Fast mixing of all-to-all quantum systems at high temperatures

arXiv:2606.26090 · quant-ph · Submitted 2026-06-24 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Quantum Radio. Generated commentary on the latest quantum physics and condensed matter papers.

Kai: Today's paper: "Fast mixing of all-to-all quantum systems at high temperatures".

Mira: This paper establishes that arbitrary all-to-all quantum k-local Hamiltonians with bounded strength interactions admit a quantum Gibbs sampler

CKG23: possessing a system-size independent spectral gap at sufficiently high temperatures.

Kai: First, who's behind it and why it matters.

Title and authors: Kai: So, let’s talk about who wrote this and what the title says. The paper is titled "Fast mixing of all-to-all quantum systems at high temperatures," and I see Thiago Bergamaschi is the lead author. Mira, what's your take on the significance of that specific title?

Mira: I think the title immediately signals a focus on two things: the nature of the interactions, which are all-to-all, and the condition under which fast mixing occurs—at high temperatures. It’s telling us that thermal effects can overcome structural limitations in how we analyze dynamics.

Lev: For someone like me working on error correction, I’m curious if this means we can apply these tools to systems that are inherently non-geometric, which is where traditional Lieb-Robinson bounds fail us. If the authors managed to prove fast mixing under these conditions, it suggests a path forward for analyzing strongly correlated systems.

Kai: Right, that's the big picture—tackling those long-range interactions we see in things like electronic structure Hamiltonians or strongly correlated models that don't fit neatly onto a simple lattice structure. It’s about finding new ways to understand how information spreads in these scenarios without relying on geometric constraints.

Mira: Precisely, and the authors are doing this by using tools like the quantum cluster expansion of Netoˇcn´y and Redig NR04 to decompose the complex evolution into manageable pieces, which is a clever way to handle that non-locality.

Lev: That decomposition sounds promising for complexity analysis because it breaks down a hard problem into smaller, localized components, even if those components are weakly coupled in some sense.

Kai: It’s about finding that structure within the chaos of all-to-all interactions and showing how high temperatures can exploit that structure to achieve rapid thermalization.

The paper's summary: Kai: Now, let’s look at the detailed summary again, because I want to make sure we’re on the same page regarding the technical steps they took.

Mira: The paper describes modeling this thermalization process using a quantum Markov semigroup generated by the thermal Lindbladian L, and the central technical achievement is establishing a spectral gap for that Lindbladian L for arbitrary all-to-all Hamiltonians at high temperatures.

Lev: That’s the core mathematical statement; proving the spectral gap for L is what mathematically guarantees that the system converges rapidly to its Gibbs state, which is what we need to move from slow mixing to fast simulation algorithms.

Kai: And this result then flows into Corollary one point four, showing a polynomial time quantum algorithm exists for estimating partition functions and global expectation values with a relative error of one plus epsilon. That’s the practical payoff we’re talking about here.

Mira: So, to put it simply, they showed that by choosing high enough temperatures, the system dynamics become fast-mixing because the Lindbladian L has a spectral gap, which immediately grants us efficient polynomial-time approximation algorithms for these quantities.

Lev: If we could actually implement this on hardware with reasonable connectivity—and I’m thinking about the required coherence times—that would be a massive validation of the theoretical framework for error correction simulations.

Kai: And that's what we need to keep pushing toward: showing that this theoretical speed translates into something achievable in terms of actual quantum computation resources.

The paper's improvements: Kai: Speaking of practical implications, let’s discuss the improvements the authors suggest or build upon within this work.

Mira: One key improvement is that they introduce "quasi-locality parameters" derived from the complex-time evolution, specifically convβ and corrβ(i, j), which measure how much a single site's operator deviates or how much pairs of operators correlate under this evolution.

Lev: Those two constants sound like the necessary ingredients to bridge the gap between the original Lindbladian L and a "pseudo-Lindbladian" generator K that actually has sharper quasi-locality properties. That’s where they build their argument for a spectral gap in Lemma one point one three.

Kai: So, by controlling these quasi-locality parameters—convβ and corrβ(i, j)—they can satisfy the two conditions required for the pseudo-Lindbladian generator K to have a spectral gap, even when dealing with all-to-all interactions.

Mira: Exactly; they show that if convβ/four plus corrβ/four is less than one, then this pseudo-Lindbladian K is gapped, which then implies the original Lindbladian L must also be gapped at sufficiently high temperatures, completing the proof of Theorem one point three.

Lev: That dependence on those parameters gives us a concrete way to analyze how system properties like interaction strength J and degree d affect whether we can achieve this fast mixing. It’s a useful diagnostic for hardware design considerations.

