11 results
Now showing1 - 10 of 11
- Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Different alloy, different story? Impact of transfer learning on predictive models for in silico screening of Mg dissolution modulators(Elsevier, 2026-09-26); ; ; ; Ideally, datasets for training machine learning models consist of thousands of versatile data samples — a scale that is seldom met in materials research to date. Transfer learning (TL) is widely used to mitigate data scarcity in available training sets by leveraging knowledge from related domains, since models pretrained on auxiliary datasets may learn structure–property relationships that are transferable from one to another. In this work, the potential of TL for predicting the performance of small organic molecule dissolution modulators is probed using corrosion inhibition responses of eleven Mg-based materials taken from the previously published ExCorr database. A significant improvement in prediction accuracy is observed when transfer learning is applied. However, for the dataset configuration studied here, the apparent cross-alloy TL gains are shown not to arise from transferable structure–activity relationship (SAR) knowledge but rather from a learned alloy-context signal embedded in the data. Seven Mg-based materials—six commercial alloys plus the high-purity variant HPMg51—form a shared-SAR cluster with strongly correlated inhibitor rankings, such that molecules labelled on one material are readily predictable for others within the cluster. Within this shared-SAR cluster, extensive overlap reduces the effective training pool from approximately 900 records to only around 180 unique molecules, placing the task firmly in a small-data regime. In this setting, an in-context large language model (Gemini 3.1 Pro) is the only approach that successfully learns meaningful SAR, with its pretrained knowledge providing the transfer that the explicit source dataset fails to deliver. Seven Mg-based materials in the ExCorr database form a shared-structure–activity relationship (SAR) cluster (pairwise Spearman r=0.67–0.92). Under molecule-disjoint filtering the cross-alloy TL gain on ExCorr splits into a null inhibitor-SAR contribution and a dominant alloy-context contribution. Cross-alloy overlap shrinks ExCorr’s training pool from ≈900records to ≈180unique molecules, only an in-context LLM (Gemini 3.1 Pro) stays predictive.Publicationtype: Journal Article7 5 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Large language models predicting the corrosion inhibition efficiency of magnesium dissolution modulators(Elsevier, 2025-06-20); ; ; ; Large language models (LLMs), such as GPT-4o, have shown promise in solving everyday tasks and addressing fundamental scientific challenges by leveraging extensive pre-trained knowledge. In this study, we investigate their potential to predict the efficiency of various organic compounds in inhibiting the corrosion of the magnesium alloy ZE41. Traditional approaches, such as Multilayer Perceptrons (MLPs), rely on non-contextual data, often necessitating large datasets and substantial effort per sample to achieve accurate predictions. These methods particularly struggle with small datasets as their training data and domain of applicability is limited to a small area of the available chemical space. LLMs can contextualize and interpret limited data points by drawing on their vast knowledge, including the chemical properties of molecules and their influence on corrosion processes in other materials (e.g. iron and aluminium). By prompting the model with a small dataset, LLMs can provide meaningful predictions without the need for extensive training. Our study demonstrates that LLMs can predict corrosion inhibition outcomes and outperform classical approaches, such as MLPs, having access to the identical number of training samples.Publicationtype: Journal Article34 23 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Searching the chemical space for effective magnesium dissolution modulators: a deep learning approach using sparse features(Nature Publishing Group, 2023-12-01); ; ; ; Small organic molecules can alter the degradation rates of the magnesium alloy ZE41. However, identifying suitable candidate compounds from the vast chemical space requires sophisticated tools. The information contained in only a few molecular descriptors derived from recursive feature elimination was previously shown to hold the potential for determining such candidates using deep neural networks. We evaluate the capability of these networks to generalise by blind testing them on 15 randomly selected, completely unseen compounds. We find that their generalisation ability is still somewhat limited, most likely due to the relatively small amount of available training data. However, we demonstrate that our approach is scalable; meaning deficiencies caused by data limitations can presumably be overcome as the data availability increases. Finally, we illustrate the influence and importance of well-chosen descriptors towards the predictive power of deep neural networks.Publicationtype: Journal Article35 40 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Predicting corrosion