Yunkai Sun
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yunkai Sun has a total of 76 co-authors. Here are the top co-authors by number of shared publications:
Giovanni Zangari
University of Virginia
14 shared publications- Giovanni Zangari
Department of Materials Science and Engineering, University of Virginia, Charlottesville, VA 22904, USA
9 shared publications - Zhenzhen Yang
Center for Steel Electrification by Electrosynthesis, Argonne National Laboratory, Lemont, IL 60439, USA
8 shared publications Brian J. Ingram
Center for Steel Electrification by Electrosynthesis
7 shared publications- Hassan Harb
Argonne National Laboratory
6 shared publications - Rajeev Surendran Assary
Argonne National Laboratory
6 shared publications
Latest Publications
- Open Access
- 24 September 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Judgment by agents, numbers by code: auditable automated speciation analysis towards a chemistry world model
Chemical speciation governs the reactivity, solubility, and toxicity of a solution, yet assembling speciation calculations from evaluated thermodynamic data remains a manual expert task whose reasoning is discarded once the figure is drawn. To mirror real-world research, automation belongs within a chemistry world model (CWM) where language agents curate and analyze data. As a first module, our agentic framework converts plain-language questions into deterministic equilibrium calculations grounded in NIST Standard Reference Database 46 (SRD-46). Agents resolve intent, curate records, adjudicate duplicates, design calculations, and estimate values beyond database coverage through induction-deduction-validation loops, while deterministic code validates every agentic commitment before it reaches the solver and compacts diagram topology into text readable without a vision stage. Across four case studies of escalating difficulty, quoted constants, patched records, and boundary coordinates reproduced their sources essentially exactly, whereas the same model without the framework missed phases or failed to build a solver. Residual errors surfaced only as auditable narration flaws, from misread topology to a sign-flipped hydrogen-evolution argument. The framework serves as a human-in-the-loop instrument delivering consensus-grade starting points, turning evaluated databases into active discovery instruments, mapping where language-model chemical reasoning still fails, and providing a general agentic template for finetuning.
- 24 September 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessJudgment by agents, numbers by code: auditable automated speciation analysis towards a chemistry world model
Chemical speciation governs the reactivity, solubility, and toxicity of a solution, yet assembling speciation calculations from evaluated thermodynamic data remains a manual expert task whose reasoning is discarded once the figure is drawn. To mirror real-world research, automation belongs within a chemistry world model (CWM) where language agents curate and analyze data. As a first module, our agentic framework converts plain-language questions into deterministic equilibrium calculations grounded in NIST Standard Reference Database 46 (SRD-46). Agents resolve intent, curate records, adjudicate duplicates, design calculations, and estimate values beyond database coverage through induction-deduction-validation loops, while deterministic code validates every agentic commitment before it reaches the solver and compacts diagram topology into text readable without a vision stage. Across four case studies of escalating difficulty, quoted constants, patched records, and boundary coordinates reproduced their sources essentially exactly, whereas the same model without the framework missed phases or failed to build a solver. Residual errors surfaced only as auditable narration flaws, from misread topology to a sign-flipped hydrogen-evolution argument. The framework serves as a human-in-the-loop instrument delivering consensus-grade starting points, turning evaluated databases into active discovery instruments, mapping where language-model chemical reasoning still fails, and providing a general agentic template for finetuning.