Kai: It’s about providing explicit criteria—like convβ/four + corrβ/four < one—that tell us when these complex systems are fast-mixing, which is much more helpful than just saying "if it works."

Conclusion: Kai: We’ve covered a lot of ground on the paper "Fast mixing of all-to-all quantum systems at high temperatures," from the initial concepts to the final bounds. So, let's wrap up by summarizing what this means for us moving forward.

Mira: The main conclusion is that arbitrary all-to-all quantum k-local Hamiltonians with bounded strength interactions admit a quantum Gibbs sampler with a system-size independent spectral gap when the temperature is sufficiently high. This means these systems are fast-mixing, which in turn guarantees the existence of fully polynomial time quantum approximation algorithms for partition functions and global expectation values.

Lev: For error correction researchers, this result suggests that we have a new theoretical foundation supporting the idea that these complex systems can be efficiently thermalized, which is a major step toward developing robust simulation techniques for non-local physics.

Kai: It really suggests that we can use this theory to build quantum simulators capable of handling these highly connected Hamiltonians in a way that scales polynomially with system size.

Mira: And the limitation they flag is that the method relies on the convergence of complex-time evolution and specific bounds derived from cluster expansions, so it stops working where those underlying assumptions about quasi-locality might break down.

Lev: That’s fair; we need to be careful that when we move this to hardware, we have a clear understanding of those limits imposed by the cluster expansion analysis.

Kai: Fantastic discussion on the paper "Fast mixing of all-to-all quantum systems at high temperatures." We're going to take these findings and start thinking about how this could influence our next round of experiments.

Department of EECS, UC Berkeley

quant-ph

Submitted: 2026-06-24

Updated: 2026-09-30

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 80/100

The gist: This paper establishes that arbitrary all-to-all quantum k-local Hamiltonians with bounded strength interactions admit a quantum Gibbs sampler [CKG23] possessing a system-size independent spectral

Key concepts

Spectral Gap
A spectral gap in this context means the rate at which the system's evolution converges to its equilibrium state is independent of the system size (n). A large gap ensures rapid mixing, meaning the quantum simulation reaches a stable state quickly.
Fast Mixing
Fast mixing describes how quickly a quantum system evolves toward its thermal equilibrium. Proving fast mixing means the time required for this convergence is polynomial in the system size, which is crucial for efficient simulations.
Cluster Expansion
This technique breaks down complex time evolution into simpler, localized pieces (clusters) based on the interaction graph. By analyzing how these clusters decay with interaction strength, researchers can derive useful quasi-locality parameters needed for the proof.

Terminology

Summary

This paper establishes that arbitrary all-to-all quantum k-local Hamiltonians with bounded strength interactions admit a quantum Gibbs sampler [CKG23] possessing a system-size independent spectral gap at sufficiently high temperatures. This result is significant because it proves that such systems are fast-mixing, which in turn implies the existence of fully-polynomial time quantum approximation algorithms for partition functions and global expectation values, generalizing existing fast-mixing results beyond the geometrically local setting where Lieb-Robinson bounds are typically required.

Fast Mixing and Fast Algorithms

The core contribution is Theorem 1.3, which proves that for an all-to-all Hamiltonian as defined in Definition 1.1, there exists a threshold temperature inverse beta, denoted as βc:= [c / (q·k·dJ)]−1, such that the Lindbladian L admits a system-size independent spectral gap. This spectral gap directly translates to rapid mixing; specifically, the evolution converges to within ε trace distance of the Gibbs state in time tmix(ε) ≤ O(n + log 1/ε), allowing for efficient state-preparation by simulation of the dynamics [CKG23, Theorem I.2]. Consequently, Corollary 1.4 follows: for any ε > 0, there is a poly(n, 1/ε) time quantum algorithm to estimate the partition function Zβ and global expectation values up to relative error 1 ± ε.

Cluster Expansions and Locality

The proof relies on revisiting noncommutative versions of Dobrushin conditions under the quantum cluster expansion of Netoˇcn´y and Redig [NR04]. The complex-time evolution, A(z) = e izHAe-izH, is decomposed into a series of connected clusters F in the interaction graph GI, as shown in Lemma 3.4: A(z) = A + Σ F AF(z), where the norm of each cluster contribution decays with the product of interaction strengths within F. This expansion yields two crucial quasi-locality parameters:

  1. The complex-time evolution constant convβ, defined as the maximum deviation of any single-site Pauli operator from itself under complex-time evolution (Definition 3.5).

