Inverse design of colloidal nanocrystal assemblies with targeted structural and optical features

Abstract

The ability to design and control the structure of nanoscale building blocks, such as colloidal nanocrystal assemblies, is central to advancing the next generation of optoelectronic and energy materials with tunable properties. Traditional bottom-up synthesis approaches provide limited control over interparticle interactions, making it difficult to reliably achieve ordered, multilayered structures with targeted geometries. In this work, a computational inverse design framework is presented that combines molecular dynamics (MD) simulations with optimization techniques to direct the assembly of layered colloidal systems. These simulations are performed using JAX-MD, a differentiable MD software framework, where particle interactions are modeled in this thesis using a combination of Morse, screened Coulomb and spring bond potentials. By embedding JAX’s automatic differentiation and gradient-based optimization into the simulation loop, interaction parameters are adaptively tuned until the resulting particle configurations match the predefined design targets. Two demonstrations illustrate the capabilities of the methodology. First, a two layered system that establishes a baseline, composed of identical particles in each layer is optimized by tuning the electrostatic screening parameter (κ) to achieve the target interlayer spacings. The second extends the system to three layers, where particles are uniform within a layer but differ across layers in size and charge. This optimization focuses on tuning κ and the particle charges in the different layers to also achieve target spacings between the various layers. This study establishes a proof of concept for using optimization-driven simulations to engineer ordered, multilayered colloidal structures with controllable geometries, offering a systematic approach to navigating complex nanoscale design problems.