When Can a Cavity Move a Mott Transition? A Spectral-Density Criterion within Gutzwiller Theory

Technical Summary: "When Can a Cavity Move a Mott Transition? A Spectral-Density Criterion within Gutzwiller Theory"

Problem Statement
The central question addressed is whether vacuum electromagnetic fluctuations in a quantized cavity environment can alter the equilibrium thermodynamic phase boundary of a bulk correlation-driven metal–insulator transition (specifically the Mott transition). While recent studies have shown that cavity fields can induce collective spectroscopic effects (such as Rabi splitting) and modify microscopic interactions, it remains unsettled whether these fields can shift the critical interaction strength ($U_c$) required for a Mott transition in the thermodynamic limit. Previous work suggests that properly normalized single modes are often irrelevant to the thermodynamic energy density, yet realistic multimode environments with confined surface modes may offer different outcomes. A rigorous criterion connecting the frequency and spatial structure of an arbitrary electromagnetic environment to the correlation-driven phase boundary has been missing.

Methodology
The authors employ the Gutzwiller variational method extended to include quantized electromagnetic fields within the Pauli–Fierz framework.

• Model: The study utilizes the half-filled Hubbard model coupled to normalized electromagnetic modes. The Hamiltonian includes the kinetic energy (scaled by coordination number $z$), on-site Coulomb repulsion $U$, and the photonic sector with light–matter coupling $\lambda$.
• Variational Approach: A variational photon displacement is introduced, mapping the Pauli–Fierz Hamiltonian onto a photon-dressed electronic Hamiltonian plus a separable residual interaction. This approach is exact within the variational manifold in the infinite-coordination limit ($z \to \infty$), recovering the analytic Brinkman–Rice quasiparticle-collapse transition.
• Spectral Analysis: The authors define a joint Pauli–Fierz spectral density, $J_{PF}(\omega, r)$, which characterizes modes by their thermodynamic weight per site ($\eta_\alpha$) and their bond-gradient content ($r_\alpha$). The bond-gradient factor $r_\alpha$ measures the spatial variation of the field across nearest-neighbor bonds.
• Validation: The analytical results are validated using finite-coordination variational Monte Carlo (VMC) on random regular graphs. This allows for an exact evaluation of the determinant state within the chosen variational family, testing the robustness of the infinite-$z$ predictions at finite coordination numbers ($z=4, 8, 12, 16$).
• Specific Geometry: The theory is applied to a surface-phonon-polariton geometry (a correlated layer above a 4H-SiC surface), utilizing macroscopic QED and the dyadic Green tensor to calculate the spectral density in the nonretarded regime.

Key Contributions and Results

1. The Spectral-Density Criterion: The paper derives a necessary condition for a cavity to shift a Mott transition: the electromagnetic environment must supply finite thermodynamic spectral weight that exhibits bond-scale spatial variation.

   • A single normalized mode contributes $O(1/N)$ to the energy density and cannot shift the phase boundary in the thermodynamic limit unless it possesses superextensive coupling or divergent susceptibility.
   • The shift in the critical interaction, $\Delta U_c$, is governed by the joint spectral density $J_{PF}(\omega, r)$ integrated against a response kernel $K(\omega, r) = \frac{4A_0 r}{\omega + 4A_0 r}$.
   • Uniform fields ($r \to 0$) do not shift the transition, regardless of intensity. Only fields that vary across the electronic bond ($r > 0$) contribute.

2. Mechanism of the Shift: The shift arises from two cooperative effects:

   • Residual Self-Polarization: Increases the cost of charge fluctuations.
   • Bandwidth Narrowing: Photon dressing reduces the effective kinetic energy scale.
   • The paper clarifies that $\Delta U_c < 0$ does not imply a reduction in the bare Coulomb integral $U$, but rather that a smaller intrinsic $U$ is sufficient to localize the photon-dressed quasiparticles.

3. Surface-Polariton Scaling: For a correlated layer at distance $d$ above a polar surface (e.g., 4H-SiC):

   • The shift is controlled by the bond-projected component of the local near-field weight.
   • A crossover in scaling behavior is predicted: for $d \gg a$ (lattice spacing), the shift scales as $d^{-5}$; for $d \sim a$, it crosses over to the local near-field law of $d^{-3}$.
   • This demonstrates that a large local density of optical states is insufficient; the field must resolve the bond scale to be effective.

4. VMC Validation:

   • Finite-coordination VMC calculations confirm the analytical critical coefficient within 5–7% for $z=4$ to $16$.
   • The calculations verify the $M/N$ extensivity law, showing that the energy shift scales linearly with the mode density $\nu_m = M/N$ and vanishes for a single normalized mode ($M=1, N \to \infty$).

Significance and Claims
The paper claims to resolve the apparent tension between large spectroscopic hybridization (e.g., Rabi splitting) and negligible thermodynamic shifts. It establishes that collective strong coupling alone is insufficient to move a Mott transition. Instead, the ability to control equilibrium phases depends on the joint spectral density of the environment, specifically its ability to provide finite weight at frequencies and spatial scales that match the electronic localization process.

The work provides a modular design rule for cavity quantum electrodynamics (QED) in correlated materials:

• It bridges correlated-electron theory (Gutzwiller/DMFT) with realistic nanophotonic electrodynamics (Green tensor calculations).
• It allows for the screening of photonic structures without arbitrary discrete-mode cutoffs by focusing on the bond-projected field correlations.
• The authors suggest that future designs should prioritize engineering momentum-, frequency-, and polarization-resolved field correlations matched to the specific many-body process, rather than simply maximizing Rabi splitting or minimizing mode volume.

The paper remains modest regarding its scope, noting that while the VMC results are robust within the Gutzwiller family, they do not constitute an unbiased solution to the finite-dimensional Hubbard model (omitting Hubbard bands and long-range magnetic order). The framework is presented as a foundation that can be extended to other response functions (e.g., DMFT, QMC) to address charge-transfer, magnetic, and superconducting instabilities.