hilbert space at fixed particle number Nel=3N_{\mathrm{el}} = 3 and are understood to include the identity in photon space. For the three-site ring, the latter contains exactly one doublon and one holon, with interaction energy UU, so that PHUP=0PH_UP = 0 and QHUQ=UQQH_UQ = UQ. A single hopping process acting on the singly occupied sector creates a doublon-holon pair, implying PHtP=0PH_tP = 0, whereas hopping within QQ can move the doublon or holon without changing the interaction energy. We therefore introduce the off-diagonal and diagonal perturbations:

Vod(τ)=PHt(τ)Q+QHt(τ)P,V_{\mathrm{od}}(\tau) = PH_t(\tau)Q + QH_t(\tau)P,
Vd(τ)=PHt(τ)P+QHt(τ)Q.(17)V_{\mathrm{d}}(\tau) = PH_t(\tau)P + QH_t(\tau)Q. \tag*{(17)}

Next, we use a time-dependent SWT W(τ)=eS(τ)W(\tau) = e^{S(\tau)} that block-diagonalizes HH order by order. We choose an anti-Hermitian, block-off-diagonal generator, S†=−SS^\dagger= -S, with PSP=QSQ=0PSP = QSQ = 0. We use O00=POPO_{00} = POP, O11=QOQO_{11} = QOQ, O10=QOPO_{10} = QOP, and O01=POQO_{01} = POQ, so that

H~rot(τ)−i∂τ=W(τ)[Hrot(τ)−i∂τ]W†(τ).(18)\widetilde{H}_{\mathrm{rot}}(\tau) - i\partial_\tau= W(\tau)[H_{\mathrm{rot}}(\tau) - i\partial_\tau]W^\dagger(\tau). \tag*{(18)}

We introduce a dimensionless bookkeeping parameter λ\lambda by writing Hrot=HU+λ(Vod+Vd)H_{\mathrm{rot}}=H_U+\lambda(V_{\mathrm{od}}+V_d) and expand the generator as S(τ)=λS1(τ)+λ2S2(τ)S(\tau)=\lambda S_1(\tau)+\lambda^2 S_2(\tau). Each power of λ\lambda tracks the corresponding order in the hopping amplitudes; we set λ=1\lambda=1 at the end. Up to third order in λ\lambda, we obtain:

H~rot(τ)=HU+λ(Vod+Vd)+λ[S1,HU]+iλS˙1+λ2[S1,Vod+Vd]+λ2[S2,HU]+λ22[S1,[S1,HU]]+iλ22(S1S˙1−S˙1S1)+iλ2S˙2+λ3[S2,Vod+Vd]+λ32[S2,[S1,HU]]+λ32[S1,[S2,HU]]+λ32[S1,[S1,Vod+Vd]]+λ36[S1,[S1,[S1,HU]]]+iλ36[S1,[S1,S˙1]]+iλ32([S1,S˙2]+[S2,S˙1])+O(λ4).(19)\begin{aligned} \widetilde{H}_{\mathrm{rot}}(\tau)={}&H_U+\lambda(V_{\mathrm{od}}+V_d)+\lambda[S_1,H_U]+i\lambda\dot{S}_1\\ &+\lambda^2[S_1,V_{\mathrm{od}}+V_d]+\lambda^2[S_2,H_U]+\frac{\lambda^2}{2}[S_1,[S_1,H_U]]+\frac{i\lambda^2}{2}(S_1\dot{S}_1-\dot{S}_1S_1)+i\lambda^2\dot{S}_2\\ &+\lambda^3[S_2,V_{\mathrm{od}}+V_d]+\frac{\lambda^3}{2}[S_2,[S_1,H_U]]+\frac{\lambda^3}{2}[S_1,[S_2,H_U]]+\frac{\lambda^3}{2}[S_1,[S_1,V_{\mathrm{od}}+V_d]]+\frac{\lambda^3}{6}[S_1,[S_1,[S_1,H_U]]]\\ &+\frac{i\lambda^3}{6}[S_1,[S_1,\dot{S}_1]]+\frac{i\lambda^3}{2}\left([S_1,\dot{S}_2]+[S_2,\dot{S}_1]\right)+O(\lambda^4). \tag*{(19)} \end{aligned}

We truncate the generator at S(τ)=λS1(τ)+λ2S2(τ)S(\tau)=\lambda S_1(\tau)+\lambda^2S_2(\tau), which suffices to obtain the projected low-energy Hamiltonian through third order in the hopping. A third-order block-off-diagonal generator contributes only through [S3,HU]+iS˙3[S_3,H_U]+i\dot{S}_3, whose projection onto the singly occupied sector vanishes. Next, we find S1S_1 and S2S_2 order by order by eliminating the transition between the two blocks up to second order:

[S1,HU]+Vod+iS˙1=0,(20)[S_1,H_U]+V_{\mathrm{od}}+i\dot{S}_1=0, \tag*{(20)}
[S2,HU]+[S1,Vd]+iS˙2=0,(21)[S_2,H_U]+[S_1,V_d]+i\dot{S}_2=0, \tag*{(21)}

which, in turn, leads to the following effective low-energy Hamiltonian:

Heff(τ)≡PH~rot(τ)P=λ22P[S1(τ),Vod(τ)]P+λ32P[S2(τ),Vod(τ)]P+O(λ4).(22)H_{\mathrm{eff}}(\tau)\equiv P\widetilde{H}_{\mathrm{rot}}(\tau)P=\frac{\lambda^2}{2}P[S_1(\tau),V_{\mathrm{od}}(\tau)]P+\frac{\lambda^3}{2}P[S_2(\tau),V_{\mathrm{od}}(\tau)]P+O(\lambda^4). \tag*{(22)}

Here PHUP=0PH_UP=0. Although the truncated transformation leaves third-order off-diagonal terms, these vanish upon projection and do not affect HeffH_{\mathrm{eff}} at this order. The generators S1S_1 and S2S_2 are conveniently obtained in Fourier space, since all operators inherit the cavity periodicity:

O(τ)=∑nO(n)e−inωcτ,(23)O(\tau)=\sum_n O^{(n)}e^{-in\omega_c\tau}, \tag*{(23)}
O(n)=1Tc∫0TcdτO(τ)einωcτ,(24)O^{(n)}=\frac{1}{T_c}\int_0^{T_c}d\tau O(\tau)e^{in\omega_c\tau}, \tag*{(24)}

so that

Vod(n)=∑j,σtjBgj(n)[cjσ†njσˉcj+1σ(1−nj+1σˉ)⏟∝Vod,10+cjσ†(1−njσˉ)cj+1σnj+1σˉ⏟∝Vod,01]+∑j,σtjB−gj(n)[cj+1σ†nj+1σˉcjσ(1−njσˉ)⏟∝Vod,10+cj+1σ†(1−nj+1σˉ)cjσnjσˉ⏟∝Vod,01],(25)\begin{aligned} V_{\mathrm{od}}^{(n)}={}&\sum_{j,\sigma}t_jB_{g_j}^{(n)} \left[ \underbrace{c_{j\sigma}^{\dagger}n_{j\bar{\sigma}}c_{j+1\sigma}(1-n_{j+1\bar{\sigma}})}_{\propto V_{\mathrm{od},10}} +\underbrace{c_{j\sigma}^{\dagger}(1-n_{j\bar{\sigma}})c_{j+1\sigma}n_{j+1\bar{\sigma}}}_{\propto V_{\mathrm{od},01}} \right]\\ &+\sum_{j,\sigma}t_jB_{-g_j}^{(n)} \left[ \underbrace{c_{j+1\sigma}^{\dagger}n_{j+1\bar{\sigma}}c_{j\sigma}(1-n_{j\bar{\sigma}})}_{\propto V_{\mathrm{od},10}} +\underbrace{c_{j+1\sigma}^{\dagger}(1-n_{j+1\bar{\sigma}})c_{j\sigma}n_{j\bar{\sigma}}}_{\propto V_{\mathrm{od},01}} \right], \tag*{(25)} \end{aligned}
Vd(n)=∑j,σtj[Bgj(n)cjσ†(1−njσˉ)cj+1σ(1−nj+1σˉ)+B−gj(n)cj+1σ†(1−nj+1σˉ)cjσ(1−njσˉ)]+∑j,σtj[Bgj(n)cjσ†njσˉcj+1σnj+1σˉ+B−gj(n)cj+1σ†nj+1σˉcjσnjσˉ](26)\begin{aligned} V_d^{(n)}={}&\sum_{j,\sigma}t_j\left[B_{g_j}^{(n)}c_{j\sigma}^{\dagger}(1-n_{j\bar{\sigma}})c_{j+1\sigma}(1-n_{j+1\bar{\sigma}}) +B_{-g_j}^{(n)}c_{j+1\sigma}^{\dagger}(1-n_{j+1\bar{\sigma}})c_{j\sigma}(1-n_{j\bar{\sigma}})\right]\\ &+\sum_{j,\sigma}t_j\left[B_{g_j}^{(n)}c_{j\sigma}^{\dagger}n_{j\bar{\sigma}}c_{j+1\sigma}n_{j+1\bar{\sigma}} +B_{-g_j}^{(n)}c_{j+1\sigma}^{\dagger}n_{j+1\bar{\sigma}}c_{j\sigma}n_{j\bar{\sigma}}\right] \tag*{(26)} \end{aligned}
S1(n)=Vod,10(n)U−nωc−Vod,01(n)U+nωc=∑j,σtj[(cjσ†cj+1σ)10Bgj(n)+(cj+1σ†cjσ)10B−gj(n)]1U−nωc−∑j,σtj[(cjσ†cj+1σ)01Bgj(n)+(cj+1σ†cjσ)01B−gj(n)]1U+nωc=∑j,σtj[cjσ†njσˉcj+1σ(1−nj+1σˉ)Bgj(n)+cj+1σ†nj+1σˉcjσ(1−njσˉ)B−gj(n)]1U−nωc−∑j,σtj[cjσ†(1−njσˉ)cj+1σnj+1σˉBgj(n)+cj+1σ†(1−nj+1σˉ)cjσnjσˉB−gj(n)]1U+nωc,(27)\begin{aligned} S_1^{(n)}={}&\frac{V_{\mathrm{od},10}^{(n)}}{U-n\omega_c}-\frac{V_{\mathrm{od},01}^{(n)}}{U+n\omega_c}\\ ={}&\sum_{j,\sigma}t_j\left[(c_{j\sigma}^{\dagger}c_{j+1\sigma})_{10}B_{g_j}^{(n)}+(c_{j+1\sigma}^{\dagger}c_{j\sigma})_{10}B_{-g_j}^{(n)}\right]\frac{1}{U-n\omega_c}\\ &-\sum_{j,\sigma}t_j\left[(c_{j\sigma}^{\dagger}c_{j+1\sigma})_{01}B_{g_j}^{(n)}+(c_{j+1\sigma}^{\dagger}c_{j\sigma})_{01}B_{-g_j}^{(n)}\right]\frac{1}{U+n\omega_c}\\ ={}&\sum_{j,\sigma}t_j\left[c_{j\sigma}^{\dagger}n_{j\bar{\sigma}}c_{j+1\sigma}(1-n_{j+1\bar{\sigma}})B_{g_j}^{(n)}+c_{j+1\sigma}^{\dagger}n_{j+1\bar{\sigma}}c_{j\sigma}(1-n_{j\bar{\sigma}})B_{-g_j}^{(n)}\right]\frac{1}{U-n\omega_c}\\ &-\sum_{j,\sigma}t_j\left[c_{j\sigma}^{\dagger}(1-n_{j\bar{\sigma}})c_{j+1\sigma}n_{j+1\bar{\sigma}}B_{g_j}^{(n)}+c_{j+1\sigma}^{\dagger}(1-n_{j+1\bar{\sigma}})c_{j\sigma}n_{j\bar{\sigma}}B_{-g_j}^{(n)}\right]\frac{1}{U+n\omega_c}, \tag*{(27)} \end{aligned}

