arXiv subjects

Santosh Adhikari

Publications and source records attributed to Santosh Adhikari.

At least 19 recordsLinked to original sources

arXiv2026

Agentic AI for Density-Functional Development: Revisiting r2SCAN

We demonstrate physics-constrained agentic development of a meta-generalized gradient approximation (meta-GGA) functional using a large language model (LLM) to assist the search and optimization of a band-gap-oriented revision of r2SCAN. No real bonded systems were fitted, preserving r2SCAN's nonempirical philosophy. Across the finalist set, band gaps and several molecular subsets improve relative to r2SCAN; the top finalist, r2SCAN+, reduces the band-gap MAE on a benchmark comprising 24 solids from 1.26 to 0.96 eV and the aggregate MAE on 329 molecular properties from 4.79 to 4.42 kcal/mol. We first curated 78 exchange and 90 correlation candidate correction terms from r2SCAN's dimensionless ingredients, spanning polynomial terms through third degree, exponentials, exponentially damped products, and ratios. Allowing each candidate to combine one to three correction terms from the exchange catalog, the correlation catalog, or both yields about 8 x 10^5 distinct forms, making exhaustive high-throughput screening impractical. We defined the search criteria for the LLM agent using r2SCAN's exact constraints, physical norms, and the targeted iso-orbital derivative response. The agent then combined these criteria with its pretrained knowledge and accumulated search feedback to propose and refine sparse forms, prioritizing terms tied to the iso-orbital response; a second LLM critic screened proposals before deterministic verification. Compared with uniform random search, the workflow learned from prior evaluations, incurred far fewer downstream rejections (0.6% versus 24.6%), and located stronger high-response candidates: 51 agentic candidates exceeded the best random-search response of 1.263, with the overall best reaching 1.331. These results show that agentic search can support density-functional development when flexible hypothesis generation is coupled to automated physical verification.

physics.chem-ph↗

arXiv2026

Guided Synthesis of EMT Zeolites by Machine Learning

Zeolites are microporous crystalline materials with diverse frameworks, widely used in industrial applications such as petroleum refining and molecular separation. Unlike most zeolites, EMT can be synthesized under mild conditions (at low temperatures and without the use of organic structure-directing agents), making it attractive for cost-effective and environmentally sustainable production. However, the specific synthesis conditions that selectively produce EMT rather than similar frameworks like FAU are not yet well established. In this work, we develop machine learning (ML) models to guide the discovery of synthesis conditions for EMT zeolites. Our dataset comprises 174 experimental synthesis attempts, recording reaction time, temperature, silica and alumina sources, Si/Al stoichiometric ratio, and other synthesis parameters. We apply both classical ML methods and pretrained foundation models to predict zeolite framework outcomes from these synthesis parameters. Feature importance analysis identifies critical parameters for EMT formation, validating known synthesis principles. Leveraging the ML models, we explore the synthesis space and identify six promising new conditions for EMT formation. Experimental validation confirms EMT crystallization in five cases, including two with Si/Al stoichiometric ratios outside the training dataset's range. Evaluation on independent literature-reported synthesis conditions further demonstrates the generalizability of the model. This work demonstrates a data-driven approach to accelerating zeolite synthesis, closing the loop between ML prediction and experimental validation.

cond-mat.mtrl-sci↗

arXiv2026

MLIPilot: LLM-Driven Auto-Research for Machine-Learned Interatomic Potentials

Constructing production-quality machine-learned interatomic potentials (MLIPs) requires balancing accuracy, dynamical stability, and computational throughput under constraints that are not captured by a single training loss. We introduce MLIPilot, an auto-research framework in which tool-calling large language models propose hypotheses, edit MLIP training code, launch HPC jobs, and accept or revert changes using a fixed, physically constrained scorecard. We evaluate MLIPilot on MACE potential optimization using both commercial and open-weight LLM agents, including GPT-5.5, GPT-4.1, Mistral-24B, and Qwen3-32B. The benchmarks span molecular and periodic settings: a QM7-derived dataset for which we generated B3LYP/6-31G(d) energies and forces, and a Cu EMT dataset with periodic copper supercells labeled by ASE's Effective Medium Theory calculator. Across these benchmarks, the strongest agents move initially constraint-violating baselines to accepted models by discovering useful training strategies, including output normalization, loss-function changes, progressive training schedules, and model-capacity adjustments. These results suggest that LLM agents can serve as autonomous operators for scientific machine-learning workflows when their search is constrained by domain-specific validation criteria, shifting part of MLIP development from manual trial-and-error toward auditable, automated experimentation.

