Timescale for macroscopic equilibration in isolated quantum systems: a rigorous derivation for free fermions

The paper rigorously proves that translation-invariant free-fermion systems on a $d$-dimensional hypercubic lattice equilibrate with respect to coarse-grained density within an optimal timescale of order $L$, regardless of the initial pure state.

The Gist

The Big Picture: Why Does a Hot Cup of Coffee Cool Down?

Imagine you have a cup of hot coffee in a perfectly sealed, magical box. No heat can escape, and no air can get in. In the real world, we know the coffee eventually cools down to room temperature and stays there. This process is called equilibration.

But here is the mystery: In the quantum world, the rules are different. Particles (like the atoms in your coffee) don't just stop moving; they bounce around forever in a complex dance. According to the laws of quantum mechanics, if you knew the exact position and speed of every single particle at the start, you could theoretically predict exactly where they would be a million years from now. They never truly "forget" their starting point.

So, how does a chaotic, never-ending quantum dance turn into a calm, steady state (equilibrium)? And how long does it take?

For nearly a century, physicists have known that quantum systems eventually look like they are in equilibrium, but they didn't have a solid answer for how long it takes. This paper finally solves that puzzle for a specific type of system.

The Experiment: A Grid of Quantum Runners

The authors (Takashi Hara and Tatsuhiko Koike) studied a simplified model to get a clear answer.

• The Setup: Imagine a giant, multi-dimensional grid (like a 3D chessboard, but much bigger). On this grid, there are "free fermions." Think of these as ghostly runners that can hop from one square to the next.
• The Rules: These runners are "free," meaning they don't bump into each other or push each other away. They just hop randomly.
• The Goal: They wanted to see how long it takes for the runners to spread out evenly across the entire grid. If you start with all the runners crowded in one corner, how long until the crowd looks the same everywhere?

The Discovery: The "Speed Limit" of Spreading

The paper proves a very specific and surprising result: The time it takes for the system to settle down is directly proportional to the size of the grid.

If your grid is 100 steps wide, it takes about 100 "time units" to settle.
If your grid is 1,000 steps wide, it takes about 1,000 "time units."

The Analogy: The Fire Alarm
Imagine a fire alarm in a massive office building.

• If the building is small (a few rooms), the sound reaches everyone almost instantly.
• If the building is a skyscraper, the sound takes time to travel from the top floor to the bottom.
• The authors proved that in this quantum world, the "information" about where the particles are spreads at a fixed speed (like sound). Therefore, the time it takes for the whole system to "know" it's time to settle down is simply the time it takes for that information to cross the building.

They call this an $O(L)$ timescale, where $L$ is the size of the system. This is considered the "optimal" or fastest possible time. You can't settle down faster than the time it takes for a signal to cross the room.

Why Is This a Big Deal?

1. It's Rigorous: Before this, many scientists guessed that systems equilibrate quickly, but they couldn't prove it mathematically for "realistic" systems (ones that aren't just made up for the sake of the math). This paper provides a rock-solid mathematical proof.
2. It's "Macroscopic": They didn't just look at two particles. They looked at a "macroscopic" view—like looking at the density of the crowd rather than tracking every single person. This is how we see the real world (we see the temperature, not the speed of every air molecule).
3. It Solves a Century-Old Question: It connects the chaotic, reversible world of quantum mechanics with the calm, irreversible world of thermodynamics (heat and cooling) in a way that explains the timing.

The "Magic" of the Proof

How did they prove this?

They used a clever trick involving time averaging. Instead of asking "Is the system settled right now?", they asked, "If we watch the system for a long time, how much of that time is it not settled?"

They found that for any starting position of the runners, the amount of time the system spends in a "messy, unbalanced" state becomes vanishingly small as the system gets bigger, provided you wait long enough (specifically, longer than the size of the grid).

The "Blindfolded" Analogy:
Imagine you are blindfolded in a dark room with a million people running around chaotically.

• At the start: You might hear a huge noise in one corner (the runners are crowded).
• After a while: The noise spreads out.
• The Proof: The authors show that if you listen for a time equal to the time it takes to walk across the room, the "loud spots" will almost disappear. The noise will be so evenly distributed that, for all practical purposes, the room is quiet and calm.

