[Paper Review] Above and Beyond the Landauer Bound: Thermodynamics of Modularity

Alexander B. Boyd, Dibyendu Mandal+1 more|arXiv (Cornell University)|Aug 9, 2017

Advanced Thermodynamics and Statistical Mechanics48 references19 citations

TL;DR

This paper identifies a new thermodynamic cost—modularity dissipation—arising from localized, modular computation that exceeds Landauer's bound, even when individual gates operate efficiently. It shows that optimal efficiency is achieved when information ratchets store global correlations via predictive or retrodictive hidden states, enabling them to match the thermodynamic efficiency of globally coordinated computations.

ABSTRACT

Information processing typically occurs via the composition of modular units, such as universal logic gates. The benefit of modular information processing, in contrast to globally integrated information processing, is that complex global computations are more easily and flexibly implemented via a series of simpler, localized information processing operations which only control and change local degrees of freedom. We show that, despite these benefits, there are unavoidable thermodynamic costs to modularity---costs that arise directly from the operation of localized processing and that go beyond Landauer's dissipation bound for erasing information. Integrated computations can achieve Landauer's bound, however, when they globally coordinate the control of all of an information reservoir's degrees of freedom. Unfortunately, global correlations among the information-bearing degrees of freedom are easily lost by modular implementations. This is costly since such correlations are a thermodynamic fuel. We quantify the minimum irretrievable dissipation of modular computations in terms of the difference between the change in global nonequilibrium free energy, which captures these global correlations, and the local (marginal) change in nonequilibrium free energy, which bounds modular work production. This modularity dissipation is proportional to the amount of additional work required to perform the computational task modularly. It has immediate consequences for physically embedded transducers, known as information ratchets. We show how to circumvent modularity dissipation by designing internal ratchet states that capture the global correlations and patterns in the ratchet's information reservoir. Designed in this way, information ratchets match the optimum thermodynamic efficiency of globally integrated computations.

Motivation & Objective

• To identify and quantify a new thermodynamic cost—modularity dissipation—arising from the use of modular information processing despite individual gates operating near Landauer's limit.
• To explain why modular computations inherently lose global correlations in the environment, which are thermodynamically valuable resources, leading to additional irreversible dissipation.
• To establish conditions under which modular systems, such as information ratchets, can achieve the minimal thermodynamic cost of globally integrated computation.
• To link the efficiency of information ratchets to their ability to store global correlations via predictive or retrodictive hidden states.

Proposed method

• Uses stochastic thermodynamics and information theory to derive an exact expression for dissipation in localized processing operations.
• Defines modularity dissipation as the difference between the change in global nonequilibrium free energy (capturing global correlations) and the local (marginal) change in free energy.
• Applies the information diagram formalism to decompose mutual information and isolate the contribution of lost correlations to irreversibility.
• Analyzes information ratchets as physical transducers that convert environmental correlations into work, showing that optimal efficiency requires hidden states to be predictive (for extractors) or retrodictive (for generators).
• Uses recursive information-theoretic arguments to prove that zero entropy production is equivalent to predictive or retrodictive behavior in ratchets.
• Demonstrates that by designing ratchet states to capture global patterns, modularity dissipation can be eliminated, achieving the thermodynamic optimum.

Experimental results

Research questions

• RQ1 What is the thermodynamic cost of modular computation that exceeds Landauer’s bound, and what causes it?
• RQ2 Why do modular computations inherently lose global correlations, and how does this loss lead to additional dissipation?
• RQ3 Under what conditions can modular systems like information ratchets achieve the thermodynamic efficiency of globally coordinated computations?
• RQ4 How are predictive and retrodictive hidden states in ratchets related to zero entropy production and minimal dissipation?
• RQ5 Can the modularity dissipation be eliminated by engineering the internal state of a ratchet to store global environmental correlations?

Key findings

• Modularity dissipation arises from the loss of global correlations during localized processing and is distinct from Landauer’s bound, representing an additional thermodynamic cost.
• The minimum dissipation in modular computation is quantified as the difference between global and local changes in nonequilibrium free energy, which captures the loss of usable correlation energy.
• For information ratchets that extract work from structured inputs, minimal dissipation requires the hidden state to be predictive of the input sequence.
• For ratchets that generate structured outputs, minimal dissipation requires the hidden state to be retrodictive of the past output sequence.
• Zero entropy production in ratchets is equivalent to the ratchet being predictive (for extractors) or retrodictive (for generators), establishing a direct link between information structure and thermodynamic efficiency.
• By designing ratchet states to store global correlations, modularity dissipation can be fully eliminated, allowing ratchets to match the thermodynamic efficiency of globally coordinated computations.