where

Bgj(n)=1Tc∫0TcdτeigjA(τ)einωcτ=e−gj2/2Tc∫0Tcdτeigja†eiωcτeigjae−iωcτeinωcτ=e−gj2/2∑k=max⁡(0,−n)∞(igj)n+2k(a†)kan+kk!(n+k)!.(28)B_{g_j}^{(n)}=\frac{1}{T_c}\int_0^{T_c}d\tau e^{ig_jA(\tau)}e^{in\omega_c\tau}=\frac{e^{-g_j^2/2}}{T_c}\int_0^{T_c}d\tau e^{ig_ja^\dagger e^{i\omega_c\tau}}e^{ig_jae^{-i\omega_c\tau}}e^{in\omega_c\tau} =e^{-g_j^2/2}\sum_{k=\max(0,-n)}^\infty\frac{(ig_j)^{n+2k}(a^\dagger)^ka^{n+k}}{k!(n+k)!}. \tag*{(28)}

The Fourier component Bg(n)B_{g}^{(n)} changes the photon number by −n-n, so that positive nn corresponds to photon annihilation. We can also extract the time-dependent transformation:

S1(τ)=∑j,σtj[cjσ†njσˉcj+1σ(1−nj+1σˉ)Bgj(τ)+cj+1σ†nj+1σˉcjσ(1−njσˉ)B‾−gj(τ)]−∑j,σtj[cjσ†(1−njσˉ)cj+1σnj+1σˉBgj+(τ)+cj+1σ†(1−nj+1σˉ)cjσnjσˉB−gj+(τ)],(29)S_1(\tau)=\sum_{j,\sigma}t_j\left[c_{j\sigma}^\dagger n_{j\bar{\sigma}}c_{j+1\sigma}(1-n_{j+1\bar{\sigma}})B_{g_j}(\tau)+c_{j+1\sigma}^\dagger n_{j+1\bar{\sigma}}c_{j\sigma}(1-n_{j\bar{\sigma}})\overline{B}_{-g_j}(\tau)\right] -\sum_{j,\sigma}t_j\left[c_{j\sigma}^\dagger(1-n_{j\bar{\sigma}})c_{j+1\sigma}n_{j+1\bar{\sigma}}B_{g_j}^{+}(\tau)+c_{j+1\sigma}^\dagger(1-n_{j+1\bar{\sigma}})c_{j\sigma}n_{j\bar{\sigma}}B_{-g_j}^{+}(\tau)\right], \tag*{(29)}

where

Bgj±(τ)=∑nBgj(n)e−inωcτU±nωc.(30)B_{g_j}^{\pm}(\tau)=\sum_n B_{g_j}^{(n)}\frac{e^{-in\omega_c\tau}}{U\pm n\omega_c}. \tag*{(30)}

This allows us to evaluate the second-order correction to the Hamiltonian in the singly occupied block:

ΔH2=12[S1(τ),Vod(τ)]00=∑nΔH2ne−inωcτ,(31)\Delta H_2=\frac{1}{2}[S_1(\tau),V_{od}(\tau)]_{00}=\sum_n\Delta H_2^n e^{-in\omega_c\tau}, \tag*{(31)}
ΔH2n=−12∑m,lδm+l,n(Vod,01(m)Vod,10(l)U+mωc+Vod,01(m)Vod,10(l)U−lωc)=−12∑m,l∑j,σ,σ′δm+l,ntj2[Bgj(m)B−gj(l)cjσ†(1−njσˉ)cj+1σnj+1σˉcj+1σ′†nj+1σˉ′cjσ′(1−njσˉ′)+B−gj(m)Bgj(l)cj+1σ†(1−nj+1σˉ)cjσnjσˉcjσ′†njσˉ′cj+1σ′(1−nj+1σˉ′)](1U+mωc+1U−lωc)=−12∑m,l∑jδm+l,ntj2(Bgj(m)B−gj(l)+B−gj(m)Bgj(l))(12−2SjzSj+1z−Sj+Sj+1−−Sj−Sj+1+)(1U+mωc+1U−lωc)=∑m,l∑jδm+l,ntj2(Bgj(m)B−gj(l)+B−gj(m)Bgj(l))(Sj⋅Sj+1−14)(1U+mωc+1U−lωc),(32)\begin{aligned} \Delta H_2^n=-\frac{1}{2}\sum_{m,l}\delta_{m+l,n}\left(\frac{V_{od,01}^{(m)}V_{od,10}^{(l)}}{U+m\omega_c}+\frac{V_{od,01}^{(m)}V_{od,10}^{(l)}}{U-l\omega_c}\right) \\ &=-\frac{1}{2}\sum_{m,l}\sum_{j,\sigma,\sigma'}\delta_{m+l,n}t_j^2\left[B_{g_j}^{(m)}B_{-g_j}^{(l)}c_{j\sigma}^\dagger(1-n_{j\bar{\sigma}})c_{j+1\sigma}n_{j+1\bar{\sigma}}c_{j+1\sigma'}^\dagger n_{j+1\bar{\sigma}'}c_{j\sigma'}(1-n_{j\bar{\sigma}'}) +B_{-g_j}^{(m)}B_{g_j}^{(l)}c_{j+1\sigma}^\dagger(1-n_{j+1\bar{\sigma}})c_{j\sigma}n_{j\bar{\sigma}}c_{j\sigma'}^\dagger n_{j\bar{\sigma}'}c_{j+1\sigma'}(1-n_{j+1\bar{\sigma}'})\right]\left(\frac{1}{U+m\omega_c}+\frac{1}{U-l\omega_c}\right) \\ &=-\frac{1}{2}\sum_{m,l}\sum_j\delta_{m+l,n}t_j^2\left(B_{g_j}^{(m)}B_{-g_j}^{(l)}+B_{-g_j}^{(m)}B_{g_j}^{(l)}\right)\left(\frac{1}{2}-2S_j^zS_{j+1}^z-S_j^+S_{j+1}^- -S_j^-S_{j+1}^+\right)\left(\frac{1}{U+m\omega_c}+\frac{1}{U-l\omega_c}\right) \\ &=\sum_{m,l}\sum_j\delta_{m+l,n}t_j^2\left(B_{g_j}^{(m)}B_{-g_j}^{(l)}+B_{-g_j}^{(m)}B_{g_j}^{(l)}\right)\left(S_j\cdot S_{j+1}-\frac{1}{4}\right)\left(\frac{1}{U+m\omega_c}+\frac{1}{U-l\omega_c}\right), \tag*{(32)} \end{aligned}

where Pjj+1P_{jj+1} is the projector onto the singlet state on the bond jj. This expression represents the photon-dressed superexchange interaction. In general, the exchange acquires an operator-valued dependence on the cavity field through the Fourier coefficients Bg(n)B_g^{(n)} and the photon-dependent energy denominators. In the low-frequency limit U≫ωcU\gg\omega_c, the photon energies may be neglected in the denominators, recovering the conventional Heisenberg exchange

ΔH2=∑jJj(Sj⋅Sj+1−1/4),(33)\Delta H_2=\sum_jJ_j\left(S_j\cdot S_{j+1}-1/4\right), \tag*{(33)}

where Jj=4tj2/UJ_j=4t_j^2/U represents the effective strength of the nearest-neighbor Heisenberg exchange interaction. Put differently, in this limit the dynamics reduces to particles tunneling back and forth between the two sites, so that the Peierls phase is fully canceled in the effective process.