  2. The pairwise correlations constant corrβ(i, j), which measures the maximum commutator norm of any pair of single-site Pauli operators under complex-time evolution (Definition 3.7).

Dobrushin Conditions for the Pseudo-Lindbladian K

The analysis utilizes a pseudo-Lindbladian generator K [BC26], which possesses sharper quasi-locality properties than the original Lindbladian L [CKG23]. Sufficient conditions for K to admit a spectral gap are established in Lemma 1.13, which requires two assumptions:

  1. Locality in temperature: for each site j, the generator Kj can be approximated by that at infinite temperature (the depolarizing semigroup) such that∥K(β)j − K(0)j∥ρ ≤ ηj.

  2. Pairwise influence: for each pair of sites i and j, the generator Kj approximately commutes with the depolarizing semigroup on another site l, satisfying [K(0)l, K(β)j]ρ ≤ κjl.

Dirichlet Form Comparison and Final Gap Proof

The final step in proving Theorem 4.1 is relating the spectral gap of K to that of L via Dirichlet form comparison arguments (Section 5). The strategy involves sequentially peeling off the frequency and time-dependence from the integral expression for the Dirichlet form of L [Lemma 5.3] by relying on the convergence of complex-time evolution. This leads to Lemma 5.4, which bounds the g(t) Dirichlet form by that of K:

X a∈S1[n] Z∞−∞ g(t) · [Aa(t), O]∥2ρdt ≥ γ · X a∈S1[n] [A, O]2ρ.

By combining the bounds derived from the cluster expansion (Lemma 3.10) and the resulting contraction in the KMS oscillator norm (Lemma 4.3), it is shown that when convβ/4 + corrβ/4 < 1, K is gapped (Corollary 4.13). This implies L is also gapped at sufficiently high temperatures, completing the proof of Theorem 1.3.

Conclusion and Outlook

The paper concludes by showing that the Dirichlet form comparison yields a final bound where E[O] ≥ 1/4 C[O] (Lemma 5.2), which simplifies to E[O] ≥ 1/2 - βJd · (c q / cβ)k · C[O]. By choosing βJd sufficiently small such that Lemma 3.

Improvements for AI systems

As a fastidious and diligent researcher, I have analyzed the provided paper, Fast mixing of all-to-all quantum systems at high temperatures, and identified several transformative capabilities for AI systems derived from its core findings.

The central breakthrough is proving that arbitrary all-to-all quantum k-local Hamiltonians admit a spectral gap in their thermal Lindbladian at sufficiently high temperatures, which implies they are fast-mixing. This leads directly to fully polynomial time quantum approximation algorithms for partition functions and global expectation values.

Here are the specific improvements and capabilities for AI systems:

)

)

)

  1. A new class of quantum simulation algorithms capable of efficiently simulating complex, non-geometric quantum systems that defy standard lattice approximations (e.g., electronic structure Hamiltonians, strongly correlated models).

  2. The ability to perform exact or high-precision thermal state preparation for arbitrary k-local interacting systems in polynomial time.

)

)

)

Specific improvements and capabilities:

  1. An efficient quantum Gibbs Sampler (CKG23) for all-to-all systems: The system can prepare the thermal Gibbs state, defined by any arbitrary k-local Hamiltonian with bounded strength interactions, in time complexity polynomial in the system size, specifically as fast as a poly(n).

  2. Fully Polynomial Time Quantum Approximation Algorithms: Because these systems are fast-mixing at high temperatures, an AI system can estimate partition functions and global expectation values (like free energy) to a desired precision (e.g., relative error 1 ± ε) in time complexity of polynomial in the system size and the required precision (poly(n, 1/ε)).

  3. Simulation of Non-Geometric Quantum Systems: The AI can now tackle Hamiltonians where interactions are not restricted to nearest neighbors or geometric lattices (e.g., systems arising from molecular simulations or complex condensed matter physics), overcoming the limitations imposed by Lieb-Robinson bounds in non-geometric settings.

  4. Efficient Thermalization Analysis: The system can analyze how quantum information scrambles under the evolution of these all-to-all Hamiltonians, providing a theoretical foundation for understanding thermalization dynamics in systems lacking traditional locality constraints.

Abstract

It is shown that arbitrary quantum k-local Hamiltonians with bounded strength interactions admit a quantum Gibbs sampler [CKG23] with a system-size independent spectral gap, at sufficiently high temperatures. As a consequence, such systems admit fully-polynomial time quantum approximation algorithms for partition functions and global expectation values.

Sources

Related papers