Source-linked AI summary
The thermodynamic meaning of negative entropy
Lídia del Rio, Johan Aberg, Renato Renner, Oscar Dahlsten, Vlatko Vedral
TL;DR
The paper asks how Landauer erasure depends on an observer’s information when that information may be quantum-mechanical. It introduces an observer-conditioned erasure process and shows that work proportional to H(S|O) suffices, so negative conditional entropy can produce net work.
Problem
The paper addresses how Landauer erasure should be related to observer-dependent uncertainty when an observer holds quantum information about the system.
Method
The paper constructs an erasure process using an observer’s classical or quantum memory and relates its work requirement to conditional entropy.
Results
Work proportional to H(S|O) is sufficient for erasure, and negative conditional entropy corresponds to a negative work cost.
Takeaways & Limitations
Negative quantum conditional entropies have a direct thermodynamic interpretation: erasure can yield work and cool the environment.
Abstract
from arXiv · showhide
Landauer's erasure principle exposes an intrinsic relation between thermodynamics and information theory: the erasure of information stored in a system, S, requires an amount of work proportional to the entropy of that system. This entropy, H(S|O), depends on the information that a given observer, O, has about S, and the work necessary to erase a system may therefore vary for different observers. Here, we consider a general setting where the information held by the observer may be quantum-mechanical, and show that an amount of work proportional to H(S|O) is still sufficient to erase S. Since the entropy H(S|O) can now become negative, erasing a system can result in a net gain of work (and a corresponding cooling of the environment).
I. PRELIMINARIES
The paper situates Landauer erasure within the connection between thermodynamics and information theory, then asks whether negative quantum uncertainty can yield work during erasure.
- Landauer’s principle links irreversible information erasure to work dissipated as heat.
- The required work decreases as an observer’s uncertainty about the system decreases.
- Quantum conditional entropy can be negative, unlike classical uncertainty measures.
- The paper asks whether negative uncertainty permits work extraction during erasure and reports that it does.
A. Physics from an information-theoretic viewpoint
The paper frames entropy as observer-dependent because observers possess different information about the same physical system, while relating this subjectivity to standard thermodynamics.
- Finite measurement, storage, and precision limit the information observers retain about physical systems.
- Different observers can therefore have substantially different knowledge of the same physical reality.
- Conditional entropy H(S|O) quantifies an observer’s uncertainty about system S.
- For a fully degenerate n-qubit system, Alice’s entropy is 0 while Bob’s is n because their information differs.
- A standard observer reconciles observer-dependent entropy with thermodynamics by knowing selected macroscopic parameters while remaining otherwise maximally uncertain.
B. Quantum knowledge
The paper extends observer information to quantum memories, where conditional entropy can become negative, and establishes a thermodynamic interpretation for that negativity.
- Observers may hold information about S in a quantum memory rather than only in classical form.
- Quasimodo’s memory maximally entangles with S and can reproduce Alice’s classical data through measurement.
- For Quasimodo, the pure joint state and maximally mixed memory give H(S|Q) = −n.
- Negative conditional entropies have operational roles in state merging and quantify violations of uncertainty bounds when quantum side information is available.
- The paper connects conditional entropy to erasure work, extending the relation to the quantum regime and interpreting negative entropy thermodynamically.
C. Information-work relation
The paper defines erasure as preparing a system in a fixed pure state and shows that its work cost depends on the observer’s information, including quantum information.
- Erasure takes a system to the pre-defined pure state |0⟩, potentially reducing uncertainty about its prior state.
- For a spin-1/2 example, the system begins degenerate and the magnetic field controls the energies of |↓⟩ and |↑⟩.
- Alice can erase a known pure state reversibly at zero energy cost, whereas Bob’s fully mixed description requires kT ln 2.
- The fully mixed-qubit protocol raises |↑⟩ while coupled to a heat bath, reducing its occupation before lowering its energy after isolation.
- The erasure work is formulated conditionally on observer memory, so different observers can require different work amounts.
- The paper generalizes the information-work relation from classical observers to observers with quantum memories.
II. THE GENERAL RELATION BETWEEN INFORMATION AND WORK
The paper studies erasure by an observer with quantum memory, allowing operations on the system and memory while preserving the memory’s information about a reference system.
- The paper places quantum-memory erasure within prior work on thermal processes, work extraction, and Maxwell’s demon.The cited literature includes settings where correlations and entanglement affect erasure and work extraction.
- The setting includes a system S, quantum memory O, heat bath at temperature T, battery, and reference system R.The observer may perform operations on S and O, without LOCC restrictions.