The third-order contribution follows analogously, but is qualitatively richer because the cyclic hopping process involves two successive virtual intermediate states containing a doublon–holon pair. Consequently, the effective chirality interaction depends on products of photon-dressed denominators, providing the natural extension of cavity-engineered superexchange to emergent scalar spin chirality. We can extract the nn-th Fourier component of S2(τ)S_2(\tau):

S2(n)=[S1(τ),Vd(τ)]10(n)U−nωc−[S1(τ),Vd(τ)]01(n)U+nωc.(34)S_2^{(n)}=\frac{[S_1(\tau),V_d(\tau)]_{10}^{(n)}}{U-n\omega_c}-\frac{[S_1(\tau),V_d(\tau)]_{01}^{(n)}}{U+n\omega_c}. \tag*{(34)}

To proceed, we must first evaluate the nn-th Fourier component of the commutator [S1(τ),Vd(τ)][S_1(\tau),V_d(\tau)]:

[S1(τ),Vd(τ)]10(n)=∑m,lδm+l,n(S1,10(m)Vd,00(l)−Vd,11(m)S1,10(l))=−∑mVd,11(m)S1,10(n−m).(35)[S_1(\tau),V_d(\tau)]_{10}^{(n)}=\sum_{m,l}\delta_{m+l,n}\left(S_{1,10}^{(m)}V_{d,00}^{(l)}-V_{d,11}^{(m)}S_{1,10}^{(l)}\right)=-\sum_m V_{d,11}^{(m)}S_{1,10}^{(n-m)}. \tag*{(35)}
[S1(τ),Vd(τ)]01(n)=∑m,lδm+l,n(S1,01(m)Vd,11(l)−Vd,00(m)S1,01(l))=∑mS1,01(m)Vd,11(n−m).(36)[S_1(\tau),V_d(\tau)]_{01}^{(n)}=\sum_{m,l}\delta_{m+l,n}\left(S_{1,01}^{(m)}V_{d,11}^{(l)}-V_{d,00}^{(m)}S_{1,01}^{(l)}\right)=\sum_m S_{1,01}^{(m)}V_{d,11}^{(n-m)}. \tag*{(36)}

Making use of the explicit form of S1(m)S_1^{(m)}, we arrive at

S2(n)=−∑m(1(U−nωc)(U−(n−m)ωc)Vd,11(m)Vod,10(n−m)−1(U+nωc)(U+mωc)Vod,01(m)Vd,11(n−m)).(37)S_2^{(n)}=-\sum_m\left(\frac{1}{(U-n\omega_c)(U-(n-m)\omega_c)}V_{d,11}^{(m)}V_{od,10}^{(n-m)}-\frac{1}{(U+n\omega_c)(U+m\omega_c)}V_{od,01}^{(m)}V_{d,11}^{(n-m)}\right). \tag*{(37)}

This allows us to write the third-order correction to the Hamiltonian projected onto the singly occupied sector as

ΔH3(τ)=12P[S2(τ),Vod(τ)]P=∑nΔH3(n)e−inωcτ,(38)\Delta H_3(\tau)=\frac{1}{2}P[S_2(\tau),V_{od}(\tau)]P=\sum_n\Delta H_3^{(n)}e^{-in\omega_c\tau}, \tag*{(38)}
ΔH3(n)=12∑r,s,tδr+s+t,nVod,01(r)Vd,11(s)Vod,10(t)(1[U+(r+s)ωc](U+rωc)+1[U−(s+t)ωc](U−tωc))≡∑jJj,j+1(n)(Sj⋅Sj+1−14)+K(n)S1⋅(S2×S3).(39)\begin{aligned} \Delta H_3^{(n)}=\frac{1}{2}\sum_{r,s,t}\delta_{r+s+t,n}V_{od,01}^{(r)}V_{d,11}^{(s)}V_{od,10}^{(t)}\left(\frac{1}{[U+(r+s)\omega_c](U+r\omega_c)}+\frac{1}{[U-(s+t)\omega_c](U-t\omega_c)}\right) \\ &\equiv\sum_j\mathcal{J}_{j,j+1}^{(n)}\left(S_j\cdot S_{j+1}-\frac{1}{4}\right)+\mathcal{K}^{(n)}S_1\cdot(S_2\times S_3). \tag*{(39)} \end{aligned}

where

Jj,j+1(n)=∑(a,b,c)∈Perm⁡(1,2,3)ϵabc(2δjb−1)Cabc(n),K(n)=2i∑(a,b,c)∈Perm⁡(1,2,3)Cabc(n),(40)\mathcal{J}_{j,j+1}^{(n)}=\sum_{(a,b,c)\in\operatorname{Perm}(1,2,3)}\epsilon_{abc}(2\delta_{jb}-1)C_{abc}^{(n)},\qquad\mathcal{K}^{(n)}=2i\sum_{(a,b,c)\in\operatorname{Perm}(1,2,3)}C_{abc}^{(n)}, \tag*{(40)}
Cabc(n)=t1t2t32∑r,s,tδr+s+t,n(Bga(r)Bgb(s)Bgc(t)−B−ga(r)B−gb(s)B−gc(t))(1[U+(r+s)ωc](U+rωc)+1[U−(s+t)ωc](U−tωc)).(41)C_{abc}^{(n)}=\frac{t_1t_2t_3}{2}\sum_{r,s,t}\delta_{r+s+t,n}\left(B_{g_a}^{(r)}B_{g_b}^{(s)}B_{g_c}^{(t)}-B_{-g_a}^{(r)}B_{-g_b}^{(s)}B_{-g_c}^{(t)}\right)\left(\frac{1}{[U+(r+s)\omega_c](U+r\omega_c)}+\frac{1}{[U-(s+t)\omega_c](U-t\omega_c)}\right). \tag*{(41)}

Here a,b,ca,b,c label bonds, ϵ123=1\epsilon_{123}=1, and site indices are understood modulo three. Both permutation sums run over Perm⁡(1,2,3)\operatorname{Perm}(1,2,3). When the photon sideband energies with appreciable weight are small compared with UU, they may be neglected in the denominators. All six ordered cyclic processes then acquire the same photon factor, depending only on g=∑jgjg=\sum_j g_j. The third-order Heisenberg corrections cancel, leaving the scalar-chirality interaction. Returning to the laboratory frame and omitting the spin-independent constant, we obtain:

Hs-ph≈∑jJjSj⋅Sj+1−Ksin⁡[g(a†+a)]S1⋅(S2×S3)+ωca†a,(42)H_{s\text{-}ph}\approx\sum_j J_jS_j\cdot S_{j+1}-K\sin[g(a^\dagger+a)]S_1\cdot(S_2\times S_3)+\omega_ca^\dagger a, \tag*{(42)}

where g=∑j=13gjg=\sum_{j=1}^{3}g_j, and K=24t1t2t3U2K=\frac{24t_1t_2t_3}{U^2} is the effective third-order exchange coupling strength.

Finally, restricting the spin dynamics to the total S=1/2S=1/2 manifold yields the effective chirality-photon Hamiltonian

Hc-ph≈12Δ⋅C−K′sin⁡[g(a†+a)]Cz+ωca†a,(43)H_{c\text{-}ph}\approx\frac{1}{2}\boldsymbol{\Delta}\cdot\mathbf{C}-K'\sin[g(a^\dagger+a)]C_z+\omega_ca^\dagger a, \tag*{(43)}
Δx=J1−J2+J32,Δy=32(J3−J2),Δz=0,K′=34K,(44)\Delta_x=J_1-\frac{J_2+J_3}{2},\qquad\Delta_y=\frac{\sqrt{3}}{2}(J_3-J_2),\qquad\Delta_z=0,\qquad K'=\frac{\sqrt{3}}{4}K, \tag*{(44)}

where we have omitted an additive constant. These operators act as Pauli matrices on the chirality degree of freedom within each fixed-SzS_z doublet of the S=1/2S=1/2 manifold and vanish in the S=3/2S=3/2 manifold. The resulting Hamiltonian coincides with the effective model derived in the main text and explicitly shows that, within the low-frequency approximation, the cavity affects only the three-spin ring exchange term, while the nearest-neighbor Heisenberg interaction retains its conventional form.

The derivation presented above also provides a systematic route beyond the low-frequency approximation employed in the main text. Retaining the full photon-frequency dependence of the virtual intermediate states leads to photon-dependent exchange and chirality couplings, whose quantitative consequences lie beyond the scope of the present work but provide a natural starting point for future studies of cavity-engineered orbital magnetism.

Spin-orbit interaction effects

We consider a general bond-dependent Dzyaloshinskii–Moriya interaction [51, 52] on a triangular spin cluster,

HDM=∑jDjj+1⋅(Sj×Sj+1),(45)H_{\mathrm{DM}}=\sum_j \mathbf{D}_{j j+1}\cdot(\mathbf{S}_j\times\mathbf{S}_{j+1}), \tag*{(45)}

where the vectors Djj+1\mathbf{D}_{j j+1} are allowed to differ from bond to bond. We assume that the isotropic exchange interaction dominates, such that the low-energy sector consists of the four-dimensional manifold with total spin S=1/2S=1/2, while the S=3/2S=3/2 quartet is separated by a finite (exchange) gap.

To unravel the structure of the DM interaction, we decompose the DM vectors into a uniform and a nonuniform contribution,

Djj+1=D0+δDjj+1,D0=13(D12+D23+D31),(46)\mathbf{D}_{j j+1}=\mathbf{D}_0+\delta\mathbf{D}_{j j+1},\qquad\mathbf{D}_0=\frac{1}{3}\left(\mathbf{D}_{12}+\mathbf{D}_{23}+\mathbf{D}_{31}\right), \tag*{(46)}

with

δD12+δD23+δD31=0.(47)\delta\mathbf{D}_{12}+\delta\mathbf{D}_{23}+\delta\mathbf{D}_{31}=0. \tag*{(47)}

The projection of the uniform component onto the S=1/2S=1/2 manifold is readily obtained from the identity

P1/2(Sj×Sj+1)P1/2=13CzS,(48)P_{1/2}\left(\mathbf{S}_j\times\mathbf{S}_{j+1}\right)P_{1/2}=\frac{1}{\sqrt{3}}C_z\mathbf{S}, \tag*{(48)}

where P1/2P_{1/2} denotes the projector onto the total-spin S=1/2S=1/2 manifold, and S=∑j=13Sj\mathbf{S}=\sum_{j=1}^{3}\mathbf{S}_j is the total spin operator. Consequently,

P1/2HDMP1/2=3Cz D0⋅S.(49)P_{1/2}H_{\mathrm{DM}}P_{1/2}=\sqrt{3}C_z\,\mathbf{D}_0\cdot\mathbf{S}. \tag*{(49)}

Choosing the spin quantization axis along D0\mathbf{D}_0 gives HSO=ΔSOCzSzH_{\mathrm{SO}}=\Delta_{\mathrm{SO}}C_zS_z, with ΔSO=3∣D0∣\Delta_{\mathrm{SO}}=\sqrt{3}|\mathbf{D}_0|. Thus, within the low-energy manifold, the uniform DM interaction reduces to a spin-chirality coupling. The nonuniform component has a vanishing projection within the S=1/2S=1/2 manifold because ∑jδDjj+1=0\sum_j\delta\mathbf{D}_{j j+1}=0. It nevertheless couples this manifold to the excited S=3/2S=3/2 quartet, generating corrections of order δD2/Δex\delta D^2/\Delta_{\mathrm{ex}}, where Δex\Delta_{\mathrm{ex}} is the exchange gap and δD\delta D denotes the characteristic magnitude of the nonuniform DM vectors. We neglect these corrections, assuming δD≪Δex\delta D\ll\Delta_{\mathrm{ex}}.

