Surya R. Kalidindi

Many modern science and technology applications require the development of new multifunctional materials with advanced, tailorable properties. The large design spaces of materials like high entropy alloys (HEAs) make existing physics-based high-throughput materials design approaches intractable, leading to the adoption of machine learning approaches. In this study, we use the recently developed Voxelized Atomic Structure (VASt) framework to model thermo-mechanical properties of MoNbTaTiVWZr-containing HEAs. We demonstrate that the VASt framework is capable of efficiently and accurately predicting complex properties that describe the material response to perturbations of the atomic structure using only the charge density field of the ground state structure. The resulting VASt structure-property relationship is then used to correlate practical HEA design space parameters with properties indicating strength, ductility, thermal response, and anisotropy. The correlations are used to identify new rules that can help guide the design of new HEAs with high strength, high ductility, and low thermal expansion.

Removed because all co-authors did not approve of submission and posting.

Coarse-grained molecular dynamics (CGMD) presents a computationally efficient alternative to all-atom (AA) simulations for modeling polymer systems, yet the accuracy of its predictions for emergent thermodynamic properties remains a critical challenge. This study builds upon our previous work, which introduced a Bayesian optimization framework for refining Martini3 force field topologies, to validate the transferability of these optimized models to multi-component systems. We demonstrate that CG force fields, optimized solely against single-chain properties (density and radius of gyration), can qualitatively estimate the complex phase behavior of polymer blends. Bayesian-optimized topologies for polyvinylidene fluoride (PVDF), polystyrene (PS), poly(methyl methacrylate) (PMMA), and polyethylene (PE) were evaluated through direct simulation of binary blend morphologies. The framework successfully predicted macroscopic phase separation in the immiscible PS-PMMA system and complete miscibility in the PVDF-PMMA system, with resulting density profiles showing excellent qualitative agreement with experimental observations. Flory-Huggins interaction parameters (χ) estimated from Hildebrand solubility parameters (𝛿) correctly rank relative miscibilities and align qualitatively with literature trends, though quantitative values are systematically overestimated due to limitations of the regular solution approximation. Our results show that optimizing for fundamental properties such as density and radius of gyration (achieving <10% error) implicitly captures the necessary thermodynamic contributions to predict complex, multi-component phase behaviors. This work establishes Bayesian-optimized CGMD as a robust and predictive tool for polymer blend thermodynamics, effectively bridging the gap between computational efficiency and predictive fidelity in materials discovery.

Establishing high-fidelity datasets and leveraging machine learning-driven analysis to uncover composition & process-structure-property (CP-S-P) relationships in γ′-strengthened Ni-based superalloys is essential for accelerating the design of high-performance alloys for demanding thermal environments. This study presents a high-throughput combinatorial framework to investigate γ′ precipitation behavior across compositionally graded IN100–IN625 alloy mixtures fabricated using Directed Energy Deposition. By systematically varying alloy composition, the γ′ evolution was examined across four processing stages: as-deposited (AD), solution heat-treated (SHT), and first -and second-step age hardening (AG1 & AG2), and the characteristics of the produced materials are evaluated in terms of nucleation thresholds, morphology, and mechanical performance. Microstructural analysis reveals that whilst no significant γ′ is observed in AD and SHT conditions, the first discernible γ′ precipitates in cuboidal shapes appeared in AG1 at approximately 62.5 wt% IN100 and at 50 wt% in AG2, as aging time increased. Automated hardness tests and small punch tests (SPT) confirmed that the tensile strength increased with γ′ volume fraction, while ductility decreased. A novelty of our data analyses approach lies in the use of Principal Component Analysis (PCA) to reduce the dimensionality of time-temperature profiles, enabling efficient and compact representation of thermal histories. These PCA features were utilized in exploring different variants of reduced-order composition-process-structure-property (CPSP) models using Gaussian Process Regression (GPR). It was observed that including the microstructural descriptors in these models provided only marginal improvements, indicating that thermal history and composition predominantly governed mechanical properties. This study demonstrates the power of combining AM-based high-throughput experiments with machine learning to accelerate alloy design and deepen understanding of γ′ evolution in Ni-based superalloys. • This study reveals γ′ precipitation thresholds and evolution in AM IN100–IN625 graded alloys. • γ′ onset shifts from 62.5 wt% (AG1) to 50 wt% IN100 (AG2) with extended aging treatments. • Increasing γ′ precursors (Al and Ti) enhance strength but reduces ductility in the alloy system. • Microhardness decreases after SHT but increases during AG due to γ′ nucleation and growth. • PCA + GPR models capture γ′ evolution and mechanical properties, establishing CPSP linkages.

