Boris N. Slautin
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Boris N. Slautin has a total of 123 co-authors. Here are the top co-authors by number of shared publications:
- Sergei V. Kalinin
University of Tennessee
21 shared publications - Vladimir Ya. Shur
School of Natural Sciences and Mathematics, Ural Federal University , 620000 Ekaterinburg
10 shared publications - Andrei Leonidovitch Kholkin
Department of Physics & CICECO-Aveiro Institute of Materials, University of Aveiro, 3810-193 Aveiro, Portugal
8 shared publications - Denis Olegovich Alikin
CICECO Aveiro Institute of Materials Physics Department Campus De Santiago University of Aveiro Aveiro Portugal
8 shared publications - Mahshid Ahmadi
Institute for Advanced Materials and Manufacturing, Department of Materials Science and Engineering, University of Tennessee, Knoxville, TN 37996, USA
6 shared publications - Yu Liu
Department of Materials Science and Engineering, University of Tennessee 1 , Knoxville, Tennessee 37996
6 shared publications
Latest Publications
- 20 November 2026
- Published by Springer Nature in Nature Machine Intelligence
Enabling lab autonomy with reward functions and executable workflows
- 20 November 2026
- Published by Springer Nature in Nature Machine Intelligence
Enabling lab autonomy with reward functions and executable workflows
- Open Access
- 14 October 2026
- Published by Royal Society of Chemistry (RSC) in Digital Discovery
- Vol. 5 (10) , 4366-4380
Accelerated materials discovery through random-library exploration by automated scanning transmission electron microscopy
Random materials libraries and automated STEM enable accelerated screening and characterization across high-dimensional compositional spaces.
- 14 October 2026
- Published by Royal Society of Chemistry (RSC) in Digital Discovery
- Vol. 5 (10) , 4366-4380
Open AccessAccelerated materials discovery through random-library exploration by automated scanning transmission electron microscopy
Random materials libraries and automated STEM enable accelerated screening and characterization across high-dimensional compositional spaces.
- Open Access
- 29 September 2026
- Published by American Chemical Society (ACS) in ChemRxiv
From Publications to Physical Knowledge Models: Transferable physical understanding for AI-native experimental science
Scientific communication has repeatedly accelerated when the objects used to preserve and exchange knowledge became easier to validate, search, and reuse. The scientific paper remains a human-readable compression of a richer scientific state: quantitative physical and observational models, their uncertainty and validity, and the evidence that supports or challenges them. We propose Physical Knowledge Models (PKMs) as explicit, versioned, composable, and transferable representations of this state. The intellectual object per se is not fundamentally new: expert mental models, comprehensive textbooks, and domain model repositories already integrate equations, mechanisms, observations, and validity judgments. The proposed advance is to make that integrated understanding machine-readable and directly connected to experimentation. Broad PKMs can be projected into reduced experiment-specific models; local evidence produces versioned knowledge proposals rather than silently overwriting community knowledge; and a Knowledge Integration Layer aligns partially overlapping models, evaluates evidence and beliefs, preserves competing branches, and exposes contradictions requiring further tests. Because the ontology itself can change, PKMs retain persistent links to original measurements, sample histories, and instrument and decision traces so that prior evidence can be reinterpreted under revised concepts and models. Coupled to experimental planning, this creates a structured route from shared knowledge to local experiment and back. Data, code, digital twins, and agents remain essential, but they become mechanisms for populating, interrogating, testing, and updating durable physical knowledge rather than substitutes for it.
- 29 September 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessFrom Publications to Physical Knowledge Models: Transferable physical understanding for AI-native experimental science
Scientific communication has repeatedly accelerated when the objects used to preserve and exchange knowledge became easier to validate, search, and reuse. The scientific paper remains a human-readable compression of a richer scientific state: quantitative physical and observational models, their uncertainty and validity, and the evidence that supports or challenges them. We propose Physical Knowledge Models (PKMs) as explicit, versioned, composable, and transferable representations of this state. The intellectual object per se is not fundamentally new: expert mental models, comprehensive textbooks, and domain model repositories already integrate equations, mechanisms, observations, and validity judgments. The proposed advance is to make that integrated understanding machine-readable and directly connected to experimentation. Broad PKMs can be projected into reduced experiment-specific models; local evidence produces versioned knowledge proposals rather than silently overwriting community knowledge; and a Knowledge Integration Layer aligns partially overlapping models, evaluates evidence and beliefs, preserves competing branches, and exposes contradictions requiring further tests. Because the ontology itself can change, PKMs retain persistent links to original measurements, sample histories, and instrument and decision traces so that prior evidence can be reinterpreted under revised concepts and models. Coupled to experimental planning, this creates a structured route from shared knowledge to local experiment and back. Data, code, digital twins, and agents remain essential, but they become mechanisms for populating, interrogating, testing, and updating durable physical knowledge rather than substitutes for it.
