Zhantao Chen

This Zenodo repository accompanies the work “An Agentic Artificially Intelligent X-ray Scientist.” It contains the source code used to support the results reported in the study. Research Square pre-print link: https://doi.org/10.21203/rs.3.rs-7456716/v1

This Zenodo repository accompanies the work “An Agentic Artificially Intelligent X-ray Scientist.” It contains the source code used to support the results reported in the study. Research Square pre-print link: https://doi.org/10.21203/rs.3.rs-7456716/v1

This Zenodo repository accompanies the work “An Agentic Artificially Intelligent X-ray Scientist.” It contains supporting data generated during the study, including AI agent conversation histories, SPEC files from real beamline experiments, and acquired experimental images. Research Square pre-print link: https://doi.org/10.21203/rs.3.rs-7456716/v1

This Zenodo repository accompanies the work “An Agentic Artificially Intelligent X-ray Scientist.” It contains supporting data generated during the study, including AI agent conversation histories, SPEC files from real beamline experiments, and acquired experimental images. Research Square pre-print link: https://doi.org/10.21203/rs.3.rs-7456716/v1

Braille literacy remains a critical challenge for Visually Impaired and Blind (VIB) individuals, with existing tools often failing to meet their diverse learning needs. There is a pressing demand for innovative, accessible, and cost-effective solutions to support independent Braille learning. This study proposes a user-centered approach to designing a holistic Braille self-learning device that addresses these gaps. By integrating the ESP32 microcontroller and Tencent Cloud’s Speech Recognition API, the device overcomes common hardware constraints while enabling an interactive learning experience. A comprehensive hardware cost analysis ensures technological advancement aligns with economic viability, facilitating the development of a market-ready prototype. Usability testing, focusing on learnability, accessibility, portability, and reliability, was conducted with VIB users, supplemented by detailed survey feedback. The results indicate promising performance across key metrics, though they also reveal opportunities for refinement in enhancing learnability and broadening compatibility with assistive technologies. This research contributes a validated framework for designing affordable and effective Braille learning tools, underscoring technology’s role in fostering self-directed learning and promoting inclusivity in education.

As sports training becomes more data-driven, traditional dart coaching based mainly on experience and visual observation is increasingly inadequate for high-precision, goal-oriented movements. Although prior studies have highlighted the importance of release parameters, joint motion, and coordination in dart throwing, most quantitative methods still focus on local variables, single-release metrics, or static template matching. These approaches offer limited support for personalized training and often overlook useful movement variability. This paper presents a data-driven dart training assistance system. The system creates a closed-loop framework spanning motion capture, feature modeling, and personalized feedback. Dart-throwing data were collected in markerless conditions using a Kinect 2.0 depth sensor and an optical camera. Eighteen kinematic features were extracted from four biomechanical dimensions: three-link coordination, release velocity, multi-joint angular configuration, and postural stability. Two modules were developed: a personalized optimal throwing trajectory model that combines historical high-quality samples with the minimum jerk criterion, and a motion deviation diagnosis and recommendation model based on z-scores and hierarchical logic. A total of 2,396 throwing samples from professional and non-professional athletes were collected. Results show that the system generates smooth personalized reference trajectories consistent with natural human movement. Case studies indicate that it can detect poor trunk stability, abnormal elbow displacement, and imbalanced velocity control, then provide targeted recommendations. The framework shifts dart evaluation from deviation from a uniform standard to deviation from an individual's optimal control range, improving personalization and interpretability for darts training and other high-precision target sports.

As sports training becomes more data-driven, traditional dart coaching based mainly on experience and visual observation is increasingly inadequate for high-precision, goal-oriented movements. Although prior studies have highlighted the importance of release parameters, joint motion, and coordination in dart throwing, most quantitative methods still focus on local variables, single-release metrics, or static template matching. These approaches offer limited support for personalized training and often overlook useful movement variability. This paper presents a data-driven dart training assistance system. The system creates a closed-loop framework spanning motion capture, feature modeling, and personalized feedback. Dart-throwing data were collected in markerless conditions using a Kinect 2.0 depth sensor and an optical camera. Eighteen kinematic features were extracted from four biomechanical dimensions: three-link coordination, release velocity, multi-joint angular configuration, and postural stability. Two modules were developed: a personalized optimal throwing trajectory model that combines historical high-quality samples with the minimum jerk criterion, and a motion deviation diagnosis and recommendation model based on z-scores and hierarchical logic. A total of 2,396 throwing samples from professional and non-professional athletes were collected. Results show that the system generates smooth personalized reference trajectories consistent with natural human movement. Case studies indicate that it can detect poor trunk stability, abnormal elbow displacement, and imbalanced velocity control, then provide targeted recommendations. The framework shifts dart evaluation from deviation from a uniform standard to deviation from an individual's optimal control range, improving personalization and interpretability for darts training and other high-precision target sports.