Collective fluctuations and chirality polaritons

Effective chirality polaritons

We consider identical molecules at zero temperature, neglecting spin–orbit coupling. The cavity coupling per molecule is g/Ng/\sqrt{N}, so that Φ^=g(a+a†)/N\hat{\Phi}=g(a+a^\dagger)/\sqrt{N}. Writing a=Nα+da=\sqrt{N}\alpha+d, with the real displacement α\alpha determined by mean-field minimization, gives

Φ^=Φ0+gN(d+d†),Φ0=2gα.(50)\hat{\Phi}=\Phi_0+\frac{g}{\sqrt{N}}(d+d^\dagger),\qquad\Phi_0=2g\alpha. \tag*{(50)}

Within a fixed spin sector, the single-molecule mean-field Hamiltonian is

hMF=Δ2Cx−K′sin⁡Φ0Cz,(51)h_{\mathrm{MF}}=\frac{\Delta}{2}C_x-K'\sin\Phi_0 C_z, \tag*{(51)}

where we rotated the in-plane chirality axes to align the splitting with CxC_x, leaving CzC_z unchanged. The ground and excited eigenstates of hMFh_{\mathrm{MF}} at the equilibrium flux Φ0\Phi_0 are denoted by ∣0⟩|0\rangle and ∣1⟩|1\rangle, with energies −E-E and EE, respectively, where

E=Δ24+K′2sin⁡2Φ0.(52)E=\sqrt{\frac{\Delta^2}{4}+K'^2\sin^2\Phi_0}. \tag*{(52)}

Although each molecule has a finite-dimensional chirality Hilbert space, dilute collective excitations admit a bosonic description. Defining the collective transition operators

Σab=∑j=1N∣a⟩j⟨b∣j,(53)\Sigma_{ab}=\sum_{j=1}^{N}|a\rangle_j\langle b|_j, \tag*{(53)}

with a,b∈{0,1}a,b\in\{0,1\}, we obtain

[Σ01N,Σ10N]=Σ00−Σ11N=1−2Σ11N.(54)\left[\frac{\Sigma_{01}}{\sqrt{N}},\frac{\Sigma_{10}}{\sqrt{N}}\right]=\frac{\Sigma_{00}-\Sigma_{11}}{N}=1-\frac{2\Sigma_{11}}{N}. \tag*{(54)}

For an excitation population much smaller than NN, this commutator approaches the canonical bosonic value. Within the permutation-symmetric sector, the saturation constraint is incorporated exactly by the Holstein–Primakoff representation [41]

Σ00=N−b†b,Σ11=b†b,(55)\Sigma_{00}=N-b^\dagger b,\qquad\Sigma_{11}=b^\dagger b, \tag*{(55)}
Σ10=b†N−b†b,Σ01=N−b†b b,(56)\Sigma_{10}=b^\dagger\sqrt{N-b^\dagger b},\qquad\Sigma_{01}=\sqrt{N-b^\dagger b}\,b, \tag*{(56)}

where [b,b†]=1[b,b^\dagger]=1 and the physical occupation satisfies b†b≤Nb^\dagger b\leq N. Expanding for b†b≪Nb^\dagger b\ll N gives Σ10≃Nb†\Sigma_{10}\simeq\sqrt{N}b^\dagger and Σ01≃Nb\Sigma_{01}\simeq\sqrt{N}b. The boson describes a collective chirality excitation, while the square roots encode the finite excitation capacity of the molecular ensemble. In this representation,

∑jhMF,j=−NE+Δαb†b,Δα=2E.(57)\sum_j h_{\mathrm{MF},j}=-NE+\Delta_\alpha b^\dagger b,\qquad\Delta_\alpha=2E. \tag*{(57)}

We stress that the product state ⨂j∣0⟩j\bigotimes_j|0\rangle_j is the vacuum of bb, while b†b^\dagger creates a symmetric collective transition from ∣0⟩|0\rangle to ∣1⟩|1\rangle. Choosing the phase of ∣1⟩|1\rangle so that ⟨0∣Cz∣1⟩\langle0|C_z|1\rangle is positive, we find

∑jCz,j=K′sin⁡Φ0E(N−2b†b)+NΔ2E(b+b†)+O(N−1/2),(58)\sum_j C_{z,j}=\frac{K'\sin\Phi_0}{E}\left(N-2b^\dagger b\right)+\sqrt{N}\frac{\Delta}{2E}(b+b^\dagger)+O(N^{-1/2}), \tag*{(58)}

where the remainder is understood for occupations of order unity. The mean-field energy per molecule is EMF=ωcα2−E\mathcal{E}_{\mathrm{MF}}=\omega_c\alpha^2-E, with the stationarity condition

ωcα=gK′2sin⁡Φ0cos⁡Φ0E.(59)\omega_c\alpha=\frac{gK'^2\sin\Phi_0\cos\Phi_0}{E}. \tag*{(59)}

Expanding the sinusoidal coupling through second order in the fluctuations, we find that the stationarity condition cancels the terms linear in d+d†d+d^\dagger. Up to an additive constant, the quadratic Hamiltonian is

Hfl=ωcd†d+Aα(d+d†)2+Δαb†b+Gα(d+d†)(b+b†),(60)H_{\mathrm{fl}}=\omega_c d^\dagger d+A_\alpha(d+d^\dagger)^2+\Delta_\alpha b^\dagger b+G_\alpha(d+d^\dagger)(b+b^\dagger), \tag*{(60)}

where

Aα=g2K′2sin⁡2Φ0Δα,Gα=−gK′Δcos⁡Φ0Δα.(61)A_\alpha=\frac{g^2K'^2\sin^2\Phi_0}{\Delta_\alpha},\qquad G_\alpha=-\frac{gK'\Delta\cos\Phi_0}{\Delta_\alpha}. \tag*{(61)}

The quadratic cavity term originates from the curvature of the sinusoidal Peierls coupling and must be retained in the ordered phase. The sign of GαG_\alpha can be changed by redefining b→−bb\to-b and does not affect the spectrum. Defining ω~c2=ωc(ωc+4Aα)\widetilde{\omega}_c^2=\omega_c(\omega_c+4A_\alpha), diagonalization gives

ω±2=ω~c2+Δα22±12(ω~c2−Δα2)2+16ωcΔαGα2.(62)\omega_\pm^2=\frac{\widetilde{\omega}_c^2+\Delta_\alpha^2}{2}\pm\frac{1}{2}\sqrt{(\widetilde{\omega}_c^2-\Delta_\alpha^2)^2+16\omega_c\Delta_\alpha G_\alpha^2}. \tag*{(62)}

Full Hubbard polaritons {#sec:Full Hubbard polaritons}

To benchmark this effective description, we construct the Gaussian fluctuation spectrum using the complete local Hubbard spectrum of a triangle. Let hHub(Φ)h_{\mathrm{Hub}}(\Phi) denote the single-triangle electronic Hamiltonian, with the relative distribution of the Peierls phases among the bonds held fixed as Φ\Phi varies. At its Hubbard mean-field saddle point Φ0=2gα\Phi_0=2g\alpha, we diagonalize

hHub(Φ0)∣m⟩=Em∣m⟩,Ωm=Em−E0.(63)h_{\mathrm{Hub}}(\Phi_0)|m\rangle=E_m|m\rangle,\qquad\Omega_m=E_m-E_0. \tag*{(63)}

The states ∣m⟩|m\rangle span the complete local Hubbard Hilbert space in the chosen symmetry sector, labeled by (Nel,Stot,Mtot)(N_{\mathrm{el}},S_{\mathrm{tot}},M_{\mathrm{tot}}). Defining

V=∂ΦhHub∣Φ0,W=∂Φ2hHub∣Φ0,(64)V=\left.\partial_{\Phi}h_{\mathrm{Hub}}\right|_{\Phi_0},\qquad W=\left.\partial_{\Phi}^{2}h_{\mathrm{Hub}}\right|_{\Phi_0}, \tag*{(64)}

the local Hamiltonian expands as

hHub(Φ0+δΦ)=hHub(Φ0)+δΦV+12(δΦ)2W+⋯ ,(65)h_{\mathrm{Hub}}(\Phi_0+\delta\Phi)=h_{\mathrm{Hub}}(\Phi_0)+\delta\Phi V+\frac{1}{2}(\delta\Phi)^2W+\cdots, \tag*{(65)}

where δΦ=gN(d+d†)\delta\Phi=\frac{g}{\sqrt{N}}(d+d^\dagger). The Hubbard mean-field energy per molecule is EMFHub=ωcα2+E0(Φ0)\mathcal{E}_{\mathrm{MF}}^{\mathrm{Hub}}=\omega_c\alpha^2+E_0(\Phi_0). Its stationarity condition,

ωcα+g⟨0∣V∣0⟩=0,(66)\omega_c\alpha+g\langle0|V|0\rangle=0, \tag*{(66)}

again cancels the terms linear in the cavity fluctuation. The collective-transition construction extends to the multilevel Hubbard spectrum [53]. At leading order, Σm0≃Nbm†\Sigma_{m0}\simeq\sqrt{N}b_m^\dagger, where bm†b_m^\dagger creates the symmetric collective excitation associated with the local transition ∣0⟩→∣m⟩|0\rangle\to|m\rangle. Here ∣0⟩|0\rangle denotes the ground state of the full local Hubbard Hamiltonian at its own mean-field saddle, within the chosen particle-number and spin sector. All expectation values below are evaluated in this state, and V(t)V(t) evolves under hHub(Φ0)h_{\mathrm{Hub}}(\Phi_0). The resulting quadratic Hamiltonian is