- Open Access
- 17 September 2026
- Published by American Chemical Society (ACS) in ChemRxiv
A traceable agentic framework for chemistry research query and analysis with ThermoML database
Reliable use of thermophysical databases in exploratory chemistry requires resolving heterogeneous identifiers, reconciling experimental conditions, and preserving numerical provenance across large, nested records. We present a hierarchical large language model (LLM) agent framework, implemented with a custom ReAct reasoning engine, that translates naturallanguage research questions into traceable retrieval and analysis workflows over the NIST ThermoML database. The framework combines hierarchical canonical identifiers, queryconditioned evidence compaction, deterministic inspection ledgers and numerical-grounding gates, and tool-assisted Redlich–Kister fitting. Five case studies are discussed in detail, representing chemistry tasks spanning data-rich and data-sparse retrieval, binary-mixture correlation, and open-ended cosolvent screening. For an ethanol–water equimolar density query, the system reduced ~190k characters of search output to 2k and precisely recovered a measured density of 859.3 kg m-3 at an ethanol mole fraction of 0.498, 298.15 K, and 100 kPa. For methanolethanol viscosity, it selected condition-aligned mixture and pure-component records, examined the fitting results at different orders, and produced a 2nd-order Redlich–Kister fit with 𝑅 = 0.982. Results for an open-ended iron-electrodeposition electrolyte design query synthesized evidence from approximately 12k measurements across 200 data blocks. However, with numerical grounding disabled for this specific case, the answer contained severe numerical hallucinations despite qualitatively coherent reasoning. These cases establish the feasibility of agent-assisted ThermoML research and identify identity management, evidence compaction, and deterministic numerical provenance as central design requirements for reliable chemistry-database research agents with some level of abstraction chemistry reasoning.
- 17 September 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessA traceable agentic framework for chemistry research query and analysis with ThermoML database
Reliable use of thermophysical databases in exploratory chemistry requires resolving heterogeneous identifiers, reconciling experimental conditions, and preserving numerical provenance across large, nested records. We present a hierarchical large language model (LLM) agent framework, implemented with a custom ReAct reasoning engine, that translates naturallanguage research questions into traceable retrieval and analysis workflows over the NIST ThermoML database. The framework combines hierarchical canonical identifiers, queryconditioned evidence compaction, deterministic inspection ledgers and numerical-grounding gates, and tool-assisted Redlich–Kister fitting. Five case studies are discussed in detail, representing chemistry tasks spanning data-rich and data-sparse retrieval, binary-mixture correlation, and open-ended cosolvent screening. For an ethanol–water equimolar density query, the system reduced ~190k characters of search output to 2k and precisely recovered a measured density of 859.3 kg m-3 at an ethanol mole fraction of 0.498, 298.15 K, and 100 kPa. For methanolethanol viscosity, it selected condition-aligned mixture and pure-component records, examined the fitting results at different orders, and produced a 2nd-order Redlich–Kister fit with 𝑅 = 0.982. Results for an open-ended iron-electrodeposition electrolyte design query synthesized evidence from approximately 12k measurements across 200 data blocks. However, with numerical grounding disabled for this specific case, the answer contained severe numerical hallucinations despite qualitatively coherent reasoning. These cases establish the feasibility of agent-assisted ThermoML research and identify identity management, evidence compaction, and deterministic numerical provenance as central design requirements for reliable chemistry-database research agents with some level of abstraction chemistry reasoning.
- Open Access
- 24 August 2026
- Published by American Chemical Society (ACS) in ChemRxiv
A Generalizable Framework for Quantifying Uncertainty and Bias in Curve Aggregation for High-Throughput Electrodeposition Data
Experimental reproducibility in electrochemistry is fundamentally limited by process uncertainty. This challenge, especially pronounced in complex systems like Fe-based multi-metal electrodeposition, is a major bottleneck for high-throughput and autonomous materials discovery. To address this, we present a generalizable curve aggregation framework that combines Gaussian Process Regression (GPR) with Functional Data Analysis (FDA) to aggregate replicate measurements while preserving within-curve correlation structure. The method represents each transient as a continuous function and iteratively optimizes curve-wise weights to emphasize consistent signals and suppress outliers. We validate the framework using potentiostatic transients from near-equiatomic CoCuFeNi electrodeposition across seven deposition conditions. The framework successfully aggregated 226 transients, quantified experimental variance and transient shape, and showed performance comparable to a more computationally intensive but physically robust first-principle measurement model. We find that uncertainty, bias, and convergence difficulty evolve with dataset size resembling a Dunning–Kruger-like pattern, where small datasets appear deceptively certain, followed by a maximum convergence difficulty peak, then converge to a stable final aggregate. For this system, the variance stabilized at approximately 𝑁 ≈ 35–40 measurements, providing a practical criterion for data sufficiency. The GPR-FDA framework is broadly applicable to any experimental field involving replicate functional data that must be aggregated without losing correlation information.
