Nonempirical Definition of the Mendeleev Numbers: Organizing the Chemical Space

Allahyari, Zahed; Oganov, Artem R.

doi:10.1021/acs.jpcc.0c07857

Cited by 24 publications

(18 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Because distance in feature space (e.g., in KRR) or latent space (i.e., in the ANN) is a critical component, prediction by ML models of new materials with compositions not well supported by training data is universally challenging. Nevertheless, RACs and other widely used representations 60 employ the nuclear charge in the representation, exaggerating differences between rows of the periodic table with respect to differences in group number 75 . While some heuristic properties in RACs (e.g., the Pauling electronegativity) more faithfully encode chemical similarity 64 , the nuclear-charge-based features are often emphasized in feature-selected subsets, highlighting their importance in model prediction.…”

Section: Introductionmentioning

confidence: 99%

Representations and strategies for transferable machine learning improve model performance in chemical discovery

Harper

Nandy

Arunachalam

et al. 2022

The Journal of Chemical Physics

View full text Add to dashboard Cite

Strategies for machine-learning(ML)-accelerated discovery that are general across materials composition spaces are essential, but demonstrations of ML have been primarily limited to narrow composition variations. By addressing the scarcity of data in promising regions of chemical space for challenging targets like open-shell transition-metal complexes, general representations and transferable ML models that leverage known relationships in existing data will accelerate discovery. Over a large set (ca. 1000) of isovalent transition-metal complexes, we quantify evident relationships for different properties (i.e., spin-splitting and ligand dissociation) between rows of the periodic table (i.e., 3d/4d metals and 2p/3p ligands). We demonstrate an extension to graph-based revised autocorrelation (RAC) representation (i.e., eRAC) that incorporates the effective nuclear charge alongside the nuclear charge heuristic that otherwise overestimates dissimilarity of isovalent complexes. To address the common challenge of discovery in a new space where data is limited, we introduce a transfer learning approach in which we seed models trained on a large amount of data from one row of the periodic table with a small number of data points from the additional row. We demonstrate the synergistic value of the eRACs alongside this transfer learning strategy to consistently improve model performance. Analysis of these models highlights how the approach succeeds by reordering the distances between complexes to be more consistent with the periodic table, a property we expect to be broadly useful for other materials domains.

show abstract

Section: Introductionmentioning

confidence: 99%

Representations and strategies for transferable machine learning improve model performance in chemical discovery

Harper

Nandy

Arunachalam

et al. 2022

The Journal of Chemical Physics

View full text Add to dashboard Cite

show abstract

“…Atomic radii and electronegativity are often quintessential for how chemistry is rationalized. 1,2 The history of quantifying the sizes of atoms under ambient conditions includes a large body of work, extending over the last one and a half-centuries (for a non-exhaustive summary of this history see references 3 and 4). One early motivation for attaining atomic and ionic sizes was to help understand X-ray diffraction patterns in terms of crystal structures, 5,6 another to provide a rationalization for metallization.…”

Section: Introductionmentioning

confidence: 99%

Relating atomic energy, radius and electronegativity through compression

2021

View full text Add to dashboard Cite

show abstract

“…Mn is becoming increasingly important in phenomenological data-driven searches for new chemical compounds. More recently, Allahyari and Oganov (2020) revealed that a one-dimensional ordering using only atomic size and electronegativity allows the computation of a non-empirical Mn and the derivation of a universal sequence of elements.…”

Section: Chemical Periodicity Periodic Systems and Periodic Tablesmentioning

confidence: 99%

The value of vague ideas in the development of the periodic system of chemical elements

Vogt

2021

Synthese

View full text Add to dashboard Cite

The exploration of chemical periodicity over the past 250 years led to the development of the Periodic System of Elements and demonstrates the value of vague ideas that ignored early scientific anomalies and instead allowed for extended periods of normal science where new methodologies and concepts are developed. The basic chemical element provides this exploration with direction and explanation and has shown to be a central and historically adaptable concept for a theory of matter far from the reductionist frontier. This is explored in the histories of Prout's hypothesis, Döbereiner Triads, element inversions necessary when ordering chemical elements by atomic weights, and van den Broeck's ad-hoc proposal to switch to nuclear charges instead. The development of more accurate methods to determine atomic weights, Rayleigh and Ramsey's gas separation and analytical techniques, Moseley's x-ray spectroscopy to identify chemical elements, and more recent accelerator-based cold fusion methods to create new elements at the end of the Periodic Table point to the importance of methodological development complementing conceptual advances. I propose to frame the crossover from physics to chemistry not as a loss of accuracy and precision but as an increased application of vague concepts such as similarity which permit classification. This approach provides epistemic flexibility to adapt to scientific anomalies and the continued growth of chemical compound space and rejects the Procrustean philosophy of reductionist physics. Furthermore, it establishes chemistry with its explanatory and operational autonomy epitomized by the periodic system of elements as a gateway to other experimental sciences.

show abstract

Nonempirical Definition of the Mendeleev Numbers: Organizing the Chemical Space

Cited by 24 publications

References 25 publications

Representations and strategies for transferable machine learning improve model performance in chemical discovery

Representations and strategies for transferable machine learning improve model performance in chemical discovery

Relating atomic energy, radius and electronegativity through compression

The value of vague ideas in the development of the periodic system of chemical elements

Contact Info

Product

Resources

About