Photovoltaics & Solar
Description
Screen perovskite, tandem and passivation materials by prediction. MatterSeek predicts photovoltaic material properties from composition.
Photovoltaics & Solar — Material System / Predictable Performance / Validation Pathway
AI for Photovoltaic Materials — Engineer Perovskite Absorbers and Passivation Layers by Prediction
It predicts the properties of absorbers, passivation layers, transport layers and encapsulation materials from chemical composition, so candidate chemistries can be ranked before a coating or annealing run is scheduled.
Absorber band structure must be tuned while thermal and moisture stability is preserved; passivation chemistry must suppress interface recombination without introducing parasitic absorption; tandem stacks add a compatibility constraint on top of both. Each of these objectives is normally tested through separate fabrication and measurement campaigns, and the combinations multiply quickly.
Describe the target composition range and the property window you need, and MatterSeek returns ranked candidates with predicted property values and uncertainty. The platform handles one to four base elements with doping-level precision of 0.02 — the granularity that alloying and A-site or B-site substitution studies actually require. Prediction completes in about one minute for two base elements, up to seven minutes for four. Results are presented as radar charts for multi-property comparison, 2D and 3D heat maps that visualize property landscapes across the composition space, and exportable data tables for downstream design review.
The engine spans optical, electronic, thermal, thermodynamic and structural property families, including melting point and electronic-structure-related properties — the quantities that govern phase stability during annealing and the electronic behaviour of an absorber or transport layer.
Shortlisted candidates move into MatterSeek's 1,200 m² autonomous laboratory. Its six process clusters cover the routes that matter for photovoltaics: clean thin-film deposition and post-treatment with multi-bath substrate cleaning, magnetron sputtering and electron-beam evaporation, plus sol–gel and supercritical drying for porous and oxide layers. Automated characterization includes XRD for phase identification, Raman, XRF and simultaneous thermal analysis, with a stated target of 300 to 500 samples per day depending on route. Prediction accuracy is 95.7% against physical measurement.
tandem and transport layers, defect-passivation chemistries, transparent conductive oxides and module encapsulation materials.
FAQ
Q: Can MatterSeek predict properties of perovskite compositions?
A: Yes. You define the composition range and the property window, and the platform returns ranked candidates with predicted values and uncertainty before any film is deposited.
Q: Which fabrication routes are available for validation?
A: Thin-film deposition with magnetron sputtering and electron-beam evaporation, sol–gel routes, and supercritical drying — all automated within the autonomous laboratory.
Q: How accurate are the predictions?
A: 95.7% accuracy against physical measurement, with uncertainty reported per prediction.
Yuyao (Beijing) Technology Co., Ltd. is a professional company specializing in the intersection of AI and advanced materials. The company has independently developed MatterSeek—a foundational large model tailored to the materials sector—and pioneered the HNN (Hierarchical Neural Network) algorithm, while building a proprietary, controllable materials experiment database.