My Customized Papers (Auto-Filtered)

[ChemRxiv] ChemTSv3: Generalizing Molecular Design via Flexible Search Space Control

ChemRxiv — Sun, 28 Dec 2025 00:00:00 GMT

Recent advances in generative artificial intelligence have enabled in silico molecular design to become a powerful approach for exploring chemical space toward specific design goals across various domains. However, in actual design workflows, determining the appropriate generation conditions, including generative strategies and reward formulations, remains difficult; thus, trial-and-error adjustments are unavoidable. Yet, most existing generation methods implicitly fix the searchable chemical space defined by the molecular representation and generation method, which significantly limits the flexibility of practical design. This paper introduces ChemTSv3, an exploration framework based on reinforcement learning with a flexible architecture that accommodates diverse design scenarios for adaptive molecular design. Specifically, molecular representations are unified as nodes, enabling, for example, string-based encodings, molecular graphs, and protein sequences to be handled within the same logic. Molecular generations and editing operations are abstracted as transitions between nodes, allowing classical graph-based modifications, sequential mutations, and even large-language-model-driven transformations to be handled within the same formulation. ChemTSv3 supports dynamic switching among molecular representations and transition types, which enables the search strategy itself to adapt to the stage and nature of the design task. ChemTSv3 enables scalable molecular generation, from drug-like small molecules to proteins, and its switching capability supports realistic change in design scenarios while allowing efficient exploration.

[ChemRxiv] Machine learning the quantum topology of chemical bonds

ChemRxiv — Sun, 28 Dec 2025 00:00:00 GMT

The chemical bonding can be characterized within quantum chemical topology (QCT), which provides a real-space description via the topological analysis of the electron density and the electron localization function (ELF). While QCT has traditionally been applied on a molecule-by-molecule basis, recent advances in machine learning (ML) and the availability of large quantum chemical datasets now enable bonding analysis at scale. Here, we integrate ELF-based topological descriptors with ML to establish a data-driven framework for mapping chemical bonding across the QM9 dataset. Wavefunctions computed at the B3LYP/6-31G(2df,p) level were used to extract ELF basin populations, which were paired with geometric and bonding descriptors to construct a bond-level dataset. Statistical analysis revealed relationships between ELF populations, bond lengths, and local chemical environments. Regression models were trained to predict ELF electron populations directly from molecular geometry. The best performance was obtained when local environmental descriptors were included, reducing the prediction error by a factor of two relative to models using only bond type and bond length. These results demonstrate real-space bonding parameters, such as bond electron populations, can be predicted from simple structural features, enabling scalable and interpretable exploration of chemical bonding across large chemical spaces.

[ChemRxiv] StereoMolGraph: Stereochemistry-Aware Molecular and Reaction Graphs

ChemRxiv — Sun, 28 Dec 2025 00:00:00 GMT

Conventional molecular graphs often are unable to reliably encode stereochemistry, especially for symmetric molecules, non-tetrahedral centers, and transition states. To overcome this, we present StereoMolGraph, an open source Python library implementing a stereochemistry-aware graph representation for molecules and condensed graphs of reactions. Our method uses permutation invariant local stereodescriptors, grounded in group theory, to provide an extensible representation of chirality. Based on this we introduce methods allowingfor robust comparison of stereoisomers, including the identification of enantiomerism and diastereomerism, and supports the of fleeting stereochemistry in transition states. We demonstrate the library’s utility for complex organic molecules and metal complexes and analysis of distinct chiral reaction pathways. With RDKit interoperability and visualization features, StereoMolGraph offers a practical and transparent tool for advanced stereochemically aware chemoinformatics workflows.