- 24 August 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessA Generalizable Framework for Quantifying Uncertainty and Bias in Curve Aggregation for High-Throughput Electrodeposition Data
Experimental reproducibility in electrochemistry is fundamentally limited by process uncertainty. This challenge, especially pronounced in complex systems like Fe-based multi-metal electrodeposition, is a major bottleneck for high-throughput and autonomous materials discovery. To address this, we present a generalizable curve aggregation framework that combines Gaussian Process Regression (GPR) with Functional Data Analysis (FDA) to aggregate replicate measurements while preserving within-curve correlation structure. The method represents each transient as a continuous function and iteratively optimizes curve-wise weights to emphasize consistent signals and suppress outliers. We validate the framework using potentiostatic transients from near-equiatomic CoCuFeNi electrodeposition across seven deposition conditions. The framework successfully aggregated 226 transients, quantified experimental variance and transient shape, and showed performance comparable to a more computationally intensive but physically robust first-principle measurement model. We find that uncertainty, bias, and convergence difficulty evolve with dataset size resembling a Dunning–Kruger-like pattern, where small datasets appear deceptively certain, followed by a maximum convergence difficulty peak, then converge to a stable final aggregate. For this system, the variance stabilized at approximately 𝑁 ≈ 35–40 measurements, providing a practical criterion for data sufficiency. The GPR-FDA framework is broadly applicable to any experimental field involving replicate functional data that must be aggregated without losing correlation information.
- Open Access
- 9 July 2026
- Published by American Chemical Society (ACS) in ChemRxiv
An Autonomous Agentic Query Framework with MCP Tools and Modernized NIST SRD-46 for Accelerating Aqueous Speciation Analysis
Large language models (LLMs) could automate reasoning over structured scientific databases in chemistry and materials science, but this requires purpose-built agentic architectures, not simply better models. We present an agentic workflow for systematically querying and extracting data from the NIST Standard Reference Database 46 (SRD-46) to perform autonomous speciation and thermodynamic analysis, with three contributions: (1) a hierarchical toolset, served by Model Context Protocol (MCP) framework, which converts natural-language scientific questions into validated SQL over a database enriched with canonical chemical identifiers, equilibrium-reaction maps, and fingerprint-similarity matrices; (2) an induction-deduction-validation reasoning loop with compactors that preserve key chemical context across multi-turn sessions under dynamic memory limits within the same session; and (3) a soft time/turn-budget hook that maintains coherent reasoning within resource constraints. The resulting framework is LLM-agnostic and spans queries from direct lookups to cross-species comparisons to open-ended tasks like estimating unknown Fe complexation constants and selecting ligands for acidic Fe electrodeposition. We benchmark four frontier models (Claude Opus 4.6, Claude Sonnet 4.6, GPT-5, and GPT-5.4) on 49 prompts across four complexity levels, repeated five times (~1,000 answers), using automated claim-level validation and regex-based tool-result matching. All models retrieve reliably on direct-lookup tasks, confirming the architecture bridges natural-language reasoning and structured database access. Beyond lookups, behaviors of models diverge: GPT-5.4 converges rapidly with high per-claim verifiability but limited exploratory depth, whereas Opus 4.6 pursues broader, cross-examined reasoning at the cost of slower convergence and more unsupported claims. This reveals a fundamental tradeoff between convergent efficiency and exploratory breadth on fuzzy scientific queries that scalar benchmarks cannot resolve. It also suggests that fuzzy scientific queries require expert-stewarded reasoning, not faster convergence criteria or higher benchmark metrics alone. Our results demonstrate that domain-expert-designed agentic architectures and evaluation frameworks are essential infrastructure for AI-driven scientific discovery, and that mixed-model deployment strategies may be needed to balance efficiency and explanatory depth for different classes of scientific questions.
