Optimal Dynamic Cooling of Multiple Qubits

Technical Summary: Optimal Dynamic Cooling of Multiple Qubits

Problem Formulation
This work addresses the closed-system dynamic cooling problem for a register of $N$ identical thermal qubits, where a subset of $M$ qubits ($M < N$) are designated as targets. The objective is to apply a global unitary transformation $U$ to redistribute the fixed spectrum of the initial thermal state such that the $M$ target qubits reach the lowest possible common local temperature. This problem is distinct from single-target cooling because it imposes two simultaneous constraints:

1. Global Optimality: The total energy of the $M$ targets must be minimized among all possible global unitaries.
2. Uniformity: All $M$ target qubits must possess identical diagonal marginals (and thus the same local temperature).

The system is isolated during the protocol; resources are limited to the initial thermal bias, the finite number of qubits, and coherent control. No external heat bath is used to reset degrees of freedom.

Methodology
The authors decompose the solution into two conceptually distinct steps:

1. Passive Rearrangement (Spectral Optimization):
   The first step utilizes the rearrangement inequality to minimize the total target energy. The eigenvalues of the initial thermal state, ordered by magnitude, are assigned to the target excitation sectors (defined by Hamming weight) in increasing order of weight. This "passive" permutation assigns the largest eigenvalues to the lowest energy sectors of the target register. This step determines the absolute lower bound for the mean target excitation, denoted as $p^*(P_1)$, regardless of whether the targets share a common temperature.

2. Symmetrization (Equal-Temperature Enforcement):
   The passive rearrangement alone does not guarantee that individual target qubits have equal populations; it only minimizes their sum. To enforce a common temperature without altering the total energy, the authors introduce a target-only complex-Hadamard transformation. Within each fixed-Hamming-weight subspace, a unitary (specifically a discrete Fourier matrix) is applied to the target strings. This operation coherently averages the ancillary blocks associated with target strings of the same weight.

   • Crucially, this transformation commutes with the total Hamiltonian and preserves the probability of each target sector.
   • It equalizes the one-qubit marginals while maintaining the passive energy minimum.

Key Contributions and Results

• Universal Optimal Protocol: The paper proves that for any $N > M$ and any initial temperature, the constrained optimum (common temperature) coincides exactly with the unconstrained passive minimum. Imposing the requirement of equal temperatures costs neither cooling depth nor additional work. The protocol is defined as $U^* = U_{sym} U_e$, where $U_e$ is the optimal passive permutation and $U_{sym}$ is the symmetrization unitary.
• Exact Cooling Curve: An exact analytical expression for the optimal target population $p^*(P_1)$ is derived based on the overlap of thermal eigenvalue shells and target sector capacities.
• Low-Temperature Limit: In the low-temperature regime, the final temperature scales as $T' \sim T / \xi^*_M$, where $\xi^*_M$ is an integer determined by the cumulative binomial coefficients of the system size. This recovers known single-target results for $M=1$ and establishes a linear improvement factor for multiple targets.
• Ancillary Resource Threshold: The authors prove that nontrivial cooling ($T' < T$) is possible if and only if $N - M \ge 2$. A single ancillary qubit is insufficient to cool the target register, even without the equal-temperature constraint.
• Implementation Criteria (Permutation vs. Coherent Control):
  • The universal protocol requires complex-Hadamard corrections, which may be circuit-expensive.
  • The paper derives an exact arithmetic criterion for when the same optimum can be achieved using a temperature-independent computational-basis permutation alone. This requires "shell-wise one-body balance," a condition weaker than full Hamming symmetry.
  • A complete classification of "islands of feasibility" (pairs of $N, M$ where permutation-only cooling is possible) is provided for $M+2 \le N \le 128$.
  • The authors identify a strict separation between Hamming-symmetric permutations and general locally balanced permutations, noting that 58 pairs in the studied range admit the latter but not the former.
  • An "Odd-$M$ Conjecture" is proposed: for odd $M$, no temperature-independent locally balanced permutation exists if $N < M^2$.
• Global vs. Parallel Strategies: The paper demonstrates that joint many-target cooling strictly outperforms parallel single-target strategies (dividing the register into disjoint machines). In the regime $M+2 \le N \le 3M-1$, the global protocol achieves cooling while parallel strategies cannot cool at all. Asymptotically, the global protocol achieves a temperature lower by a factor of $M$ compared to parallel strategies.

Significance and Claims
The paper claims to solve the fundamental thermodynamic limit of cooling multiple qubits in a closed system. Its primary significance lies in establishing that equal-temperature cooling is not a compromise; it is an attainable representative of the unconstrained optimum. The work separates thermodynamic performance from implementation complexity, showing that while the universal optimum is always achievable via coherent control, specific system sizes allow for simpler, permutation-only implementations.

The authors emphasize that the discrete structure of the spectrum dictates the implementation (requiring coherent symmetrization in many cases) but does not obstruct the optimum itself. The results provide a rigorous benchmark for quantum refrigeration protocols, highlighting the cooperative advantage of global unitaries over independent local operations and defining the precise resource thresholds (ancillary qubits) required for nontrivial cooling.