[ChemRxiv] GPU-Accelerated Analytic Coulomb- and Exchange Gradients for Hartree Fock and Density Functional Theory

cond-mat updates on arXiv.org — Mon, 29 Dec 2025 05:00:00 GMT

arXiv:2512.21758v1 Announce Type: new +Abstract: The separation of colloidal particles from fluids is essential to ensure a safe global supply of drinking water yet, in the case of microscopic particles, it remains a highly energy-intensive process when using traditional filtration methods. Water cleaning through diffusiophoresis $\unicode{x2014}$spontaneous colloid migration in chemical gradients$\unicode{x2014}$ effectively circumvents the need for physical filters, representing a promising alternative. This separation process is typically realized in internal flows where a cross-channel electrolyte gradient drives particle accumulation at walls, with colloid separation slowly increasing in the streamwise direction. However, the maximum separation efficiency, achieved sufficiently downstream as diffusiophoretic migration (driving particle accumulation) is balanced by Brownian motion (inducing diffusive spreading), has not yet been characterized. In this work, we develop a theory to predict this upper bound, and derive the colloid separation efficiency by analyzing the asymptotic structure of the governing equations. We find that the mechanism by which the chemical permeates in the channel, as well as the reaction kinetics governing its dissociation into ions, play key roles in the process. Moreover, we identify four distinct regimes in which separation is controlled by different scaling laws involving a Damk\"ohler and a P\'eclet number, which measure the ratio of reaction kinetics to ion diffusion and of diffusiophoresis to Brownian motion, respectively. We also confirm the scaling of one of these regimes using microfluidic experiments where separation is driven by CO$_\text{2}$ gradients. Our results shed light on pathways towards new, more efficient separations, and are also applicable to quantify colloidal accumulation in the presence of chemical gradients in more general situations.

[cond-mat updates on arXiv.org] Temperature- and Pressure-Dependent Vibrational Properties and Phase Stability of Pristine and Sb-Doped Vacancy-Ordered Double Perovskite

cond-mat updates on arXiv.org — Mon, 29 Dec 2025 05:00:00 GMT

arXiv:2512.21810v1 Announce Type: new +Abstract: Understanding lattice dynamics and structural transitions in vacancy-ordered double perovskites is crucial for developing lead-free optoelectronic materials, yet the role of dopants in modulating these properties remains poorly understood. We investigate Sb-doped Cs$_2$TiCl$_6$ through temperature-dependent Raman spectroscopy (4 to 273 K), high-pressure studies (0 to 30 GPa), powder XRD, and photoluminescence measurements. Sb doping dramatically improves phase purity, eliminating all impurity-related Raman modes present in pristine and Bi-doped samples while retaining only the three fundamental [TiCl$_6$]$^{2-}$ octahedral vibrations. This enhanced purity reveals a previously unobserved structural phenomenon: Sb-doped samples (2\% doped and 3\%) incorporated) exhibit a sharp anomaly at 100 K marked by the emergence of a new Raman mode M$_1$ at 314--319 cm$^{-1}$ and abrupt changes in the temperature coefficient $\chi$ (factor of 2--8$\times$ change) and anharmonic constant $A$ across this threshold. No such transition occurs in pristine Cs$_2$TiCl$_6$, indicating Sb-dopant-induced order-disorder transformation. The enhanced phonon anharmonicity in Sb-doped samples directly manifests in photoluminescence: self-trapped exciton emission at 448 nm shows 19\% broader FWHM (164.73 nm) compared to Bi-doped samples (138.2 nm), confirming stronger electron-phonon coupling. High-pressure measurements reveal structural robustness to 30 GPa with no phase transitions. These findings establish that strategic Sb doping not only improves material quality but also enables a novel low-temperature structural transition, providing fundamental insights into dopant-mediated phase control in vacancy-ordered perovskites for next-generation optoelectronic devices.

[cond-mat updates on arXiv.org] Incorporating rank-free coupling and external field via an amplitude-only modulated spatial photonic Ising machine