inhibition efficiencies of small organic molecules using data-driven techniquesSelecting effective corrosion inhibitors from the vast chemical space is not a trivial task, as it is essentially infinite. Fortunately, machine learning techniques have shown great potential in generating shortlists of inhibitor candidates prior to large-scale experimental testing. In this work, we used the corrosion responses of 58 small organic molecules on the magnesium alloy AZ91 and utilized molecular descriptors derived from their geometry and density functional theory calculations to encode their molecular information. Statistical methods were applied to select the most relevant features to the target property for support vector regression and kernel ridge regression models, respectively, to predict the behavior of untested compounds. The performance of the two supervised learning approaches were compared and the robustness of the data-driven models were assessed by experimental blind testing.Publicationtype: Journal ArticleScopus© Citations 2 39 87 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Data-driven selection of electrolyte additives for aqueous magnesium batteries(2022-09-23); ; ; ; Aqueous primary Mg-air batteries have considerable potential as energy sources for sea applications and portable devices. However, some challenges at the anode-electrolyte interface related to self-corrosion, aging of the electrolyte and the chunk-effect have to be solved to improve the discharge potential of the battery as well as the utilization efficiency of the anode material. Aside from alloying, an effective strategy to mitigate self-corrosion and battery failure is the use of electrolyte additives. Selecting useful additives from the vast chemical space of possible compounds is not a trivial task. Fortunately, data-driven quantitative structure-property relationship (QSPR) models can facilitate efficient searches for promising battery booster candidates. Here, the robustness and predictive performance of two QSPR models are evaluated using an active design of experiments approach. We also present a multi-objective optimization method that allows to identify new electrolyte additives that can boost the battery anode performance with respect to a target application, thus accelerating the discovery of advanced battery systems.Publicationtype: Journal Article88 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Predicting the inhibition efficiencies of magnesium dissolution modulators using sparse machine learning models(Nature Publishing Group, 2021-12); ; ; ; The degradation behaviour of magnesium and its alloys can be tuned by small organic molecules. However, an automatic identification of effective organic additives within the vast chemical space of potential compounds needs sophisticated tools. Herein, we propose two systematic approaches of sparse feature selection for identifying molecular descriptors that are most relevant for the corrosion inhibition efficiency of chemical compounds. One is based on the classical statistical tool of analysis of variance, the other one based on random forests. We demonstrate how both can—when combined with deep neural networks—help to predict the corrosion inhibition efficiencies of chemical compounds for the magnesium alloy ZE41. In particular, we demonstrate that this framework outperforms predictions relying on a random selection of molecular descriptors. Finally, we point out how autoencoders could be used in the future to enable even more accurate automated predictions of corrosion inhibition efficiencies.Publicationtype: Journal ArticleScopus© Citations 16 151 252 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Exploring structure-property relationships in magnesium dissolution modulators(Macmillan Publishers Limited, part of Springer Nature, 2021-01-08); ; ; ; Small organic molecules that modulate the degradation behavior of Mg constitute benign and useful materials to modify the service environment of light metal materials for specific applications. The vast chemical space of potentially effective compounds can be explored by machine learning-based quantitative structure-property relationship models, accelerating the discovery of potent dissolution modulators. Here, we demonstrate how unsupervised clustering of a large number of potential Mg dissolution modulators by structural similarities and sketch-maps can predict their experimental performance using a kernel ridge regression model. We compare the prediction accuracy of this approach to that of a prior artificial neural networks study. 