Diffractive deep neural network adjoint assist or (DNA)2: a fast and efficient nonlinear diffractive neural network implementation.

The recent advent of diffractive deep neural networks or D2NNs has opened new avenues for the design and optimization of multi-functional optical materials; despite the effectiveness of the D2NN approach, there is a need for making these networks as well as the design algorithms more general and computationally efficient. The work demonstrated in this paper brings significant improvements to both these areas by introducing an algorithm that performs inverse design on fully nonlinear diffractive deep neural network - assisted by an adjoint sensitivity analysis which we term (DNA)2. As implied by the name, the procedure optimizes the parameters associated with the diffractive elements including both linear and nonlinear amplitude and phase contributions as well as the spacing between planes via adjoint sensitivity analysis. The computation of all gradients can be obtained in a single GPU compatible step. We demonstrate the capability of this approach by designing several types of three layered D2NN to classify 8800 handwritten digits taken from the MNIST database. In all cases, the D2NN was able to achieve a minimum 94.64% classification accuracy with 192 minutes or less of training.

Uncovering the mystery of ferroelectricity in zero dimensional nanoparticles.

It is generally accepted that chemically synthesized nanoparticles lose their ferroelectricity (spontaneous polarization) as the particles become smaller. In contrast, ball-milled ferroelectric nanoparticles have an enhanced ferroelectric response at remarkably small sizes (≤10 nm). Although prior theory suggests that surface stress influences ferroelectricity, the source of such a stress and how it physically influences ferroelectricity in zero-dimensional nanoparticles has remained a mystery. In this paper, we demonstrate that the top-down approach of wet ball-milling not only results in fragmented materials on the nanoscale, but it also is responsible for a mechanochemical synthesis of metal carboxylates forming at the nanoparticles' surface. We prove that the presence of such a compound with a particular type of binding mode chemisorbed at the nanoparticles' surface is responsible for producing surface stress. This surface stress results in a stabilization and dramatic enhancement of the spontaneous polarization, which is 5 times greater than that of the bulk material and 650 times greater than what is measured in materials fabricated using standard chemical synthesis techniques. The results of this study have further led to the development of a new process that produces ferroelectric nanoparticles (≤10 nm) with uniform shape and size using a combination of wet chemistry and mechanochemical synthesis.

Molecular catalysis at polarized interfaces created by ferroelectric BaTiO3.

The local environment at polarized solid-liquid interfaces provides a unique medium for chemical reactions that could be exploited to control the selectivity of non-faradaic reactions. Polarized interfaces are commonly prepared by applying a voltage to an electrode in an electrolyte solution, but it is challenging to achieve high surface charge densities while suppressing faradaic reactions. Ferroelectric materials have permanent surface charge densities that arise from the dipole moments of ferroelectric domains and can be used to create polarized solid-liquid interfaces without applying a voltage. We studied the effects of ferroelectric oxides on the selectivity of a Rh porphyrin-catalyzed carbene rearrangement. The addition of ferroelectric BaTiO3 nanoparticles to the reaction solution changed the product ratio in the same direction and by a similar magnitude as performing the reaction at an electrode-electrolyte interface polarized by a voltage. The results demonstrate that colloidal suspensions of BaTiO3 nanoparticles act as a dispersible polarized interface that can influence the selectivity of non-faradaic reactions.