cond-mat updates on arXiv.org — Mon, 29 Dec 2025 05:00:00 GMT

arXiv:2512.21587v1 Announce Type: cross +Abstract: Ising machines have emerged as effective solvers for combinatorial optimization problems, such as NP-hard problems, machine learning, and financial modeling. Recent spatial photonic Ising machines (SPIMs) excel in multi-node optimization and spin glass simulations, leveraging their large-scale and fully connected characteristics. However, existing laser diffraction-based SPIMs usually sacrifice time efficiency or spin count to encode high-rank spin-spin coupling and external fields, limiting their scalability for real-world applications. Here, we demonstrate an amplitude-only modulated rank-free spatial photonic Ising machine (AR-SPIM) with 200 iterations per second. By re-formulating an arbitrary Ising Hamiltonian as the sum of Hadamard products, followed by loading the corresponding matrices/vectors onto an aligned amplitude spatial light modulator and digital micro-mirrors device, we directly map a 797-spin Ising model with external fields (nearly 9-bit precision, -255 to 255) into an incoherent light field, eliminating the need for repeated and auxiliary operations. Serving as encoding accuracy metrics, the linear coefficient of determination and Pearson correlation coefficient between measured light intensities and Ising Hamiltonians exceed 0.9800, with values exceed 0.9997 globally. The AR-SPIM achieves less than 0.3% error rate for ground-state search of biased Max-cut problems with arbitrary ranks and weights, enables complex phase transition observations, and facilitates scalable spin counts for sparse Ising problems via removing zero-valued Hadamard product terms. This reconfigurable AR-SPIM can be further developed to support large-scale machine-learning training and deployed for practical applications in discrete optimization and quantum many-body simulations.

[cond-mat updates on arXiv.org] Accelerating Scientific Discovery with Autonomous Goal-evolving Agents

cond-mat updates on arXiv.org — Mon, 29 Dec 2025 05:00:00 GMT

arXiv:2512.21782v1 Announce Type: cross +Abstract: There has been unprecedented interest in developing agents that expand the boundary of scientific discovery, primarily by optimizing quantitative objective functions specified by scientists. However, for grand challenges in science , these objectives are only imperfect proxies. We argue that automating objective function design is a central, yet unmet requirement for scientific discovery agents. In this work, we introduce the Scientific Autonomous Goal-evolving Agent (SAGA) to amend this challenge. SAGA employs a bi-level architecture in which an outer loop of LLM agents analyzes optimization outcomes, proposes new objectives, and converts them into computable scoring functions, while an inner loop performs solution optimization under the current objectives. This bi-level design enables systematic exploration of the space of objectives and their trade-offs, rather than treating them as fixed inputs. We demonstrate the framework through a broad spectrum of applications, including antibiotic design, inorganic materials design, functional DNA sequence design, and chemical process design, showing that automating objective formulation can substantially improve the effectiveness of scientific discovery agents.

[cond-mat updates on arXiv.org] Correlated Terahertz phonon-ion interactions control ion conduction in a solid electrolyte

cond-mat updates on arXiv.org — Mon, 29 Dec 2025 05:00:00 GMT

arXiv:2305.01632v4 Announce Type: replace +Abstract: Ionic conduction in solids that exceeds 1 mS/cm is predicted to involve coupled phonon-ion interactions in the crystal lattice. Here, we use theory and experiment to measure the possible contribution of coupled phonon-ion hopping modes which enhance Li+ migration in Li0.5La0.5TiO3 (LLTO). The ab initio calculations predict that the targeted excitation of individual TiO6 rocking modes greatly increases the Li+ jump rate as compared to the excitation of vibrational modes associated with heating. Experimentally, coherently driving TiO6 rocking modes via terahertz (THz) illumination leads to a ten-fold decrease in the differential impedance compared to the excitation of acoustic and optical phonons. Additionally, we differentiate the ultrafast responses of LLTO due to ultrafast heating and THz-range vibrations using laser-driven spectroscopy (LUIS), finding a unique long-lived response for the THz-range excitation. These findings provide new insights into coupled ion migration mechanisms, indicating the important role of THz-range coupled phonon-ion hopping modes in enabling fast ion conduction at room temperature.

[cond-mat updates on arXiv.org] Fluctuation theorems with optical tweezers: theory and practice

cond-mat updates on arXiv.org — Mon, 29 Dec 2025 05:00:00 GMT

arXiv:2503.20894v2 Announce Type: replace +Abstract: Fluctuation theorems, such as the Jarzynski equality and the Crooks relation, are effective tools connecting non-equilibrium work statistics and equilibrium free energy differences. However, detailed hands-on, reproducible protocols for implementing and analyzing these relations in real experiments remain scarce. This tutorial provides an end-to-end workflow for measuring, validating, and applying fluctuation theorems using a single-beam optical tweezers setup. It introduces the foundational ideas and consolidates practical calibration (PSD-based trap stiffness and position sensitivity), protocol design (forward/reverse finite-time drives over multiple amplitudes and durations), and robust estimators for free-energy difference and dissipated work, highlighting finite-sampling and rare-event effects. We demonstrate the procedures using an extensive set of measured trajectories under different conditions and provide openly accessible datasets and Python code, enabling new researchers or educators to reproduce the results with minimal effort. Beyond pedagogical validation, we discuss how these recipes translate to broader soft-matter and mesoscopic contexts. By combining user-friendly instruments with clear and transparent analysis, this work promotes the education and reliable adoption of stochastic thermodynamic methods in the curricula of physics and chemistry, as well as among emerging research teams.

