🔬 condensed matter

Thermalization in a closed quantum system from randomized dynamics

This paper proposes a distinct mechanism for thermalization in closed quantum systems under strong random perturbations, where the chaotic nature of eigenvectors and total energy constraints directly generate a canonical ensemble for both local and global observables without relying on the Eigenstate Thermalization Hypothesis, a microcanonical ensemble, or subsystem-bath partitions.

Original authors: Nikolay V. Gnezdilov, Andrei I. Pavlov

Published 2026-08-12

📖 7 min read🧠 Deep dive

CC BY 4.0

Original authors: Nikolay V. Gnezdilov, Andrei I. Pavlov

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the universe as a giant, incredibly complex machine made of tiny, dancing particles. For over a century, scientists have had a secret rulebook for predicting how these particles behave when they get hot or cold: statistical mechanics. This rulebook says that if you wait long enough, a system will settle into a comfortable "thermal equilibrium," like a cup of coffee cooling down to room temperature. But here's the puzzle: in the quantum world, where particles act like waves and follow strict, unbreakable laws of motion, there is no "room temperature" to cool down to. The system is closed; it's a self-contained universe. So, how does a perfectly isolated quantum system ever learn to act like it's hot or cold?

Traditionally, physicists have solved this by pretending the system is part of a bigger party. They imagine a small group of particles (the system) hanging out with a massive crowd (the "bath"). The crowd acts as a thermal bath, soaking up or giving away energy until the small group settles down. This idea, known as the Eigenstate Thermalization Hypothesis (ETH), works great for local observations—like checking the temperature of just one corner of a room. But it hits a wall when you try to look at the entire room as a single unit. If the whole room is the system, who is the bath? And what if you can't split the system into a "part" and a "rest"? This is the mystery that a new study from Dartmouth College and the Karlsruhe Institute of Technology sets out to solve, offering a fresh, chaotic twist on how heat emerges from the quantum cold.

The Chaotic Shuffle: A New Way to Heat Up

Usually, to get a quantum system to act "thermal," scientists rely on a weak nudge. They gently shake the system, hoping it settles into a state where only a tiny slice of its energy levels matters. Think of it like tapping a bell lightly; it rings in a specific, predictable way. But in this new paper, the authors, Nikolay Gnezdilov and Andrei Pavlov, decide to stop tapping gently and start shaking the table violently.

They propose a method where they take a quantum system—specifically a chain of 11 tiny magnets (spins) arranged in a line, known as the transverse-field Ising model—and hit it with a massive, random storm of interactions. Imagine you have a deck of cards representing the energy states of your system. In the old way, you'd shuffle the deck just a little, keeping the cards mostly in order. In this new approach, they throw the deck into a tornado. The "random interaction" is so strong that it scrambles every single energy level, mixing the entire spectrum of possibilities together.

Here is the magic trick: Even though the system is being thrown into chaos, the authors found that if you look at the average result of many different random storms, a beautiful pattern emerges. It's like if you threw a handful of sand into the air a thousand times; each grain flies in a different, chaotic direction, but if you average where they all land, they form a perfect, smooth hill.

In this study, the "sand grains" are the probabilities of the system being in different energy states. When the system is hit with these strong random perturbations, the way it occupies these energy states follows a famous statistical pattern called the Porter-Thomas distribution. This is a mathematical description of how chaotic systems behave. The authors discovered that the average of this chaotic distribution is exactly the Gibbs distribution—the same formula that describes how heat is shared in a cup of coffee or a star in the sky.

The Temperature of the Initial State

So, where does the temperature come from? In the old "bath" method, the temperature is set by the environment. In this new, closed-system method, the temperature is set by the initial state of the system.

The authors set up their experiment with a specific starting point: a state where the spins are ordered in a particular way. They then imposed a strict rule: the total energy of the system must stay the same as it was at the start. Because the random storm mixes everything up, the system doesn't just stay in one energy level; it spreads out. But because the total energy is locked in, the system has no choice but to settle into a specific "thermal" distribution that matches that energy.

