Using neural networks to infer acceleration for an atom-based accelerometer

Lopes, Eric

Student Work

Using neural networks to infer acceleration for an atom-based accelerometer

Public

Atom based inertial sensing is widely acknowledged by the atomic physics and precision measurements communities as having enormous potential in applications of both scientific and practical interest. One of the promising approaches to many sensor applications can be addressed by systems based on the physics of optical atomic lattices. An optical lattice is created by interfering laser beams to form a two-dimensional lattice. Such a lattice can be used to trap and manipulate cold atoms. The present project is devoted to the theoretical analysis of one of such interferometric geometries. Atoms in optical lattices achieve their sensing capability by first preparing them in the quantum-mechanical ground state of the lattice. Subsequently the wavefunction of the atoms are caused to undergo a series of transformations. That series of transformations correspond to those of an interferometer (splitting, propagation, reflection, propagation, and recombination). Each of those wavefunction transformations is achieved by modulating the optical lattice. We introduce and analyse a specific optical lattice modulation protocol that realises the interferometric cycle. The results can not be interpreted analytically, so the analysis relies on machine learning techniques. Neural networks ware employed to interpret the results, and are shown to be successful for inferring the acceleration of the lattice within multiple ranges of acceleration.

This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.

Creator