HflHub=ωcd†d+AHub(d+d†)2+∑m≠0Ωmbm†bm+(d+d†)∑m≠0(Gmbm†+Gm∗bm),(67)H_{\mathrm{fl}}^{\mathrm{Hub}}=\omega_c d^\dagger d+A_{\mathrm{Hub}}(d+d^\dagger)^2+\sum_{m\ne0}\Omega_m b_m^\dagger b_m+(d+d^\dagger)\sum_{m\ne0}(G_m b_m^\dagger+G_m^*b_m), \tag*{(67)}

with

AHub=g22⟨0∣W∣0⟩,Gm=g⟨m∣V∣0⟩.(68)A_{\mathrm{Hub}}=\frac{g^2}{2}\langle0|W|0\rangle,\qquad G_m=g\langle m|V|0\rangle. \tag*{(68)}

All local excited states within the chosen sector are retained, and transitions with vanishing matrix elements have Gm=0G_m=0. Equivalently, the electronic contribution can be expressed through the zero-temperature retarded susceptibility

χVVR(ω)=−i∫0∞dt ei(ω+i0+)t⟨[V(t),V(0)]⟩=∑m≠02Ωm∣⟨0∣V∣m⟩∣2(ω+i0+)2−Ωm2.(69)\chi_{VV}^{R}(\omega)=-i\int_{0}^{\infty}dt\,e^{i(\omega+i0^+)t}\langle[V(t),V(0)]\rangle=\sum_{m\ne0}\frac{2\Omega_m|\langle0|V|m\rangle|^2}{(\omega+i0^+)^2-\Omega_m^2}. \tag*{(69)}

Eliminating the collective electronic modes yields the equation for the cavity-active excitation frequencies,

ω2−ωc2−2ωcg2[⟨0∣W∣0⟩+χVVR(ω)]=0.(70)\omega^2-\omega_c^2-2\omega_cg^2\left[\langle0|W|0\rangle+\chi_{VV}^{R}(\omega)\right]=0. \tag*{(70)}

Both the static curvature term and the dynamical susceptibility are required. Their zero-frequency combination obeys

d2E0dΦ2∣Φ0=⟨0∣W∣0⟩+χVVR(0),(71)\left.\frac{d^2E_0}{d\Phi^2}\right|_{\Phi_0}=\langle0|W|0\rangle+\chi_{VV}^{R}(0), \tag*{(71)}

providing a consistency check between the fluctuation kernel and the mean-field energy curvature. For the effective chirality Hamiltonian, these expressions reduce to

Vα=−K′cos⁡Φ0 Cz,Wα=K′sin⁡Φ0 Cz,(72)V_\alpha=-K'\cos\Phi_0\,C_z,\qquad W_\alpha=K'\sin\Phi_0\,C_z, \tag*{(72)}

with a single active transition of energy Δα\Delta_\alpha. Consequently,

g22⟨Wα⟩=Aα,g2χVαVαR(ω)=2ΔαGα2(ω+i0+)2−Δα2.(73)\frac{g^2}{2}\langle W_\alpha\rangle=A_\alpha,\qquad g^2\chi_{V_\alpha V_\alpha}^{R}(\omega)=\frac{2\Delta_\alpha G_\alpha^2}{(\omega+i0^+)^2-\Delta_\alpha^2}. \tag*{(73)}

Equation (70) then becomes

(ω2−ω~c2)(ω2−Δα2)−4ωcΔαGα2=0,(74)(\omega^2-\widetilde{\omega}_c^2)(\omega^2-\Delta_\alpha^2)-4\omega_c\Delta_\alpha G_\alpha^2=0, \tag*{(74)}

recovering Eq. (62). The Hubbard calculation therefore supplies a microscopic benchmark that retains all local electronic transitions, while treating the collective fluctuations at Gaussian order.

Mean-field comparison and current-inferred chirality

For the mean-field comparison in Figure 2(a), we retain all 20 states of the three-electron Hubbard Hilbert space. We denote by hHub(Φ)h_{\mathrm{Hub}}(\Phi) the electronic Hamiltonian at a dimensionless total Peierls phase Φ\Phi around the triangle. The individual bond phases are written as ϕj=ηjΦ\phi_j=\eta_j\Phi, with ∑jηj=1\sum_j\eta_j=1 and the coefficients ηj\eta_j held fixed as Φ\Phi varies. The Hubbard mean-field displacement α∗\alpha_\ast minimizes EMF(α)=ωcα2+E0(2gα)\mathcal{E}_{\mathrm{MF}}(\alpha)=\omega_c\alpha^2+E_0(2g\alpha), where E0(Φ)E_0(\Phi) is the lowest eigenvalue of hHub(Φ)h_{\mathrm{Hub}}(\Phi). The corresponding electronic ground state is ∣0⟩|0\rangle, with equilibrium phase Φ∗=2gα∗\Phi_\ast=2g\alpha_\ast.

To characterize the orbital response, we introduce the operator conjugate to the total Peierls phase,

I^Hub(Φ)≡∂hHub(Φ)∂Φ=i∑j,σηjtj[eiηjΦcjσ†cj+1σ−e−iηjΦcj+1σ†cjσ].\hat{\mathcal{I}}_{\mathrm{Hub}}(\Phi)\equiv\frac{\partial h_{\mathrm{Hub}}(\Phi)}{\partial\Phi}=i\sum_{j,\sigma}\eta_jt_j\left[e^{i\eta_j\Phi}c_{j\sigma}^{\dagger}c_{j+1\sigma}-e^{-i\eta_j\Phi}c_{j+1\sigma}^{\dagger}c_{j\sigma}\right].

This operator has units of energy. With the bond-current convention used in the main text, the physical current on bond jj is I^j=(e/ℏ)∂ϕjhHub\hat{I}_j=(e/\hbar)\partial_{\phi_j}h_{\mathrm{Hub}}, so that I^Hub=(ℏ/e)∑jηjI^j\hat{\mathcal{I}}_{\mathrm{Hub}}=(\hbar/e)\sum_j\eta_j\hat{I}_j. For equally distributed phases, ηj=1/3\eta_j=1/3, this is proportional to the average of the three bond-current operators. In a stationary electronic state, charge conservation requires equal current expectation values on all three oriented bonds. Consequently, the physical circulating current is (e/ℏ)IHub(e/\hbar)\mathcal{I}_{\mathrm{Hub}}, where

IHub(Φ∗)=⟨0∣I^Hub(Φ∗)∣0⟩.\mathcal{I}_{\mathrm{Hub}}(\Phi_\ast)=\langle0|\hat{\mathcal{I}}_{\mathrm{Hub}}(\Phi_\ast)|0\rangle.

The third-order effective Hamiltonian contains the flux-dependent term −K′sin⁡Φ Cz-K'\sin\Phi\,C_z, giving the low-energy current relation Ieff=−K′cos⁡Φ ⟨Cz⟩\mathcal{I}_{\mathrm{eff}}=-K'\cos\Phi\,\langle C_z\rangle. We therefore define the current-inferred Hubbard chirality as

⟨Cz(I)⟩=−IHub(Φ∗)K′cos⁡Φ∗,K′=63t1t2t3U2.\langle C_z^{(I)}\rangle=-\frac{\mathcal{I}_{\mathrm{Hub}}(\Phi_\ast)}{K'\cos\Phi_\ast},\qquad K'=\frac{6\sqrt{3}t_1t_2t_3}{U^2}.

where cos⁡Φ∗≠0\cos\Phi_\ast\ne0. The superscript (I)(I) emphasizes that this quantity is inferred from the circulating current through the leading ring-exchange relation. Higher-order strong-coupling corrections are retained in the full Hubbard current and can produce deviations from the effective-model chirality.

We compare ⟨Cz(I)⟩\langle C_z^{(I)}\rangle with ⟨Cz⟩\langle C_z\rangle evaluated at the effective chirality model’s independently determined mean-field minimum. Both observables within each model therefore refer to the same self-consistent state. Finally, the Hellmann-Feynman relation gives ∂ΦE0=IHub\partial_\Phi E_0=\mathcal{I}_{\mathrm{Hub}}, so the Hubbard mean-field stationarity condition reads

ωcα∗+gIHub(Φ∗)=0.\omega_c\alpha_\ast+g\mathcal{I}_{\mathrm{Hub}}(\Phi_\ast)=0.

This relation provides a numerical consistency check between the photon displacement and the circulating current.

Numerical convergence

For Figure 3(a), we locate the finite-size pseudocritical coupling from the maximum slope of the parity-even photon amplitude αfl\alpha_{\mathrm{fl}} defined in the main text. This diagnostic is motivated by photonic observables used in finite-size studies of the Dicke transition [42, 44]. The subtraction removes the vacuum contribution ⟨X2⟩vac=1\langle X^2\rangle_{\mathrm{vac}}=1, and the normalization gives αfl=∣α∣\alpha_{\mathrm{fl}}=|\alpha| for a real coherent state ∣Nα⟩|\sqrt{N}\alpha\rangle. At N=200N=200, we checked photon-cutoff and coupling-grid convergence at Δ=0.0205\Delta=0.0205, 0.150.15, and 0.300.30, along the microscopic parameter trajectory used in Figure 3(a). Comparing photon cutoffs nph=50n_{\mathrm{ph}}=50, 7070, and 100100 changes the maximum-slope pseudocritical couplings by less than 3×10−73\times10^{-7}. Refining the coupling grid from 80 to 160 intervals shifts these couplings by less than 0.04%0.04\%. The scans cover g∈[0.70gcMF,1.45gcMF]g\in[0.70g_c^{\mathrm{MF}},1.45g_c^{\mathrm{MF}}]. These checks establish numerical stability at the tested finite-size points.

nphn_{\mathrm{ph}}gc(N)g_c(N)Grid intervalsgc(N)g_c(N)
500.86500233400.86725648
700.86500241800.86527168
1000.865002311600.86500231

Table 1. Photon-cutoff and coupling-grid checks at N=200N=200 and Δ=0.0205\Delta=0.0205. The cutoff comparison uses the 160-interval grid; the grid comparison uses nph=100n_{\mathrm{ph}}=100.

