Augmenting Imaginary-Time Evolution with Local Geometric Information
This work introduces Augmented Imaginary-Time Evolution (AITE) to replace standard imaginary-time evolution (ITE). AITE uses higher-order statistical information, specifically skewness of the energy distribution, to guide the evolution along locally optimal geometric directions. This results in superlinear convergence and a finite-time extinction of the energy error, which is faster than ITE's asymptotic exponential decay.