The design of structural&functional materials for specialized applications is being fueled by rapid advancements in materials synthesis, characterization, manufacturing, with sophisticated computational materials modeling frameworks that span a wide spectrum of length&time scales in the mesoscale between atomistic&continuum approaches. This is leading towards a systems-based design methodology that will replace traditional empirical approaches, embracing the principles of the Materials Genome Initiative. However, several gaps remain in this framework as it relates to advanced structural materials:(1) limited availability&access to high-fidelity experimental&computational datasets, (2) lack of co-design of experiments&simulation aimed at computational model validation,(3) lack of on-demand access to verified and validated codes for simulation and for experimental analyses,&(4) limited opportunities for workforce training and educational outreach. These shortcomings stifle major innovations in structural materials design. This paper describes plans for a community-driven research initiative that addresses current gaps based on best-practice recommendations of leaders in mesoscale modeling, experimentation&cyberinfrastructure obtained at an NSF-sponsored workshop dedicated to this topic. The proposal is to create a hub for Mesoscale Experimentation and Simulation co-Operation (hMESO)-that will (I) provide curation and sharing of models, data,&codes, (II) foster co-design of experiments for model validation with systematic uncertainty quantification,&(III) provide a platform for education&workforce development. It will engage experimental&computational experts in mesoscale mechanics and plasticity, along with mathematicians and computer scientists with expertise in algorithms, data science, machine learning,&large-scale cyberinfrastructure initiatives.

The capabilities of additive manufacturing have facilitated the design and production of mechanical metamaterials with diverse unit cell geometries. Establishing linkages between the vast design space of unit cells and their effective mechanical properties is critical for the efficient design and performance evaluation of such metamaterials. However, physics-based simulations of metamaterial unit cells across the entire design space are computationally expensive, necessitating a materials informatics framework to efficiently capture complex structure-property relationships. In this work, principal component analysis of two-point correlation functions is performed to extract the salient features from a large dataset of randomly generated 2D metamaterials. Physics-based simulations are performed using a fast Fourier transform (FFT)-based homogenization approach to efficiently compute the homogenized effective elastic stiffness across the extensive unit cell designs. Subsequently, Gaussian process regression is used to generate reduced-order surrogates, mapping unit cell designs to their homogenized effective elastic constants. It is demonstrated that the adopted workflow enables a high-value low-dimensional representation of the voluminous stochastic metamaterial dataset, facilitating the construction of robust structure-property maps. Finally, an uncertainty-based active learning framework is utilized to train a surrogate model with a significantly smaller number of data points compared to the original full dataset. It is shown that a dataset as small as <a:math xmlns:a="http://www.w3.org/1998/Math/MathML"> <a:mrow> <a:mn>0.61</a:mn> <a:mo>%</a:mo> </a:mrow> </a:math> of the entire dataset is sufficient to generate accurate and robust structure-property maps.

Molecular Dynamics (MD) simulations are vital for predicting the physical and chemical properties of molecular systems across various ensembles. While All-Atom (AA) MD provides high accuracy, its computational cost has spurred the development of Coarse-Grained MD (CGMD), which simplifies molecular structures into representative beads to reduce expense but sacrifice precision. CGMD methods like Martini3, calibrated against experimental data, generalize well across molecular classes but often fail to meet the accuracy demands of domain-specific applications. This work introduces a Bayesian Optimization-based approach to refine Martini3 topologies—specifically the bonded interaction parameters within a given coarse-grained mapping—for specialized applications, ensuring accuracy and efficiency. The resulting optimized CG potential accommodates any degree of polymerization, offering accuracy comparable to AA simulations while retaining the computational speed of CGMD. By bridging the gap between efficiency and accuracy, this method advances multiscale molecular simulations, enabling cost-effective molecular discovery for diverse scientific and technological fields.