- The memory contents are preserved during erasure because the joint state ρ_OR remains unchanged while R is not touched.This condition protects information the memory may hold about systems other than S.
A. A special case
A maximally entangled memory can enable erasure with a net work gain while preserving the memory’s correlations with a reference system. The gain is explained by consuming entanglement between the memory and S.
- A special case: A memory qubit Q1 is maximally entangled with S, while Q2 is entangled with reference qubit R; the memory-reference state must remain intact.The reduced state of Q1 is fully mixed because of its entanglement with S.
- A special case: 2kT ln 2 is extracted from the entangled pure state of Q1 and S, followed by an erasure costing kT ln 2, yielding a net gain of kT ln 2.The first step leaves Q1 and S fully mixed and preserves the joint memory-reference state.
- A special case: H(S|Q) = −1 produces a negative erasure cost, W(S|Q) = −kT ln 2.The example connects negative conditional entropy directly to work extraction during erasure.
- A special case: The gain cannot be reused indefinitely because erasure removes the entanglement between S and Q, leaving H(S|Q) = 0.The energy gain comes from the heat bath, while the global entropy increases and the memory-reference state remains protected.
B. Single-shot erasure
The single-shot result bounds erasure work using smooth conditional max-entropy, with a small failure probability. Negative conditional entropy therefore permits work extraction beyond the maximally entangled example.
- B. Single-shot erasure: Theorem 1 guarantees an erasure process whose work cost is bounded by conditional max-entropy except with a small probability.The bound applies to a system S conditioned on quantum memory O at bath temperature T.
- B. Single-shot erasure: The ε-smooth max-entropy is a single-shot generalization of von Neumann entropy that reduces to it in a thermodynamic limit.This quantity is the entropy measure used in the theorem’s work-cost bound.
- B. Single-shot erasure: δ = 3% entails an additional work price of approximately 20 kT ln 2 beyond the entropy-determined consumption.The correction term can be chosen small, especially for large systems.
- B. Single-shot erasure: Hε_max(S|O) < 0 allows an observer to erase S with negative work cost, extracting work using correlations beyond maximal entanglement.The theorem allows observers to exploit more general correlations between S and O.
- B. Single-shot erasure: The proof also yields a work-extraction process that keeps the memory intact while leaving the final state of S arbitrary.This result is stated as a corollary for an n-qubit system and memory O.
C. Thermodynamic limit
In the thermodynamic limit, repeated erasure of many independent copies removes fluctuations and connects smooth max-entropy to von Neumann entropy. The resulting work-cost rate is bounded by conditional entropy.
- C. Thermodynamic limit: The thermodynamic limit is modeled by erasing a large collection of independent systems and evaluating the average work cost rate.This construction treats thermal fluctuations through repeated copies.
- C. Thermodynamic limit: Information compression isolates a subsystem S1 whose size grows with correlations between S and O and shrinks with correlations between S and Γ.S1 is purified by an equal-sized subsystem P in the remaining systems, forming a fully entangled state.
- C. Thermodynamic limit: The quantum Asymptotic Equipartition Property makes smooth max-entropy converge toward von Neumann entropy for many identical copies.The convergence lets the single-shot theorem determine the asymptotic work-cost rate.
- C. Thermodynamic limit: The erasure work-cost rate satisfies ¯w(S|O) ≤ H(S|O) kT ln 2.The bound is the thermodynamic-limit counterpart of the single-shot relation.
III. OUTLINE OF THE PROOF
The proof implements erasure through three stages: compressing correlations into an approximately pure entangled subsystem, extracting work from it, and erasing S. The resulting work consumption is bounded by conditional smooth max-entropy.
- Proof steps: The explicit erasure process first compresses correlations between S and O into an approximately pure subsystem of S ⊗ O.The subsystem contains approximately n − Hmax(S|O) qubits and is formed using decoupling results.
- Proof steps: The observer extracts roughly [n − Hmax(S|O)] kT ln 2 of work from the resulting pure entangled state.The extraction yields kT ln 2 per qubit.
- Proof steps: The observer then erases S using n kT ln 2 of work, again at kT ln 2 per qubit.This is the final erasure step after work extraction.
- Work bound: The complete process has work balance (ℓ−n) kT ln 2, and its work consumption satisfies W(S|O) ≤ [Hεmax(S|O) + ∆]kT ln 2.The logarithmic correction is usually negative when δ and ε are chosen small.