physics.chem-ph↗

arXiv2025

Cartesian atomic moment machine learning interatomic potentials

Machine learning interatomic potentials (MLIPs) have substantially advanced atomistic simulations in materials science and chemistry by balancing accuracy and computational efficiency. While leading MLIPs rely on representing atomic environments using spherical tensors, Cartesian representations offer potential advantages in simplicity and efficiency. Here, we introduce the Cartesian Atomic Moment Potential (CAMP), an approach to building MLIPs entirely in Cartesian space. CAMP constructs atomic moment tensors from neighboring atoms and employs tensor products to incorporate higher body-order interactions, providing a complete description of local atomic environments. Integrated into a graph neural network (GNN) framework, CAMP enables physically motivated, systematically improvable potentials. The model demonstrates excellent performance across diverse systems, including periodic structures, small organic molecules, and two-dimensional materials, achieving accuracy, efficiency, and stability in molecular dynamics simulations that rival or surpass current leading models. CAMP provides a powerful tool for atomistic simulations to accelerate materials understanding and discovery.

cond-mat.mtrl-sci↗

arXiv2024

Uncertainty Quantification and Propagation in Atomistic Machine Learning

Machine learning (ML) offers promising new approaches to tackle complex problems and has been increasingly adopted in chemical and materials sciences. Broadly speaking, ML models employ generic mathematical functions and attempt to learn essential physics and chemistry from a large amount of data. Consequently, because of the limited physical or chemical principles in the functional form, the reliability of the predictions is oftentimes not guaranteed, particularly for data far out of distribution. It is critical to quantify the uncertainty in model predictions and understand how the uncertainty propagates to downstream chemical and materials applications. Herein, we review existing uncertainty quantification (UQ) and uncertainty propagation (UP) methods for atomistic ML under a united framework of probabilistic modeling. We first categorize the UQ methods, with the aim to elucidate the similarities and differences between them. We also discuss performance metrics to evaluate the accuracy, precision, calibration, and efficiency of the UQ methods and techniques for model recalibration. With these metrics, we survey existing benchmark studies of the UQ methods using molecular and materials datasets. Furthermore, we discuss UP methods to propagate the uncertainty obtained from ML models in widely used materials and chemical simulation techniques, such as molecular dynamics and microkinetic modeling. We also provide remarks on the challenges and future opportunities of UQ and UP in atomistic ML.

cond-mat.mtrl-sci↗

arXiv2024

VTON-IT: Virtual Try-On using Image Translation

Virtual Try-On (trying clothes virtually) is a promising application of the Generative Adversarial Network (GAN). However, it is an arduous task to transfer the desired clothing item onto the corresponding regions of a human body because of varying body size, pose, and occlusions like hair and overlapped clothes. In this paper, we try to produce photo-realistic translated images through semantic segmentation and a generative adversarial architecture-based image translation network. We present a novel image-based Virtual Try-On application VTON-IT that takes an RGB image, segments desired body part, and overlays target cloth over the segmented body region. Most state-of-the-art GAN-based Virtual Try-On applications produce unaligned pixelated synthesis images on real-life test images. However, our approach generates high-resolution natural images with detailed textures on such variant images.

cs.CV↗

arXiv2024

A Review of Sustainable Practices in Road Freight Transport

Sustainable road freight transport becomes indispensable in the field of transportation and logistics. The new technological change, the environmental impacts, and social responsibility laid freight road transport in front of various challenges, which makes the sustainable practices a vital solution in the sector. This paper aims to provide a theoretical research findings in sustainable road freight transport. The methodology discusses the road freight transport sustainability indicators among the literature studies realized in different countries in the world. The review analysis the studies and practical applications from various countries. The result exposes that the sustainability dimensions such as economic, social, environment was discussed in different cases, which prove the efforts of many countries to reduce environmental impact, improve economic efficiency, support social well-being, and expand technological innovations to achieve a sustainable transport system.

physics.soc-ph↗

arXiv2024

Leveraging Domain Adaptation for Accurate Machine Learning Predictions of New Halide Perovskites

