Deep learning for retrieval of the internuclear distance in a molecule from interference patterns in photoelectron momentum distributions
N. I. Shvetsov-Shilovski, Manfred Lein
Abstract
We use a convolutional neural network to retrieve the internuclear distance in the two-dimensional ${\mathrm{H}}_{2}^{+}$ molecule ionized by a strong few-cycle laser pulse based on the photoelectron momentum distribution. We show that a neural network trained on a relatively small dataset consisting of a few thousand images can predict the internuclear distance with an absolute error less than $0.1$ a.u. Deep learning allows us to retrieve more than one parameter from a given momentum distribution. Specifically, we used a convolutional neural network to retrieve both the internuclear distance and the laser intensity. We study the effect of focal averaging, and we find that the convolutional neural network trained using the focal averaged electron momentum distributions also shows a good performance in reconstructing the internuclear distance.