- 9 July 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessAn Autonomous Agentic Query Framework with MCP Tools and Modernized NIST SRD-46 for Accelerating Aqueous Speciation Analysis
Large language models (LLMs) could automate reasoning over structured scientific databases in chemistry and materials science, but this requires purpose-built agentic architectures, not simply better models. We present an agentic workflow for systematically querying and extracting data from the NIST Standard Reference Database 46 (SRD-46) to perform autonomous speciation and thermodynamic analysis, with three contributions: (1) a hierarchical toolset, served by Model Context Protocol (MCP) framework, which converts natural-language scientific questions into validated SQL over a database enriched with canonical chemical identifiers, equilibrium-reaction maps, and fingerprint-similarity matrices; (2) an induction-deduction-validation reasoning loop with compactors that preserve key chemical context across multi-turn sessions under dynamic memory limits within the same session; and (3) a soft time/turn-budget hook that maintains coherent reasoning within resource constraints. The resulting framework is LLM-agnostic and spans queries from direct lookups to cross-species comparisons to open-ended tasks like estimating unknown Fe complexation constants and selecting ligands for acidic Fe electrodeposition. We benchmark four frontier models (Claude Opus 4.6, Claude Sonnet 4.6, GPT-5, and GPT-5.4) on 49 prompts across four complexity levels, repeated five times (~1,000 answers), using automated claim-level validation and regex-based tool-result matching. All models retrieve reliably on direct-lookup tasks, confirming the architecture bridges natural-language reasoning and structured database access. Beyond lookups, behaviors of models diverge: GPT-5.4 converges rapidly with high per-claim verifiability but limited exploratory depth, whereas Opus 4.6 pursues broader, cross-examined reasoning at the cost of slower convergence and more unsupported claims. This reveals a fundamental tradeoff between convergent efficiency and exploratory breadth on fuzzy scientific queries that scalar benchmarks cannot resolve. It also suggests that fuzzy scientific queries require expert-stewarded reasoning, not faster convergence criteria or higher benchmark metrics alone. Our results demonstrate that domain-expert-designed agentic architectures and evaluation frameworks are essential infrastructure for AI-driven scientific discovery, and that mixed-model deployment strategies may be needed to balance efficiency and explanatory depth for different classes of scientific questions.
- Open Access
- 1 June 2026
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 7 (3) , 030503
The ballad of LLM agents: philosophical reasoning for chemistry
Large language models (LLMs) show remarkable potential for scientific reasoning but often produce unreliable or scientifically unactionable outputs when faced with multi-step logic, domain grounding, and interpretability challenges, especially in complex fields like chemistry and materials science. Here, we introduce a framework of philosophical reasoning agents, inspired by canonical thinkers such as Socrates, Descartes, Kant, and Hume, to guide LLM behavior via structured prompt engineering. These agents embody distinct reasoning paradigms (dialectical inquiry, deductive logic, rule-based judgment, and empirical validation) and are evaluated across multiple chemistry subdomains, physical, analytical, general, inorganic, and organic chemistry, using the ChemBench benchmark. Our agentic prompting approach yields substantial accuracy gains on open-ended numerical chemistry questions, with gains of +11.5 percentage points for GPT-4o with Hume, +4.5 percentage points for GPT-5 with Kant, and +21.8 percentage points for GPT-5.1 with Socrates at the strict 1% error threshold, relative to the corresponding base models. Beyond accuracy, we observe benchmark-level model–agent performance patterns, suggesting that different prompting styles interact differently with each base model. These findings demonstrate that embedding philosophy-of-science principles into multi-agent frameworks can improve and produce interpretable, adaptive, and domain-aligned scientific LLMs.
