🔬 condensed matter

Thermalization packets and optimal ice cubes

This paper introduces "thermalization packets" (or "ice cubes") as auxiliary systems prepared with specific microscopic configurations to accelerate a target system's relaxation to equilibrium by suppressing its slowest decay mode, revealing that counterintuitively, the optimal cooling packet is not necessarily the coldest and can exhibit Mpemba-like effects.

Original authors: Israel Klich, Marija Vucelja

Published 2026-08-27

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Israel Klich, Marija Vucelja

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

In the physical world, everything eventually settles down. A hot cup of coffee cools to room temperature, a spinning top slows to a halt, and a shaken bottle of soda eventually stops fizzing. This process of returning to a calm, balanced state is called relaxation, and it is governed by the laws of thermodynamics. Usually, if we want to speed up this process, we make the environment colder or more aggressive. We might plunge a hot object into ice water or blow cold air across it. The assumption has always been that the colder the helper, the faster the cooling. But this intuition, while useful for everyday tasks, misses a deeper layer of how systems actually behave. Scientists have long known that complex systems do not relax at a single, uniform speed. Instead, they have different "modes" of slowing down, some of which are incredibly sluggish. These slow modes act like heavy anchors, dragging the system down and preventing it from reaching equilibrium quickly, regardless of how cold the surroundings are.

A team of researchers at the University of Virginia and the Max Planck Institute for the Physics of Complex Systems has proposed a new way to think about this problem. They suggest that instead of just making a cooling agent colder, we should prepare it with a specific internal structure that cancels out those sluggish anchors. They call these specially prepared helpers "thermalization packets," or more casually, "ice cubes." The core idea is that by carefully tuning the initial state of this auxiliary system before it ever touches the target, we can eliminate the slowest part of the relaxation process entirely. This approach trades the effort of preparing the helper in advance for a significant reduction in the waiting time required for the target to settle. It turns out that the best "ice cube" is not necessarily the coldest one available; in fact, making it too cold can sometimes make the whole process slower.

The researchers demonstrated this concept using several different models, ranging from simple pairs of quantum bits to more complex chains of interacting spins. In their simplest example, involving two tiny quantum systems, they showed that if you prepare the helper system at just the right temperature, it can completely remove the slowest mode of relaxation. When this happens, the target system relaxes at a much faster rate, determined only by the quicker modes that remain. The team found that for a wide range of conditions, there is always a specific temperature for the helper that achieves this perfect cancellation. This temperature is usually different from the temperature of the surrounding environment and, crucially, different from the temperature of the target system itself.

What makes this finding surprising is the relationship between temperature and speed. Common sense dictates that a colder object should cool a warmer one faster. However, the researchers found that if you take the helper system and make it colder than this specific "perfect" temperature, the slow mode reappears, and the relaxation slows down again. This creates a counterintuitive situation where a moderately cold helper works better than an extremely cold one. This behavior is a form of the Mpemba effect, a phenomenon where a system starting further from equilibrium can reach equilibrium faster than one starting closer, but here it applies to the preparation of the cooling agent rather than the cooling agent itself. The researchers showed that this effect is not a fluke of their simple models; it appears in more complex systems of interacting spins as well, where they could calculate the exact point where the slow mode vanishes.

The study also looked at what happens when we need the system to reach equilibrium within a specific, short amount of time, rather than waiting for it to settle naturally. In these cases, the optimal preparation changes. If the deadline is very tight, the best strategy shifts toward using an even colder helper, one that might be colder than the "perfect" long-term solution. This is because a colder helper can force the system to overshoot the target and then correct itself, effectively canceling out the slow drift at that specific moment. The researchers calculated that for certain deadlines, the best helper would need to be at absolute zero, though in many practical cases, a finite, very cold temperature is sufficient.

Beyond the specific mechanics of cooling, the work suggests a broader principle for controlling how systems evolve. The researchers showed that the line of "perfect" preparations connects the ordinary state of equilibrium to a special point where the system behaves in a highly unusual way, known as a strong Mpemba point. This means that the same mathematical logic that allows us to design a perfect ice cube also explains why some systems can cool down anomalously fast under the right conditions. The findings are based on exact mathematical solutions and simulations of well-defined physical models, providing a clear proof that this spectral matching strategy works. While the paper focuses on theoretical models, the authors note that these ideas could be tested in real-world platforms, such as quantum computers or classical stochastic systems, where auxiliary states can be prepared independently and then coupled to a target.

Ultimately, this research reframes the problem of cooling. It moves the focus from simply extracting heat to engineering the initial conditions of the tools we use. By treating the cooling agent as a resource that can be spectrally tuned, rather than just a cold object, we gain a new lever to control the speed of relaxation. The "perfect ice cube" is not defined by how cold it is, but by how well its internal structure matches the specific dynamics of the system it is meant to cool. This insight opens the door to designing auxiliary systems that are optimized not for their temperature, but for their ability to silence the slowest, most stubborn parts of a system's journey toward balance.

Technical Summary: Thermalization Packets and Optimal Ice Cubes

Problem Statement
Conventional cooling and thermalization strategies typically focus on maximizing heat extraction or minimizing the temperature of an auxiliary system (coolant) to accelerate a target system's relaxation toward equilibrium. This paper challenges the assumption that the coldest possible auxiliary system is always optimal. Instead, the authors propose a distinct strategy: preparing an auxiliary system (a "thermalization packet" or colloquially, an "ice cube") in a specific non-equilibrium state before coupling it to the target system. The goal is not merely to absorb energy, but to spectrally match the packet's preparation to the target system to suppress the slowest dynamical modes of the combined system, thereby minimizing the time required to reach a prescribed thermal state.

