Presentation Information

[8p-A23-11]Spontaneous piston-phase alignment in genetic-algorithm wavefront shaping of an optical phased array

Ryusei Takahashi1, Naoki Yamato2, 〇Mamoru Hashimoto1 (1.IST Hokkaido Univ., 2.DDIn2 Kyushu Univ.)

Keywords:

Piston phase,Wavefront shaping,Genetic algorithm

In fiber bundle endoscopy, wavefront shaping via a spatial light modulator corrects inter-core phase differences, thereby focusing light coherently at the distal end. This optimization possesses a gauge degree of freedom called the piston phase: a uniform offset added to all segment phases leaves focusing intensity invariant. Sequential algorithms handle this implicitly via a reference segment; partitioning algorithms resolve it explicitly each iteration. We report that genetic algorithm (GA) optimization resolves this degeneracy spontaneously, without explicit treatment.Two metrics quantify phase alignment. σPSM measures inter-group relative phase via the phase-shift method (measurement only, not fed back to individuals), capturing piston-phase alignment in both simulation and experiment. σLPE measures residual phase errors relative to the known target distribution, governing focusing intensity and accessible only in simulation.Standard GA was applied in noiseless and noisy simulations, as well as experimentally, using a 10,000-core fiber bundle. Two-stage convergence emerged: σPSM converges within tens of generations, while σLPE decreases gradually over hundreds of generations. An intermediate state in which σPSM is small but σLPE remains large confirms this picture, as reproduced experimentally.When parents differ in piston phase, crossover yields offspring with destructive interference, reducing fitness. Piston phase mismatch thus incurs selective pressure through crossover, driving early population-wide alignment. This ordering of a gauge-invisible degree of freedom emerges from the interplay of recombination and selection, robust against noise.