Please login first
Exploring chemical space using computational methods
* 1 , 1, 2 , 1 , 1 , 3, 4, 5, 6 , 1 , 1, 2 , 1
1  Institute of Surface Science, Helmholtz-Zentrum Geesthacht, Geesthacht, Germany
2  Institute of Polymer and Composites, Hamburg University of Technology, Hamburg, Germany
3  La Trobe Institute for Molecular Science, La Trobe University, Bundoora, Australia
4  Monash Institute of Pharmaceutical Sciences, Monash University, Parkville, Australia
5  CSIRO Data61, Pullenvale, Australia
6  School of Pharmacy, University of Nottingham, Nottingham, United Kingdom


As the lightest structural engineering metal, magnesium (Mg) is a promising base material for the development of advanced technologies in transport, medical as well as in battery applications. A prerequisite to unlock the full potential of Mg–based materials is gaining control over their corrosion behaviour due to the relatively high chemical reactivity of Mg whereas each application field imposes unique requirements on this challenge. Corrosion prevention is essential in transport applications to avoid material failure. Bone implants require a degradation rate that is tailored to a specific injury whereas constant dissolution of the anode material is required to boost the efficiency of Mg-air primary batteries. Fortunately, small organic molecules have shown great potential to control the dissolution properties of pure Mg materials and its alloys.1 However, the vast space of small molecules with potentially useful dissolution modulating properties (inhibitors or accelerators) renders conventional experimental discovery methods too time- and resource-consuming. Consequently, computer-assisted selection prior to experimental investigations of the most promising candidates is of great benefit in the search for effective corrosion modulating additives.

Here, we demonstrate how unsupervised clustering of potential Mg dissolution modulators based on structural similarities and sketch-maps2 can quantitatively predict their experimental performance when combined with a kernel ridge regression (KRR) model.3 The prediction accuracy of the KRR model is compared to an artificial neural network (ANN) model that was trained on a combination of atomistic and structural molecular descriptors.4 Furthermore, we confirm the robustness of our data-driven model by blind prediction of the dissolution modulating performance of 10 untested compounds. Finally, a workflow is presented that facilitates an automated selection of compounds with promising properties by screening of a large database comprised of commercially available substances.

[1] S. V. Lamaka, B. Vaghefinazari, D. Mei, R.P. Petrauskas, D. Höche, M. L. Zheludkevich, Corros. Sci. 2017, 128, 224-240.

[2] M. Ceriotti, G. A. Tribello, M. Parrinello, Proc. Natl Acad. Sci. USA 2011, 108, 13023.

[3] T. Würger, D. Mei, B. Vaghefinazari, D. A. Winkler, S. V. Lamaka, M. L. Zheludkevich, R. H. Meißner, C. Feiler, npj. Mater. Degrad. 2021, 5, 2.

[4] C. Feiler, D. Mei, B. Vaghefinazari, T. Würger, R. H.Meißner, B. J. C. Luthringer-Feyerabend, D. A. Winkler, M. L. Zheludkevich, S. V. Lamaka, Corros. Sci. 2020, 163, 108245.

Keywords: Magnesium; Dissolution Modulators; Quantitative Structure-Property Relationships; Dimensionality Reduction; Database Screening