Topology optimization finds the best shape for a structure under given loads. Multi-objective topology optimization finds the set of best shapes when objectives conflict — stiffness versus weight, thermal versus mechanical performance. The Pareto frontier is this set: every point on it is optimal in the sense that improving one objective requires worsening another.
Oellerich and Yamada (arXiv:2603.22739) recast the level set function — the mathematical surface that defines the shape — as a generalized coordinate of a fictitious material system, then derive its equations of motion from Hamilton's principle. The optimization becomes physics. The level set evolves according to a damped wave equation, with gradient information providing the driving force and damping preventing oscillation.
For multiple objectives, the weighting factors between them become coordinates on a geometric structure. The Pareto frontier lives on a subset of this space, and the dynamics can be extended to let the weights evolve alongside the shape, exploring the frontier adaptively rather than requiring the designer to specify weights in advance.
The payoff is a stable, uniform sampling of the Pareto frontier that scales to higher-dimensional objective spaces — three or more competing objectives, where exhaustive parameter sweeps become impractical.
The structural insight is that treating optimization as dynamics — with momentum, damping, and energy — gives you properties that purely gradient-based methods lack. Momentum carries the solution through flat regions. Damping prevents oscillation near equilibria. The wave equation doesn't just find stationary points; it moves through the space with physical intuition. The fictitious physics solves a real engineering problem by making the search process behave like a physical system that naturally settles into equilibria.