arXiv subjects

Saaketh Desai

Publications and source records attributed to Saaketh Desai.

15 recordsLinked to original sources

arXiv2026

Materials Behavior as Mechanism Ensembles: A Probabilistic Framework for Emergent Behaviors

Materials behavior is often treated as a deterministic mapping from structure to properties, yet many important phenomena emerge from the conditional activation of multiple mechanisms across scales. This is especially evident in fatigue of metals, where crack growth is typically modeled as monotonic and irreversible process, despite evidence that local microstructure, loading history, and competing unit processes can shift the balance among propagation, arrest, and self-healing. Here we present a probabilistic framework that describes materials behavior as an ensemble of constituent mechanisms whose activation, interaction, and evolution determine emergent outcomes. The framework connects mechanism activation, state evolution, and macroscopic observables in a probabilistic way. In the case of fatigue crack propagation, it reframes damage tolerance as an inference problem over mechanism competition and provides a basis for integrating multiscale simulation, multimodal characterization, and machine learning. The same logic extends to other physical and chemical systems suggesting a portable framework for any system in which emergent behavior reflects mechanism competition under changing conditions. The broader ambition of this perspective review is a shift from correlating structure and performance after the fact to identifying, in advance, the conditions that make desired emergent behavior probable.

cond-mat.mtrl-sci↗

arXiv2026

LLM-AutoSciLab: Closed-Loop Scientific Discovery via Active Experimentation with LLMs

Scientific discovery is a closed-loop process in which hypotheses guide data acquisition and observations refine the hypothesis space. Yet most approaches reduce discovery to supervised learning over fixed datasets, where limited observations can support multiple plausible mechanisms that fit locally but fail to generalize. Thus, the key challenge is selecting informative observations to resolve uncertainty, shifting the focus from static inference to adaptive data acquisition. To address this, we propose LLM-AutoSciLab, a closed-loop framework that couples hypothesis generation with hypothesis-conditioned experiment selection and mechanism refinement. Rather than fitting models to passively collected data, LLM-AutoSciLab iteratively proposes plausible hypotheses, selects informative experiments to distinguish or refine them, and updates its state using the resulting evidence. To evaluate dynamic, closed-loop scientific discovery with active data acquisition, we introduce ActiveSciBench, comprising two datasets: ActiveSciBench-Chem with 57 enzyme-kinetics tasks and ActiveSciBench-GRN with 45 gene-regulatory-network tasks. These datasets model discovery as a budget-constrained process requiring adaptive experiment design, variable selection, and recovery of true mechanisms. Across NewtonBench, ActiveSciBench-Chem, and ActiveSciBench-GRN, LLM-AutoSciLab outperforms prior methods, achieving 67.6% and 35.1% symbolic accuracy on NewtonBench and ActiveSciBench-Chem, respectively, and 31.1% exact graph recovery on ActiveSciBench-GRN. Moreover, hypothesis-guided experimentation is 2-5x more sample-efficient than the strongest competing baselines. Code and data are available at: https://github.com/scientific-discovery/LLM-AutoSciLab

cs.LG↗

arXiv2026

LLEMA: Evolutionary Search with LLMs for Multi-Objective Materials Discovery

Materials discovery requires navigating vast chemical and structural spaces while satisfying multiple, often conflicting, objectives. We present LLM-guided Evolution for MAterials discovery (LLEMA), a unified framework that couples the scientific knowledge embedded in large language models with chemistry-informed evolutionary rules and memory-based refinement. At each iteration, an LLM proposes crystallographically specified candidates under explicit property constraints; a surrogate-augmented oracle estimates physicochemical properties; and a multi-objective scorer updates success/failure memories to guide subsequent generations. Evaluated on 14 realistic tasks that span electronics, energy, coatings, optics, and aerospace, LLEMA discovers candidates that are chemically plausible, thermodynamically stable, and property-aligned, achieving higher hit rates and improved Pareto front quality relative to generative and LLM-only baselines. Ablation studies confirm the importance of rule-guided generation, memory-based refinement, and surrogate prediction. By enforcing synthesizability and multi-objective trade-offs, LLEMA provides a principled approach to accelerating practical materials discovery. Project website: https://scientific-discovery.github.io/llema-project/

