Ivan Cole¹, Patrick Keil², Pablo Ordejon³, Ernane De Freitas Martins¹*³, Jose Castillo Robles¹*³,
Steffen Jeschke¹, Rou Jun Toh¹, Jim Josel Qiushi Deng¹, Paul White¹, Xiaobo Chen¹,
Chathumini Samarawickrama¹, Philip Eiden⁴, Milan Patel¹

¹ RMIT, University Melbourne , Australia;

   ² For BASF Coatings , Munster, Germany

³ICN2, Barcelona, Spain

⁴BASF, Ludwigshafen, Germany

EXTENDED ABSTRACT: "Inverse Design,, is a grand challenge for the design of materials systems in general and corrosion inhibitors. In inverse design, the features that render a material system functionally effective are identified and then data bases of materials are searched for materials with these features. In the case of the use of small molecules as corrosion inhibitors, molecular forms and properties (molecular attributes) that give rise to good inhibition properties are selected by statistical, Quantitative Structure-Activity Relationships (QSAR) or machine learning. Having identified these attributes, molecular databases are searched to extract promising inhibitor candidates. This approach is particularly promising as not only can it be used to sort through the tens of possible inhibitors, but it can virtually explore new families not previously considered as candidates. However, there are a number of challenges limiting the successful application of inverse design to corrosion inhibitor selection. In many works, the selection of molecular attributes does not appear to model the observed variation in inhibitor efficiency in a robust way so that while reasonable test models can be developed their predictive capability is limited. Secondary very large data bases of both molecular attributes and electrochemical properties are required to drive current QSAR models and these are expensive and time-consuming to build. Lastly, there are some fundamental limits to our understanding of small molecular surface interactions in aqueous media particularly when the surface is charged. This paper will present recent work on modelling the double layer or electrified interface in aqueous systems, the development of robotic electrochemistry to help build large data bases and recent large data studies that have refined the definition of molecular attributes. In particular the application of hybrid quantum mechanics/molecular mechanics simulation QM/MM simulation and the incorporation of Non-equilibrium green function (NEGF) through the program Transiesta, allow modelling of larger structures , direct modelling of the electrified interfaces under an applied and incorporation of solvent effects . This leads to an enhanced understanding of charge distributions that arise both at the interface and across surface binding molecules while also permitting the calculation of the energy associated with the formation of the first surface layer. This deeper insight will allow us to refine our definition of molecular attributes.To address some of the above problems with the QSAR approach we are addressing the possibilities that arise when ^evolutionary algorithms" are combined with robotic experimentation. In this approach rather than construct large data bases, one begins with a constrained selection of possible inhibitors that are tested via robotic electrochemistry and then a refined set of molecules is derived by "mutating" the original selection based on this testing results and the mutated selection is then tested and the process continues until an optimum selection of inhibitors is obtained