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 the automated discovery of chemicals with desired dissolution modulating properties from a commercial database. We subsequently prove this concept by blind validation of five chemicals.Publicationtype: Journal ArticleScopus© Citations 15 150 138 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , ATR-FTIR in Kretschmann configuration integrated with electrochemical cell as in situ interfacial sensitive tool to study corrosion inhibitors for magnesium substrates(Elsevier, 2020-06-10); ; ; ; Integrated attenuated total reflection – Fourier transform infrared spectroscopy (ATR-FTIR) – Electrochemical impedance spectroscopy (EIS) measurements were used to simultaneously follow chemisorption mechanisms of organic inhibitors as well as their corrosion inhibition efficiency towards magnesium based substrates. Four carboxylic compounds, i.e. 2,5-pyridinedicarboxylic acid (PDC), 3-methylsalicylic acid (MSA), sodium salicylate (SS) and fumaric acid (FA), were selected based on their promising inhibiting capacities and were all shown to chemisorb at the MgO/Mg(OH)2 surface by carboxylate bond formation. Orientation analysis using polarized infrared light showed that carboxylate bonds established using aliphatic carboxylate compound aligned perpendicular to the magnesium surface, whereas carboxylate bonds with aromatic compounds were oriented in plane with the magnesium surface. This different orientation is associated to the involvement of π-interactions in the MgO/Mg(OH)2 – aromatic carboxylate adsorption. Additionally, DFT calculations revealed that the addition of hetero-atoms (i.e. N or OH) in the molecular structure contributes to increased adsorption energies, indicating that next to carboxylate groups also these hetero-atoms are involved in interfacial interactions. Integrating the ATR-FTIR setup with an electrochemical cell allowing for simultaneous EIS measurements lead to two surface phenomena determining the inhibition efficiency. Surface hydroxylation processes on one hand forming a MgO/Mg(OH)2 layer on one hand, and the chemisorption of carboxylate compounds on the other hand. The inhibition efficiency was found to increase in following order: FA < PDC < MSA and was mainly associated to the formation of a MgO/Mg(OH)2 layer. SS was shown to act as a corrosion accelerator rather than a corrosion inhibitor. Despite its high sensitivity for water, both surface processes could be followed in situ by means of ATR-FTIR. Simultaneously, protective properties of the formed films could be quantified by means of EIS. Consequently, integrated ATR-FTIR – EIS methodology has shown to be highly valuable for gaining in-situ insights in the inhibition mechanism, while quantifying the inhibition efficiency. This was even possible for highly active metal substrate as magnesium, although further developments are suggested if one aims to quantify electrochemical constants related to corrosion and other surface processes measured at the low frequencies (i.e. < 1 Hz).Publicationtype: Journal ArticleScopus© Citations 39 180 219 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , In silico screening of modulators of magnesium dissolutionThe vast number of small molecules with potentially useful dissolution modulating properties (inhibitors or accelerators) renders currently used experimental discovery methods time- and resource-consuming. Fortunately, emerging computer-assisted methods can explore large areas of chemical space with less effort. Here we show how density functional theory calculations and machine learning methods can work synergistically to generate robust and predictive models that recapitulate experimentally-derived corrosion inhibition efficiencies of small organic compounds for pure magnesium. We further validate our methods by predicting a priori the corrosion modulation properties of seven hitherto untested small molecules and confirm the prediction in subsequent experiments.Publicationtype: Journal ArticleScopus© Citations 41 372 330 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Data science based mg corrosion engineering(2019-03-08); ; ; ; Magnesium exhibits a high potential for a variety of applications in areas such as transport, energy and medicine. However, untreated magnesium alloys are prone to corrosion, restricting their practical application. Therefore, it is necessary to develop new approaches that can prevent or control corrosion and degradation processes in order to adapt to the specific needs of the application. One potential solution is using corrosion inhibitors which are capable of drastically reducing the degradation rate as a result of interactions with the metal surface or components of the corrosive medium. As the sheer number of potential dissolution modulators makes it impossible to obtain a detailed atomistic understanding of the inhibition mechanisms for each additive, other measures for inhibition prediction are required. For this purpose, a concept is presented that combines corrosion experiments, machine learning, data mining, density functional theory calculations and molecular dynamics to estimate corrosion inhibition properties of still untested molecules. Concomitantly, this approach will provide a deeper understanding of the fundamental mechanisms behind the prevention of corrosion events in magnesium-based materials and enables more accurate continuum corrosion simulations. The presented concept facilitates the search for molecules with a positive or negative effect on the inhibition efficiency and could thus significantly contribute to the better control of magnesium / electrolyte interface properties. © 2019 Würger, Feiler, Musil, Feldbauer, Höche, Lamaka, Zheludkevich and Meißner.Publicationtype: Journal ArticleScopus© Citations 36 650 392