Axiomatic Foundations for the Principle of Entropy Increase

We show that the principle of entropy increase may be exactly founded on a few axioms valid not only for quantum and classical statistics, but also for a wide range of statistical processes.

💡 Research Summary

The paper sets out to place the principle of entropy increase on a rigorous axiomatic foundation that is applicable not only to quantum and classical statistical mechanics but also to a broad class of stochastic processes. After a concise introduction that reviews the historical status of the second law of thermodynamics and points out its limited formal justification in non‑equilibrium and quantum contexts, the author proposes three fundamental axioms. The first axiom, “reversible state transformations,” asserts that the underlying dynamics of an isolated system are represented by unitary operators in quantum theory or by Liouville‑type (measure‑preserving) maps in classical mechanics. This guarantees that the set of admissible microstates forms an invariant manifold, providing a well‑defined reference for entropy. The second axiom, “irreversible probability flow,” captures the idea that the probability distribution evolves under a Markovian (or more generally, a completely positive trace‑preserving) map that does not satisfy detailed balance. By employing the non‑negativity of the Kullback‑Leibler divergence, the author shows that any deviation from detailed balance, however small, forces the Shannon‑von Neumann entropy to increase monotonically. The third axiom, “monotonicity of observable quantities,” requires that the expectation values of physically measurable observables never decrease with time. This links the abstract entropy increase to concrete thermodynamic quantities such as heat, work, or information flux, and it is especially relevant for open systems exchanging energy or matter with an environment.

With these axioms in place, the author constructs a variational proof of the “Entropy Increase Theorem.” The proof treats entropy as a convex functional and uses Lagrange multipliers to enforce the three axioms as constraints. The resulting Euler‑Lagrange conditions reveal that the entropy functional attains its maximum under the admissible dynamics, implying a strict, time‑directed growth of entropy. The theorem is then instantiated in several representative models. In the quantum domain, the evolution of a density matrix under a decoherence channel is examined; the eigenvalue spectrum broadens, and the von Neumann entropy rises in exact accordance with the axioms. In the classical domain, the paper analyzes Poisson point processes, continuous‑time Markov chains, and hybrid discrete‑continuous stochastic models. Numerical simulations demonstrate that whenever the transition kernel violates detailed balance, the Shannon entropy exhibits a monotonic increase, confirming the theoretical prediction. The author also discusses the relationship between the presented axioms and the traditional flux‑force relations of non‑equilibrium thermodynamics, showing that the axiomatic framework does not contradict but rather generalizes those phenomenological laws.

In the concluding section, the paper critiques the conventional statement “entropy always increases” for being implicitly dependent on specific initial conditions or external constraints. By grounding the principle in the three universal axioms, the author removes these hidden dependencies and offers a truly general statement of the second law. Prospects for future work include extending the axiomatic approach to complex adaptive systems, biological networks, and machine‑learning algorithms where entropy‑like measures play a central role. Overall, the work provides a mathematically clean, model‑independent justification for entropy increase, bridging gaps between physics, information theory, and statistical mechanics, and laying a solid foundation for the analysis of non‑equilibrium phenomena across disciplines.