cs.LG↗

arXiv2026

Boltzmann Reinforcement Learning for Noise resilience in Analog Ising Machines

Analog Ising machines (AIMs) have emerged as a promising paradigm for combinatorial optimization, utilizing physical dynamics to solve Ising problems with high energy efficiency. However, the performance of traditional optimization and sampling algorithms on these platforms is often limited by inherent measurement noise. We introduce BRAIN (Boltzmann Reinforcement for Analog Ising Networks), a distribution learning framework that utilizes variational reinforcement learning to approximate the Boltzmann distribution. By shifting from state-by-state sampling to aggregating information across multiple noisy measurements, BRAIN is resilient to Gaussian noise characteristic of AIMs. We evaluate BRAIN across diverse combinatorial topologies, including the Curie-Weiss and 2D nearest-neighbor Ising systems. We find that under realistic 3\% Gaussian measurement noise, BRAIN maintains 98\% ground state fidelity, whereas Markov Chain Monte Carlo (MCMC) methods degrade to 51\% fidelity. Furthermore, BRAIN reaches the MCMC-equivalent solution up to 192x faster under these conditions. BRAIN exhibits $\mathcal{O}(N^{1.55})$ scaling up to 65,536 spins and maintains robustness against severe measurement uncertainty up to 40\%. Beyond ground state optimization, BRAIN accurately captures thermodynamic phase transitions and metastable states, providing a scalable and noise-resilient method for utilizing analog computing architectures in complex optimizations.

cs.LG↗

arXiv2025

Learning interpretable surface elasticity properties from bulk properties

Surface elasticity is central to understanding the mechanics and stability of surfaces and interfaces. It is characterized by quantities such as surface tension, residual surface stress, and surface stiffness, however their analytical expressions are typically difficult to derive from atomistic data, and depend strongly on modeling choices. This work presents a neural network-based equation learner which combines customized activation functions and connection-based pruning to discover parsimonious, closed-form equations for surface elasticity from atomistic simulations. Applying the method to seven face centered cubic (FCC) metals, our equation learner uncovers interpretable equations that describe both low-Miller index and high-Miller index surface properties, capturing long-tail property distributions accurately. The discovered expressions are decoupled into two components: a universal, geometry-driven orientation function, and material-specific baseline coefficients. We find that lower-order properties such as surface tension are fundamentally geometry dependent, while higher-order properties such as surface stress and elasticity show more complex geometry and material dependence. We also relate material dependent coefficients to bulk properties, forming a clear map from bulk material properties to surface elasticity. Overall, this approach demonstrates that interpretable neurosymbolic machine learning can bridge the gap between atomistic simulations and physical laws, enabling the discovery of generalizable structure-property relationships for materials science phenomena such as surface elasticity.

cond-mat.mtrl-sci↗

arXiv2025

A broadband silicon photonic-integrated-circuit based RF spectrum analyzer with 10 MHz spectral resolution

Designing miniaturized optical spectrometers is an increasingly active area of research as spectrometers are crucial components for a wide range of applications including chemical and material analysis, medical diagnostics, classical and quantum sensing, characterization of light sources, and radio frequency (RF) spectrum analysis. Among these applications, designing on-chip spectrometers for RF spectrum analysis is particularly challenging since it requires combining high resolution and large bandwidth with a fast update rate. Existing chip-scale spectrometers cannot achieve the resolution required for RF analysis, setting aside challenges in maintaining a fast update rate and broad bandwidth. In this work, we address these challenges by introducing a silicon photonic integrated circuit (PIC)-based RF spectrum analyzer that combines an ultra-high-resolution speckle spectrometer with an interferometric RF-to-optical encoding scheme. The PIC-based speckle spectrometer uses a path-mismatched multimode interferometer with inverse designed splitters to compensate for waveguide loss, enabling a record-high resolution of 100 MHz (0.8 pm at a wavelength of 1550 nm). To further improve the resolution of the overall RF spectrum analyzer, we modify the RF-to-optical encoding scheme by directing the RF signal through a path mismatched interferometer and encoding the outputs of the RF interferometer on separate optical carriers. This further reduces the RF spectral correlation width of the combined system, enabling the RF spectrum analyzer to resolve RF tones separated by 10 MHz across a bandwidth of 10 GHz. Since this approach operates as a single-shot spectrometer, it can support fast update rates, providing a path to compact, persistent wideband RF spectrum analysis.

physics.optics↗

arXiv2024

AutoSciLab: A Self-Driving Laboratory For Interpretable Scientific Discovery

Advances in robotic control and sensing have propelled the rise of automated scientific laboratories capable of high-throughput experiments. However, automated scientific laboratories are currently limited by human intuition in their ability to efficiently design and interpret experiments in high-dimensional spaces, throttling scientific discovery. We present AutoSciLab, a machine learning framework for driving autonomous scientific experiments, forming a surrogate researcher purposed for scientific discovery in high-dimensional spaces. AutoSciLab autonomously follows the scientific method in four steps: (i) generating high-dimensional experiments (x \in R^D) using a variational autoencoder (ii) selecting optimal experiments by forming hypotheses using active learning (iii) distilling the experimental results to discover relevant low-dimensional latent variables (z \in R^d, with d << D) with a 'directional autoencoder' and (iv) learning a human interpretable equation connecting the discovered latent variables with a quantity of interest (y = f(z)), using a neural network equation learner. We validate the generalizability of AutoSciLab by rediscovering a) the principles of projectile motion and b) the phase transitions within the spin-states of the Ising model (NP-hard problem). Applying our framework to an open-ended nanophotonics challenge, AutoSciLab uncovers a fundamentally novel method for directing incoherent light emission that surpasses the current state-of-the-art (Iyer et al. 2023b, 2020).