We combine graph neural networks (GNN) with an inexpensive and reliable structure generation approach based on the bond-valence method (BVM) to train accurate machine learning models for screening 222,960 halide perovskites using statistical estimates of the DFT/PBE formation energy (Ef), and the PBE and HSE band gaps (Eg). The GNNs were fined tuned using domain adaptation (DA) from a source model, which yields a factor of 1.8 times improvement in Ef and 1.2 - 1.35 times improvement in HSE Eg compared to direct training (i.e., without DA). Using these two ML models, 48 compounds were identified out of 222,960 candidates as both stable and that have an HSE Eg that is relevant for photovoltaic applications. For this subset, only 8 have been reported to date, indicating that 40 compounds remain unexplored to the best of our knowledge and therefore offer opportunities for potential experimental examination.

cond-mat.mtrl-sci↗

arXiv2023

Performance of ChatGPT on USMLE: Unlocking the Potential of Large Language Models for AI-Assisted Medical Education

Artificial intelligence is gaining traction in more ways than ever before. The popularity of language models and AI-based businesses has soared since ChatGPT was made available to the general public via OpenAI. It is becoming increasingly common for people to use ChatGPT both professionally and personally. Considering the widespread use of ChatGPT and the reliance people place on it, this study determined how reliable ChatGPT can be for answering complex medical and clinical questions. Harvard University gross anatomy along with the United States Medical Licensing Examination (USMLE) questionnaire were used to accomplish the objective. The paper evaluated the obtained results using a 2-way ANOVA and posthoc analysis. Both showed systematic covariation between format and prompt. Furthermore, the physician adjudicators independently rated the outcome's accuracy, concordance, and insight. As a result of the analysis, ChatGPT-generated answers were found to be more context-oriented and represented a better model for deductive reasoning than regular Google search results. Furthermore, ChatGPT obtained 58.8% on logical questions and 60% on ethical questions. This means that the ChatGPT is approaching the passing range for logical questions and has crossed the threshold for ethical questions. The paper believes ChatGPT and other language learning models can be invaluable tools for e-learners; however, the study suggests that there is still room to improve their accuracy. In order to improve ChatGPT's performance in the future, further research is needed to better understand how it can answer different types of questions.

cs.CY↗

arXiv2023

Interpretable machine learning to understand the performance of semi local density functionals for materials thermochemistry

This study investigates the use of machine learning (ML) to correct the enthalpy of formation (Hf) from two separate DFT functionals, PBE and SCAN, to the experimental Hf across 1011 solid-state compounds. The ML model uses a set of 25 properties that characterize the electronic structure as calculated using PBE and SCAN. The ML model significantly decreases the error in PBE-calculated Hf values from an mean absolute error (MAE) of 195 meV/atom to an MAE = 80 meV/atom when compared to the experiment. For PBE, the PDP+GAM analysis shows compounds with a high ionicity (I), i.e., I>0.22, have errors in Hf that are twice as large as compounds having I < 0.22 (246 meV/atom compared to 113 meV/atom). Conversely, no analogous trend is observed for SCAN-calculated Hfs, which explains why the ML model for PBE can more easily correct the systematic error in calculated Hfs for PBE but not for SCAN. Although the literature suggests PBE is reliable for intermetallics but less so for oxides and halides, our analysis reveals intermetallics pose a challenge for PBE only when the charge transfer is significant (I >0.22). Meanwhile, oxides and halides may be described accurately by PBE for systems in which charge transfer is relatively low (I < 0.22).

cond-mat.mtrl-sci↗

arXiv2021

Self-interaction corrected Kohn-Sham effective potentials using the density-consistent effective potential method

Density functional theory (DFT) and beyond-DFT methods are often used in combination with photoelectron spectroscopy to obtain physical insights into the electronic structure of molecules and solids. The Kohn-Sham eigenvalues are not electron removal energies except for the highest occupied orbital. The eigenvalues of the highest occupied molecular orbitals often underestimate the electron removal or ionization energies due to the self-interaction (SI) errors in approximate density functionals. In this work, we adapt and implement the density-consistent effective potential(DCEP) method of Kohut, Ryabinkin, and Staroverov to obtain SI corrected local effective potentials from the SI corrected Fermi-Löwdin orbitals and density in the FLOSIC scheme. The implementation is used to obtain the density of states (photoelectron spectra) and HOMO-LUMO gaps for a set of molecules and polyacenes. Good agreement with experimental values is obtained compared to a range of SI uncorrected density functional approximations.

physics.chem-ph↗

arXiv2021

Opening band gaps of low-dimensional materials at the meta-GGA level of density functional approximations