All 18 sampled values of Δ\Delta were evaluated through N=240N = 240. The extrapolated estimates in Figure 3(a) are obtained from unweighted fits to

gc(N)=gc(∞)+rN−2/3,(75)g_c(N) = g_c(\infty) + rN^{-2/3}, \tag*{(75)}

using N={120,160,200,240}N = \{120, 160, 200, 240\}, except at Δ=0.15\Delta= 0.15 and 0.300.30, where we use N={200,240,320,400}N = \{200, 240, 320, 400\}. The latter gives gc(∞)=1.78003g_c(\infty) = 1.78003 and 2.110322.11032, respectively, compared with mean-field values 1.781741.78174 and 2.108192.10819. Increasing the photon cutoff from 70 to 100 changes gc(400)g_c(400) by less than 4×10−84 \times10^{-8} at these points. We assess extrapolation sensitivity by varying the fitting window and including a subleading term sN−4/3sN^{-4/3} in fits with at least four sizes. All plotted estimates use the leading four-size fits, with unconstrained intercepts. The variation between fits characterizes extrapolation sensitivity rather than a statistical confidence interval.

For the biased entropy curves in Figure 3(b), we include Hbias=hN(a+a†)H_{\mathrm{bias}} = h\sqrt{N}(a + a^\dagger) with h=10−4h = 10^{-4} and compare photon cutoffs nph=60n_{\mathrm{ph}} = 60, 80, and 100 for N=40N = 40, 100, and 200. The maximum absolute entropy differences over the sampled curves are below 1.5×10−81.5 \times10^{-8}, and the sampled peak locations remain unchanged. Their resolution is limited by the coupling-grid spacing, δg≃0.0156\delta g \simeq0.0156.

For the Hubbard-chirality comparison in Figure 3(c), we use t1=t2=1t_1 = t_2 = 1, t3=1.05t_3 = 1.05, U=30U = 30, and ωc=0.1\omega_c = 0.1, with a selecting field Hbias=hNXH_{\mathrm{bias}} = h\sqrt{N}X, where h=10−4h = 10^{-4}. Both models are evaluated at 49 equally spaced coupling points over 0≤g≤20 \leq g \leq2. We plot ∣⟨X⟩∣/(2N)\lvert\langle X\rangle\rvert/(2\sqrt{N}). The plotted curves use photon cutoffs nph=40n_{\mathrm{ph}} = 40, 20, 20, 12 for N=1,2,3,4N = 1, 2, 3, 4, respectively. Increasing the cutoffs from 20 to 40 for N=1N = 1 and from 10 to 20 for N=2,3N = 2, 3 changes the displacement by less than 4×10−84 \times10^{-8}. For N=4N = 4, increasing the cutoff from 8 to 12 changes it by less than 3×10−83 \times10^{-8}. The differences between the Hubbard and chirality curves therefore persist after photon-cutoff refinement.

For Figure 3(d), we consider N=20N = 20 in a fixed, fully polarized spin sector, with a selecting-field magnitude h=10−5h = 10^{-5}. Increasing the photon cutoff from nph=30n_{\mathrm{ph}} = 30 to 70 changes the plotted displacement by at most 3.4×10−103.4 \times10^{-10}. Refining the coupling-grid spacing from 0.01 to 0.005 shifts the maximum-slope location by at most 6.7×10−56.7 \times10^{-5}. At zero SOI, this location remains sensitive to the selecting-field strength: changing hh by a factor of ten shifts it by approximately 0.12–0.13. The panel therefore characterizes the finite-size response at h=10−5h = 10^{-5} rather than a thermodynamic phase boundary.

Temperature effects on the superradiant phase transition

We now include the projected DM interaction and extend the mean-field treatment to finite temperature. We choose the spin quantization axis along D0\mathbf{D}_0 and retain ΔSO=3∣D0∣\Delta_{\mathrm{SO}} = \sqrt{3}\lvert\mathbf{D}_0\rvert as defined above. With the collective scaling introduced above, ⟨a⟩=N α\langle a\rangle= \sqrt{N}\,\alpha, the single-molecule Hamiltonian reads

hMF(α)=Δ2Cx−K′sin⁡(2gα)Cz+ΔSOCzSz.(76)h_{\mathrm{MF}}(\alpha) = \frac{\Delta}{2}C_x - K'\sin(2g\alpha)C_z + \Delta_{\mathrm{SO}}C_zS_z. \tag*{(76)}

Here ΔSO\Delta_{\mathrm{SO}} is the coefficient of the projected spin-chirality coupling, rather than the full molecular excitation gap. Since SzS_z is conserved, the Hamiltonian separates into two chirality blocks labeled by σ=2Sz=±1\sigma= 2S_z = \pm1:

hσ(α)=(qσ(α)Δ/2Δ/2−qσ(α)),qσ(α)=−K′sin⁡(2gα)+σΔSO2.(77)h_\sigma(\alpha) = \begin{pmatrix} q_\sigma(\alpha) & \Delta/2 \\ \Delta/2 & -q_\sigma(\alpha) \end{pmatrix}, \qquad q_\sigma(\alpha) = -K'\sin(2g\alpha) + \sigma\frac{\Delta_{\mathrm{SO}}}{2}. \tag*{(77)}

The corresponding eigenvalues are ±Eσ(α)\pm E_\sigma(\alpha), with

Eσ(α)=(Δ2)2+qσ2(α).(78)E_\sigma(\alpha) = \sqrt{\left(\frac{\Delta}{2}\right)^2 + q_\sigma^2(\alpha)}. \tag*{(78)}

The chirality excitation gap within each spin sector is 2Eσ(α)2E_\sigma(\alpha).

We assume kBT≪Δexk_{\mathrm{B}}T \ll\Delta_{\mathrm{ex}}, so that thermal occupation of the excited S=3/2S = 3/2 quartet can be neglected. Then, tracing over all four states of the projected molecular Hilbert space gives the equilibrium free energy per molecule,

F(α)=ωcα2−1βln⁡Zel(α),Zel(α)=2∑σ=±1cosh⁡[βEσ(α)],(79)\mathcal{F}(\alpha) = \omega_c\alpha^2 - \frac{1}{\beta}\ln Z_{\mathrm{el}}(\alpha), \qquad Z_{\mathrm{el}}(\alpha) = 2\sum_{\sigma=\pm1}\cosh[\beta E_\sigma(\alpha)], \tag*{(79)}

where β=(kBT)−1\beta= (k_{\mathrm{B}}T)^{-1}. This trace assumes equilibrium populations of both spin sectors. Because Eσ(−α)=E−σ(α)E_\sigma(-\alpha) = E_{-\sigma}(\alpha), the free energy is even in α\alpha and is analytic at finite temperature.

Unconstrained spin polarization

To determine the local stability of the normal state, we introduce

Eref=Eσ(0)=(Δ2)2+(ΔSO2)2.(80)E_{\mathrm{ref}} = E_{\sigma}(0) = \sqrt{\left(\frac{\Delta}{2}\right)^2 + \left(\frac{\Delta_{\mathrm{SO}}}{2}\right)^2}. \tag*{(80)}

Expanding around α=0\alpha= 0 yields

F(α)=F(0)+[ωc−2K′2g2χeff(T,ΔSO)]α2+O(α4),(81)\mathcal{F}(\alpha) = \mathcal{F}(0) + \left[\omega_c - 2K^{\prime2}g^2\chi_{\mathrm{eff}}(T,\Delta_{\mathrm{SO}})\right]\alpha^2 + O(\alpha^4), \tag*{(81)}

where

χeff(T,ΔSO)=(Δ/2)2Eref3tanh⁡(βEref)+β(ΔSO/2)2Eref2.(82)\chi_{\mathrm{eff}}(T,\Delta_{\mathrm{SO}}) = \frac{(\Delta/2)^2}{E_{\mathrm{ref}}^3}\tanh(\beta E_{\mathrm{ref}}) + \beta\frac{(\Delta_{\mathrm{SO}}/2)^2}{E_{\mathrm{ref}}^2}. \tag*{(82)}

Here χeff\chi_{\mathrm{eff}} is the static susceptibility to a field coupled to CzC_z. Its first term describes virtual transitions between the chirality levels, while the second arises from redistribution between the two spin sectors and has a Curie-like temperature dependence. The normal state loses local stability at

gc(T,ΔSO)=ωc2K′2χeff(T,ΔSO).(83)g_c(T,\Delta_{\mathrm{SO}}) = \sqrt{\frac{\omega_c}{2K^{\prime2}\chi_{\mathrm{eff}}(T,\Delta_{\mathrm{SO}})}}. \tag*{(83)}

This is the equilibrium transition coupling when the transition is continuous. For ΔSO=0\Delta_{\mathrm{SO}} = 0 and T→0T \to0, it reduces to

gc(0,0)=ωcΔ2∣K′∣.(84)g_c(0,0) = \frac{\sqrt{\omega_c\Delta}}{2\lvert K^{\prime}\rvert}. \tag*{(84)}

For nonzero ΔSO\Delta_{\mathrm{SO}}, the Curie contribution instead dominates as T→0T \to0, giving

gc(T,ΔSO)≃Eref∣K′∣ΔSO2ωckBT.(85)g_c(T,\Delta_{\mathrm{SO}}) \simeq\frac{E_{\mathrm{ref}}}{\lvert K^{\prime}\rvert\Delta_{\mathrm{SO}}}\sqrt{2\omega_c k_{\mathrm{B}}T}. \tag*{(85)}

At strictly zero temperature, minimizing over the two spin sectors gives

FT=0(α)=ωcα2−(Δ2)2+[ΔSO2+∣K′sin⁡(2gα)∣]2.(86)\mathcal{F}_{T=0}(\alpha) = \omega_c\alpha^2 - \sqrt{\left(\frac{\Delta}{2}\right)^2 + \left[\frac{\Delta_{\mathrm{SO}}}{2} + \lvert K^{\prime}\sin(2g\alpha)\rvert\right]^2}. \tag*{(86)}

For ΔSO>0\Delta_{\mathrm{SO}} > 0 and g≥0g \ge0, its small-displacement expansion is

FT=0(α)−FT=0(0)=−∣K′∣gΔSOEref∣α∣+O(α2).(87)\mathcal{F}_{T=0}(\alpha) - \mathcal{F}_{T=0}(0) = -\frac{\lvert K^{\prime}\rvert g\Delta_{\mathrm{SO}}}{E_{\mathrm{ref}}}\lvert\alpha\rvert+ O(\alpha^2). \tag*{(87)}

The negative linear cusp dominates the quadratic cavity cost, so any nonzero coupling destabilizes α=0\alpha= 0 and favors symmetry-related displaced minima. Finite temperature rounds the cusp and restores a finite normal-state instability threshold. The sinusoidal and linearized couplings give the same local instability condition because ∂αsin⁡(2gα)∣α=0=2g\partial_\alpha\sin(2g\alpha)|_{\alpha=0} = 2g. Away from the normal state, their free energies differ, and the full sinusoidal dependence determines the equilibrium displacement and the fluctuation spectrum.