[cond-mat updates on arXiv.org] Rewards-based image analysis in microscopy

cond-mat updates on arXiv.org — Mon, 29 Dec 2025 05:00:00 GMT

arXiv:2502.18522v2 Announce Type: replace-cross +Abstract: Imaging and hyperspectral data analysis is central to progress across biology, medicine, chemistry, and physics. The core challenge lies in converting high-resolution or high-dimensional datasets into interpretable representations that enable insight into the underlying physical or chemical properties of a system. Traditional analysis relies on expert-designed, multistep workflows, such as denoising, feature extraction, clustering, dimensionality reduction, and physics-based deconvolution, or on machine learning (ML) methods that accelerate individual steps. Both approaches, however, typically demand significant human intervention, including hyperparameter tuning and data labeling. Achieving the next level of autonomy in scientific imaging requires designing effective reward-based workflows that guide algorithms toward best data representation for human or automated decision-making. Here, we discuss recent advances in reward-based workflows for image analysis, which capture key elements of human reasoning and exhibit strong transferability across various tasks. We highlight how reward-driven approaches enable a shift from supervised black-box models toward explainable, unsupervised optimization on the examples of Scanning Probe and Electron Microscopies. Such reward-based frameworks are promising for a broad range of applications, including classification, regression, structure-property mapping, and general hyperspectral data processing.

[ChemRxiv] A Protocol to Identify Large Language Model Use in Undergraduate Chemistry Essays

ChemRxiv — Mon, 29 Dec 2025 00:00:00 GMT

Large language modules (LLMs) such as Chat-GPT have been widely adopted by chemistry undergraduate university students as a learning tool, but few methods exist to measure the scope of their influence on essay writing. This report introduces a protocol based on monitoring the frequency of specific words used excessively by LLMs amongst a sample of around 1000 chemistry student essays between 2018 and 2025. More than 50 key words known to be favored by LLMs were found to have simultaneously increased in frequency amongst essays submitted during the last two years. When the lists of key words were applied to specific essays, the findings indicated 13-29% of essays submitted during 2025 relied heavily on text generated by LLMs. This protocol offers a simple method to generate an approximate scope of LLM uptake amongst a cohort, with potential applications in defining higher education AI policy and assessment strategies.

[ChemRxiv] Graph machine learning can estimate drug concentrations in whole blood from forensic screening results

ChemRxiv — Mon, 29 Dec 2025 00:00:00 GMT

LC-HRMS is widely used in forensic toxicology for broad-scope screening. When a newly emerging or rarely encountered compound is tentatively identified, toxicologists must decide whether it may be relevant to the case and, if so, quantify it. Acquiring reference material for quantification is costly and time-consuming. A fast semi-quantitative estimate would help prioritize only compounds above the toxic threshold. This study presents a machine-learning framework that estimates drug concentrations in whole blood using molecular structure information and LC-HRMS signal intensities. Using a dataset of 191 drugs spiked into whole blood at multiple concentration levels, we trained and evaluated several machine-learning models. Standard models, including random forests, achieved moderate performance. In contrast, a recently published graph neural network (GNN) leveraging atomic features and global molecular properties consistently produced the highest accuracy. Under cross-validation, the GNN predicted signal-to-concentration ratio for 79\% of all molecules, corresponding to concentration estimates between 50-200\% of true value. Toxicological thresholds often span multiple orders of magnitude, making this precision acceptable for application. The GNN model was additionally evaluated on an external benchmark dataset of ionization efficiencies (logIE), where it outperformed the current state of the art. Overall, the results demonstrate the feasibility of using graph-based machine learning to estimate drug concentrations in whole blood without reference material for prioritization. This is a practical and implementation-ready machine learning tool that can support decision-making in toxicological evaluation, particularly for newly emerging or rarely encountered drugs. The GNN model is open source and the dataset used for training and testing the models is publicly available.