The Catch (and the Limit)

The paper notes one important limitation: This works for density (where the particles are), but not for momentum (how fast they are moving).

• Density: If you start with a crowd in the corner, they will spread out and fill the room.
• Momentum: If you start with everyone running in the exact same direction, they will keep running in that direction forever. They never "forget" their speed.

This is a crucial distinction. In our real world, things usually equilibrate in both position and speed because particles bump into each other (interact). In this specific "free" model, they don't bump, so speed never equilibrates. However, for the "position" part—which is what we usually mean when we say a gas has spread out—the math holds up perfectly.

The Bottom Line

This paper is a milestone because it finally puts a stopwatch on quantum thermalization. It tells us that for a system of particles that don't interact, the time it takes to reach a steady state is simply the time it takes for a signal to cross the system.

It's a beautiful confirmation that even in the weird, fuzzy world of quantum mechanics, there are strict, predictable limits to how fast things can change, and those limits are tied directly to the size of the universe you are looking at.

Technical Summary

1. Problem Statement

The paper addresses a fundamental gap in the foundations of quantum statistical mechanics: while it is well-established that isolated quantum systems evolve toward equilibrium (thermalization) for macroscopic observables, the timescale required for this equilibration has rarely been rigorously derived for physically realistic, short-ranged systems.

• Context: Previous works (e.g., von Neumann, Tasaki, Goldstein et al.) proved that systems equilibrate for "sufficiently long" times but did not specify the order of magnitude of this time in terms of system size $L$.
• Specific Question: For a system of free fermions on a $d$-dimensional lattice of linear size $L$, what is the precise timescale $\tau$ required for the system to equilibrate with respect to macroscopic density observables, starting from an arbitrary pure initial state?
• Goal: To establish a rigorous, optimal upper bound on the equilibration timescale and prove that the fraction of time the system spends in a non-equilibrium state vanishes in the thermodynamic limit ($L \to \infty$).

2. Methodology and Setup

Physical Model

• System: $N$ spinless free fermions on a $d$-dimensional hypercubic lattice $\Lambda$ of size $L^d$ with periodic boundary conditions.
• Hamiltonian: Uniform nearest-neighbor hopping:
  $$\hat{H} = \sum_{\|x-y\|=1} \hat{c}^\dagger_x \hat{c}_y$$
• Observables: Macroscopic number density operators $\hat{\rho}_c$ defined over boxes $B(c)$ of volume $l^d$ (where $l$ is macroscopic but $n = L/l$ is $O(1)$).
  $$\hat{\rho}_c = \frac{1}{l^d} \sum_{x \in B(c)} \hat{c}^\dagger_x \hat{c}_x$$
• Equilibrium Definition: The system is in equilibrium if the expectation value of $\hat{\rho}_c$ is close to the average density $\bar{\rho} = N/L^d$ for all boxes. The authors define a "non-equilibrium subspace" $\mathcal{H}_{neq}$ where deviations exceed a small threshold $\epsilon \bar{\rho}$.

Mathematical Framework

The authors analyze the time evolution $|\Psi(t)\rangle = e^{-i\hat{H}t}|\Psi(0)\rangle$ and focus on the fraction of time the system spends in the non-equilibrium subspace.

1. Time Averaging: Instead of analyzing pointwise time evolution, they utilize a specific time-average function $f_\tau(t) = \frac{1}{\pi\tau}(\frac{\sin(t/\tau)}{t/\tau})^2$. This function is concentrated around $[-\tau, \tau]$ and allows for the use of Fourier transform properties to decouple energy differences.
2. Reduction to Single-Particle Physics: By exploiting the free-fermion nature and translation symmetry, the problem of the $N$-particle density fluctuation is reduced to analyzing the time evolution of a single-particle operator in momentum space. The density operator is expressed as a sum over momentum modes involving energy differences $\tilde{E}_{\beta, m} = E_\beta - E_{\beta+m}$.
3. Key Identity: The authors employ a specific identity for the time average of oscillating terms $e^{-it(E' - E'')}$:
   $$[e^{-it(E' - E'')}]_\tau = \frac{\tau}{2} \int dE \, \mathbb{I}[|E' - E| \leq 1/\tau] \, \mathbb{I}[|E - E''| \leq 1/\tau]$$
   This converts the time average into an integral over energy, effectively counting the number of states (density of states) within an energy window of width $1/\tau$.
4. Bounding Operator Norms: The proof relies on bounding the operator norm of the squared deviation $(\Delta \rho)^2$. This reduces to estimating the "number of states" $|\Omega_m(E)|$ (the count of momentum indices $\beta$ such that the energy difference $\tilde{E}_{\beta, m}$ falls within a narrow window near $E$).

