Computational Reverse-Engineering Analysis for Scattering Experiments (CREASE) with Machine Learning Enhancement to Determine Structure of Nanoparticle Mixtures and Solutions

Abstract

We present a new open-source, machine learning (ML) enhanced computational method for experimentalists to quickly analyze high-throughput small-angle scattering results from multicomponent nanoparticle mixtures and solutions at varying compositions and concentrations to obtain reconstructed 3D structures of the sample. This new method is an improvement over our original computational reverse-engineering analysis for scattering experiments (CREASE) method (ACS Materials Au 2021, 1 (22), 140−156), which takes as input the experimental scattering profiles and outputs a 3D visualization and structural characterization (e.g., real space pair-correlation functions, domain sizes, and extent of mixing in binary nanoparticle mixtures) of the nanoparticle mixtures. The new gene-based CREASE method reduces the computational running time by >95% as compared to the original CREASE and performs better in scenarios where the original CREASE method performed poorly. Furthermore, the ML model linking features of nanoparticle solutions (e.g., concentration, nanoparticles' tendency to aggregate) to a computed scattering profile is generic enough to analyze scattering profiles for nanoparticle solutions at conditions (nanoparticle chemistry and size) beyond those that were used for the ML training. Finally, we demonstrate application of this new gene-based CREASE method for analysis of small-angle X-ray scattering results from a nanoparticle solution with unknown nanoparticle aggregation and small-angle neutron scattering results from a binary nanoparticle assembly with unknown mixing/segregation among the nanoparticles.

Modeling Structural Colors from Disordered One-Component Colloidal Nanoparticle-Based Supraballs Using Combined Experimental and Simulation Techniques

“…We apply the recently developed genetic algorithm-based CREASE method 24 to reconstruct the supraball structure in (B) SAXS plot of I as a function of q for silica (top, yellow dots) and melanin (bottom, blue dots) supraballs overlaid with the CREASE output structures' scattering profile for silica (top, black line) and melanin (bottom; gray line) and the analytical scattering model for silica (top, green curve) and melanin (bottom; red curve). (C) Comparative plot of structure factor, S(q), between that calculated from the CREASE reconstruction for silica (left, black) and melanin (right, gray), and that from the "sticky" hard sphere S(q) model for silica (left, green) and melanin (right, red).…”

“…Before concluding this work, we provide pertinent information on utilizing the CREASE-FDTD approach and potential avenues for future work. For a researcher to utilize this CREASE-FDTD approach on a similarly sized system of interest (∼65 000 nanoparticles), they would require ∼1 h on a single core for the machine learning CREASE method to converge the genes 24 and generate the 3D nanoparticle structure for those converged genes and ∼8−10 h on 14 cores for the FDTD calculation on the reconstructed structure. Thus, in less than half a day, a researcher could determine their material's structure and validate that the reconstructed structure possesses the desired optical properties, enabling rapid development of structure−color relationships.…”

“…We apply the recently developed genetic algorithm-based CREASE method to reconstruct the supraball structure in Figure (see Supporting Information for details). In Figure A, we show that the CREASE method takes the experimental scattering intensity profile as input and outputs a 3D structure, whose computed scattering profile closely matches the experimental input.…”

“…We note that the CREASE method can converge those genes even if the mean diameter and dispersity are not known precisely. CREASE then converts those genes into the 3D nanoparticle structure and calculates the computed scattering intensity I comp ( q ) with the Debye scattering equation , or with the use of machine learning . CREASE calculates the fitness for each individual which quantifies the quality of the match between I comp ( q ) and I expt ( q ).…”

“…Here, we present a systematic computational approach to predict structural color generation from disordered colloidal nanoparticle assemblies. This approach combines (a) the recently developed computational reverse-engineering analysis for scattering experiments (CREASE) tool 24 to reconstruct the 3D structure of assembled particles given an intensity profile from small-angle scattering measurements and (b) FDTD to calculate the color reflectance spectra from the reconstructed particle assembly structure. We demonstrate that this approach provides simulated reflectance spectra in close agreement with experimental reflectance profiles.…”