Recent advances in Foundation Models for Materials Science are poised to revolutionize the discovery, manufacture, and design of novel materials with tailored properties and responses. Although great strides have been made, successes have been restricted to materials classes where multi-million sample data repositories can be readily curated (e.g., atomistic structures). Unfortunately, for many structural and functional materials (e.g., mesoscale structured metal alloys), such datasets are too costly or prohibitive to construct; instead, datasets are limited to very few examples. To address this challenge, we introduce a novel machine learning approach for learning from hyper-sparse, complex spatial data in scientific domains. Our core contribution is a physics-driven data augmentation scheme that leverages an ensemble of local generative models, trained on as few as five experimental observations, and coordinates them through a novel diversity curation strategy to generate a large-scale, physically diverse dataset. We utilize this framework to construct PolyMicros, the first Foundation Model for polycrystalline materials (a structural material class important across a broad range of industrial and scientific applications). We demonstrate the utility of PolyMicros by zero-shot solving several long standing challenges related to accelerating 3D experimental microscopy. Finally, we make both our models and datasets openly available to the community.

The capabilities of additive manufacturing have facilitated the design and production of mechanical metamaterials with diverse unit cell geometries. Establishing linkages between the vast design space of unit cells and their effective mechanical properties is critical for the efficient design and performance evaluation of such metamaterials. However, physics-based simulations of metamaterial unit cells across the entire design space are computationally expensive, necessitating a materials informatics framework to efficiently capture complex structure-property relationships. In this work, principal component analysis of 2-point correlation functions is performed to extract the salient features from a large dataset of randomly generated 2D metamaterials. Physics-based simulations are performed using a fast Fourier transform (FFT)-based homogenization approach to efficiently compute the homogenized effective elastic stiffness across the extensive unit cell designs. Subsequently, Gaussian process regression is used to generate reduced-order surrogates, mapping unit cell designs to their homogenized effective elastic constant. It is demonstrated that the adopted workflow enables a high-value low-dimensional representation of the voluminous stochastic metamaterial dataset, facilitating the construction of robust structure-property maps. Finally, an uncertainty-based active learning framework is utilized to train a surrogate model with a significantly smaller number of data points compared to the original full dataset. It is shown that a dataset as small as $0.61\%$ of the entire dataset is sufficient to generate accurate and robust structure-property maps.

Decades of progress in manufacturing automation have yielded technologies for highly customizable configurations of shape and materials. High‐mix, low‐volume methods, like additive manufacturing, require significant process optimization to utilize new materials or achieve certain complex geometric targets. Uncovering feasible parameters can require costly trial‐and‐error testing or artificially limiting the design process space. During parameter optimization, intuition often guides professionals who perform experiments, troubleshoot, or decide when to declare an effort untenable. With modern data‐driven tools, previously generated data can be used to create a model representation of intuition that may be used to guide further optimization. This work explores this concept through a case study of direct ink write 3D printing. By presampling a design space of freestanding geometries, we investigate the ability of a system to autonomously optimize various shapes within a target geometric class. The tradeoffs and benefits of presampling with traditional design of experiments and a targeted active learning scheme are examined in comparison to optimizing from no initial data. Leveraging a presampling scheme in broader open‐ended parameterization problems has the potential to accelerate adapting to new materials, enable autonomous systems to self‐define capabilities, and improve development time of future manufacturing methodologies.