IV. CONCLUSIONS
The paper gives conditional entropies a direct thermodynamic meaning by using an observer’s quantum information to erase a system. Negative conditional entropy corresponds to negative erasure work, while quantum memory can enable greater work extraction than classical memory.
- Conclusions: The erasure work cost depends on the conditional entropy describing an observer’s uncertainty about the system.The authors introduce an erasure process that uses the quantum information held by the observer.
- Conclusions: Negative conditional entropy corresponds to a negative work cost for erasure.Thus, erasing can yield work rather than require it in the relevant quantum setting.
- Conclusions: An observer with a quantum memory can extract twice as much work from a system as an observer with a classical memory.This comparison directly links the information representation available to the observer with extractable work.
- Conclusions: The results connect information theory and statistical mechanics and suggest transferring concepts between the two fields.The paper relates work extraction bounds to a thermodynamic derivation and interpretation of the data processing inequality.
- Conclusions: Discord quantifies the difference between conditional uncertainties for quantum and measured classical memories.It is defined as δ(S|O) = H(S|OQ)−H(S|OC).
A. Applications
The applications and formal setting extend conditional-entropy work bounds to memory erasure, smooth entropies, and asymptotic von Neumann entropy. The framework assumes controlled quantum operations, a heat bath, a battery, and preservation of the memory-reference state.
- Applications: Erasing part of a memory while preserving the rest requires work upperbounded by a conditional entropy, which can be much smaller than the erased part’s entropy.This bound applies to a common computation scenario in which the remaining memory must stay intact.
- Scope and limitation: The result requires almost perfect control of the quantum systems involved.The authors present it as a theoretical ideal limit for minimally heat-generating reversible computation.
- Formal setting: The setting includes S, a quantum memory O, a heat bath at temperature T, a battery, and a reference system R, with a pure initial global state.The composed system S ⊗ O has a fully degenerate Hamiltonian.
- Formal setting: Allowed operations include subsystem unitaries, energy-level manipulation, and coupling to the heat bath or battery, while operations on the reference system are forbidden.These restrictions define the physical processes used in the erasure model.
- Formal setting: Successful erasure takes S to a predefined pure state while preserving the joint state of O and R.The work cost is defined as the difference between the battery’s initial and final charge.
- Smooth entropies: The main theorem uses smooth max-entropy because the work-entropy relation is valid independently of the underlying quantum-state structure.A von Neumann entropy relation follows under additional assumptions.
- Smooth entropies: For i.i.d. states, the work-entropy relation asymptotically also holds for the von Neumann entropy.This follows when the relevant smooth min- and max-entropies coincide in the asymptotic setting.
Appendix C: Information Compression
The appendix proves that correlations between a system and an observer’s memory enable reversible information compression, supporting erasure whose work cost depends on conditional entropy. The construction decouples part of S, purifies it within S⊗O, and uses the resulting pure subsystem for work extraction or erasure.
- Information compression: Correlations between S and O allow local reversible transformations on S to create a subsystem of S⊗O that is close to pure.Here S is the system being erased, O is the observer’s memory, and Γ contains the battery, heat bath, and reference system.
- Information compression: Theorem 2 establishes that an ℓ-qubit subsystem of S⊗O can be made δ-close to pure by applying a local unitary on S.Its proof first decouples an ℓ/2-qubit subsystem of S from Γ, then uses global purity to find a same-sized purifying subsystem in S⊗O.
- Information compression: 3% failure probability costs 10 qubits in S1 and increases erasure work consumption by 20kT ln 2.The erasure process fails with maximum probability δ.
- Information compression: The decoupling bound is optimal: no unitary on S can decouple more than m qubits from Γ.The decoupled subsystem’s size depends on correlations between S and O, quantified by the smooth conditional max-entropy.
- Work extraction and erasure: A fully mixed ℓ-qubit subsystem can be erased using ℓkT ln 2 of work, while a pure ℓ-qubit subsystem can yield ℓkT ln 2 of work when randomized.The pure-to-mixed process leaves the Hamiltonian unchanged; inversion produces erasure.
- Work extraction and erasure: For quantum observers, erasure work is proportional to H(S|O), so negative conditional entropy corresponds to a work yield rather than a work cost.Entanglement can simultaneously erase a qubit and convert kT ln 2 of heat into work, but the initial entanglement is consumed and cannot support a repeating cycle.
- Scope: The procedures described are limited to algorithms with classical input and output; quantum-input or quantum-output applications remain open.Suggested applications include physical-system simulation and tomography-type procedures.