Inverse materials design — starting from a target property and searching for structures that can realize it — holds great promise for accelerating the discovery of advanced materials. However, its value in closed-loop discovery depends not only on prediction performance, but also on whether expensive first-principles results are independently validated, provenance-recorded, and admitted as feedback only when evidence is sufficient. This is especially critical for composite properties such as carrier mobility, where a final scalar value can hide intermediate quantities, fit quality, convergence history, and workflow assumptions.
Now, researchers have developed InvDesMobility, a reliability-gated first-principles feedback framework that addresses these challenges. The work, authored by Li, Wen-Kao; Gao, Ze-Feng; Guo, Peng-Jie; Wei, Lu; and Zhong-Yi, is detailed in a paper on arXiv. According to the authors, InvDesMobility integrates multi-agent automated DFT (density functional theory), evidence stratification, generative structure proposal, acquisition ranking, and auditable release.
Framework Components and Workflow
The framework processes candidate materials through a structured pipeline. Using 516 2DMatPedia-derived candidates, the workflow produced 280 QC-passed materials and 573 retained carrier-direction seed channels after channel-level reliability gating. These records were split into two feedback objects:
- Relaxed structures updated the generative model.
- Retained mobility channels trained the acquisition model and set validation priority.
Quantitative Results
Over multiple iterations, InvDesMobility screened 2.4 × 10^6 structures, submitted 102 candidates for DFT validation, and retained 86 reliability-gated generated channels across 41 formulas. The following table summarizes the key metrics:
| Metric | Value |
|---|---|
| 2DMatPedia-derived candidates | 516 |
| QC-passed materials | 280 |
| Retained carrier-direction seed channels after gating | 573 |
| Structures screened | 2.4 × 10^6 |
| Candidates submitted for DFT validation | 102 |
| Retained reliability-gated generated channels (41 formulas) | 86 |
Significance and Availability
The main contribution, as stated by the authors, is not a fixed list of high-mobility materials, but a transferable feedback contract that makes closed-loop inverse design both useful and auditable when learning from expensive calculated properties. All source data, retained feedback records, and workflows are available at an online repository, accompanied by an evidence website.
Implications for Enterprise Technology
For technology decision-makers, this framework demonstrates a principled approach to integrating AI with computationally expensive simulations. The reliability gating mechanism ensures that only validated data feeds back into the model, reducing the risk of propagating errors. Such methodologies could be adapted to other domains where high-cost evaluations are combined with machine learning, such as drug discovery or catalysis design.