In case the distribution of mechanical material properties is inhomogeneous, for example in welded joints, an indentation based measure can be adopted to obtain local estimates. However, the property precision and accuracy is often lacking and a next step to reveal the one-to-one relation between instrumented macroindentation and material stress strain curve was set, proposing semi-analytical model improvements and deploying Monte Carlo based numerical model simulations for verification. Utilizing experimental data for DH36 and S355 steel, high-precision Young’s modulus estimates were obtained within 95 [%] reliability of the uniaxial tensile test reference values and less than a 10 [%] error. For high reliability, the number of indentation tests must be sufficiently large. Contact radius information is at least required to obtain the yield strength and several formulations provide estimates within ∼ 20 [%] error for a large range of material properties. A significantly more accurate value can be obtained if the Young’s modulus is not involved, providing an error within 10 [%]. Introducing a proportional plastic strain offset criterion rather than the well-established constant one to obtain an indentation based yield strength estimate, the strength-to-stiffness ratio dependency was eliminated. Accuracy is improved to a ∼ 20 [%] error range with 95 [%] reliability. Using the material proof strength rather than the yield strength even reduces the 95 [%] reliability error estimate to ∼ 15 [%]. Experimental instrumented macroindentation based yield strength estimates are within 10 [%] error compared to the uniaxial tensile test values. • Proposed novel spherical macroindentation based local mechanical properties model. • Improved indentation based strength-to-stiffness independent yield strength estimate. • Explored precision and accuracy of three contact radius models. • Verified model using Monte Carlo FE simulations. • Validated model using tensile tests and macroindentation tests.

Conventional yield criteria for anisotropic plasticity rely on linear transformations of the stress tensor to map the directional dependence of critical stress tensors at yield onset onto a unit sphere in stress space. These linear transformations are made material specific by a number of anisotropic parameters, which need to be determined by experimental procedures for each material. One drawback of this approach is that these anisotropic parameters cannot be explicitly expressed as functions of the crystallographic texture. Hence, any change in the texture of a material, as it occurs during cold deformation, requires a complete re-parametrization of the yield function. In this work, we present a data-oriented yield criterion based on Support Vector Classification (SVC) that is an explicit function of the crystallographic texture. This texture-dependency is achieved by including the coefficients of the general spherical harmonics (GSH) series expansion of the orientation distribution function (ODF) to the feature space of the machine learning model. The capabilities of the proposed yield criterion are demonstrated by training the model on a dataset containing micromechanical data from over 8000 distinct cubic-orthorhombic textures. The trained SVC combines the efficiency of classical phenomenological models with the flexibility of elaborate CP models. It provides a path to efficient hierarchical materials modeling as the anisotropy of the macroscopic yield onset is explicitly linked to the crystallographic texture.

The design of structural &amp; functional materials for specialized applications is being fueled by rapid advancements in materials synthesis, characterization, manufacturing, with sophisticated computational materials modeling frameworks that span a wide spectrum of length &amp; time scales in the mesoscale between atomistic &amp; continuum approaches. This is leading towards a systems-based design methodology that will replace traditional empirical approaches, embracing the principles of the Materials Genome Initiative. However, several gaps remain in this framework as it relates to advanced structural materials:(1) limited availability &amp; access to high-fidelity experimental &amp; computational datasets, (2) lack of co-design of experiments &amp; simulation aimed at computational model validation,(3) lack of on-demand access to verified and validated codes for simulation and for experimental analyses, &amp; (4) limited opportunities for workforce training and educational outreach. These shortcomings stifle major innovations in structural materials design. This paper describes plans for a community-driven research initiative that addresses current gaps based on best-practice recommendations of leaders in mesoscale modeling, experimentation &amp; cyberinfrastructure obtained at an NSF-sponsored workshop dedicated to this topic. The proposal is to create a hub for Mesoscale Experimentation and Simulation co-Operation (hMESO)-that will (I) provide curation and sharing of models, data, &amp; codes, (II) foster co-design of experiments for model validation with systematic uncertainty quantification, &amp; (III) provide a platform for education &amp; workforce development. It will engage experimental &amp; computational experts in mesoscale mechanics and plasticity, along with mathematicians and computer scientists with expertise in algorithms, data science, machine learning, &amp; large-scale cyberinfrastructure initiatives.