The recent TASK meta-GGA density functional [Phys. Rev. Research, 1, 033082 (2019)] is constructed with an enhanced nonlocality in the generalized Kohn-Sham scheme, and therefore harbors great opportunities for band gap prediction. Although this approximation was found to yield excellent band gaps of bulk solids, this accuracy cannot be straightforwardly transferred to low-dimensional materials. The reduced screening of these materials results in larger band gaps compared to their bulk counterparts, as an additional barrier to overcome. In this work, we demonstrate how the alteration of exact physical constraints in this functional affects the band gaps of monolayers and nanoribbons, and present accurate band gaps competing with the HSE06 approximation. In order to achieve this goal, we have modified the TASK functional (a) by changing the tight upper-bound for one or two-electron systems ($h_X^0$) from 1.174 to 1.29 (b) by changing the limit of interpolation function $f_X (α\rightarrow \infty$) of the TASK functional that interpolates the exchange enhancement factor $F_X (s,α)$ from $α=$ 0 to 1. The resulting modified TASK (mTASK) was tested for various materials from 3D to 2D to 1D (nanoribbons), and was compared with the results of the higher-level hybrid functional HSE06 or with the G$_0$W$_0$ approximation within many-body perturbation theory. We find that mTASK greatly improves the band gaps and band structures of 2D and 1D systems, without significantly affecting the accuracy of the original TASK for the bulk 3D materials, when compared to the PBE-GGA and SCAN meta-GGA. We further demonstrate the applicability of mTASK by assessing the band structures of TMD nanoribbons with respect to various bending curvatures.

cond-mat.mtrl-sci↗

arXiv2021

Resolving the structure-energy dilemma at organic-inorganic interfaces: Adsorption of benzene, thiophene, and xenon over coinage metal surfaces

Semilocal (SL) density functional approximations (DFAs) are widely applied but have limitations due to their inability to incorporate long-range van der Waals (vdW) interaction. Non-local functionals (vdW-DF, VV10, rVV10) or empirical methods (DFT+D, DFT+vdW, DFT+MBD) are used with SL-DFAs to account for such missing interaction. The physisorption of a molecule on the surface of the coinage metals (Cu, Ag, and Au) is a typical example of systems where vdW interaction is significant. However, it is difficult to find a general method that reasonably describes both adsorption energy and geometry of even the simple prototypes of cyclic and heterocyclic aromatic molecules like benzene (C6H6) and thiophene (C4H4S) respectively, with reasonable accuracy. In this work, we present an alternative scheme based on Zaremba-Kohn's theory, called DFT+vdW-dZK. We show that, unlike other popular methods, DFT+vdW-dZK and particularly SCAN+vdW-dZK gives an accurate description of the physisorption of a rare-gas atom (Xe) and two small albeit diverse prototype organic molecules on the (111) surfaces of the coinage metals.

cond-mat.mtrl-sci↗

arXiv2020

Formation energy puzzle in intermetallic alloys: Random phase approximation fails to predict accurate formation energies

We performed density functional calculations to estimate the formation energies of intermetallic alloys. We used two semilocal approximations, the generalized gradient approximation (GGA) by Perdew-Burke-Ernzerhof (PBE) and the strongly constrained and appropriately normed (SCAN) meta-GGA. In addition, we utilized two nonlocal DFT functionals, the hybrid HSE06, and the state-of-the-art random phase approximation (RPA). The nonlocal functionals such as HSE06 and RPA yield accurate formation energies of binary alloys with completely-filled d-band metals, where semilocal functionals underperform. The accuracy at the nonlocal functionals is greatly reduced when a partially-filled d-band metal is present in an alloy, while PBE-GGA outperforms in these cases. We show that the accurate prediction of formation energies by any DFT method depends on its ability to predict the accurate electronic properties, e.g., valence d-band contribution to the density of states (DOS). The SCAN meta-GGA often corrects the PBE-DOS, however, it does not provide accurate formation energies compared to PBE. This is assumed to be due to the lack of proper error cancellation that should be expected due to the similar bulk nature of both alloys and their constituents, which may improve with the modification of meta-GGA ingredients. RPA yields too negative formation energies of alloys with partially-filled d-band metals. RPA results can be corrected by restoring the exchange-correlation kernel, thereby improving the short-range electron-electron correlation in metallic densities.

cond-mat.mtrl-sci↗

arXiv2020

The Fermi-Löwdin self-interaction correction for ionization energies of organic molecules