- 1 June 2026
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 7 (3) , 030503
Open AccessThe ballad of LLM agents: philosophical reasoning for chemistry
Large language models (LLMs) show remarkable potential for scientific reasoning but often produce unreliable or scientifically unactionable outputs when faced with multi-step logic, domain grounding, and interpretability challenges, especially in complex fields like chemistry and materials science. Here, we introduce a framework of philosophical reasoning agents, inspired by canonical thinkers such as Socrates, Descartes, Kant, and Hume, to guide LLM behavior via structured prompt engineering. These agents embody distinct reasoning paradigms (dialectical inquiry, deductive logic, rule-based judgment, and empirical validation) and are evaluated across multiple chemistry subdomains, physical, analytical, general, inorganic, and organic chemistry, using the ChemBench benchmark. Our agentic prompting approach yields substantial accuracy gains on open-ended numerical chemistry questions, with gains of +11.5 percentage points for GPT-4o with Hume, +4.5 percentage points for GPT-5 with Kant, and +21.8 percentage points for GPT-5.1 with Socrates at the strict 1% error threshold, relative to the corresponding base models. Beyond accuracy, we observe benchmark-level model–agent performance patterns, suggesting that different prompting styles interact differently with each base model. These findings demonstrate that embedding philosophy-of-science principles into multi-agent frameworks can improve and produce interpretable, adaptive, and domain-aligned scientific LLMs.
- Open Access
- 16 April 2026
- Published by The Electrochemical Society in ECS Advances
- Vol. 5 (2)
Experimental Benchmarking for High-Reproducibility, Cross-Institutional Evaluation of Iron Redox Electrochemistry
We present a practical case study standardizing experimental protocols between collaborators with the goal of understanding ferrous iron (Fe2+) chemistry and improving the iron deposition reaction for energy-efficient, electrochemical iron production. The study of iron reactions can be difficult, as aqueous iron electrolytes exhibit complex behaviors that can lead to differing interpretation of ostensibly similar experiments. The question we want to answer: are we studying the same chemistry? Our protocols address inherent challenges such as the tendency for Fe2+ to spontaneously oxidize to ferric iron (Fe3+) and the production of hydrogen at the potentials of interest. Our standardized protocol, executed by four collaborators in different labs and institutions, yields high-reproducibility results, and identified glassy carbon electrode surface quality and Fe3+ impurities in the salt as key factors with outsized effects on cyclic voltammetry measurements. The process of developing the protocols helped to troubleshoot underlying issues that created poorer reversibility and reproducibility. This study highlights the fact that even nominally straightforward electrochemical systems can yield vastly different outcomes due to small differences in experimental preparation and serves as a useful example for creating transparent and achievable standards for the generation of reliable datasets that can be widely used and shared.
- 16 April 2026
- Published by The Electrochemical Society in ECS Advances
- Vol. 5 (2)
Open AccessExperimental Benchmarking for High-Reproducibility, Cross-Institutional Evaluation of Iron Redox Electrochemistry
We present a practical case study standardizing experimental protocols between collaborators with the goal of understanding ferrous iron (Fe2+) chemistry and improving the iron deposition reaction for energy-efficient, electrochemical iron production. The study of iron reactions can be difficult, as aqueous iron electrolytes exhibit complex behaviors that can lead to differing interpretation of ostensibly similar experiments. The question we want to answer: are we studying the same chemistry? Our protocols address inherent challenges such as the tendency for Fe2+ to spontaneously oxidize to ferric iron (Fe3+) and the production of hydrogen at the potentials of interest. Our standardized protocol, executed by four collaborators in different labs and institutions, yields high-reproducibility results, and identified glassy carbon electrode surface quality and Fe3+ impurities in the salt as key factors with outsized effects on cyclic voltammetry measurements. The process of developing the protocols helped to troubleshoot underlying issues that created poorer reversibility and reproducibility. This study highlights the fact that even nominally straightforward electrochemical systems can yield vastly different outcomes due to small differences in experimental preparation and serves as a useful example for creating transparent and achievable standards for the generation of reliable datasets that can be widely used and shared.
- Open Access
- 1 January 2026
- Published by Royal Society of Chemistry (RSC) in Digital Discovery
An autonomous agentic query framework with MCP tools and modernized NIST SRD-46 for accelerating aqueous speciation analysis
Snapshot of the core contents discussed in this work. Detailed benchmarking data and results are available in Zenodo codespace.
- 1 January 2026
- Published by Royal Society of Chemistry (RSC) in Digital Discovery
Open AccessAn autonomous agentic query framework with MCP tools and modernized NIST SRD-46 for accelerating aqueous speciation analysis
Snapshot of the core contents discussed in this work. Detailed benchmarking data and results are available in Zenodo codespace.