Fixed magnetization at zero temperature

If the total spin polarization is constrained to vanish, the two spin sectors have equal populations. Their ground-state energies are averaged before minimizing over the cavity displacement:

Ffixed(α)=ωcα2−12[E+(α)+E−(α)].(88)\mathcal{F}_{\mathrm{fixed}}(\alpha) = \omega_c\alpha^2 - \frac{1}{2}\left[E_{+}(\alpha) + E_{-}(\alpha)\right]. \tag*{(88)}

The resulting energy remains even in α\alpha and, for Δ>0\Delta> 0, is analytic near α=0\alpha= 0. Expanding the ground-state energy to quadratic order gives

Ffixed(α)=Ffixed(0)+[ωc−4g2K′2Δ2(Δ2+ΔSO2)3/2]α2+O(α4).(89)\mathcal{F}_{\mathrm{fixed}}(\alpha) = \mathcal{F}_{\mathrm{fixed}}(0) + \left[\omega_c - \frac{4g^2K^{\prime2}\Delta^2}{(\Delta^2+\Delta_{\mathrm{SO}}^2)^{3/2}}\right]\alpha^2 + O(\alpha^4). \tag*{(89)}

The instability therefore occurs at

gc(ΔSO)=gc(0)[1+(ΔSOΔ)2]3/4.(90)g_c(\Delta_{\mathrm{SO}}) = g_c(0)\left[1+\left(\frac{\Delta_{\mathrm{SO}}}{\Delta}\right)^2\right]^{3/4}. \tag*{(90)}

This is the normal-state instability threshold and equals the transition coupling when the transition is continuous.

Experimental estimates for the orbital light-matter coupling

The zero-point current fluctuations of a superconducting LC resonator are given by [54]

Izpf=ℏωc2L.(91)I_{\mathrm{zpf}}=\sqrt{\frac{\hbar\omega_c}{2L}}. \tag*{(91)}

Approximating the inductive element as a circular loop of radius RR, the magnetic field at its center follows from the Biot-Savart law as

Bzpf≃μ0Izpf2R.(92)B_{\mathrm{zpf}}\simeq\frac{\mu_0 I_{\mathrm{zpf}}}{2R}. \tag*{(92)}

A molecular triangle of area A△A_{\triangle}, oriented approximately parallel to the loop, then experiences the zero-point flux

Φzpf≃BzpfA△,(93)\Phi_{\mathrm{zpf}}\simeq B_{\mathrm{zpf}}A_{\triangle}, \tag*{(93)}

leading to the dimensionless Peierls coupling

gi=2πΦzpf,iΦe,Φe=he.(94)g_i=2\pi\frac{\Phi_{\mathrm{zpf},i}}{\Phi_e},\qquad\Phi_e=\frac{h}{e}. \tag*{(94)}

For an ensemble of molecular triangles, the collective coupling to the resonator mode is characterized by

g=∑i=1Ngi2.(95)g=\sqrt{\sum_{i=1}^{N}g_i^2}. \tag*{(95)}

Here Φzpf,i\Phi_{\mathrm{zpf},i} is the physical zero-point magnetic flux through triangle ii, while gig_i is the corresponding dimensionless Peierls-phase amplitude. The phase operator entering its ring-exchange interaction is

Φ^cav,i=gi(a+a†).(96)\hat{\Phi}_{\mathrm{cav},i}=g_i(a+a^\dagger). \tag*{(96)}

For approximately uniform coupling, gi≃gmolg_i\simeq g_{\mathrm{mol}} and g≃gmolNg\simeq g_{\mathrm{mol}}\sqrt{N}. Consequently, Φ^cav,i=Φ^cav/N\hat{\Phi}_{\mathrm{cav},i}=\hat{\Phi}_{\mathrm{cav}}/\sqrt{N}, with Φ^cav=g(a+a†)\hat{\Phi}_{\mathrm{cav}}=g(a+a^\dagger), recovering the collective convention used in the main text.

To obtain a conservative estimate, we consider a circular superconducting loop with representative parameters

R∼25 μm,L∼1 nH,ωc∼2π×5 GHz,(97)R\sim25\ \mu\mathrm{m},\qquad L\sim1\ \mathrm{nH},\qquad\omega_c\sim2\pi\times5\ \mathrm{GHz}, \tag*{(97)}

together with a molecular area A△∼1 nm2A_{\triangle}\sim1\ \mathrm{nm}^2. These parameters give:

Izpf=ℏωc2L∼4×10−8 A,(98)I_{\mathrm{zpf}}=\sqrt{\frac{\hbar\omega_c}{2L}}\sim4\times10^{-8}\ \mathrm{A}, \tag*{(98)}

and hence

Bzpf∼10−9 T.(99)B_{\mathrm{zpf}}\sim10^{-9}\ \mathrm{T}. \tag*{(99)}

The corresponding single-molecule Peierls coupling is therefore of order

gmol∼10−12.(100)g_{\mathrm{mol}}\sim10^{-12}. \tag*{(100)}

For an ensemble containing N∼1012N \sim10^{12} molecular triangles, this yields

g≃gmolN∼10−6.(101)g \simeq g_{\mathrm{mol}}\sqrt{N} \sim10^{-6}. \tag*{(101)}

Assuming a molecular number density of order 1 nm−31\ \mathrm{nm}^{-3}, N∼1012N \sim10^{12} corresponds to a volume of approximately 103 μm310^{3}\ \mathrm{\mu m}^{3}. This could be accommodated in platelets with lateral dimensions of 10×1010 \times10 and 20×20 μm220 \times20\ \mathrm{\mu m}^{2} and thicknesses of approximately 1010 and 2.5 μm2.5\ \mathrm{\mu m}, respectively. The homogeneous-field approximation should be assessed over this sample volume. In the normal phase, the collective transition coupling is

Δcav=∣K′∣g∣⟨1∣Cz∣0⟩∣.(102)\Delta_{\mathrm{cav}} = \lvert K' \rvert g \lvert\langle1 \lvert C_z \rvert0 \rangle\rvert. \tag*{(102)}

For the SOI-free model, ∣⟨1∣Cz∣0⟩∣=1\lvert\langle1 \lvert C_z \rvert0 \rangle\rvert= 1, giving Δcav=∣K′∣g\Delta_{\mathrm{cav}} = \lvert K' \rvert g. For the illustrative range

K′ℏ∼10–100 GHz,(103)\frac{K'}{\hbar} \sim10\text{--}100\ \mathrm{GHz}, \tag*{(103)}

the circular-loop estimate gives

Δcavℏ∼10–100 kHz.(104)\frac{\Delta_{\mathrm{cav}}}{\hbar} \sim10\text{--}100\ \mathrm{kHz}. \tag*{(104)}

This corresponds to an equivalent temperature scale

ΔcavkB∼0.5–5 μK.(105)\frac{\Delta_{\mathrm{cav}}}{k_B} \sim0.5\text{--}5\ \mathrm{\mu K}. \tag*{(105)}

For comparison, molecular electron-spin ensembles can exhibit MHz-scale inhomogeneous broadening [55], although the linewidth of the chirality transition considered here requires a separate material-specific assessment. These estimates provide a baseline for a simple circular-loop geometry. Increasing the collective orbital coupling requires optimizing the zero-point magnetic field over the molecular sample. Narrow inductive constrictions and other field-focusing structures offer possible routes [47], but their benefits must be evaluated through g=∑igi2g = \sqrt{\sum_i g_i^2}, taking into account the spatial field profile and the resulting inductance of the resonator. If the collective coupling itself could be enhanced by a factor of 1010--100100, the estimated interaction scale would increase to approximately 0.10.1--10 MHz10\ \mathrm{MHz}. Whether this permits resolved chirality polaritons depends on the cavity and chirality-transition linewidths.