It's like a game of musical chairs where the music is a chaotic storm. If you start with a specific number of people (energy), and you force them to dance randomly, they will eventually spread out across the room in a very specific pattern that depends entirely on how many people you started with. The authors showed that this pattern is the canonical ensemble—the gold standard of thermal physics—without needing to split the system into a "part" and a "bath."

Proving It Works: From Global to Local

To prove this wasn't just a mathematical fluke, the team ran simulations on their 11-spin chain. They looked at two types of things:

Global Observables: They checked the "occupation" of every single energy state in the system. In traditional physics, the idea that the whole system follows a thermal distribution is often considered impossible or requires a bath. But here, the global projectors (which tell you how much of the system is in a specific state) lined up perfectly with the thermal prediction.
Local Observables: They also looked at how the spins talked to each other over distance (spin-spin correlations). In a thermal system, these connections should fade away exponentially as you move further apart, like a whisper dying out in a noisy room. The authors found that the distance over which the spins could "hear" each other (the correlation length) shrank exactly as the temperature rose.

The results were striking. When they increased the initial energy of the system, the "temperature" went up, and the correlation length got shorter, just as standard thermodynamics predicts. They even showed that this works whether you calculate the result by looking at the final energy states (exact diagonalization) or by watching the system evolve in real-time (real-time propagation). In both cases, the system relaxed to a thermal state in a finite amount of time, roughly proportional to $1/J$ (where $J$ is the strength of the random interaction).

Why This Matters

This paper doesn't just solve a theoretical puzzle; it opens a door for quantum computers. Currently, simulating thermal states on a quantum computer is hard because you usually have to simulate a huge "bath" along with your system, which eats up all your precious qubits. This new method suggests you can generate a thermal state for the entire system just by letting it evolve under strong, random dynamics for a short time. You don't need to build a fake environment; you just need to shake the system hard enough and let the chaos do the work.

The authors are careful to note that this is a simulation-based discovery. They demonstrated it for a specific 1D chain of 11 spins. However, because the mechanism relies on the universal properties of chaotic systems (the Porter-Thomas distribution) rather than the specific details of the magnets, they believe this approach could work for much larger and more complex systems, provided they are chaotic and have a fixed total energy.

In short, Gnezdilov and Pavlov have shown that you don't need a thermal bath to make a quantum system hot. You just need to introduce enough chaos, lock the energy, and let the system average itself out. The heat isn't coming from outside; it's emerging from the beautiful, statistical order hidden inside the quantum noise.

Technical Summary: Thermalization in a Closed Quantum System from Randomized Dynamics

Problem Statement
The emergence of statistical mechanics from quantum dynamics in closed systems is conventionally explained through the Eigenstate Thermalization Hypothesis (ETH), quantum chaos, and canonical typicality. In standard frameworks, thermal behavior for local observables arises when a subsystem is treated as a bath for the rest of the system, effectively tracing out degrees of freedom in the thermodynamic limit. However, for finite-size systems—particularly those relevant to current and near-term quantum processors—this subsystem-bath partition is often artificial or inapplicable. Furthermore, conventional thermalization typically relies on weak perturbations that restrict contributing states to a narrow microcanonical window, limiting the scope to local observables. There is a need for a mechanism that generates canonical ensemble behavior for both local and global observables in finite, closed quantum systems without requiring a subsystem-bath partition or invoking ETH.

