Finite-volume and thermodynamic-limit lower bounds for the coherently driven Bose--Hubbard model via the Lieb--Yamazaki method

Technical Summary: Finite-volume and Thermodynamic-limit Lower Bounds for the Coherently Driven Bose–Hubbard Model via the Lieb–Yamazaki Method

Problem Statement
The paper addresses the challenge of obtaining rigorous lower bounds for the ground-state energy of the coherently driven Bose–Hubbard model. While variational trial states readily provide upper bounds, rigorous lower bounds are difficult to establish in the presence of both quartic on-site interactions and a coherent source term (which breaks particle-number conservation). Existing methods, such as elementary one-site or bond-cluster estimates, often lack explicit dependence on model parameters (detuning, hopping, drive) or fail to provide a stable framework for analyzing observables and excitation spectra in the thermodynamic limit. The authors aim to construct a rigorous lower bound that remains stable under the thermodynamic limit, retains explicit parameter dependence, and yields a solvable auxiliary system for deriving macroscopic observables.

Methodology
The authors apply the Lieb–Yamazaki positive-operator decomposition to the driven Bose–Hubbard Hamiltonian on a periodic hypercubic lattice in arbitrary dimension $d \ge 1$. The methodology proceeds through the following steps:

1. Positive-Operator Decomposition: The dimensionless Hamiltonian $h$ is decomposed into a sum of positive semidefinite operators $w^2 = \sum B_k^\dagger B_k$ and a remainder Hamiltonian $h_1$. The operators $B_k$ are defined using real variational parameters $K$ and $g$:
   $$B_k = a_k^\dagger a_k - K - g(a_k - a_k^\dagger)$$
   This yields the operator inequality $h \ge h_1$.

2. Diagonalization of the Remainder: The remainder $h_1$ is a quadratic Hamiltonian. It is diagonalized via:

   • A uniform real displacement $\lambda$ to eliminate linear terms.
   • A Bogoliubov transformation to diagonalize the quadratic terms.
     This process defines a "bounding Hamiltonian" $h_1$ whose ground state is a Gaussian state $|0_B\rangle$.

3. Stability Conditions: The validity of the decomposition and the diagonalization requires specific stability constraints on the parameters $K$ and $g$. Specifically, the Bogoliubov transformation is canonical and yields a positive spectrum only if $\omega_\mu > 2g^2$ for all momentum modes $\mu$, where $\omega_\mu$ depends on $K, g$, the hopping $j$, and the detuning $\nu_0$.

4. Thermodynamic Limit: The authors prove the existence of the thermodynamic limit for the lower-bound energy density $e_{0,l}^{(d)}$ and the continuous excitation spectrum $\Omega(q)$ by converting lattice sums into Brillouin-zone integrals.

5. Temple Inequality Analysis: The paper investigates the applicability of Temple's inequality to refine the lower bound using the variance of the full Hamiltonian in the Gaussian state. It establishes that while the inequality is rigorous in finite volume, the extensive nature of the energy variance relative to the intensive spectral gap makes the condition increasingly restrictive as system size grows.

Key Contributions and Results

• Rigorous Lower Bound (Theorem II.3): The authors derive an explicit, rigorous lower bound for the ground-state energy in any dimension $d \ge 1$. The bound is parameterized by $K$ and $g$ subject to a stability constraint. The thermodynamic limit of this bound is expressed as a closed-form integral over the Brillouin zone.
• Excitation Spectrum: The bounding Hamiltonian $h_1$ possesses a continuous excitation band $\Omega(q)$.
  • In the stable region, the spectrum has a finite energy gap and a quadratic dispersion (massive quasiparticles).
  • At the boundary of the admissible region, the gap closes, and the dispersion becomes linear ($\Omega(q) \propto |q|$) with a slope determined by the pairing parameter $g$ and hopping $j$. The authors emphasize that this linear slope characterizes the bounding Hamiltonian, not necessarily the full driven model.
• Macroscopic Observables: Within the Gaussian ground state $|0_B\rangle$ of the bounding Hamiltonian, the authors derive exact expressions for:
  • Order parameter (displacement $\lambda$).
  • Filling fraction (particle density).
  • Energy densities (kinetic and on-site).
  • Particle-number variance and the second-order correlation function $g^{(2)}(0)$.
  • Anomalous averages.
    All these quantities possess well-defined thermodynamic limits.
• Antibunching: The analysis shows that a sufficiently strong coherent drive induces antibunching ($g^{(2)}(0) < 1$) and sub-Poissonian statistics within the auxiliary Gaussian state. This is driven by the displacement term overcoming the fluctuations.
• Numerical Assessment:
  • The authors compare the Lieb–Yamazaki bounds with elementary site and bond-cluster lower bounds. They find that while the bond-cluster bound is often sharper numerically, the Lieb–Yamazaki construction offers the advantage of an analytically solvable auxiliary state and explicit parameter dependence.
  • Numerical optimization of the variational parameters $(K, g)$ reveals that the relative gap between the lower bound and the variational upper bound (evaluated at the same parameters) is narrowest for strong drives ($f \gtrsim 3$) and moderate hopping.
  • Finite-size effects for a system of size $M=4$ are found to be small compared to the width of the energy bracket, indicating that the "looseness" of the bound is primarily due to the limitations of the Gaussian ansatz rather than finite-size artifacts.
  • The Temple inequality condition ($M^d \Delta_M < \Omega(0)$) is shown to fail for large systems unless the trial state is extremely close to the exact ground state, limiting its utility as a general thermodynamic correction.

Significance and Claims
The paper claims to establish a variational family of rigorous lower bounds for the driven Bose–Hubbard model that is valid in arbitrary dimensions and possesses a well-defined thermodynamic limit. The primary significance lies in the construction of an explicitly solvable auxiliary Gaussian state ($h_1$) that provides:

1. A certified lower threshold for the ground-state energy.
2. Analytical expressions for observables and correlation functions that are exact within the bounding state.
3. A framework to identify regimes of drive-induced antibunching.

The authors are modest regarding the precision of the bounds. They acknowledge that the Gaussian ansatz cannot capture strong non-Gaussian correlations (e.g., near the stability boundary or in the Mott regime of the undriven model, which is not reproduced here due to symmetry breaking). Consequently, the bounds are not claimed to be uniformly optimal; in fact, elementary bond-cluster bounds are numerically sharper in several test cases. The value of the work is presented as providing a rigorous, analytically tractable lower bound with explicit dependence on physical parameters, rather than a replacement for high-precision numerical methods like DMRG or tensor networks. The results characterize the properties of the bounding Hamiltonian (e.g., its gapless boundary slope), and the authors explicitly state that whether these features correspond to the full interacting model requires independent verification via finite-size scaling or other numerical techniques.