- Open Access
- 1 January 2026
- Published by Royal Society of Chemistry (RSC) in Journal of Materials Chemistry A
From molecular reorganization to nanoscale heterogeneity: the role of DMF in aqueous Fe electrodeposition electrolytes
DMF leaves the Fe( ii ) aquo shell and the electrode surface unchanged. It acts in the bulk, slowing transport and suppressing Fe( iii ) hydrolysis, widening the stability window of aqueous Fe electrodeposition electrolytes.
- 1 January 2026
- Published by Royal Society of Chemistry (RSC) in Journal of Materials Chemistry A
Open AccessFrom molecular reorganization to nanoscale heterogeneity: the role of DMF in aqueous Fe electrodeposition electrolytes
DMF leaves the Fe( ii ) aquo shell and the electrode surface unchanged. It acts in the bulk, slowing transport and suppressing Fe( iii ) hydrolysis, widening the stability window of aqueous Fe electrodeposition electrolytes.
- Open Access
- 24 December 2025
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 6 (4) , 045073
Towards philosophical reasoning with agentic LLMs: Socratic method for scientific assistance
As large language models (LLMs) become central tools in science, improving their reasoning capabilities is critical for meaningful and trustworthy applications. We introduce a Socratic agent for scientific reasoning, implemented through a structured system prompt that guides LLMs via classical principles of inquiry. Unlike typical prompt engineering or retrieval-based methods, our approach leverages definition, analogy, hypothesis elimination, and other Socratic techniques to generate more coherent, critical, and domain-aware responses. We evaluate the agent across diverse scientific domains and benchmark it on the abstraction and reasoning corpus challenge dataset, achieving 97.15% under a fixed prompting protocol and without fine-tuning or external tools. Expert evaluation shows improved reasoning depth, clarity, and adaptability over conventional LLM outputs, suggesting that structured prompting rooted in philosophical reasoning can improve the scientific utility of language models.
- 24 December 2025
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 6 (4) , 045073
Open AccessTowards philosophical reasoning with agentic LLMs: Socratic method for scientific assistance
As large language models (LLMs) become central tools in science, improving their reasoning capabilities is critical for meaningful and trustworthy applications. We introduce a Socratic agent for scientific reasoning, implemented through a structured system prompt that guides LLMs via classical principles of inquiry. Unlike typical prompt engineering or retrieval-based methods, our approach leverages definition, analogy, hypothesis elimination, and other Socratic techniques to generate more coherent, critical, and domain-aware responses. We evaluate the agent across diverse scientific domains and benchmark it on the abstraction and reasoning corpus challenge dataset, achieving 97.15% under a fixed prompting protocol and without fine-tuning or external tools. Expert evaluation shows improved reasoning depth, clarity, and adaptability over conventional LLM outputs, suggesting that structured prompting rooted in philosophical reasoning can improve the scientific utility of language models.
- 24 November 2025
- Published by The Electrochemical Society in ECS Meeting Abstracts
- Vol. MA2025-02 (22) , 1317
Exploration of the Electrodeposition of Smooth Near-Equiatomic CoCuFeNi from Glycine-Citrate Electrolytes
In this presentation, we will introduce the complete work of exploration of electrodeposition of smooth near-equiatomic CoCuFeNi multi-principal element alloy. The films with target composition (error Figure caption: (a) surface quality and (b) composition of the electrodeposited smooth near-equiatomic CoCuFeNi MPEA. Figure 1
- 24 November 2025
- Published by The Electrochemical Society in ECS Meeting Abstracts
- Vol. MA2025-02 (22) , 1317
Exploration of the Electrodeposition of Smooth Near-Equiatomic CoCuFeNi from Glycine-Citrate Electrolytes
In this presentation, we will introduce the complete work of exploration of electrodeposition of smooth near-equiatomic CoCuFeNi multi-principal element alloy. The films with target composition (error Figure caption: (a) surface quality and (b) composition of the electrodeposited smooth near-equiatomic CoCuFeNi MPEA. Figure 1