[Paper Review] Poincare Recurrence, Zermelo's Second Law Paradox, and Probabilistic Origin in Statistical Mechanics

P. D. Gujrati|ArXiv.org|Mar 7, 2008

Force Microscopy Techniques and Applications3 citations

TL;DR

This paper resolves Zermelo's paradox by demonstrating that Poincaré recurrence does not violate the second law of thermodynamics because entropy depends on the probability distribution of microstates, not just the recurrence of a single state. The authors show that probabilistic behavior in statistical mechanics arises from external noise, not system preparation, and that entropy remains constant at equilibrium, never decreasing in isolated stochastic systems.

ABSTRACT

We show that Poincare recurrence does not mean that the entropy will eventually decrease, contrary to the claim of Zermelo, and that the probabilitistic origin in statistical physics must lie in the external noise, and not the preparation of the system.

Motivation & Objective

• To resolve Zermelo's paradox, which claims that Poincaré recurrence implies entropy must decrease, violating the second law.
• To clarify that the probabilistic nature of entropy in statistical mechanics originates not from system preparation but from stochastic interactions with the environment.
• To demonstrate that entropy remains constant after equilibrium is reached in isolated stochastic systems, even if the initial microstate recurs.
• To distinguish between deterministic Poincaré recurrence (with probability 1) and stochastic recurrence (with probability ≈1/W), showing the latter does not imply entropy reversal.

Proposed method

• Uses the Gibbs ensemble formulation to define entropy as S(t) = Σ p_i(t) u_i(t), where u_i(t) = -ln p_i(t), and p_i(t) is the probability of microstate i.
• Applies Poincaré recurrence theorem to deterministic systems confined in finite phase space, showing recurrence occurs with probability 1.
• Contrasts deterministic recurrence with stochastic evolution, where microstate transitions occur due to external noise, leading to p_i(t) < 1 even for the initial state.
• Analyzes the entropy evolution in stochastic systems using the ensemble approach, showing that S(t) > 0 even when p_0(t) > 0, due to contributions from all microstates.
• Calculates the probability of true recurrence (all replicas in microstate 0 simultaneously) as W^{-N} → 0 as N → ∞, proving it is effectively impossible.
• Establishes that in equilibrium, p_i(t) → 1/W for all i, so S(t) = ln W, and entropy remains constant thereafter, never decreasing.

Experimental results

Research questions

• RQ1 Does Poincaré recurrence imply a violation of the second law of thermodynamics, as claimed by Zermelo?
• RQ2 What is the true origin of probabilistic behavior in statistical mechanics—system preparation or environmental noise?
• RQ3 Can the entropy of an isolated stochastic system decrease after reaching equilibrium, even if the initial microstate recurs?
• RQ4 Why does the recurrence of a single microstate not imply a return to low-entropy states, given that entropy is a function of probabilities?
• RQ5 What is the difference between deterministic Poincaré recurrence and stochastic recurrence in terms of entropy evolution?

Key findings

• Poincaré recurrence in deterministic systems does not violate the second law because entropy remains constant during the recurrence cycle.
• The recurrence of the initial microstate in stochastic systems occurs with probability p_0(t) < 1, so entropy does not revert to zero, even if the state is revisited.
• The entropy of a system in equilibrium is S(t) = ln W, and remains constant thereafter, never decreasing, regardless of microstate recurrence.
• The probability of true recurrence—where all replicas are simultaneously in the initial microstate—is W^{-N}, which vanishes as N → ∞, making it effectively impossible.
• The probabilistic origin of entropy in statistical mechanics arises from external noise, not from the method of system preparation.
• Entropy is determined by the full probability distribution over all microstates, not just the probability of a single state, so recurrence of one state does not imply entropy decrease.