[ChemRxiv] ChemTSv3: Generalizing Molecular Design via Flexible Search Space Control

ChemRxiv — Sun, 28 Dec 2025 00:00:00 GMT

[ChemRxiv] Machine learning the quantum topology of chemical bonds

ChemRxiv — Sun, 28 Dec 2025 00:00:00 GMT

[ChemRxiv] StereoMolGraph: Stereochemistry-Aware Molecular and Reaction Graphs

ChemRxiv — Sun, 28 Dec 2025 00:00:00 GMT

[ChemRxiv] GPU-Accelerated Analytic Coulomb- and Exchange Gradients for Hartree Fock and Density Functional Theory

ChemRxiv — Sun, 28 Dec 2025 00:00:00 GMT

We present a GPU-accelerated software package for the evaluation of analytic two-electron energy and gradient contributions in Hartree-Fock (HF) and Density Functional theory (DFT) calculations. The implementation is provided as a Python library with a C++ backend, enabling straightforward integration into modern computational chemistry and drug-discovery workflows. The code supports single-point energy and nuclear gradient evaluations on both single- and multi-GPU systems, and employs MPI-based parallelization with dynamic load balancing in multi-node environments. We report comprehensive benchmarks demonstrating favorable scaling with respect to system size, as well as high throughput for batched evaluations relevant to molecular dynamics, geometry optimization, and large-scale virtual screening. Parallel execution of a single system was carried out on up to 24 A100 GPUs. The implementation builds on optimized GPU-enabled variants of the LibintX and GauXC libraries to efficiently compute density-fitted Coulomb, semi-numerical exchange, and exchange--correlation contributions.

[ACS Nano: Latest Articles (ACS Publications)] [ASAP] Machine Learning−Accelerated Discovery of Proton-Conducting 2D Materials for Proton Exchange Membranes

ACS Nano: Latest Articles (ACS Publications) — Sat, 27 Dec 2025 14:37:43 GMT

ACS Nano

DOI: 10.1021/acsnano.5c15161

[ScienceDirect Publication: Computational Materials Science] An enhanced machine learning and computational screening framework for synthesizable single-phase high-entropy spinel oxides

ScienceDirect Publication: Computational Materials Science — Fri, 26 Dec 2025 18:29:22 GMT

Publication date: February 2026

Source: Computational Materials Science, Volume 263

Author(s): Mahalaxmi Chandramohan, Hridhya Vinod, Meenal Deo

[ScienceDirect Publication: Materials Today Physics] Anomalous temperature evolution of lattice anharmonicity and thermal transport in orthorhombic SnSe

ScienceDirect Publication: Materials Today Physics — Fri, 26 Dec 2025 18:29:17 GMT

Publication date: Available online 26 December 2025

Source: Materials Today Physics

Author(s): Tianxiang Jiang, Wujie Qiu, Haijuan Zhang, Jifen Wang, Kunpeng Zhao, Huaqing Xie

[ScienceDirect Publication: Materials Today Physics] Machine learning aided bandgap and defect engineering of mixed halide perovskites for photovoltaic applications

ScienceDirect Publication: Materials Today Physics — Fri, 26 Dec 2025 18:29:17 GMT

Publication date: Available online 26 December 2025

Source: Materials Today Physics

Author(s): Ayush Kumar Pandey, Vivek Pandey, Abhishek Tewari

[Journal of Chemical Theory and Computation: Latest Articles (ACS Publications)] [ASAP] Automated Machine Learning Pipeline: Large Language Models-Assisted Automated Data set Generation for Training Machine-Learned Interatomic Potentials

Journal of Chemical Theory and Computation: Latest Articles (ACS Publications) — Fri, 26 Dec 2025 18:25:53 GMT

Journal of Chemical Theory and Computation

DOI: 10.1021/acs.jctc.5c01610

[The Journal of Physical Chemistry Letters: Latest Articles (ACS Publications)] [ASAP] Machine Learning Photodynamics Unveils a Controlled H2 Loss Channel in the Methaniminium Cation

The Journal of Physical Chemistry Letters: Latest Articles (ACS Publications) — Fri, 26 Dec 2025 17:51:53 GMT

