The promise and pitfalls of AI for molecular and materials synthesis

David, Nicholas; Sun, Wenhao; Coley, Connor W.

doi:10.1038/s43588-023-00446-x

Cited by 17 publications

(8 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Overall, our text-mined dataset provides explicit post hoc empirical validation of the MTC hypothesis, and furthermore highlights the value of such text-mined datasets in not only providing data to train machine-learning models 35 , but also in empirically validating new mechanistic theories 36 .…”

Section: Evaluation Of Mtc From Literature Aqueous Synthesis Recipesmentioning

confidence: 80%

Optimal thermodynamic conditions to minimize kinetic by-products in aqueous materials synthesis

Wang,

Sun,

Cruse

et al. 2024

Nat. Synth

Self Cite

View full text Add to dashboard Cite

Phase diagrams offer substantial predictive power for materials synthesis by identifying the stability regions of target phases. However, thermodynamic phase diagrams do not offer explicit information regarding the kinetic competitiveness of undesired by-product phases. Here we propose a quantitative and computable thermodynamic metric to identify synthesis conditions under which the propensity to form kinetically competing by-products is minimized. We hypothesize that thermodynamic competition is minimized when the difference in free energy between a target phase and the minimal energy of all other competing phases is maximized. We validate this hypothesis for aqueous materials synthesis through two empirical approaches: first, by analysing 331 aqueous synthesis recipes text-mined from the literature; and second, by systematic experimental synthesis of LiIn(IO3)4 and LiFePO4 across a wide range of aqueous electrochemical conditions. Our results show that even for synthesis conditions that are within the stability region of a thermodynamic Pourbaix diagram, phase-pure synthesis occurs only when thermodynamic competition with undesired phases is minimized.

show abstract

Section: Evaluation Of Mtc From Literature Aqueous Synthesis Recipesmentioning

confidence: 80%

Optimal thermodynamic conditions to minimize kinetic by-products in aqueous materials synthesis

Wang,

Sun,

Cruse

et al. 2024

Nat. Synth

Self Cite

View full text Add to dashboard Cite

show abstract

“…Though subject to error, natural language processing could also be used to mine existing miscibility data from the scientific literature (to our knowledge, only one other sizable miscibility data set of 34 polymers and surfactants has been generated). On the other hand, the advantage of using a single data set is that it avoids interlab variations, such as variations in purity or protocol, that might make the data less useful for machine learning . The relatively simple nature of the experimental process (mixing aqueous solutions) and characterization (via microscopic image analysis) lends itself to laboratory automation for generating a large data set.…”

Section: Discussionmentioning

confidence: 99%

“…On the other hand, the advantage of using a single data set is that it avoids interlab variations, such as variations in purity or protocol, that might make the data less useful for machine learning. 64 The relatively simple nature of the experimental process (mixing aqueous solutions) and characterization (via microscopic image analysis 21 ) lends itself to laboratory automation for generating a large data set.…”

Section: ■ Conclusionmentioning

confidence: 99%

Data-Driven Imputation of Miscibility of Aqueous Solutions via Graph-Regularized Logistic Matrix Factorization

Behnoudfar,

Simon,

Schrier

2023

J. Phys. Chem. B

View full text Add to dashboard Cite

Aqueous, two-phase systems (ATPSs) may form upon mixing two solutions of independently water-soluble compounds. Many separation, purification, and extraction processes rely on ATPSs. Predicting the miscibility of solutions can accelerate and reduce the cost of the discovery of new ATPSs for these applications. Whereas previous machine learning approaches to ATPS prediction used physicochemical properties of each solute as a descriptor, in this work, we show how to impute missing miscibility outcomes directly from an incomplete collection of pairwise miscibility experiments. We use graph-regularized logistic matrix factorization (GR-LMF) to learn a latent vector of each solution from (i) the observed entries in the pairwise miscibility matrix and (ii) a graph where each node is a solution and edges are relationships indicating the general category of the solute (i.e., polymer, surfactant, salt, protein). For an experimental data set of the pairwise miscibility of 68 solutions from Peacock et al. [ACS Appl. Mater. Interfaces 2021, 13, 11449–11460], we find that GR-LMF more accurately predicts missing (im)miscibility outcomes of pairs of solutions than ordinary logistic matrix factorization and random forest classifiers that use physicochemical features of the solutes. GR-LMF obviates the need for features of the solutions and solutions to impute missing miscibility outcomes, but it cannot predict the miscibility of a new solution without some observations of its miscibility with other solutions in the training data set.

show abstract

“…Artificial intelligence holds great promise to accelerate the chemical sciences. 1–4 Over the last decade, we have witnessed ground-breaking advances in machine learning for de novo molecular design, 5–10 synthesis planning, 11–17 and reaction outcome prediction. 18–26 Recently, research has focused on sequential model-based optimisation algorithms, particularly Bayesian optimisation (BO), to identify optimal conditions for chemical reactions effectively.…”

Section: Introductionmentioning

confidence: 99%

Bayesian optimisation for additive screening and yield improvements – beyond one-hot encoding

Ranković,

Griffiths,

Moss

et al. 2024

Digital Discovery

View full text Add to dashboard Cite

show abstract

The promise and pitfalls of AI for molecular and materials synthesis

Cited by 17 publications

References 21 publications

Optimal thermodynamic conditions to minimize kinetic by-products in aqueous materials synthesis

Optimal thermodynamic conditions to minimize kinetic by-products in aqueous materials synthesis

Data-Driven Imputation of Miscibility of Aqueous Solutions via Graph-Regularized Logistic Matrix Factorization

Bayesian optimisation for additive screening and yield improvements – beyond one-hot encoding

Contact Info

Product

Resources

About