- 17 August 2026
- Published by American Chemical Society (ACS) in ACS Nano
- Vol. 20 (34) , 23740-23747
Bayesian Conavigation of a Computational Physical Model and Atomic Force Microscopy Experiment to Autonomously Survey a Combinatorial Materials Library
Building autonomous experiment workflows requires transcending data-driven surrogate models to incorporate and dynamically refine physical theory during exploration. Here we demonstrate the fully automated experimental realization of Bayesian conavigation─a framework in which an autonomous agent simultaneously runs a physical experiment and a computationally expensive physical model. Using an automated Atomic Force Microscopy (AFM) platform coupled to a kinetic Monte Carlo (kMC) model of thin-film growth, the system infers a set of effective bond energies for the (CrTaWV)x–Mo(1–x) pseudobinary combinatorial library, progressively adjusting the kMC parameters to decrease the epistemic disparity between simulation and experiment. This real-time theoretical refinement enables the kMC model to capture the behavior of the specific materials system and reveals the mechanistic role of heterobonding in governing surface diffusion. Together, these results establish conavigation as a general strategy for tightly integrating physical models with autonomous experimental platforms to produce interpretable and continually self-correcting theoretical modeling of complex materials systems.
- 17 August 2026
- Published by American Chemical Society (ACS) in ACS Nano
- Vol. 20 (34) , 23740-23747
Bayesian Conavigation of a Computational Physical Model and Atomic Force Microscopy Experiment to Autonomously Survey a Combinatorial Materials Library
Building autonomous experiment workflows requires transcending data-driven surrogate models to incorporate and dynamically refine physical theory during exploration. Here we demonstrate the fully automated experimental realization of Bayesian conavigation─a framework in which an autonomous agent simultaneously runs a physical experiment and a computationally expensive physical model. Using an automated Atomic Force Microscopy (AFM) platform coupled to a kinetic Monte Carlo (kMC) model of thin-film growth, the system infers a set of effective bond energies for the (CrTaWV)x–Mo(1–x) pseudobinary combinatorial library, progressively adjusting the kMC parameters to decrease the epistemic disparity between simulation and experiment. This real-time theoretical refinement enables the kMC model to capture the behavior of the specific materials system and reveals the mechanistic role of heterobonding in governing surface diffusion. Together, these results establish conavigation as a general strategy for tightly integrating physical models with autonomous experimental platforms to produce interpretable and continually self-correcting theoretical modeling of complex materials systems.
- Open Access
- 17 August 2026
- Published by Royal Society of Chemistry (RSC) in Materials Advances
- Vol. 7 (16) , 8419-8429
High-quality (PEA) 2 (FA)Pb 2 I 7 Ruddlesden–Popper hybrid perovskite single crystals: growth, structural and optical properties
This work reports the growth of 2D (PEA) 2 (FA)Pb 2 I 7 single crystals via mild solvent and characterization of the structure–property relationship.
- 17 August 2026
- Published by Royal Society of Chemistry (RSC) in Materials Advances
- Vol. 7 (16) , 8419-8429
Open AccessHigh-quality (PEA) 2 (FA)Pb 2 I 7 Ruddlesden–Popper hybrid perovskite single crystals: growth, structural and optical properties
This work reports the growth of 2D (PEA) 2 (FA)Pb 2 I 7 single crystals via mild solvent and characterization of the structure–property relationship.