3. Key Contributions and Technical Results

The Main Theorem (Theorem 1)

The authors prove that for any initial state $|\Psi(0)\rangle$, the fraction of time the system spends in a non-equilibrium state during the interval $[0, \tau]$ vanishes in the thermodynamic limit ($L \to \infty$) provided:
$$\tau > L \times g(L)$$
where $g(L)$ is any function diverging as $L \to \infty$.

• Result: The equilibration timescale is $O(L)$.
• Optimality: This scaling is optimal because the Lieb-Robinson bound implies that information (and thus equilibration) cannot propagate faster than a finite velocity, requiring at least time $O(L)$ to traverse the system.

Technical Theorem (Theorem 1')

A more precise bound is established for the fraction of non-equilibrium time:
$$\frac{1}{2\tau} |T_{\delta, |\Psi(0)\rangle}^{[-\tau, \tau]}| \leq K_d \frac{n^{3d}}{(\epsilon \bar{\rho})^2} \delta^{1/2}(\tau, L)$$
where $\delta(\tau, L)$ is a small function decaying as $\tau/L$ and $L$ increase.

Key Lemmas and Estimates

• Lemma 5 (1D Density of States): A rigorous upper bound on the number of states $|\omega_m(x, \delta)|$ for 1D free fermions with energy differences near a specific value. The bound scales as $L\delta / \sqrt{1-x^2}$.
• Lemma 6 & 7 (Integral Bounds): Bounds on the integral $I_m = \int |\Omega_m(E)|^2 dE$, showing that for large $\tau$, these integrals scale favorably (involving logarithmic factors divided by $\tau$).
• Lemma 8 (Summation Bound): Combines the weight factors $\tilde{w}(m)$ (from the spatial averaging of the box) with the state-count bounds to show the total operator norm decays as $O(\sqrt{\frac{\log(\tau/L)}{\tau/L}})$.

4. Significance and Implications

1. First Rigorous Timescale for Realistic Systems: This is the first work to rigorously establish a realistic equilibration timescale ($O(L)$) for an isolated, physically natural quantum system (free fermions with short-range hopping) without relying on random matrix theory or abstract assumptions.
2. Macroscopic vs. Microscopic: The proof highlights that while microscopic observables (like momentum occupation numbers) may never equilibrate (e.g., if starting in a momentum eigenstate), macroscopic observables (spatial density) do equilibrate rapidly. This distinction is crucial for the validity of statistical mechanics.
3. No Randomness Required: The derivation does not assume any randomness in the Hamiltonian or initial state. It holds for deterministic, translation-invariant systems, even with degenerate energy spectra.
4. Optimality: By matching the lower bound implied by the Lieb-Robinson velocity, the result confirms that $O(L)$ is the fundamental physical limit for macroscopic equilibration in such systems.
5. Methodological Advance: The technique of using a specific time-average kernel to convert time evolution into an energy-window counting problem (density of states) provides a powerful tool for analyzing equilibration in other non-interacting or weakly interacting systems.

Conclusion

Hara and Koike provide a definitive answer to the timescale of macroscopic equilibration for free fermions. They demonstrate that starting from any pure state, the system relaxes to a state where macroscopic density fluctuations are negligible within a time proportional to the system size $L$. This rigorously validates the expectation that normal macroscopic systems with conserved quantities (like particle number) thermalize on the timescale of the system's light-crossing time (or sound-crossing time), bridging the gap between abstract quantum ergodicity and concrete physical reality.