Synthesis of metal alloy catalysts using high-throughput experiments and machine learning optimization
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
The periodic table comprises over a hundred elements, offering numerous possibilities \nfor the discovery of novel materials that have superior properties and could therefore \nbe used to address current technological and societal challenges. However, exploring \nthe extensive range of combinations are resource-intensive: slow and costly, \nparticularly for materials significantly affected by the synthesis procedures. In this \nfinal year project, a workflow for the high throughput synthesis of multimetallic alloys \nis presented. The two-step workflow is comprised by a liquid mixing step and an \nannealing step. An acceleration factor of 2.4 relative to the traditional auto combustion \nsol gel synthesis method is achieved by synthesizing 24 samples in 620 minutes. To \nevaluate the effectiveness of this methodology and with the assistance of previous \ncomputational work carried out by collaborators at Meta AI, copper and three other \ncopper alloys, namely binary Cu-Ag, Cu-Zn, and ternary Cu-Zn-Ag, are synthesized, \ndue to their predicted promising use in CO2 reduction. The synthesized samples show \nhomogeneously distributed elemental composition and high phase purity. The catalytic \nperformance is evaluated by collaborators at the University of Toronto. The initial \nfindings from measuring pure Cu, which serves as a baseline, demonstrate consistent \nperformance when compared to commercially available Cu nanoparticles. Crucially, \nthe Faradaic efficiencies show different results compared to Cu nanoparticles. Firstly, \na substantial amount of H2 gas is produced, accompanied by low CO. This is due to \nthe large amount of carbon in our powders, stemming from the annealing step, and \nlarge particle size of the pure Cu. To guide future experiments and optimize the \nFaradaic efficiencies, the experimental data collected in this project is used to deploy \na Bayesian Optimization (BO) algorithm. Specifically, q-Noisy Expected \nHypervolume Improvement based Bayesian Optimization (qNEHVI-BO) model is \nimplemented, providing insight to guide the next experimental steps to achieve dry \nsamples and minimize the absolute difference between the obtained composition and \nthe target.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.003 | 0.003 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it