- Open Access
- 26 July 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Utility of Control: Scanning Probe Microscopy as a Testbed for Tool Creation in Autonomous Experimental Science
The central task of physical characterization tools is to render unknown physical environments observable and explorable. Traditionally, progress in this field has been framed in terms of the physical instrumentation, including novel modalities, better probes, detectors, and electronics, stability, resolution, and sensitivity. However, there is an equally important control problem: how the instrument should act, at what time scale, and with what level of complexity to extract useful scientific information. Here, we introduce the utility of control as a general principle for autonomous experimentation, defined as maximizing useful scientific outcome while minimizing the required control bandwidth, action-space dimensionality, command description length, and fragility under uncertainty. This naturally defines a Pareto problem in which learning, robustness, physical cost, and control complexity must be balanced. We discuss these issues using scanning probe microscopy as an example and show how they motivate the design of low-dimensional, physically meaningful control planes and operational primitives that compress complex low-level actions into short semantic commands while preserving the degrees of freedom needed for discovery. This perspective connects bottom-up optimization, top-down agentic planning, and tool creation, where optimizers operate within selected control planes, agents choose which primitives should be used and why, and tool creation builds new compressed primitives that generate scientific utility without exposing unnecessary low-level bandwidth. We argue that the development of the next generation of autonomous microscopes and physical characterization tools more broadly will be defined by utility of control.
- 26 July 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessUtility of Control: Scanning Probe Microscopy as a Testbed for Tool Creation in Autonomous Experimental Science
The central task of physical characterization tools is to render unknown physical environments observable and explorable. Traditionally, progress in this field has been framed in terms of the physical instrumentation, including novel modalities, better probes, detectors, and electronics, stability, resolution, and sensitivity. However, there is an equally important control problem: how the instrument should act, at what time scale, and with what level of complexity to extract useful scientific information. Here, we introduce the utility of control as a general principle for autonomous experimentation, defined as maximizing useful scientific outcome while minimizing the required control bandwidth, action-space dimensionality, command description length, and fragility under uncertainty. This naturally defines a Pareto problem in which learning, robustness, physical cost, and control complexity must be balanced. We discuss these issues using scanning probe microscopy as an example and show how they motivate the design of low-dimensional, physically meaningful control planes and operational primitives that compress complex low-level actions into short semantic commands while preserving the degrees of freedom needed for discovery. This perspective connects bottom-up optimization, top-down agentic planning, and tool creation, where optimizers operate within selected control planes, agents choose which primitives should be used and why, and tool creation builds new compressed primitives that generate scientific utility without exposing unnecessary low-level bandwidth. We argue that the development of the next generation of autonomous microscopes and physical characterization tools more broadly will be defined by utility of control.
- 14 July 2026
- Published by AIP Publishing in Journal of Applied Physics
- Vol. 140 (2)
Attention-based explainability for structure–property relationships
Machine learning methods are emerging as a universal paradigm for constructing correlative structure–property relationships in materials science based on multimodal characterization. However, this necessitates the development of methods for the physical interpretability of the resulting correlative models. Here, we demonstrate the potential of attention-based neural networks for revealing structure–property relationships and the underlying physical mechanisms, using the ferroelectric properties of PbTiO3 thin films as a case study. Through the analysis of attention scores, we disentangle the influence of distinct domain patterns on the polarization switching process. The attention-based Transformer model is explored both as a direct interpretability tool and as a surrogate for explaining representations learned via unsupervised machine learning, enabling the identification of physically grounded correlations. We compare attention-derived interpretability scores with classical SHapley Additive exPlanations analysis and show that, in contrast to applications in natural language processing, attention mechanisms in materials science exhibit high efficiency in highlighting meaningful structural features.
- 14 July 2026
- Published by AIP Publishing in Journal of Applied Physics
- Vol. 140 (2)
Attention-based explainability for structure–property relationships
Machine learning methods are emerging as a universal paradigm for constructing correlative structure–property relationships in materials science based on multimodal characterization. However, this necessitates the development of methods for the physical interpretability of the resulting correlative models. Here, we demonstrate the potential of attention-based neural networks for revealing structure–property relationships and the underlying physical mechanisms, using the ferroelectric properties of PbTiO3 thin films as a case study. Through the analysis of attention scores, we disentangle the influence of distinct domain patterns on the polarization switching process. The attention-based Transformer model is explored both as a direct interpretability tool and as a surrogate for explaining representations learned via unsupervised machine learning, enabling the identification of physically grounded correlations. We compare attention-derived interpretability scores with classical SHapley Additive exPlanations analysis and show that, in contrast to applications in natural language processing, attention mechanisms in materials science exhibit high efficiency in highlighting meaningful structural features.