References

  1. [1]K. Rzążewski, K. Wódkiewicz, and W. Żakowicz, Phase transitions, two-level atoms, and the A² term, Phys. Rev. Lett. 35, 432 (1975).DOI
  2. [2]I. Bialynicki-Birula and K. Rzążewski, No-go theorem concerning the superradiant phase transition in atomic systems, Phys. Rev. A 19, 301 (1979).
  3. [3]K. Gawȩdzki and K. Rzążewski, No-go theorem for the superradiant phase transition without dipole approximation, Phys. Rev. A 23, 2134 (1981).DOI
  4. [4]P. Nataf and C. Ciuti, No-go theorem for superradiant quantum phase transitions in cavity QED and counter-example in circuit QED, Nat. Commun. 1, 72 (2010).DOI
  5. [5]G. M. Andolina, F. M. D. Pellegrino, V. Giovannetti, A. H. MacDonald, and M. Polini, Cavity quantum electrodynamics of strongly correlated electron systems: A no-go theorem for photon condensation, Phys. Rev. B 100, 121109(R) (2019).DOI
  6. [6]P. Nataf, T. Champel, G. Blatter, and D. M. Basko, Rashba cavity QED: A route towards the superradiant quantum phase transition, Phys. Rev. Lett. 123, 207402 (2019).DOI
  7. [7]G. M. Andolina, F. M. D. Pellegrino, V. Giovannetti, A. H. MacDonald, and M. Polini, Theory of photon condensation in a spatially varying electromagnetic field, Phys. Rev. B 102, 125137 (2020).DOI
  8. [8]D. Guerci, P. Simon, and C. Mora, Superradiant phase transition in electronic systems and emergent topological phases, Phys. Rev. Lett. 125, 257604 (2020).
  9. [9]V. K. Kozin, E. Thingstad, D. Loss, and J. Klinovaja, Cavity-enhanced superconductivity via band engineering, Phys. Rev. B 111, 035410 (2025).DOI
  10. [10]J. H. Condon, Nonlinear de Haas–van Alphen effect and magnetic domains in beryllium, Phys. Rev. 145, 526 (1966).DOI
  11. [11]D. Guerci, P. Simon, and C. Mora, Moiré lattice effects on the orbital magnetic response of twisted bilayer graphene and Condon instability, Phys. Rev. B 103, 224436 (2021).DOI
  12. [12]G. Manzanares, T. M. Champel, D. M. Basko, and P. Nataf, Superradiant quantum phase transition for Landau polaritons with Rashba and Zeeman couplings, Phys. Rev. B 105, 245304 (2022).DOI
  13. [13]Z. Bacciconi, G. M. Andolina, T. Chanda, G. Chiriacò, M. Schirò, and M. Dalmonte, First-order photon condensation in magnetic cavities: A two-leg ladder model, Sci-Post Phys. 15, 113 (2023).
  14. [14]A. Mercurio, G. M. Andolina, F. M. D. Pellegrino, O. Di Stefano, P. Jarillo-Herrero, C. Felser, F. H. L. Koppens, S. Savasta, and M. Polini, Photon condensation, Van Vleck paramagnetism, and chiral cavities, Phys. Rev. Res. 6, 013303 (2024).DOI
  15. [15]M. Parisi, E. Paladino, G. A. Falci, G. M. Andolina, S. Savasta, M. Polini, and F. M. D. Pellegrino, Interplay between photon condensation and electron-electron interactions in molecular systems, Phys. Rev. B 114, 225416 (2026).DOI
  16. [16]O. I. Motrunich, Orbital magnetic field effects in spin liquid with spinon Fermi sea: Possible application to κ-(ET)2Cu2(CN)3, Phys. Rev. B 73, 155115 (2006).DOI
  17. [17]L. N. Bulaevskii, C. D. Batista, M. V. Mostovoy, and D. I. Khomskii, Electronic orbital currents and polarization in Mott insulators, Phys. Rev. B 78, 024402 (2008).DOI
  18. [18]X. G. Wen, F. Wilczek, and A. Zee, Chiral spin states and superconductivity, Phys. Rev. B 39, 11413 (1989).pubmed.ncbi.nlm.nih.gov/9947970
  19. [19]V. Kalmeyer and R. B. Laughlin, Equivalence of the resonating-valence-bond and fractional quantum Hall states, Phys. Rev. Lett. 59, 2095 (1987).pubmed.ncbi.nlm.nih.gov/10035416
  20. [20]S. Kitamura, T. Oka, and H. Aoki, Probing and controlling spin chirality in Mott insulators by circularly polarized laser, Phys. Rev. B 96, 014406 (2017).DOI
  21. [21]M. Claassen, H.-C. Jiang, B. Moritz, and T. P. Devereaux, Dynamical time-reversal symmetry breaking by photo-induced chiral spin liquids in frustrated Mott insulators, Nat. Commun. 8, 1192 (2017).
  22. [22]F. L. Pedrocchi, S. Chesi, and D. Loss, Quantum memory coupled to cavity modes, Phys. Rev. B 83, 115415 (2011).DOI
  23. [23]M. A. Sentef, J. Li, F. Künzel, and M. Eckstein, Quantum to classical crossover of Floquet engineering in correlated quantum systems, Phys. Rev. Res. 2, 033033 (2020).
  24. [24]E. Viñas Boström, A. Sriram, M. Claassen, and A. Rubio, Controlling the magnetic state of the proximate quantum spin liquid α-RuCl3 with an optical cavity, npj Comput. Mater. 9, 202 (2023).DOI
  25. [25]L. Grundwall, X. Cheng, E. Viñas Boström, M. Ruggen-thaler, M. H. Michael, D. M. Kennes, and A. Rubio, Cavity control of strongly correlated electrons beyond resonant coupling (2026), arXiv:2603.18933 [quant-ph].
  26. [26]A. Chiocchetta, D. Kiese, C. P. Zelle, F. Piazza, and S. Diehl, Cavity-induced quantum spin liquids, Nat. Commun. 12, 5901 (2021).DOI
  27. [27]L. Weber, E. Viñas Boström, M. Claassen, A. Rubio, and D. M. Kennes, Cavity-renormalized quantum criticality in a honeycomb bilayer antiferromagnet, Commun. Phys. 6, 247 (2023).
  28. [28]C. Wei, L. Yang, and Q.-D. Jiang, Cavity-vacuum-induced chiral spin liquids in kagome lattices: Tuning and probing topological quantum phases via cavity quantum electrodynamics, Phys. Rev. Lett. 135, 236901 (2025).DOI
  29. [29]J. Li, D. Golez, G. Mazza, A. J. Millis, A. Georges, and M. Eckstein, Electromagnetic coupling in tight-binding models for strongly correlated light and matter, Phys. Rev. B 101, 205140 (2020).DOI
  30. [30]O. Dmytruk and M. Schiró, Gauge fixing for strongly correlated electrons coupled to quantum light, Phys. Rev. B 103, 075131 (2021).
  31. [31]Gauge transformations in a quantized light-matter system act on both electronic and photonic degrees of freedom [30, 50]. Redistributing the bond phases therefore requires transforming the remaining Hamiltonian consistently to preserve gauge equivalence.
  32. [32]A. H. MacDonald, S. M. Girvin, and D. Yoshioka, t/U expansion for the Hubbard model, Phys. Rev. B 37, 9753 (1988).pubmed.ncbi.nlm.nih.gov/9944372
  33. [33]See Supplemental Material for the Schrieffer–Wolff derivation, collective-fluctuation theory, finite-temperature analysis, and numerical details.
  34. [34]At the order considered here, the effective spin Hamiltonian depends only on the total operator-valued flux threading the triangle and is therefore insensitive to the particular distribution of Peierls phases among the bonds.
  35. [35]M. Trif, F. Troiani, D. Stepanenko, and D. Loss, Spin-electric coupling in molecular magnets, Phys. Rev. Lett. 101, 217201 (2008).DOI
  36. [36]M. Trif, F. Troiani, D. Stepanenko, and D. Loss, Spin electric effects in molecular antiferromagnets, Phys. Rev. B 82, 045429 (2010).DOI
  37. [37]Y. Taguchi, Y. Ohara, H. Yoshizawa, N. Nagaosa, and Y. Tokura, Spin chirality, Berry phase, and anomalous Hall effect in a frustrated ferromagnet, Science 291, 2573 (2001).pubmed.ncbi.nlm.nih.gov/11283363
  38. [38]M. F. Islam, K. P. K. Withanage, C. M. Canali, and M. R. Pederson, First-principles study of spin-electric coupling in a Cu3 single-molecule magnet, Phys. Rev. B 82, 155446 (2010).
  39. [39]J. F. Nossa, M. F. Islam, M. R. Pederson, and C. M. Canali, Electric control of spin states in frustrated triangular molecular magnets, Phys. Rev. B 107, 245402 (2023).DOI
  40. [40]A. Cini, M. Böhme, B. Kintzel, M. Perfetti, W. Plass, R. Sessoli, and M. Fittipaldi, Electric Control of magnetic exchange in a molecular spin triangle, Nat. Commun. 16, 6564 (2025).DOI
  41. [41]T. Holstein and H. Primakoff, Field dependence of the intrinsic domain magnetization of a ferromagnet, Phys. Rev. 58, 1098 (1940).DOI
  42. [42]C. Emary and T. Brandes, Chaos and the quantum phase transition in the Dicke model, Phys. Rev. E 67, 066203 (2003).DOI
  43. [43]A. Baksic and C. Ciuti, Controlling discrete and continuous symmetries in “Superradiant” phase transitions with circuit QED systems, Phys. Rev. Lett. 112, 173601 (2014).pubmed.ncbi.nlm.nih.gov/24836245
  44. [44]J. Vidal and S. Dusuel, Finite-size scaling exponents in the Dicke model, Europhys. Lett. 74, 817 (2006).DOI
  45. [45]G. Vidal, J. I. Latorre, E. Rico, and A. Kitaev, Entanglement in quantum critical phenomena, Phys. Rev. Lett. 90, 227902 (2003).pubmed.ncbi.nlm.nih.gov/12857342
  46. [46]N. Lambert, C. Emary, and T. Brandes, Entanglement and the phase transition in single-mode superradiance, Phys. Rev. Lett. 92, 073602 (2004).pubmed.ncbi.nlm.nih.gov/14995850
  47. [47]I. Gimeno, W. Kersten, M. C. Pallarés, P. Hermosilla, M. J. Martínez-Pérez, M. D. Jenkins, A. Angerer, C. Sánchez-Azqueta, D. Zueco, J. Majer, A. Lostao, and F. Luis, Enhanced molecular spin-photon coupling at superconducting nanoconstrictions, ACS Nano 14, 8707 (2020).
  48. [48]J. D. Broz, J. C. Hoke, E. Acuna, and J. R. Petta, Leakage-protected idle operation of a triangular exchange-only spin qubit, PRX Quantum 7, 020324 (2026).
  49. [49]V. John, C. X. Yu, B. van Straaten, E. A. Rodríguez-Mena, M. Rodriguez, S. D. Oosterhout, L. E. A. Stehouwer, G. Scappucci, M. Rimbach-Russ, S. Bosco, F. Borsoi, Y.-M. Niquet, and M. Veldhorst, Robust and localised control of a 10-spin qubit array in germanium, Nat. Commun. 16, 10560 (2025).DOI
  50. [51]I. Dzyaloshinsky, A thermodynamic theory of “weak” ferromagnetism of antiferromagnetics, J. Phys. Chem. Solids 4, 241 (1958).DOI
  51. [52]T. Moriya, Anisotropic superexchange interaction and weak ferromagnetism, Phys. Rev. 120, 91 (1960).DOI
  52. [53]Z. Kurucz and K. Mølmer, Multilevel Holstein–Primakoff approximation and its application to atomic spin squeezing and ensemble quantum memories, Phys. Rev. A 81, 032314 (2010).DOI
  53. [54]S. M. Girvin, Circuit QED: Superconducting qubits coupled to microwave photons, in Quantum Machines: Measurement and Control of Engineered Quantum Systems, Les Houches Summer School, Vol. 96, edited by M. H. Devoret, B. Huard, R. J. Schoelkopf, and L. F. Cugliandolo (Oxford University Press, Oxford, 2014) pp. 113–256.DOI
  54. [55]S. Lenz, D. König, D. Hunger, and J. J. van Slageren, Room-temperature quantum memories based on molecular electron spin ensembles, Adv. Mater. 33, 2101673 (2021).DOI

Paper details

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

views · 7

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