Chemical space: limits, evolution and modelling of an object bigger than our universal library

Restrepo, Guillermo

doi:10.1039/d2dd00030j

Cited by 20 publications

(31 citation statements)

References 98 publications

(231 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…These results are shown on the right-hand side of Fig. 6: some of the formally sp 2 carbons are very similar to diamond-like environments, suggesting that these are in fact dangling-bond sp 3 environments, further corroborated by their high local energies. This kind of structure is not made as obvious in the SOAP map in Fig.…”

Section: Embedding and Visualisationmentioning

confidence: 57%

“…The vast size of compositional and configurational chemical space means that physical experiments will quickly reach their limits for these tasks. [1][2][3] Digital "experiments", powered by large datasets and machine learning (ML) models, provide high-throughput approaches to chemical discovery, and can help to answer questions that their physical counterparts on their own can not. [4][5][6][7] However, because ML methods generally rely on large datasets rather than on empirical physical knowledge, they require new insight into the methodology itself -one example in this context is the active research into interpretability and explainability of ML models.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Synthetic data enable experiments in atomistic machine learning

2023

View full text Add to dashboard Cite

show abstract

Section: Embedding and Visualisationmentioning

confidence: 57%

Section: Introductionmentioning

confidence: 99%

Synthetic data enable experiments in atomistic machine learning

2023

View full text Add to dashboard Cite

show abstract

“…Although this seems unlikely given the exponential growth of the CS, doubling its size about each 16 years 6 , if the chemical community sets up to diversify the known chemistry at the historical speeds it has done it, as discussed in ref. 44 , it may be possible in a matter of two decades to change the shape of the PS. But there are some caveats.…”

Section: Resultsmentioning

confidence: 99%

The six stages of the convergence of the periodic system to its final structure

et al. 2023

Self Cite

View full text Add to dashboard Cite

The periodic system encodes order and similarity among chemical elements arising from known substances at a given time that constitute the chemical space. Although the system has incorporated new elements, the connection with the remaining space is still to be analysed, which leads to the question of how the exponentially growing space has affected the periodic system. Here we show, by analysing the space between 1800 and 2021, that the system has converged towards its current stable structure through six stages, respectively characterised by the finding of elements (1800–1826), the emergence of the core structure of the system (1826–1860), its organic chemistry bias (1860–1900) and its further stabilisation (1900–1948), World War 2 new chemistry (1948–1980) and the system final stabilisation (1980–). Given the self-reinforced low diversity of the space and the limited chemical possibilities of the elements to be synthesised, we hypothesise that the periodic system will remain largely untouched.

show abstract

“…[36] large biases stemming from the lack of structural and chemical diversity in the available data. These biases, ultimately of anthropogenic nature, [39,40] lead unfortunately to a poor generalization error. In fact, even if the error in test sets is of the order of 20-30 meV atom −1 , the actual error during high-throughput searches can be easily one order of magnitude larger if the available training data is not representative of the actual material space.…”

Section: Introductionmentioning

confidence: 99%

Machine‐Learning‐Assisted Determination of the Global Zero‐Temperature Phase Diagram of Materials

et al. 2023

View full text Add to dashboard Cite

Crystal‐graph attention neural networks have emerged recently as remarkable tools for the prediction of thermodynamic stability. The efficacy of their learning capabilities and their reliability is however subject to the quantity and quality of the data they are fed. Previous networks exhibit strong biases due to the inhomogeneity of the training data. Here a high‐quality dataset is engineered to provide a better balance across chemical and crystal‐symmetry space. Crystal‐graph neural networks trained with this dataset show unprecedented generalization accuracy. Such networks are applied to perform machine‐learning‐assisted high‐throughput searches of stable materials, spanning 1 billion candidates. In this way, the number of vertices of the global T = 0 K phase diagram is increased by 30% and find more than ≈150 000 compounds with a distance to the convex hull of stability of less than 50 meV atom−1. The discovered materials are then accessed for applications, identifying compounds with extreme values of a few properties, such as superconductivity, superhardness, and giant gap‐deformation potentials.

show abstract

Chemical space: limits, evolution and modelling of an object bigger than our universal library

Abstract: Chemical space entails substances endowed with a notion of nearness that comes in two flavours: similarity and synthetic reachability. What is the maximum size for the chemical space? Is there...

Cited by 20 publications

References 98 publications

Synthetic data enable experiments in atomistic machine learning

Synthetic data enable experiments in atomistic machine learning

The six stages of the convergence of the periodic system to its final structure

Machine‐Learning‐Assisted Determination of the Global Zero‐Temperature Phase Diagram of Materials

Contact Info

Product

Resources

About