Enabling high throughput deep reinforcement learning with first principles to investigate catalytic reaction mechanisms DOI Creative Commons
Tian Lan, Huan Wang, Qi An

et al.

Nature Communications, Journal Year: 2024, Volume and Issue: 15(1)

Published: July 25, 2024

Exploring catalytic reaction mechanisms is crucial for understanding chemical processes, optimizing conditions, and developing more effective catalysts. We present a reaction-agnostic framework based on high-throughput deep reinforcement learning with first principles (HDRL-FP) that offers excellent generalizability investigating reactions. HDRL-FP introduces generalizable representation of reactions constructed solely from atomic positions, which are subsequently mapped to first-principles-derived potential energy landscapes. By leveraging thousands simultaneous simulations single GPU, enables rapid convergence the optimal path at low cost. Its effectiveness demonstrated through studies hydrogen nitrogen migration in Haber-Bosch ammonia synthesis Fe(111) surface. Our findings reveal Langmuir-Hinshelwood mechanism shares same transition state as Eley-Rideal H NH

Language: Английский

In Pursuit of the Exceptional: Research Directions for Machine Learning in Chemical and Materials Science DOI
Joshua Schrier, Alexander J. Norquist,

Tonio Buonassisi

et al.

Journal of the American Chemical Society, Journal Year: 2023, Volume and Issue: 145(40), P. 21699 - 21716

Published: Sept. 27, 2023

Exceptional molecules and materials with one or more extraordinary properties are both technologically valuable fundamentally interesting, because they often involve new physical phenomena compositions that defy expectations. Historically, exceptionality has been achieved through serendipity, but recently, machine learning (ML) automated experimentation have widely proposed to accelerate target identification synthesis planning. In this Perspective, we argue the data-driven methods commonly used today well-suited for optimization not realization of exceptional molecules. Finding such outliers should be possible using ML, only by shifting away from traditional ML approaches tweak composition, crystal structure, reaction pathway. We highlight case studies high-Tc oxide superconductors superhard demonstrate challenges ML-guided discovery discuss limitations automation task. then provide six recommendations development capable discovery: (i) Avoid tyranny middle focus on extrema; (ii) When data limited, qualitative predictions direction than interpolative accuracy; (iii) Sample what can made how make it defer optimization; (iv) Create room (and look) unexpected while pursuing your goal; (v) Try fill-in-the-blanks input output space; (vi) Do confuse human understanding model interpretability. conclude a description these integrated into workflows, which enable materials.

Language: Английский

Citations

46

Dataset Design for Building Models of Chemical Reactivity DOI Creative Commons
Priyanka Raghavan, Brittany C. Haas, Madeline E. Ruos

et al.

ACS Central Science, Journal Year: 2023, Volume and Issue: 9(12), P. 2196 - 2204

Published: Dec. 8, 2023

Models can codify our understanding of chemical reactivity and serve a useful purpose in the development new synthetic processes via, for example, evaluating hypothetical reaction conditions or silico substrate tolerance. Perhaps most determining factor is composition training data whether it sufficient to train model that make accurate predictions over full domain interest. Here, we discuss design datasets ways are conducive data-driven modeling, emphasizing idea set diversity generalizability rely on choice molecular representation. We additionally experimental constraints associated with generating common types chemistry how these considerations should influence dataset building.

Language: Английский

Citations

42

Open-Source Machine Learning in Computational Chemistry DOI Creative Commons
Alexander Hagg, Karl N. Kirschner

Journal of Chemical Information and Modeling, Journal Year: 2023, Volume and Issue: 63(15), P. 4505 - 4532

Published: July 19, 2023

The field of computational chemistry has seen a significant increase in the integration machine learning concepts and algorithms. In this Perspective, we surveyed 179 open-source software projects, with corresponding peer-reviewed papers published within last 5 years, to better understand topics being investigated by approaches. For each project, provide short description, link code, accompanying license type, whether training data resulting models are made publicly available. Based on those deposited GitHub repositories, most popular employed Python libraries identified. We hope that survey will serve as resource learn about or specific architectures thereof identifying accessible codes topic basis. To end, also include for generating fundamental learning. our observations considering three pillars collaborative work, open data, source (code), models, some suggestions community.

Language: Английский

Citations

28

Re-evaluating Retrosynthesis Algorithms with Syntheseus DOI

Krzysztof Maziarz,

Austin Tripp, Guoqing Liu

et al.

Faraday Discussions, Journal Year: 2024, Volume and Issue: unknown

Published: Jan. 1, 2024

Automated synthesis planning has recently re-emerged as a research area at the intersection of chemistry and machine learning. Despite appearance steady progress, we argue that imperfect benchmarks inconsistent comparisons mask systematic shortcomings existing techniques, unnecessarily hamper progress. To remedy this, present library with an extensive benchmarking framework, called SYNTHESEUS, which promotes best practice by default, enabling consistent meaningful evaluation single-step multi-step algorithms. We demonstrate capabilities SYNTHESEUS re-evaluating several previous retrosynthesis algorithms, find ranking state-of-the-art models changes in controlled experiments. end guidance for future works this area, call on community to engage discussion how improve planning.

Language: Английский

Citations

16

Research progress of machine learning in the field of photocatalysis applications DOI
Kun Li, Haoyuan Du, Lei Liu

et al.