Learning to transform conditional probability densities over time is a fundamental challenge spanning probabilistic modeling and the natural sciences. This task is paramount when forecasting the evolution of stochastic nonlinear dynamical systems in biological and physical domains. While flow-based models can predict the temporal evolution of probability distributions, existing approaches often assume discrete conditioning with samples that are paired across time, limiting their scientific applicability where frequently only sparse data with unpaired continuous conditioning is available. We propose Conditional Variable Flow Matching (CVFM), a framework for learning flows transforming conditional distributions with amortization across the continuous space of conditional densities. CVFM addresses the high-variance instability of prior methods by jointly sampling flows over state and conditioning variables, utilizing a conditioning mismatch kernel alongside a conditional Wasserstein distance to reweight the conditional optimal transport objective. Collectively, these advances allow for learning dynamics from sparse unpaired measurements of state-condition across time. We evaluate CVFM on conditional mapping benchmarks and a case study modeling the temporal evolution of materials internal structure during manufacturing processes, observing improved performance and convergence characteristics over existing conditional variants. Code is available at https://github.com/agenerale/conditional-variable-flow-matching.

Engineering problems frequently require solution of governing equations with spatially-varying discontinuous coefficients. Even for linear elliptic problems, mapping large ensembles of coefficient fields to solutions can become a major computational bottleneck using traditional numerical solvers. Furthermore, machine learning methods such as neural operators struggle to fit these maps due to sharp transitions and high contrast in the coefficient fields and a scarcity of informative training data. In this work, we focus on a canonical problem in computational mechanics: prediction of local elastic deformation fields over heterogeneous material structures subjected to periodic boundary conditions. We construct a hybrid approximation for the coefficient-to-solution map using a Thermodynamically-informed Iterative Neural Operator (TherINO). Rather than using coefficient fields as direct inputs and iterating over a learned latent space, we employ thermodynamic encodings -- drawn from the constitutive equations -- and iterate over the solution space itself. Through an extensive series of case studies, we elucidate the advantages of these design choices in terms of efficiency, accuracy, and flexibility. We also analyze the model's stability and extrapolation properties on out-of-distribution coefficient fields and demonstrate an improved speed-accuracy tradeoff for predicting elastic quantities of interest.

Algorithmic materials discovery is a multidisciplinary domain that integrates insights from specialists in alloy design, synthesis, characterization, experimental methodologies, computational modeling, and optimization. Central to this effort is a robust data management system paired with an interactive work platform. This platform should empower users to not only access others’ data but also integrate their analyses, paving the way for sophisticated data pipelines. To realize this vision, there is a need for an integrative collaboration platform, streamlined data sharing and analysis tools, and efficient communication channels. Such a collaborative mechanism should transcend geographical barriers, facilitating remote interaction and fostering a challenge-response dynamic. To further enhance precision and interoperability in this multifaceted research landscape, we must explore innovative ways to refine these processes and improve the integration of expertise and data across diverse domains. In this paper, we present our ongoing efforts in addressing the critical challenges related to an accelerated materials discovery framework as a part of the High-Throughput Materials Discovery for Extreme Conditions (HTMDEC) Initiative. Our BIRDSHOT (Batch-wise Improvement in Reduced Materials Design Space using a Holistic Optimization Technique) Center has successfully harnessed various tools and strategies, including the utilization of cloud-based storage, a standardized sample naming convention, a structured file system, the implementation of sample travelers, a robust sample tracking method, and the incorporation of knowledge graphs for efficient data management. Additionally, we present the development of a data collection platform, reinforcing seamless collaboration among our team members. In summary, this paper provides an illustration and insight into the various elements of an efficient and effective workflow within an accelerated materials discovery framework while highlighting the dynamic and adaptable nature of the data management tools and sharing platforms.

We develop an active workflow for calibrating microstructure–property relationships when a large dataset of microstructures is available, but the cost associated with evaluating the properties associated is high.