The Journal of Physical Chemistry Letters

DOI: 10.1021/acs.jpclett.5c03196

[The Journal of Physical Chemistry Letters: Latest Articles (ACS Publications)] [ASAP] Solvent Structure and Dynamics Controlled Memristive Ion Transport in Å-Scale Channels

The Journal of Physical Chemistry Letters: Latest Articles (ACS Publications) — Fri, 26 Dec 2025 16:50:38 GMT

The Journal of Physical Chemistry Letters

DOI: 10.1021/acs.jpclett.5c03397

[The Journal of Physical Chemistry Letters: Latest Articles (ACS Publications)] [ASAP] Design of Open Ion Transport Channels in Thick Electrodes through Directional Freezing-Assisted 3D Printing: Enhancing the Electrochemical Performance of High-Load Electrodes

The Journal of Physical Chemistry Letters: Latest Articles (ACS Publications) — Fri, 26 Dec 2025 16:49:57 GMT

The Journal of Physical Chemistry Letters

DOI: 10.1021/acs.jpclett.5c02968

[The Journal of Physical Chemistry C: Latest Articles (ACS Publications)] [ASAP] Machine Learning Analysis of Dimensionality Effects on Bandgap Prediction

The Journal of Physical Chemistry C: Latest Articles (ACS Publications) — Fri, 26 Dec 2025 16:06:02 GMT

The Journal of Physical Chemistry C

DOI: 10.1021/acs.jpcc.5c05232

[Wiley: Advanced Materials: Table of Contents] Plasma Design of Alloy‐Based Gradient Solid Electrolyte Interphase on Lithium Metal Anodes for Energy Storage

Wiley: Advanced Materials: Table of Contents — Fri, 26 Dec 2025 14:02:31 GMT

Advanced Materials, EarlyView.

[Wiley: Advanced Functional Materials: Table of Contents] Pixelation‐Free, Monolithic Iontronic Pressure Sensors Enabling Large‐Area Simultaneous Pressure and Position Recognition via Machine Learning

Wiley: Advanced Functional Materials: Table of Contents — Fri, 26 Dec 2025 14:01:16 GMT

Advanced Functional Materials, EarlyView.

[Wiley: Advanced Functional Materials: Table of Contents] Machine Learning‐Enhanced Smart Interactive Glove Utilizing Flexible Gradient Ridge Architecture Iontronic Capacitive Sensor

Wiley: Advanced Functional Materials: Table of Contents — Fri, 26 Dec 2025 09:52:42 GMT

Advanced Functional Materials, EarlyView.

[ACS Nano: Latest Articles (ACS Publications)] [ASAP] Machine Learning-Accelerated Development of Li-Rich Layered Oxides with High Reversible Capacity and Coulombic Efficiency

ACS Nano: Latest Articles (ACS Publications) — Fri, 26 Dec 2025 09:21:05 GMT

ACS Nano

DOI: 10.1021/acsnano.5c16117

[Wiley: Advanced Science: Table of Contents] Mendelian Randomization and Double Machine Learning Modeling Reveal Brain Imaging‐Derived Phenotypes as Functional Contributors to 18 Autoimmune Inflammatory Diseases

Wiley: Advanced Science: Table of Contents — Fri, 26 Dec 2025 06:23:13 GMT

Advanced Science, EarlyView.

[Wiley: Carbon Energy: Table of Contents] Data‐Driven Design of Scalable Perovskite Film Fabrication via Machine Learning–Guided Processing

Wiley: Carbon Energy: Table of Contents — Fri, 26 Dec 2025 06:22:51 GMT

Carbon Energy, EarlyView.

[Nature Communications] Inferring fine-grained migration patterns across the United States

Nature Communications — Fri, 26 Dec 2025 00:00:00 GMT

Nature Communications, Published online: 26 December 2025; doi:10.1038/s41467-025-68019-2

This study releases a very high-resolution migration dataset that reveals trends that shape daily life: rising moves into high-income neighborhoods, racial gaps in upward mobility, and wildfire-driven moves.

[Communications Materials] Machine learning interatomic potential reveals hydrogen embrittlement origins at general grain boundaries in α-iron

Communications Materials — Fri, 26 Dec 2025 00:00:00 GMT

Communications Materials, Published online: 26 December 2025; doi:10.1038/s43246-025-01042-4

Hydrogen embrittlement is an issue that alloys used in the energy sector must overcome. Here, a machine learning interatomic potential for iron-hydrogen is reported, with large-scale molecular dynamics simulations revealing that hydrogen can suppress >111 < /2 dislocation emission at grain boundaries.

