Maxwells Demon and Data Compression

In an asymmetric Szilard engine model of Maxwell’s demon, we show the equivalence between information theoretical and thermodynamic entropies when the demon erases information optimally. The work gain by the engine can be exactly canceled out by the work necessary to reset demon’s memory after optimal data compression a la Shannon before the erasure.

💡 Research Summary

The paper investigates the relationship between information entropy and thermodynamic entropy using a Maxwell‑demon‑type Szilard engine, extending the classic symmetric model to an asymmetric configuration and incorporating optimal data compression. In the traditional symmetric Szilard engine, a single molecule confined in a one‑dimensional cylinder is separated by a centrally placed partition. The demon measures the molecule’s side, records a binary bit (0 or 1) in a memory modeled as a single‑molecule gas, and then allows the molecule to push the partition, extracting work (W = k_{B}T\ln2). The memory is subsequently reset by compressing its gas volume by a factor of two, which costs exactly the same amount of work, thereby preserving the second law: the entropy decrease of the engine ((-k_{B}\ln2)) is compensated by the entropy increase due to information erasure ((+k_{B}\ln2)).

The authors then consider an asymmetric Szilard engine where the partition divides the volume into fractions (p) and (1-p). The work extracted per cycle becomes (W_{\text{eng}} = k_{B}T,S(p)) with (S(p) = -p\ln p -(1-p)\ln(1-p)), i.e., the binary Shannon entropy multiplied by (k_{B}T). If the demon erases its memory after each cycle, the required work is (\Delta W = k_{B}T\ln2 - k_{B}T,S(p) \ge 0), leaving a net loss.

To eliminate this loss, the demon accumulates the outcomes of (N) cycles, forming an (N)-bit string containing (pN) zeros and ((1-p)N) ones. By Shannon’s noiseless coding theorem, the string can be compressed to a length (N H(p)) where (H(p) = -p\log_{2}p -(1-p)\log_{2}(1-p)) is the information entropy in bits. The compression is assumed reversible and therefore cost‑free. Erasing the compressed string now requires work \