Methodology
The authors propose a distinct thermalization mechanism based on strong random perturbations within a closed system. The approach involves:

System Definition: A target Hamiltonian $H_0$ (specifically the 1D transverse-field Ising model with open boundary conditions) is perturbed by an auxiliary random interaction $V^{(j)}$ .
Random Interaction: The interaction $V^{(j)}$ is drawn from a Gaussian Unitary Ensemble (GUE) with zero mean and large variance. The zero mean imposes a global constraint on the total energy, conserving the average energy of the initial state. The large variance places the system in a strong-perturbation regime, ensuring the total Hamiltonian $H = H_0 + V^{(j)}$ belongs to the Random Matrix Theory (RMT) universality class and mixes degrees of freedom across the entire many-body spectrum.
Dynamics and Averaging: The system evolves unitarily under $H$ from a specific initial product state $|\psi_{in}\rangle$ . The authors perform an ensemble average over $M$ realizations of the random interaction $V^{(j)}$ .
Statistical Analysis: They analyze the long-time average of observable expectation values. By decomposing the initial state into the eigenbasis of the perturbed Hamiltonian, they examine the statistics of the overlap coefficients $|c^{(j)}_\alpha|^2$ and the eigenstate expectation values $O^{(j)}_{\alpha\alpha}$ .

Key Contributions and Results
The paper demonstrates that averaging over randomized dynamics yields a canonical ensemble for the entire finite system, independent of subsystem partitioning.

Emergence of the Gibbs Distribution: The authors show that the average occupation probability of the eigenstates, $w(E_\alpha) = \langle |c^{(j)}_\alpha|^2 \rangle_J$ , follows the Gibbs distribution $e^{-\beta E_\alpha}/Z$ . This distribution arises as the first moment of the Porter-Thomas (PT) distribution, which characterizes the chaotic overlaps of eigenstates with the initial state in RMT systems.
Temperature Determination: The inverse temperature $\beta$ is uniquely determined by the global constraint on the total energy (the initial state energy $E_{in}$ ), rather than by a subsystem-bath energy exchange.
Global and Local Observables: Unlike ETH, which typically applies to local observables, this mechanism generates canonically distributed expectation values for both:
- Global Observables: Specifically, projectors onto the eigenstates of the unperturbed Hamiltonian ( $O = |\phi_\nu\rangle\langle\phi_\nu|$ ). The occupation probabilities of these global states match canonical predictions.
- Local Observables: Spatially resolved spin-spin correlation functions ( $\langle \sigma^y_i \sigma^y_{i'} \rangle$ ). The correlation length $\xi$ exhibits a temperature-dependent decay ( $\xi \propto \beta$ ) consistent with statistical mechanics predictions for the thermodynamic limit, even in the finite-size system studied ( $N=11$ ).
Numerical Verification: Using exact diagonalization and real-time propagation for the 1D TFIM, the authors confirm that the correlation length scales linearly with the inverse temperature, matching the theoretical prediction $\xi = 4c\beta/\pi$ within error bars, despite finite-size offsets.

Significance and Claims
The paper claims to provide a direct derivation of the canonical ensemble from the universal properties of chaotic eigenstates in the presence of a global energy constraint, without invoking the microcanonical ensemble or a subsystem-bath partition.

Distinct Mechanism: The approach differs fundamentally from conventional thermalization. Instead of tracing out a bath in a typical eigenstate, the canonical distribution emerges from the classical averaging of evolved observables over random interaction realizations.
Universality: The results rely on the system belonging to the RMT universality class and satisfying the global energy constraint. The authors argue that the mechanism should apply generally to any system satisfying these conditions, regardless of microscopic details.
Practical Utility: The authors suggest that this approach is particularly relevant for quantum simulation. Since the thermal state can be prepared in finite time ( $t \sim 1/J$ ) via randomized dynamics, it offers a pathway for thermal state preparation on quantum hardware for few-body systems where traditional bath engineering is difficult. The method allows for the generation of thermal observables without partitioning the system, making it suitable for simulating thermal behavior in programmable quantum processors operating far from the thermodynamic limit.

The Chaotic Shuffle: A New Way to Heat Up

The Temperature of the Initial State

Proving It Works: From Global to Local

Why This Matters

Technical Summary: Thermalization in a Closed Quantum System from Randomized Dynamics

More like this