- Open Access
- 6 July 2026
- Published by Springer Nature in npj Computational Materials
- Vol. 12 (1) , 231
Integrating Predictive and Generative Capabilities by Latent Space Design via the DKL-VAE Model
: We introduce a deep Kernel learning variational autoencoder (DKL-VAE) framework that integrates the generative power of a variational autoencoder (VAE) with the predictive nature of deep Kernel learning (DKL). The VAE learns a latent representation of high-dimensional data, enabling the generation of novel structures, while DKL refines this latent space by structuring it in alignment with target properties through Gaussian process (GP) regression. This approach preserves the generative capabilities of the VAE while enhancing its latent space for GP-based property prediction. We evaluate the framework on two datasets: a structured card dataset with predefined variational factors and the QM9 molecular dataset, where enthalpy serves as the target function for optimization. The model demonstrates high-precision property prediction and enables the generation of novel out-of-training subset structures with desired characteristics. The DKL-VAE framework offers a promising approach for high-throughput material discovery and molecular design, balancing structured latent space organization with generative flexibility.
- 6 July 2026
- Published by Springer Nature in npj Computational Materials
- Vol. 12 (1) , 231
Open AccessIntegrating Predictive and Generative Capabilities by Latent Space Design via the DKL-VAE Model
: We introduce a deep Kernel learning variational autoencoder (DKL-VAE) framework that integrates the generative power of a variational autoencoder (VAE) with the predictive nature of deep Kernel learning (DKL). The VAE learns a latent representation of high-dimensional data, enabling the generation of novel structures, while DKL refines this latent space by structuring it in alignment with target properties through Gaussian process (GP) regression. This approach preserves the generative capabilities of the VAE while enhancing its latent space for GP-based property prediction. We evaluate the framework on two datasets: a structured card dataset with predefined variational factors and the QM9 molecular dataset, where enthalpy serves as the target function for optimization. The model demonstrates high-precision property prediction and enables the generation of novel out-of-training subset structures with desired characteristics. The DKL-VAE framework offers a promising approach for high-throughput material discovery and molecular design, balancing structured latent space organization with generative flexibility.
- Open Access
- 10 April 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Lignin-stabilized polystyrene dispersions as nanoplastic reference materials
In natural aquatic systems, nanoplastics rapidly acquire organic coatings or so-called eco-coronas that govern their environmental behavior. In contrast, laboratory-made reference nanoplastics rely on artificial surfactants, forming non-natural coatings that enable small size and colloidal stability while failing to replicate these environmentally relevant interfaces. To address this limitation, we developed environmentally relevant polystyrene (PS) nanoplastic models by incorporating three lignin biopolymers, including softwood kraft lignin (SKL), hardwood kraft lignin (HKL), and acetylated softwood kraft lignin (ACL), through anti-solvent precipitation. By combining gravimetric analysis, machine learning-assisted particle sizing, and high-resolution interfacial analysis, we propose mechanisms governing dispersion stability. Systems that formed thick or continuous layers around PS particles showed poor performance, as ACL induced bridging-driven aggregation (~42% recovery), while HKL promoted strong self-association (~22% recovery). In contrast, an optimal 2% SKL formulation achieved high mass recovery (~90%, compared to ~46% for the uncoated PS benchmark) and produced uniform, well-dispersed particles (median size ~110 nm). Notably, SKL formed a thin stabilizing layer together with a distinct formation of lignin nanoparticle surrounding the PS cores. This interfacial architecture enhanced electrosteric stabilization while maintaining colloidal stability. At higher SKL loadings, excess lignin disrupted this balance, leading to reduced stability likely due to depletion flocculation caused by enrichment of free lignin nanoparticles in the continuous phase. Overall, this work provides a mechanistic basis for designing environmentally relevant nanoplastics and highlights the potential of lignin as a sustainable material for stabilizing and studying such systems.