Interfacial thermal and acoustic phenomena have an important role in quantum science and technology, including in spintronic and spincaloritronic materials and devices. Simultaneous measurements of the low-temperature thermal and acoustic properties of a metal/insulator heterostructure reveal distinct dynamics in the characteristic phonon frequency ranges of acoustic and thermal transport. The measurements probed a heterostructure consisting of a thin film of Pt on the ferrimagnetic insulator gadolinium iron garnet (Gd3Fe5O12, GdIG) grown epitaxially on a gadolinium gallium garnet substrate. Ultrafast structural dynamics within the Pt layer were tracked using time-resolved ultrafast x-ray diffraction and analyzed to probe interfacial acoustic and thermal properties. The rapid heating of the Pt layer by a 400 nm wavelength femtosecond-duration optical pulse produced transient structural changes that provided the stimulus for these measurements. Rapid heating produced a broadband acoustic pulse that was partially reflected by the Pt/GdIG interface. Temporal frequencies up to 740 GHz, corresponding to angular frequencies of several THz, were detected in a wavelet analysis of the acoustic oscillations of the strain in the Pt layer. The structural results were analyzed to determine (i) the acoustic damping coefficient and phonon mean free path in Pt at frequencies of hundreds of GHz and (ii) the Grüneisen anharmonicity parameter. The thermal conductance of the Pt/GdIG interface was tracked using the slower, tens-of-picosecond-scale, dynamics of the initial cooling of the heated Pt layer. Analysis using a model based on the Boltzmann transport equation shows that the phonon transmission is lower at the phonon frequencies relevant to thermal transport than for subterahertz regime acoustics.

Interfacial thermal and acoustic phenomena have an important role in quantum science and technology, including in spintronic and spincaloritronic materials and devices. Simultaneous measurements of the low-temperature thermal and acoustic properties of a metal/insulator heterostructure reveal distinct dynamics in the characteristic phonon frequency ranges of acoustic and thermal transport. The measurements probed a heterostructure consisting of a thin film of Pt on the ferrimagnetic insulator gadolinium iron garnet (Gd3Fe5O12, GdIG) grown epitaxially on a gadolinium gallium garnet substrate. Ultrafast structural dynamics within the Pt layer were tracked using time-resolved ultrafast x-ray diffraction and analyzed to probe interfacial acoustic and thermal properties. The rapid heating of the Pt layer by a 400 nm wavelength femtosecond-duration optical pulse produced transient structural changes that provided the stimulus for these measurements. Rapid heating produced a broadband acoustic pulse that was partially reflected by the Pt/GdIG interface. Temporal frequencies up to 740 GHz, corresponding to angular frequencies of several THz, were detected in a wavelet analysis of the acoustic oscillations of the strain in the Pt layer. The structural results were analyzed to determine (i) the acoustic damping coefficient and phonon mean free path in Pt at frequencies of hundreds of GHz and (ii) the Grüneisen anharmonicity parameter. The thermal conductance of the Pt/GdIG interface was tracked using the slower, tens-of-picosecond-scale, dynamics of the initial cooling of the heated Pt layer. Analysis using a model based on the Boltzmann transport equation shows that the phonon transmission is lower at the phonon frequencies relevant to thermal transport than for subterahertz regime acoustics.

Abstract Significant challenges exist in efficient data analysis of most advanced experimental and observational techniques because the collected signals often include unwanted contributions, such as background and signal distortions, that can obscure the physically relevant information of interest. To address this, we have developed a self-supervised machine-learning approach for source separation using a dual implicit neural representation framework that jointly trains two neural networks: one for approximating distortions of the physical signal of interest and the other for learning the effective background contribution. Our method learns directly from the raw data by minimizing a reconstruction-based loss function without requiring labeled data or pre-defined dictionaries. We demonstrate the effectiveness of our framework by considering a challenging case study involving large-scale simulated, as well as experimental, momentum-energy-dependent inelastic neutron scattering data in a four-dimensional parameter space, characterized by heterogeneous background contributions and unknown distortions to the target signal. The method is found to successfully separate physically meaningful signals from a complex or structured background even when the signal characteristics vary across all four dimensions of the parameter space. An analytical approach that informs the choice of the regularization parameter is presented. Our method offers a versatile framework for addressing source separation problems across diverse domains, ranging from superimposed signals in astronomical measurements to structural features in biomedical image reconstructions.