[iScience] Entropy-based byte patching transformer for self-supervised pretraining of SMILES data

iScience — Fri, 26 Dec 2025 00:00:00 GMT

Efficient molecular representations are critical for improving the performance and generalization of large language models in chemical learning. Transformer-based architectures have advanced molecular representation learning, yet capturing localized and hierarchical chemical structures remains challenging. We introduce the SMILES Byte-Patch Transformer (SMiBPT), an adaptive model that dynamically segments SMILES and DeepSMILES strings into chemically meaningful substructures through entropy-based byte patching.

[ScienceDirect Publication: Solid State Ionics] Anion engineering enhances the electrochemical performance of Li₇P₃S₁₁ solid electrolyte

ScienceDirect Publication: Solid State Ionics — Thu, 25 Dec 2025 18:28:52 GMT

Publication date: 15 February 2026

Source: Solid State Ionics, Volume 435

Author(s): D. Narsimulu, Ramkumar Balasubramaniam, Kwang-Sun Ryu

[ScienceDirect Publication: Science Bulletin] Machine learning for spectral precision: a new horizon in radiative cooling material design

ScienceDirect Publication: Science Bulletin — Thu, 25 Dec 2025 18:28:50 GMT

Publication date: 30 December 2025

Source: Science Bulletin, Volume 70, Issue 24

Author(s): Xinpeng Hu, Mingxiang Liu, Xuemei Fu, Guangming Tao, Xiang Lu, Jinping Qu

[cond-mat updates on arXiv.org] Towards predictive atomistic simulations of SiC crystal growth

cond-mat updates on arXiv.org — Thu, 25 Dec 2025 05:00:00 GMT

arXiv:2512.20804v1 Announce Type: new Abstract: Simulations of SiC crystal growth using molecular dynamics (MD) have become popular in recent years. They, however, simulate very fast deposition rates, to reduce computational costs. Therefore, they are more akin to surface sputtering, leading to abnormal growth effects, including thick amorphous layers and large defect densities. A recently developed method, called the minimum energy atomic deposition (MEAD), tries to overcome this problem by depositing the atoms directly at the minimum energy positions, increasing the time scale. We apply the MEAD method to simulate SiC crystal growth on stepped C-terminated 4H substrates with 4{\deg} and 8{\deg} off-cut angle. We explore relevant calculations settings, such as amount of equilibration steps between depositions and influence of simulation cell sizes and bench mark different interatomic potentials. The carefully calibrated methodology is able to replicate the stable step-flow growth, which was so far not possible using conventional MD simulations. Furthermore, the simulated crystals are evaluated in terms of their dislocations, surface roughness and atom mobility. Our methodology paves the way for future high fidelity investigations of surface phenomena in crystal growth.

My Customized Papers (Auto-Filtered)

[ChemRxiv] ChemTSv3: Generalizing Molecular Design via Flexible Search Space Control

[ChemRxiv] Machine learning the quantum topology of chemical bonds

[ChemRxiv] StereoMolGraph: Stereochemistry-Aware Molecular and Reaction Graphs

[ChemRxiv] GPU-Accelerated Analytic Coulomb- and Exchange Gradients for Hartree Fock and Density Functional Theory

[cond-mat updates on arXiv.org] Temperature- and Pressure-Dependent Vibrational Properties and Phase Stability of Pristine and Sb-Doped Vacancy-Ordered Double Perovskite

[cond-mat updates on arXiv.org] Incorporating rank-free coupling and external field via an amplitude-only modulated spatial photonic Ising machine

[cond-mat updates on arXiv.org] Accelerating Scientific Discovery with Autonomous Goal-evolving Agents

[cond-mat updates on arXiv.org] Correlated Terahertz phonon-ion interactions control ion conduction in a solid electrolyte

[cond-mat updates on arXiv.org] Fluctuation theorems with optical tweezers: theory and practice

[cond-mat updates on arXiv.org] Rewards-based image analysis in microscopy

[ChemRxiv] A Protocol to Identify Large Language Model Use in Undergraduate Chemistry Essays