- 10 April 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessLignin-stabilized polystyrene dispersions as nanoplastic reference materials
In natural aquatic systems, nanoplastics rapidly acquire organic coatings or so-called eco-coronas that govern their environmental behavior. In contrast, laboratory-made reference nanoplastics rely on artificial surfactants, forming non-natural coatings that enable small size and colloidal stability while failing to replicate these environmentally relevant interfaces. To address this limitation, we developed environmentally relevant polystyrene (PS) nanoplastic models by incorporating three lignin biopolymers, including softwood kraft lignin (SKL), hardwood kraft lignin (HKL), and acetylated softwood kraft lignin (ACL), through anti-solvent precipitation. By combining gravimetric analysis, machine learning-assisted particle sizing, and high-resolution interfacial analysis, we propose mechanisms governing dispersion stability. Systems that formed thick or continuous layers around PS particles showed poor performance, as ACL induced bridging-driven aggregation (~42% recovery), while HKL promoted strong self-association (~22% recovery). In contrast, an optimal 2% SKL formulation achieved high mass recovery (~90%, compared to ~46% for the uncoated PS benchmark) and produced uniform, well-dispersed particles (median size ~110 nm). Notably, SKL formed a thin stabilizing layer together with a distinct formation of lignin nanoparticle surrounding the PS cores. This interfacial architecture enhanced electrosteric stabilization while maintaining colloidal stability. At higher SKL loadings, excess lignin disrupted this balance, leading to reduced stability likely due to depletion flocculation caused by enrichment of free lignin nanoparticles in the continuous phase. Overall, this work provides a mechanistic basis for designing environmentally relevant nanoplastics and highlights the potential of lignin as a sustainable material for stabilizing and studying such systems.
- Open Access
- 5 February 2026
- Published by AIP Publishing in APL Machine Learning
- Vol. 4 (1)
Bringing ML to the real world: Rewards are all we need
Realizing the promise of artificial intelligence (AI) to accelerate scientific progress and deliver technological impact depends on how effectively AI can be integrated into real-world decision-making processes. As Peter Norvig states, “Somewhat remarkably, almost all AI research until very recently has assumed that the performance measure can be exactly and correctly specified in the form of a utility or reward function” [S. J. Russell and P. Norvig, Artificial Intelligence: A Modern Approach (Pearson, 2020)]. Once the reward function is known, any problem can be formulated as an optimization or search problem—the areas well explored within the AI community. However, while the long-term objectives of a specific activity are often well defined, constructing short-term rewards that remain aligned with those goals and consistent with real-world constraints remains a major unresolved challenge. Such alignment has been achieved in domains such as chess, Go, and supervised machine learning problems, where objectives are well defined and easily simulated. However, no universal solution exists for defining intermediate rewards for complex, evolving scientific goals, which remains an open challenge. Correspondingly, the key to operationalizing automated instruments, integrating multi-instrument self-driving laboratories, and building geographically distributed research facilities is to generate experiment-aligned, probabilistic, domain-specific reward functions. These rewards must be consistent with long-term experimental objectives while remaining actionable on the timescales of decision-making on laboratory tools in microscopy and materials synthesis labs—proper reward definition operationalizes scientific intent. Here, we review the extant reward structures in the physical sciences and summarize opportunities for reward design informed by physical principles, human heuristics, and LLM-based reasoning.
- 5 February 2026
- Published by AIP Publishing in APL Machine Learning
- Vol. 4 (1)
Open AccessBringing ML to the real world: Rewards are all we need
Realizing the promise of artificial intelligence (AI) to accelerate scientific progress and deliver technological impact depends on how effectively AI can be integrated into real-world decision-making processes. As Peter Norvig states, “Somewhat remarkably, almost all AI research until very recently has assumed that the performance measure can be exactly and correctly specified in the form of a utility or reward function” [S. J. Russell and P. Norvig, Artificial Intelligence: A Modern Approach (Pearson, 2020)]. Once the reward function is known, any problem can be formulated as an optimization or search problem—the areas well explored within the AI community. However, while the long-term objectives of a specific activity are often well defined, constructing short-term rewards that remain aligned with those goals and consistent with real-world constraints remains a major unresolved challenge. Such alignment has been achieved in domains such as chess, Go, and supervised machine learning problems, where objectives are well defined and easily simulated. However, no universal solution exists for defining intermediate rewards for complex, evolving scientific goals, which remains an open challenge. Correspondingly, the key to operationalizing automated instruments, integrating multi-instrument self-driving laboratories, and building geographically distributed research facilities is to generate experiment-aligned, probabilistic, domain-specific reward functions. These rewards must be consistent with long-term experimental objectives while remaining actionable on the timescales of decision-making on laboratory tools in microscopy and materials synthesis labs—proper reward definition operationalizes scientific intent. Here, we review the extant reward structures in the physical sciences and summarize opportunities for reward design informed by physical principles, human heuristics, and LLM-based reasoning.