Methodology
The authors model the combined system (target $S$ + packet $C$ ) coupled to a joint thermal bath at temperature $T_b$ . The dynamics are governed by a continuous-time generator $W$ acting on the state space, where the evolution follows $\partial_t |p(t)\rangle = W |p(t)\rangle$ .

Spectral Analysis: The relaxation is analyzed via the eigenvalues $\lambda_k$ and eigenvectors of $W$ . The asymptotic relaxation time is dominated by the slowest non-zero mode ( $\lambda_2$ ). The amplitude of this mode, $a_2$ , is determined by the projection of the initial coupled state $|p(0)\rangle$ onto the left eigenvector $\langle u_2|$ .
Optimization Criteria:
- Perfect Packet: A preparation where $a_2 = 0$ , completely eliminating the slowest relaxation mode.
- Optimal Packet: A preparation that minimizes $|a_2|$ within a class of allowed preparations (e.g., Boltzmann distributions at various temperatures).
Coupling Mechanism: The study assumes "bath-mediated kinetic coupling." The target and packet do not interact directly via energy terms but are coupled through the bath, allowing for joint transitions (e.g., exchange moves) while maintaining a factorized equilibrium state.
Models: The framework is demonstrated using three exactly solvable models:
1. Complete-graph Metropolis dynamics.
2. A minimal two-qubit model with exchange interactions.
3. A boundary-coupled interacting Ising-spin system (6 spins in the target, 4 in the packet).

Key Contributions and Results

Existence of Perfect Boltzmann Packets:
The authors prove that under generic conditions, perfect packets exist among thermal preparations (Boltzmann states) near equilibrium. Using the implicit function theorem, they show that for a target at temperature $T_s$ near the bath temperature $T_b$ , there exists a specific packet temperature $T_p^*$ such that the slow-mode amplitude vanishes ( $a_2=0$ ). This defines a "perfect-packet contour" in the $(T_s, T_p)$ plane.
The "Packet" Mpemba Effect:
A central finding is that the optimal packet is not necessarily the coldest possible one. The paper demonstrates that cooling a packet beyond the specific temperature that cancels the slow mode can actually restore a non-zero slow-mode amplitude, thereby slowing down relaxation. This counterintuitive behavior constitutes a "packet analog" of the Mpemba effect, where a state farther from equilibrium (in terms of temperature) relaxes faster than a state closer to it, due to spectral matching rather than thermal gradient magnitude.
Finite-Time vs. Asymptotic Optimization:
The paper distinguishes between asymptotic optimization (eliminating the slowest mode entirely) and finite-time optimization. For a specific readout time $t_f$ , the optimal packet may be colder than the asymptotically perfect packet. This is because a colder packet can induce a transient cancellation between the slow and fast decay modes at the specific deadline, even if it does not eliminate the slow mode permanently.
Coexistence with Strong Mpemba Points:
In modified two-qubit models, the authors show that the contour of perfect packets ( $a_2=0$ ) can intersect the line of equal temperatures ( $T_s = T_p$ ) at points other than the bath equilibrium. These intersections correspond to "strong Mpemba" or "strong inverse-Mpemba" points, where a system prepared at a uniform non-bath temperature has zero overlap with the slowest mode. This suggests that the optimal "ice cube" can, under specific conditions, be hotter than the bath.
Demonstration in Interacting Systems:
The concept is extended to an interacting Ising spin chain. Exact diagonalization of a 10-spin system (6 target + 4 packet) confirms that a packet prepared at a specific inverse temperature ( $\beta_p^* \approx 1.44$ ) eliminates the slowest mode, accelerating the asymptotic relaxation of the relative entropy by a factor of $\approx 1.8$ compared to a standard bath-temperature packet.

Significance and Claims
The paper claims to introduce a new paradigm for controlling nonequilibrium relaxation: trading preparation effort (spectral matching of an auxiliary system) for reduced waiting time.

Conceptual Shift: The work reframes cooling not just as heat extraction but as a spectral engineering problem. The "ice cube" is defined not by its temperature or heat capacity alone, but by its microscopic preparation relative to the target.
Generalizability: The authors assert that perfect packets are generically present near equilibrium and that the framework applies to both cooling and heating protocols.
Relation to Other Protocols: The paper positions thermalization packets as distinct from, yet complementary to, existing methods like heat-bath algorithmic cooling or reservoir engineering. While those methods often use auxiliaries as entropy sinks or engineered reservoirs, this approach treats the auxiliary as a prepared thermodynamic resource designed to cancel slow modes in the population sector.
Limitations and Scope: The authors note that the existence of perfect packets relies on the slow mode being non-degenerate and sensitive to the packet's coordinates. They treat the preparation time and energetic cost as "sunk costs," focusing the optimization strictly on the subsequent relaxation dynamics. They acknowledge that future work could incorporate preparation work and entropy production as explicit thermodynamic resources.

The paper concludes that thermalization packets provide a route to accelerating relaxation by engineering finite auxiliary resources that are spectrally matched to the coupled dynamics, rather than by modifying the bath or the target system itself.

Technical Summary: Thermalization Packets and Optimal Ice Cubes

More like this