A new AI can 'discover' the fundamental laws of thermodynamics just by watching how materials move and change temperature.
March 31, 2026
Original Paper
A Convex Route to Thermomechanics: Learning Internal Energy and Dissipation
arXiv · 2603.28707
The Takeaway
Instead of humans deriving the complex equations for how a new material reacts to heat and stress, this neural network learns the underlying physics on its own. It is mathematically forced to obey the Second Law of Thermodynamics, ensuring the 'physics' it learns is real and not just a simulation.
From the abstract
We present a physics-based neural network framework for the discovery of constitutive models in fully coupled thermomechanics. In contrast to classical formulations based on the Helmholtz energy, we adopt the internal energy and a dissipation potential as primary constitutive functions, expressed in terms of deformation and entropy. This choice avoids the need to enforce mixed convexity--concavity conditions and facilitates a consistent incorporation of thermodynamic principles. In this contribu