Journal of Industrial and Engineering Chemistry, Journal Year: 2025, Volume and Issue: unknown

Published: April 1, 2025

Language: Английский

Citations

2

Designing Chemical Reaction Arrays Using Phactor and ChatGPT DOI
Babak Mahjour, Jillian Hoffstadt, Tim Cernak

et al.

Organic Process Research & Development, Journal Year: 2023, Volume and Issue: 27(8), P. 1510 - 1516

Published: Aug. 1, 2023

High-throughput experimentation is a common practice in the optimization of chemical synthesis. Chemists design reaction arrays to optimize yield couplings between building blocks. Popular reactions used pharmaceutical research include amide coupling, Suzuki and Buchwald–Hartwig coupling. We show how artificial intelligence (AI) language model ChatGPT can automatically formulate for these based on literature corpus it was trained on. Critically, we showcase results be directly translated into inputs management software phactor, which enables automated execution analysis assays. This workflow experimentally demonstrated, with modest excellent yields products obtained each instance first attempt.

Language: Английский

Citations

22

Fine-tuning GPT-3 for machine learning electronic and functional properties of organic molecules DOI Creative Commons
Zikai Xie,

Xenophon Evangelopoulos,

Ömer H. Omar

et al.

Chemical Science, Journal Year: 2023, Volume and Issue: 15(2), P. 500 - 510

Published: Dec. 5, 2023

We evaluate the effectiveness of fine-tuning GPT-3 for prediction electronic and functional properties organic molecules. Our findings show that fine-tuned can successfully identify distinguish between chemically meaningful patterns, discern subtle differences among them, exhibiting robust predictive performance molecular properties. focus on assessing models' resilience to information loss, resulting from absence atoms or chemical groups, noise we introduce

Language: Английский

Citations

21

When Yield Prediction Does Not Yield Prediction: An Overview of the Current Challenges DOI Creative Commons
Varvara Voinarovska, Mikhail A. Kabeshov, Dmytro Dudenko

et al.

Journal of Chemical Information and Modeling, Journal Year: 2023, Volume and Issue: 64(1), P. 42 - 56

Published: Dec. 20, 2023

Machine Learning (ML) techniques face significant challenges when predicting advanced chemical properties, such as yield, feasibility of synthesis, and optimal reaction conditions. These stem from the high-dimensional nature prediction task myriad essential variables involved, ranging reactants reagents to catalysts, temperature, purification processes. Successfully developing a reliable predictive model not only holds potential for optimizing high-throughput experiments but can also elevate existing retrosynthetic approaches bolster plethora applications within field. In this review, we systematically evaluate efficacy current ML methodologies in chemoinformatics, shedding light on their milestones inherent limitations. Additionally, detailed examination representative case study provides insights into prevailing issues related data availability transferability discipline.

Language: Английский

Citations

21

Machine Learning-Guided Development of Trialkylphosphine Ni(I) Dimers and Applications in Site-Selective Catalysis DOI
Teresa M. Karl, Samir Bouayad‐Gervais,

Julian A. Hueffel

et al.

Journal of the American Chemical Society, Journal Year: 2023, Volume and Issue: 145(28), P. 15414 - 15424

Published: July 6, 2023

Owing to the unknown correlation of a metal’s ligand and its resulting preferred speciation in terms oxidation state, geometry, nuclearity, rational design multinuclear catalysts remains challenging. With goal accelerate identification suitable ligands that form trialkylphosphine-derived dihalogen-bridged Ni(I) dimers, we herein employed an assumption-based machine learning approach. The workflow offers guidance space for desired without (or only minimal) prior experimental data points. We experimentally verified predictions synthesized numerous novel dimers as well explored their potential catalysis. demonstrate C–I selective arylations polyhalogenated arenes bearing competing C–Br C–Cl sites under 5 min at room temperature using 0.2 mol % newly developed dimer, [Ni(I)(μ-Br)PAd2(n-Bu)]2, which is so far unmet with alternative dinuclear or mononuclear Ni Pd catalysts.

Language: Английский

Citations

19

Smart Dope: A Self‐Driving Fluidic Lab for Accelerated Development of Doped Perovskite Quantum Dots DOI Creative Commons
Fazel Bateni, Sina Sadeghi, Negin Orouji

et al.

Advanced Energy Materials, Journal Year: 2023, Volume and Issue: 14(1)

Published: Nov. 12, 2023

Abstract Metal cation‐doped lead halide perovskite (LHP) quantum dots (QDs) with photoluminescence yields (PLQYs) higher than unity, due to cutting phenomena, are an important building block of the next‐generation renewable energy technologies. However, synthetic route exploration and development highest‐performing QDs for device applications remain challenging. In this work, Smart Dope is presented, which a self‐driving fluidic lab (SDFL), accelerated synthesis space autonomous optimization LHP QDs. Specifically, multi‐cation doping CsPbCl 3 using one‐pot high‐temperature chemistry reported. continuously synthesizes multi‐cation‐doped high‐pressure gas‐liquid segmented flow format enable continuous experimentation minimal experimental noise at reaction temperatures up 255°C. offers multiple functionalities, including mechanistic studies through digital twin QD modeling, closed‐loop discovery, on‐demand manufacturing high‐performing Through these developments, autonomously identifies optimal Mn‐Yb co‐doped PLQY 158%, highest reported value class date. illustrates power SDFLs in accelerating discovery emerging advanced materials.

Language: Английский

Citations

19