arXiv Science⌕ Search

arXiv subjects

Justin Tahmassebpur

Publications and source records attributed to Justin Tahmassebpur.

6 recordsLinked to original sources

Inverse design of superconductors and semiconductors with differentiable foundation models

Designing a crystal for a target property requires choosing its elements, stoichiometry, and atomic structure together, yet gradients exist only for the continuous structural variables. We introduce differentiable Equiformer fast effective-atom theory (DEFEAT), which lets a pretrained EquiformerV3 foundation model accept continuous species weights and site occupancies without retraining on fractional data, thereby enabling gradient-based materials optimization. Gradients from fine-tuned property models and a machine-learned energy penalty then guide updates to composition, atom count, atomic positions, and lattice vectors in a single optimization. In a search for phonon-mediated superconductors, density-functional perturbation theory on 45 structurally novel designs gives isotropic Allen-Dynes critical temperatures above 10 K for 39; the highest is 24.0 K for ZrTiNb2CN3, and H4LiPd3 (23.2 K) lies on the static DFT hull. In searches for photovoltaic absorbers, SCAN calculations confirm ten novel crystal structures with band gaps within 20% of a 1.34 eV target on their respective raw or empirically corrected scales, all within 70 meV/atom of the DFT hull, and show that the systematic gap error of the training functional propagates into the designs unless the objective corrects for it. Quasiharmonic free energies from machine-learned potentials further show that finite-temperature competition with solid phases can reorder candidates relative to the static hull.

cond-mat.mtrl-sci↗

Hereditary Effective Atom Theory for Large-Scale Atomic Structure Prediction

Predicting the atomic structure of large systems requires simultaneous exploration of atomic positions, species, and, generally, the number of atoms. The combinatorial growth of possible arrangements can make this search difficult even for standard global optimization methods. We introduce Hereditary Effective Atom Theory (HEAT), which integrates Effective Atom Theory (EAT) into a genetic algorithm. HEAT combines EAT's continuous relaxation of atomic positions, species, and population with the global search capabilities of genetic algorithms. Central to the method is a crossover scheme that directly superimposes two parent structures and relaxes the combined structure. HEAT outperforms conventional genetic-algorithm baselines in recovering a missing 240-atom H sublattice, predicting a 364-atom edge-dislocation core, and reconstructing a 1150-atom interior region of twisted-bilayer $\mathrm{WSe_2}$ from experimental ptychographic constraints. These results demonstrate how continuous effective-atom relaxation improves the exploration of large combinatorial structure spaces.

cond-mat.mtrl-sci↗

Neural networks become centroid classifiers when trained on heavily corrupted inputs

Learning from imperfect data is a central theme in machine learning, connecting practical questions of robustness to fundamental questions of learnability. Here we examine attribute noise: learning from corrupted inputs while keeping the labels intact, a setting that has received considerably less analytical attention than its label-noise counterpart. We consider two types of corruption models: additive noise and replacement noise. Through experiments with multi-layer perceptrons (MLPs) on corrupted classification datasets, we find that neural networks remain robust, maintaining well-above-chance accuracy even when inputs are >90% corrupted -- far beyond human recognition. To understand this robustness, we analyze infinite-width networks in the heavy-corruption regime using a mean-field-inspired approach and derive a leading-order decision rule for the classification outcome: the network implements a prototype rule, the nearest-class-mean, assigning each test point to the class whose training-set average it most closely resembles. This leading-order decision rule is universal across a broad range of MLP architectures, holding for any depth, as well as a wide class of activation functions and noise distributions. The same centroid mechanism closely matches finite-width network behavior in our experiments and provides an interpretable and analytically tractable account of why learning can succeed even when individual training examples carry almost no signal.

cs.LG↗

Effective Atom Theory: Gradient-Driven Materials Design

We introduce Effective Atom Theory (EAT), which embeds discrete atomic identities into a continuous differentiable space. During optimization, sites pass through temporary effective mixtures, then are driven to fully discrete, physically realizable atoms. This enables gradient-based optimization in DFT and MLIPs, yielding orders-of-magnitude acceleration over discrete methods. EAT recovers the exact ground state of NaCl in a 128 atom cell, reconstructs the Cu-Pd phase diagram, optimizes high-entropy oxides for oxygen evolution, and discovers bonding topologies as demonstrated for SrTiO3.

cond-mat.mtrl-sci↗

A geometric basis for materials families in inorganic solids

The thermodynamic stability of inorganic solids spans a vast compositional space, yet materials scientists have long organized their intuition around a manageable number of materials families. Here we show that this organization has a precise geometric basis. The formation-energy convex hull of all inorganic compounds from the Materials Project, spanning 92-dimensional elemental composition space, is captured to near DFT accuracy by a polyhedron with only seven facets. Each facet corresponds to a family of materials sharing similar chemical potentials. This low-dimensional structure is not merely an economical description of energies: without retraining or structural input, the same framework reproduces trends in DFT-calculated defect energies and elemental spatial correlations in high-entropy nanoparticles. These results reveal that a small number of material families, corresponding to geometric features of composition-energy space, govern bulk stability, defect energetics, and elemental mixing, and provide a unified, interpretable framework for rapid screening across diverse materials systems.

cond-mat.mtrl-sci↗

How Turbulent Jets Can Disperse Virus Clouds in Poorly Ventilated Spaces

We show that enhanced turbulent mixing can be used to mitigate airborne COVID-19 transmission by dispersing virus-laden clouds in enclosed, poorly ventilated spaces. A simple system of fan-driven turbulent jets is designed so as to minimize peak concentrations of passive contaminants on time scales short compared to the room ventilation time. Standard Reynolds-average and similarity methods are used, and combinations of circular and radial wall jets are considered. The turbulent diffusivity and contaminant mixing time are calculated. Results indicate that this approach can significantly reduce peak virus cloud concentrations, especially in small spaces with low occupancy, such as restrooms. Turbulent mixing is, of course, ineffective in the absence of ventilation.

physics.flu-dyn↗