arXiv subjects

Salman N. Salman

Publications and source records attributed to Salman N. Salman.

4 recordsLinked to original sources

arXiv2026

Out-of-Distribution Inverse Design of Elastic Networks with Differentiable Graph Neural Network Molecular Dynamics

Machine-learning-based inverse design can accelerate the discovery of materials with targeted properties, but conventional structure--property models often require large training datasets and generalize poorly beyond their training distribution. Here, we present a differentiable inverse design framework based on a graph neural network molecular dynamics simulator. By combining a short dynamical initialization with physics-based refinement during simulation, the framework enables optimization well beyond the conditions represented in the training data. Using disordered elastic networks, we show that a simulator trained only on non-auxetic systems with Poisson's ratios between 0.1 and 0.4 can design strongly auxetic networks with values as low as -0.3. The framework also produces stress--strain responses outside the range observed during training, creates localized mechanical defects absent from the training data, and generalizes across system size, enabling optimization of networks tested up to 5000 nodes despite being trained only on networks with fewer than 200 nodes.

physics.chem-ph↗

arXiv2026

Enabling Structure-Only Initialization and Out-of-Distribution Generalization in GNN-based Molecular Dynamics Simulators

Machine learning-based simulators offer the potential to model the dynamics of complex systems more efficiently than classical approaches, while retaining differentiability, a key property for materials design. Graph neural network (GNN)-based simulators have shown strong performance across a range of physical domains, including molecular dynamics. However, their reliance on temporal context for accurate prediction limits their use in inverse design settings, where simulations must be initialized from a single static configuration. Moreover, inverse design requires robust out-of-distribution (OOD) generalization, as candidate structures typically lie outside the training domain. Here, we address both challenges by introducing two complementary strategies that enable stable and accurate structure-only initialization of GNN-based simulations. To directly target OOD generalization, we propose an inference-time physics-based optimization framework that constrains model predictions to remain physically consistent during rollout. In addition, we introduce a differentiable, GNN-based barostat that enables accurate tracking of system dimensions and pressure, critical for capturing macroscopic responses and supporting OOD generalization. We evaluate these approaches in the context of uniaxial compression of disordered elastic networks spanning a broad range of geometries, Poisson ratios, and microscopic behaviors. We find that, together, these methods substantially improve rollout stability and enable reliable OOD generalization, including regimes with distinct, more complex dynamics than those in the training data. These results show that, when properly initialized and constrained, GNN-based simulators can serve as efficient and generalizable tools for materials discovery and structural optimization, advancing their use in materials, molecular, and dynamical system design.

physics.chem-ph↗

arXiv2026

A Non Linear Spectral Graph Neural Network Simulator for More Stable and Accurate Rollouts

Molecular dynamics (MD) simulations are a central tool in science and engineering enabling the study of dynamical behavior and the link between microscopic structure and macroscopic function. Their high computational cost, however, has motivated extensive efforts to develop accelerated alternatives. A promising approach is the use of machine-learning-based simulators that allow for substantially larger time steps than conventional MD. Among these, graph neural network (GNN)-based methods have been found to be especially attractive given that they naturally encode the inductive bias of interacting particle systems, however current architectures remain limited in accuracy and stability. In particular, standard message-passing schemes struggle to efficiently propagate long-range information. Here, we investigate whether spectral-GNN simulators can overcome these limitations by explicitly representing a simulated system in a global eigenmode basis, and therefore better capture long-range and collective behavior. Focusing on disordered elastic networks, model systems for complex materials and biological structures such as proteins, we compare spatial, linear spectral, and nonlinear spectral GNN architectures. We find that while spectral representations alone are not sufficient, nonlinear spectral models substantially outperform alternatives. By learning both the time-dependent dynamics and the mixing of eigenmodes, these models more accurately capture the system's slow, global modes, which dominate macroscopic behavior. This leads to a marked reduction in systematic particle-position error and significantly improved prediction of global physical properties.

physics.chem-ph↗

arXiv2025

Dynamical Data for More Efficient and Generalizable Learning: A Case Study in Disordered Elastic Networks

Machine learning models often require large datasets and struggle to generalize beyond their training distribution. These limitations pose significant challenges in scientific and engineering contexts, where generating exhaustive datasets is often impractical and the goal is frequently to discover novel solutions outside the training domain. In this work, we explore the use of dynamical data through a graph neural network-based simulator to enable efficient system-to-property learning and out-of-distribution prediction in the context of uniaxial compression of two-dimensional disordered elastic networks. We find that the simulator can learn the underlying physical dynamics from a small number of training examples and accurately reproduce the temporal evolution of unseen networks. Notably, the simulator is able to accurately predict emergent properties such as the Poisson's ratio and its dependence on strain, even though it was not explicitly trained for this task. In addition, it generalizes well across variations in system temperature, strain amplitude, and most significantly, Poisson's ratios beyond the training range. These findings suggest that using dynamical data to train machine learning models can support more data efficient and generalizable approaches for materials and molecular design, especially in data-scarce settings.

physics.chem-ph↗