cs.LG↗

arXiv2024

Self-driving lab discovers principles for steering spontaneous emission

We developed an autonomous experimentation platform to accelerate interpretable scientific discovery in ultrafast nanophotonics, targeting a novel method to steer spontaneous emission from reconfigurable semiconductor metasurfaces. Controlling spontaneous emission is crucial for clean-energy solutions in illumination, thermal radiation engineering, and remote sensing. Despite the potential of reconfigurable semiconductor metasurfaces with embedded sources for spatiotemporal control, achieving arbitrary far-field control remains challenging. Here, we present a self-driving lab (SDL) platform that addresses this challenge by discovering the governing equations for predicting the far-field emission profile from light-emitting metasurfaces. We discover that both the spatial gradient (grating-like) and the curvature (lens-like) of the local refractive index are key factors in steering spontaneous emission. The SDL employs a machine-learning framework comprising: (1) a variational autoencoder for generating complex spatial refractive index profiles, (2) an active learning agent for guiding experiments with real-time closed-loop feedback, and (3) a neural network-based equation learner to uncover structure-property relationships. The SDL demonstrated a four-fold enhancement in peak emission directivity (up to 77%) over a 72° field of view within ~300 experiments. Our findings reveal that combinations of positive gratings and lenses are as effective as negative lenses and gratings for all emission angles, offering a novel strategy for controlling spontaneous emission beyond conventional Fourier optics.

physics.optics↗

arXiv2023

Rethinking materials simulations: Blending direct numerical simulations with neural operators

Direct numerical simulations (DNS) are accurate but computationally expensive for predicting materials evolution across timescales, due to the complexity of the underlying evolution equations, the nature of multiscale spatio-temporal interactions, and the need to reach long-time integration. We develop a new method that blends numerical solvers with neural operators to accelerate such simulations. This methodology is based on the integration of a community numerical solver with a U-Net neural operator, enhanced by a temporal-conditioning mechanism that enables accurate extrapolation and efficient time-to-solution predictions of the dynamics. We demonstrate the effectiveness of this framework on simulations of microstructure evolution during physical vapor deposition modeled via the phase-field method. Such simulations exhibit high spatial gradients due to the co-evolution of different material phases with simultaneous slow and fast materials dynamics. We establish accurate extrapolation of the coupled solver with up to 16.5$\times$ speed-up compared to DNS. This methodology is generalizable to a broad range of evolutionary models, from solid mechanics, to fluid dynamics, geophysics, climate, and more.

cs.LG↗

arXiv2021

Active learning and molecular dynamics simulations to find high melting temperature alloys

Active learning (AL) can drastically accelerate materials discovery; its power has been shown in various classes of materials and target properties. Prior efforts have used machine learning models for the optimal selection of physical experiments or physics-based simulations. However, the latter efforts have been mostly limited to the use of electronic structure calculations and properties that can be obtained at the unit cell level and with negligible noise. We couple AL with molecular dynamics simulations to identify multiple principal component alloys (MPCAs) with high melting temperatures. Building on cloud computing services through nanoHUB, we present a fully autonomous workflow for the efficient exploration of the high dimensional compositional space of MPCAs. We characterize how uncertainties arising from the stochastic nature of the simulations and the acquisition functions used to select simulations affect the convergence of the approach. Interestingly, we find that relatively short simulations with significant uncertainties can be used to efficiently find the desired alloys as the random forest models used for AL average out fluctuations.

cond-mat.mtrl-sci↗

arXiv2021

Sim2Ls: FAIR simulation workflows and data

Just like the scientific data they generate, simulation workflows for research should be findable, accessible, interoperable, and reusable (FAIR). However, while significant progress has been made towards FAIR data, the majority of science and engineering workflows used in research remain poorly documented and often unavailable, involving ad hoc scripts and manual steps, hindering reproducibility and stifling progress. We introduce Sim2Ls (pronounced simtools) and the Sim2L Python library that allow developers to create and share end-to-end computational workflows with well-defined and verified inputs and outputs. The Sim2L library makes Sim2Ls, their requirements, and their services discoverable, verifies inputs and outputs, and automatically stores results in a globally-accessible simulation cache and results database. This simulation ecosystem is available in nanoHUB, an open platform that also provides publication services for Sim2Ls, a computational environment for developers and users, and the hardware to execute runs and store results at no cost. We exemplify the use of Sim2Ls using two applications and discuss best practices towards FAIR simulation workflows and associated data.

