🤖 AI Summary
Conventional neuromorphic computing lacks a rigorous physical foundation linking thermodynamic processes to neural computation. Method: This work introduces the “thermodynamic neuron” paradigm for physics-intrinsic brain-inspired computing, leveraging autonomous quantum heat engines—nonequilibrium heat currents in few-body qubit systems driven by multiple thermal reservoirs—to perform logic and neural operations. Using open quantum system theory and nonequilibrium steady-state analysis, we construct a tunable quantum master equation model for heat transport and design a finite-size reservoir thermometry protocol for output readout. Contribution/Results: We demonstrate, for the first time at the single-neuron level, linearly separable logic gates—including NOT, NOR, and 3-MAJORITY—and prove that a single thermodynamic neuron can implement any linearly separable function. Moreover, networks of such neurons achieve Turing-universal computational capability. This work establishes a strict physical correspondence between thermodynamics and artificial neurons, providing the first principle-based pathway toward low-power, non-von Neumann, physics-native neuromorphic computing.
📝 Abstract
We develop a physics-based model for classical computation based on autonomous quantum thermal machines. These machines consist of few interacting quantum bits (qubits) connected to several environments at different temperatures. Heat flows through the machine are here exploited for computing. The process starts by setting the temperatures of the environments according to the logical input. The machine evolves, eventually reaching a nonequilibrium steady state, from which the output of the computation can be determined via the temperature of an auxilliary finite-size reservoir. Such a machine, which we term a “thermodynamic neuron,” can implement any linearly separable function, and we discuss explicitly the cases of NOT, 3-MAJORITY, and NOR gates. In turn, we show that a network of thermodynamic neurons can perform any desired function. We discuss the close connection between our model and artificial neurons (perceptrons) and argue that our model provides an alternative physics-based analog implementation of neural networks, and more generally a platform for thermodynamic computing.