Scientific data acquisition continues to outpace storage and analysis capabilities, making voxel-based representations increasingly intractable. Implicit neural representations (INRs) offer a promising solution by encoding signals through coordinate-based neural networks, serving as surrogates of data, with computational and storage requirements scaling with network complexity rather than data dimensionality. However, smaller INRs struggle to faithfully represent the multi-scale structures, high-frequency information, and fine textures that constitute a large proportion of scientific measurements. We propose WIEN-INR, a theoretically-guided hierarchical INR framework that distributes modeling across resolution scales and enables improved representation capacity through a novel enhancement network to recover subtle details. This multi-scale architecture allows smaller networks to retain the full spectrum of information while preserving training efficiency and lowering storage cost. Evaluated on distinct raw experimental measurements across scales and complexities, WIEN-INR represents a practical step toward broader adoption of neural representations in scientific workflows, delivering compact, robust, and high-fidelity representations.

<title>Abstract</title> Executing experimental tasks in both normal research laboratories and large-scale scientific facilities often requires extensive human supervision, remaining a key challenge on the path to fully autonomous, artificial intelligence (AI)-driven science. We present the first demonstration of an AI agent which plans and executes experimental tasks, analyzes results, and iterates to achieve a scientific goal. Based on existing large language models and enhanced with the model context protocol, our AI agent was guided and tested using an in-house built virtual experimental setup which mirrors those which exist at large-scale X-ray scattering facilities, specifically here a six-circle diffractometer. It successfully transferred the knowledge to a real beamline and handled an experiment at a synchrotron X-ray source, where it correctly identified reference reflections and determined the orientation matrix---an essential first step in any type of single crystal scattering experiment. Our AI agent responded effectively to unexpected experimental conditions, demonstrating adaptive problem-solving and showing readiness for addressing practical experimental situations. Our study provides a significant step toward autonomous operation across diverse experimental environments.

Significant challenges exist in efficient data analysis of most advanced experimental and observational techniques because the collected signals often include unwanted contributions--such as background and signal distortions--that can obscure the physically relevant information of interest. To address this, we have developed a self-supervised machine-learning approach for source separation using a dual implicit neural representation framework that jointly trains two neural networks: one for approximating distortions of the physical signal of interest and the other for learning the effective background contribution. Our method learns directly from the raw data by minimizing a reconstruction-based loss function without requiring labeled data or pre-defined dictionaries. We demonstrate the effectiveness of our framework by considering a challenging case study involving large-scale simulated as well as experimental momentum-energy-dependent inelastic neutron scattering data in a four-dimensional parameter space, characterized by heterogeneous background contributions and unknown distortions to the target signal. The method is found to successfully separate physically meaningful signals from a complex or structured background even when the signal characteristics vary across all four dimensions of the parameter space. An analytical approach that informs the choice of the regularization parameter is presented. Our method offers a versatile framework for addressing source separation problems across diverse domains, ranging from superimposed signals in astronomical measurements to structural features in biomedical image reconstructions.

Thermoelectric materials offer a promising pathway to directly convert waste heat to electricity. However, achieving high performance remains challenging due to intrinsic trade-offs between electrical conductivity, the Seebeck coefficient, and thermal conductivity, which are further complicated by the presence of defects. This review explores how artificial intelligence (AI) and machine learning (ML) are transforming thermoelectric materials design. Advanced ML approaches including deep neural networks, graph-based models, and transformer architectures, integrated with high-throughput simulations and growing databases, effectively capture structure-property relationships in a complex multiscale defect space and overcome the "curse of dimensionality". This review discusses AI-enhanced defect engineering strategies such as composition optimization, entropy and dislocation engineering, and grain boundary design, along with emerging inverse design techniques for generating materials with targeted properties. Finally, it outlines future opportunities in novel physics mechanisms and sustainability, highlighting the critical role of AI in accelerating the discovery of thermoelectric materials.