(Semi)-local density functional approximations (DFAs) suffer from self-interaction error (SIE). When the first ionization energy (IE) is computed as the negative of the highest-occupied orbital (HO) eigenvalue, DFAs notoriously underestimate them compared to quasi-particle calculations. The inaccuracy for the HO is attributed to SIE inherent in DFAs. We assessed the IE based on Perdew-Zunger self-interaction corrections on 14 small to moderate-sized organic molecules relevant in organic electronics and polymer donor materials. Though self-interaction corrected DFAs were found to significantly improve the IE relative to the uncorrected DFAs, they overestimate. However, when the self-interaction correction is interiorly scaled using a function of the iso-orbital indicator zσ, only the regions where SIE is significant get a correction. We discuss these approaches and show how these methods significantly improve the description of the HO eigenvalue for the organic molecules.

physics.chem-ph↗

arXiv2020

Understanding plasmon dispersion in nearly-free-electron metals: the relevance of exact constraints for novel exchange-correlation kernels within time-dependent density functional theory

Small-wavevector excitations in Coulomb-interacting systems can be decomposed into the high-energy collective longitudinal plasmon and the low-energy single-electron excitations. At the critical wavevector and corresponding frequency where the plasmon branch merges with the single-electron excitation region, the collective energy of the plasmon dissipates into single electron-hole excitations. The jellium model provides a reasonable description of the electron-energy-loss spectrum (EELS) of metals close to the free-electron limit. The random phase approximation (RPA) is exact in the high-density limit but can capture the plasmonic dispersion reasonably even for densities with rs > 1. RPA and all beyond-RPA methods investigated here, result in a wrong infinite plasmon lifetime for a wavevector smaller than the critical one where the plasmon dispersion curve runs into particle-hole excitations. Exchange-correlation kernel corrections to RPA modify the plasmon dispersion curve. There is however a large difference in the construction and form of the kernels investigated earlier. Our current work introduces recent model exchange-only and exchange-correlation kernels and discusses the relevance of some exact constraints in the construction of the kernel. We show that, because the plasmon dispersion samples a range of wavevectors smaller than the range sampled by the correlation energy, different kernels can make a strong difference for the correlation energy and a weak difference for the plasmon dispersion. This work completes our understanding about the plasmon dispersion in realistic metals, such as Cs, where a negative plasmon dispersion has been observed. We find only positive plasmon dispersion in jellium at the density for Cs.

cond-mat.mes-hall↗

arXiv2019

Molecule-surface interaction from van der Waals-corrected semilocal density functionals: the example of thiophene on transition-metal surfaces

Semi-local density functional approximations are widely used. None of them can capture the long-range van der Waals (vdW) attraction between separated subsystems, but they differ remarkably in the extent to which they capture intermediate-range vdW effects responsible for equilibrium bonds between neighboring small closed-shell subsystems. The local density approximation (LDA) often overestimates this effect, while the Perdew-Burke-Ernzerhof (PBE) generalized gradient approximation (GGA) underestimates it. The strongly-constrained and appropriately normed (SCAN) meta-GGA often estimates it well. All of these semi-local functionals require an additive non-local correction such as the revised Vydrov-Van Voorhis 2010 (rVV10) to capture the long-range part. This work reports adsorption energies and the corresponding geometry of the aromatic thiophene (C$_4$H$_4$S) bound to transition metal surfaces. The adsorption process requires a genuine interplay of covalent and weak binding and requires a simultaneously accurate description of surface and adsorption energies with the correct prediction of the adsorption site. All these quantities must come from well balanced short and long-range correlation effects for a universally applicable method for weak interactions with chemical accuracy.

cond-mat.mtrl-sci↗

arXiv2019

Treating different bonding situations: Revisiting Au-Cu alloys using the random phase approximation

The ground state equilibrium properties of copper-gold alloys have been explored with the state of art random phase approximation (RPA). Our estimated lattice constants agree with the experiment within a mean absolute percentage error (MAPE) of 1.4 percent. Semi-local functionals such as the generalized gradient approximation (GGA) of Perdew, Burke, and Ernzerhof (PBE) and strongly constrained and appropriately normed (SCAN) fail to provide accurate bulk moduli, which indicate their inability to describe the system in a stretched or compressed state with respect to the equilibrium geometry. We find that the non-locality present in RPA is able to describe the transition between two delocalized electron densities (bulk elemental constituents to crystallized alloys), as required to provide accurate formation energies. Based on our results, we conclude that it is difficult to find a universal density functional which can give accurate results for a wide range of properties of intermetallic alloys. However, RPA can capture different bonding situations, often better than other density functionals. It gives accurate results for a wide range of ground state properties for the alloys, generated from metals with completely filled d-shells.

cond-mat.mtrl-sci↗