[ChemRxiv] Graph machine learning can estimate drug concentrations in whole blood from forensic screening results

[ChemRxiv] ChemTSv3: Generalizing Molecular Design via Flexible Search Space Control

[ChemRxiv] Machine learning the quantum topology of chemical bonds

[ChemRxiv] StereoMolGraph: Stereochemistry-Aware Molecular and Reaction Graphs

[ChemRxiv] GPU-Accelerated Analytic Coulomb- and Exchange Gradients for Hartree Fock and Density Functional Theory

[ACS Nano: Latest Articles (ACS Publications)] [ASAP] Machine Learning−Accelerated Discovery of Proton-Conducting 2D Materials for Proton Exchange Membranes

[ScienceDirect Publication: Computational Materials Science] An enhanced machine learning and computational screening framework for synthesizable single-phase high-entropy spinel oxides

[ScienceDirect Publication: Materials Today Physics] Anomalous temperature evolution of lattice anharmonicity and thermal transport in orthorhombic SnSe

[ScienceDirect Publication: Materials Today Physics] Machine learning aided bandgap and defect engineering of mixed halide perovskites for photovoltaic applications

[Journal of Chemical Theory and Computation: Latest Articles (ACS Publications)] [ASAP] Automated Machine Learning Pipeline: Large Language Models-Assisted Automated Data set Generation for Training Machine-Learned Interatomic Potentials

[The Journal of Physical Chemistry Letters: Latest Articles (ACS Publications)] [ASAP] Machine Learning Photodynamics Unveils a Controlled H2 Loss Channel in the Methaniminium Cation

[The Journal of Physical Chemistry Letters: Latest Articles (ACS Publications)] [ASAP] Solvent Structure and Dynamics Controlled Memristive Ion Transport in Å-Scale Channels

[The Journal of Physical Chemistry Letters: Latest Articles (ACS Publications)] [ASAP] Design of Open Ion Transport Channels in Thick Electrodes through Directional Freezing-Assisted 3D Printing: Enhancing the Electrochemical Performance of High-Load Electrodes

[The Journal of Physical Chemistry C: Latest Articles (ACS Publications)] [ASAP] Machine Learning Analysis of Dimensionality Effects on Bandgap Prediction

[Wiley: Advanced Materials: Table of Contents] Plasma Design of Alloy‐Based Gradient Solid Electrolyte Interphase on Lithium Metal Anodes for Energy Storage

[Wiley: Advanced Functional Materials: Table of Contents] Pixelation‐Free, Monolithic Iontronic Pressure Sensors Enabling Large‐Area Simultaneous Pressure and Position Recognition via Machine Learning

[Wiley: Advanced Functional Materials: Table of Contents] Machine Learning‐Enhanced Smart Interactive Glove Utilizing Flexible Gradient Ridge Architecture Iontronic Capacitive Sensor

[ACS Nano: Latest Articles (ACS Publications)] [ASAP] Machine Learning-Accelerated Development of Li-Rich Layered Oxides with High Reversible Capacity and Coulombic Efficiency

[Wiley: Advanced Science: Table of Contents] Mendelian Randomization and Double Machine Learning Modeling Reveal Brain Imaging‐Derived Phenotypes as Functional Contributors to 18 Autoimmune Inflammatory Diseases

[Wiley: Carbon Energy: Table of Contents] Data‐Driven Design of Scalable Perovskite Film Fabrication via Machine Learning–Guided Processing

[Nature Communications] Inferring fine-grained migration patterns across the United States

[Communications Materials] Machine learning interatomic potential reveals hydrogen embrittlement origins at general grain boundaries in α-iron

[iScience] Entropy-based byte patching transformer for self-supervised pretraining of SMILES data

[ScienceDirect Publication: Solid State Ionics] Anion engineering enhances the electrochemical performance of Li7P3S11 solid electrolyte

[ScienceDirect Publication: Science Bulletin] Machine learning for spectral precision: a new horizon in radiative cooling material design

[cond-mat updates on arXiv.org] Towards predictive atomistic simulations of SiC crystal growth

[ScienceDirect Publication: Solid State Ionics] Anion engineering enhances the electrochemical performance of Li₇P₃S₁₁ solid electrolyte