Computation time and thermodynamic uncertainty relation of Brownian circuits
We analyze a token-based Brownian circuit in which Brownian particles, coined `tokens,' move randomly by exploiting thermal fluctuations, searching for a path in multi-token state space corresponding to the solution of a given problem. The circuit can evaluate a Boolean function with a unique solution. However, its computation time varies with each run. We numerically calculate the probability distributions of Brownian adders' computation time, given by the first-passage time, and analyze the thermodynamic uncertainty relation and the thermodynamic cost based on stochastic thermodynamics. The computation can be completed in finite time without environment entropy production, i.e., without wasting heat to the environment. The thermodynamics cost is paid through error-free output detection and the resets of computation cycles. The signal-to-noise ratio quantifies the computation time's predictability, and it is well estimated by the mixed bound, which is approximated by the square root of the number of token detections. The thermodynamic cost tends to play a minor role in token-based Brownian circuits in computation cycles. This contrasts with the logically reversible Brownian Turing machine, in which the entropy production increases logarithmically with the size of the state space, and thus worsens the mixed bound.