The development of X-ray free-electron lasers (XFELs) has opened numerous opportunities to probe atomic structure and ultrafast dynamics of various materials. Single-particle imaging (SPI) with XFELs enables the investigation of biological particles in their alternative physiological states with unparalleled temporal resolution while circumventing the need for cryogenic conditions or crystallization. However, reconstructing real-space structures from reciprocal-space X-ray diffraction data is highly challenging due to the absence of phase and orientation information, which is further complicated by weak scattering signals and considerable fluctuations in photon numbers of measured scattering signals. In this work, we present an end-to-end, self-supervised machine-learning approach to recover particle orientations and estimate reciprocal-space intensities from diffraction images alone. Through comprehensive benchmarks over simulation data, our method demonstrates great robustness under demanding experimental conditions with enhanced reconstruction capabilities compared with conventional algorithms. We further demonstrate the flexibility and modularity of our method by incorporating symmetry constraints to resolve the structures of highly symmetric particles and applying it to real experimental data. The presented method introduces a novel approach to perform SPI reconstructions with XFELs.

Scattering measurements using electrons, neutrons, or photons are essential for obtaining microscopic insights into materials. However, limited facility availability and high-dimensional scattering data necessitate more efficient experimental steering techniques. Here, we report a machine learning method that guides scattering data collection and facilitates real-time estimation of model parameters, given a reliable forward model to simulate experimental signals. We employ implicit neural representations as efficient surrogates that link model parameters with simulated spectroscopies. This enables a Bayesian optimal experimental design framework to estimate the probability distributions of parameters from high-dimensional scattering data. We demonstrate the proposed method using inelastic neutron scattering with simulated and real experimental data, highlighting the method's ability to provide real-time parameter estimation with quantified uncertainties and to deliver informed experimental guidance that reduces experimental time while maximizing scientific output. This approach paves the way for accelerated discoveries in condensed matter through scattering measurements.

Inspired by advancements in natural language processing, we utilize self-supervised learning and an equivariant graph neural network to develop a unified platform for training generative models capable of generating inorganic crystal structures, as well as efficiently adapting to downstream tasks in material property prediction. To mitigate the challenge of evaluating the reliability of generated structures during training, we employ a generative adversarial network (GAN) with its discriminator being a cost-effective reliability evaluator, significantly enhancing model performance. We demonstrate the utility of our model in optimizing crystal structures under predefined conditions. Without external properties acquired experimentally or numerically, our model further displays its capability to help understand inorganic crystal formation by grouping chemically similar elements. This paper extends an invitation to further explore the scientific understanding of material structures through generative models, offering a fresh perspective on the scope and efficacy of machine learning in material science.

The increasing volumes of data produced at light sources such as the Linac Coherent Light Source (LCLS) enable the direct observation of materials and molecular assemblies at the length and timescales of molecular and atomic motion. This exponential increase in the scale and speed of data production is prohibitive to traditional analysis workflows that rely on scientists tuning parameters during live experiments to adapt data collection and analysis. User facilities will increasingly rely on the automated delivery of actionable information in real time for rapid experiment adaptation which presents a considerable challenge for data acquisition, data processing, data management, and workflow orchestration. In addition, the desire from researchers to accelerate science requires rapid analysis, dynamic integration of experiment and theory, the ability to visualize results in near real-time, and the introduction of ML and AI techniques. We present the LCLS-II Data System architecture which is designed to address these challenges via an adaptable data reduction pipeline (DRP) to reduce data volume on-thefly, online monitoring analysis software for real-time data visualization and experiment feedback, and the ability to scale to computing needs by utilizing local and remote compute resources, such as the ASCR Leadership Class Facilities, to enable quasi-real-time data analysis in minutes. We discuss the overall challenges facing LCLS, our ongoing work to develop a system responsive to these challenges, and our vision for future developments.

Abstract Advanced experimental measurements are crucial for driving theoretical developments and unveiling novel phenomena in condensed matter and materials physics, which often suffer from the scarcity of large-scale facility resources, such as x-ray or neutron scattering centers. To address these limitations, we introduce a methodology that leverages the Bayesian optimal experimental design paradigm to efficiently uncover key quantum spin fluctuation parameters from x-ray photon fluctuation spectroscopy (XPFS) data. Our method is compatible with existing theoretical simulation pipelines and can also be used in combination with fast machine learning surrogate models in the event that real-time simulations are unfeasible. Our numerical benchmarks demonstrate the superior performance in predicting model parameters and in delivering more informative measurements within limited experimental time. Our method can be adapted to many different types of experiments beyond XPFS and spin fluctuation studies, facilitating more efficient data collection and accelerating scientific discoveries.