cs.DL↗

arXiv2020

Parsimonious neural networks learn interpretable physical laws

Machine learning is playing an increasing role in the physical sciences and significant progress has been made towards embedding domain knowledge into models. Less explored is its use to discover interpretable physical laws from data. We propose parsimonious neural networks (PNNs) that combine neural networks with evolutionary optimization to find models that balance accuracy with parsimony. The power and versatility of the approach is demonstrated by developing models for classical mechanics and to predict the melting temperature of materials from fundamental properties. In the first example, the resulting PNNs are easily interpretable as Newton's second law, expressed as a non-trivial time integrator that exhibits time-reversibility and conserves energy, where the parsimony is critical to extract underlying symmetries from the data. In the second case, the PNNs not only find the celebrated Lindemann melting law, but also new relationships that outperform it in the pareto sense of parsimony vs. accuracy.

cs.LG↗

arXiv2020

Tuning martensitic transformations via coherent second phases in nanolaminates using free energy landscape engineering

We explore the possibilities and limitations of using a coherent second phase to engineer the thermo-mechanical properties of a martensitic alloy by modifying the underlying free energy landscape that controls the transformation. We use molecular dynamics simulations of a model atomistic system where the properties of a coherent, nanoscale second phase can be varied systematically. With a base martensitic material that undergoes a temperature-induced transformation from a cubic austenite to a monoclinic martensite, the simulations show a significant ability to engineer the transformation temperatures, from a ~50% reduction to a ~200% increase, with 50 at. % of the cubic second phase. We establish correlations between the properties of the second phase and the transformation characteristics and microstructure, via the free energy landscape of the two-phase systems. Coherency stresses have a strong influence on the martensitic variants observed and can even cause the non-martensitic second phase to undergo a transformation. Reducing the stiffness of second phase increases the transformation strain and modifies the martensitic microstructure, increasing the volume fraction of the transformed material. This increase in transformation strain is accompanied by a significant increase in the Af and thermal hysteresis, while the Ms remains unaltered. Our findings on the tunability of martensitic transformations can be used for informed searches of second phases to achieve desired material properties, such as achieving room temperature, lightweight shape memory alloys.

cond-mat.mtrl-sci↗

arXiv2020

Implementing a neural network interatomic model with performance portability for emerging exascale architectures

The two main thrusts of computational science are more accurate predictions and faster calculations; to this end, the zeitgeist in molecular dynamics (MD) simulations is pursuing machine learned and data driven interatomic models, e.g. neural network potentials, and novel hardware architectures, e.g. GPUs. Current implementations of neural network potentials are orders of magnitude slower than traditional interatomic models and while looming exascale computing offers the ability to run large, accurate simulations with these models, achieving portable performance for MD with new and varied exascale hardware requires rethinking traditional algorithms, using novel data structures, and library solutions. We re-implement a neural network interatomic model in CabanaMD, an MD proxy application, built on libraries developed for performance portability. Our implementation shows significantly improved on-node scaling in this complex kernel as compared to a current LAMMPS implementation, across both strong and weak scaling. Our single-source solution results in improved performance in many cases, with thread-scalability enabling simulations up to 21 million atoms on a single CPU node and 2 million atoms on a single GPU. We also explore parallelism and data layout choices (using flexible data structures called AoSoAs) and their effect on performance, seeing up to ~25% and ~10% improvements in performance on a GPU simply by choosing the right level of parallelism and data layout, respectively.

physics.comp-ph↗

arXiv2017

Molecular Modeling of the Microstructure Evolution during the Carbonization of PAN-Based Carbon Fibers

Development of high strength carbon fibers (CFs) requires an understanding of the relationship between the processing conditions, microstructure and resulting properties. We developed a molecular model that combines kinetic Monte Carlo (KMC) and molecular dynamics (MD) techniques to predict the microstructure evolution during the carbonization process of carbon fiber manufacturing. The model accurately predicts the cross-sectional microstructure of carbon fibers, predicting features such as graphitic sheets and hairpin structures that have been observed experimentally. We predict the transverse modulus of the resulting fibers and find that the modulus is slightly lower than experimental values, but is up to an order of magnitude lower than ideal graphite. We attribute this to the perfect longitudinal texture of our simulated structures, as well as the chain sliding mechanism that governs the deformation of the fibers, rather than the van der Waals interaction that governs the modulus for graphite. We also observe that high reaction rates result in porous structures that have lower moduli.

cond-mat.mtrl-sci↗