The development of X-ray Free Electron Lasers (XFELs) has opened numerous opportunities to probe atomic structure and ultrafast dynamics of various materials. Single Particle Imaging (SPI) with XFELs enables the investigation of biological particles in their natural physiological states with unparalleled temporal resolution, while circumventing the need for cryogenic conditions or crystallization. However, reconstructing real-space structures from reciprocal-space x-ray diffraction data is highly challenging due to the absence of phase and orientation information, which is further complicated by weak scattering signals and considerable fluctuations in the number of photons per pulse. In this work, we present an end-to-end, self-supervised machine learning approach to recover particle orientations and estimate reciprocal space intensities from diffraction images only. Our method demonstrates great robustness under demanding experimental conditions with significantly enhanced reconstruction capabilities compared with conventional algorithms, and signifies a paradigm shift in SPI as currently practiced at XFELs.

Abstract Understanding the nature and origin of collective excitations in materials is of fundamental importance for unraveling the underlying physics of a many-body system. Excitation spectra are usually obtained by measuring the dynamical structure factor, S ( Q , ω ), using inelastic neutron or x-ray scattering techniques and are analyzed by comparing the experimental results against calculated predictions. We introduce a data-driven analysis tool which leverages ‘neural implicit representations’ that are specifically tailored for handling spectrographic measurements and are able to efficiently obtain unknown parameters from experimental data via automatic differentiation. In this work, we employ linear spin wave theory simulations to train a machine learning platform, enabling precise exchange parameter extraction from inelastic neutron scattering data on the square-lattice spin-1 antiferromagnet La 2 NiO 4 , showcasing a viable pathway towards automatic refinement of advanced models for ordered magnetic systems.

The interplay between magnetism and electronic band topology enriches topological phases and has promising applications. However, the role of topology in magnetic fluctuations has been elusive. Here, we report evidence for topology stabilized magnetism above the magnetic transition temperature in magnetic Weyl semimetal candidate CeAlGe. Electrical transport, thermal transport, resonant elastic X-ray scattering, and dilatometry consistently indicate the presence of locally correlated magnetism within a narrow temperature window well above the thermodynamic magnetic transition temperature. The wavevector of this short-range order is consistent with the nesting condition of topological Weyl nodes, suggesting that it arises from the interaction between magnetic fluctuations and the emergent Weyl fermions. Effective field theory shows that this topology stabilized order is wavevector dependent and can be stabilized when the interband Weyl fermion scattering is dominant. Our work highlights the role of electronic band topology in stabilizing magnetic order even in the classically disordered regime.

Advanced experimental measurements are crucial for driving theoretical developments and unveiling novel phenomena in condensed matter and material physics, which often suffer from the scarcity of facility resources and increasing complexities. To address the limitations, we introduce a methodology that combines machine learning with Bayesian optimal experimental design (BOED), exemplified with x-ray photon fluctuation spectroscopy (XPFS) measurements for spin fluctuations. Our method employs a neural network model for large-scale spin dynamics simulations for precise distribution and utility calculations in BOED. The capability of automatic differentiation from the neural network model is further leveraged for more robust and accurate parameter estimation. Our numerical benchmarks demonstrate the superior performance of our method in guiding XPFS experiments, predicting model parameters, and yielding more informative measurements within limited experimental time. Although focusing on XPFS and spin fluctuations, our method can be adapted to other experiments, facilitating more efficient data collection and accelerating scientific discoveries.

The observation and description of collective excitations in solids is a fundamental issue when seeking to understand the physics of a many-body system. Analysis of these excitations is usually carried out by measuring the dynamical structure factor, S(Q, $ω$), with inelastic neutron or x-ray scattering techniques and comparing this against a calculated dynamical model. Here, we develop an artificial intelligence framework which combines a neural network trained to mimic simulated data from a model Hamiltonian with automatic differentiation to recover unknown parameters from experimental data. We benchmark this approach on a Linear Spin Wave Theory (LSWT) simulator and advanced inelastic neutron scattering data from the square-lattice spin-1 antiferromagnet La$_2$NiO$_4$. We find that the model predicts the unknown parameters with excellent agreement relative to analytical fitting. In doing so, we illustrate the ability to build and train a differentiable model only once, which then can be applied in real-time to multi-dimensional scattering data, without the need for human-guided peak finding and fitting algorithms. This prototypical approach promises a new technology for this field to automatically detect and refine more advanced models for ordered quantum systems.