An independent research platform for computational materials science, powered by the SOST ecosystem. Uses dedicated ML models and optimization algorithms — not ConvergenceX mining.
The SOST Materials Discovery Platform is an independent research initiative within the SOST ecosystem. It uses dedicated machine learning models, public materials databases, and optimization algorithms to predict, generate, and evaluate novel material compositions. The platform is not an extension of ConvergenceX mining — it operates as a separate computational service, accessible via SOST escrow payments.
The platform combines multiple computational approaches for materials discovery, each using purpose-built algorithms independent of the ConvergenceX mining engine.
| ML Models | Graph neural networks for property prediction // CGCNN, MEGNet, ALIGNN |
| Optimization | Evolutionary algorithms and numerical optimization // composition search, structure relaxation |
| DFT Integration | API access to VASP, Quantum ESPRESSO, GPAW // ab initio validation |
| Compute Donation | Miners donate idle CPU cycles between blocks // voluntary, opt-in community compute |
Researchers publish materials discovery problems. The SOST mining community contributes idle CPU time to run simulations, earning rewards from the PoPC Pool. All results are published on-chain as immutable open data, creating a permanent public materials database.
Approximately 150,000 known materials sourced from leading open-access databases. Provides the training data for prediction models and the reference set for inverse search.
| Materials Project | Inorganic crystal structures + computed properties |
| JARVIS | DFT-computed properties, 2D materials, heterostructures |
| AFLOW | Autonomous high-throughput materials discovery library |
Specify desired material properties (band gap, hardness, thermal conductivity, etc.) and receive a ranked list of candidate compositions from the database. Enables property-driven discovery rather than composition-driven trial and error.
Machine learning models trained on the Layer 1 database predict physical, electronic, and mechanical properties of hypothetical compositions that have never been synthesized. Enables rapid screening of candidate materials before expensive DFT or experimental validation.
An evolutionary algorithm generates novel material candidates by mutating and recombining known compositions. Fitness is evaluated using Layer 3 prediction models. Evolutionary algorithms and numerical optimization drive the search toward compositions with target properties.
The system improves from feedback: validated predictions refine ML models, user-submitted experimental data enriches the database, and discovered materials become seeds for further generative exploration. A self-improving loop that increases accuracy and coverage over time.
Access to the Materials Discovery Engine requires a SOST escrow deposit using the same mechanism as PoPC Model B. Escrow deposits are denominated in USD and converted to SOST at the current market rate at the time of deposit. This protects users from volatility — whether SOST is worth $1 or $10,000, the access cost remains stable in real terms.
| Tier 1 | $10 USD equivalent in SOST · 1 month — Layers 1–3 // Search + prediction |
| Tier 2 | $20 USD equivalent in SOST · 2 months — Layers 1–4 // + generative discovery |
| Tier 3 | $50 USD equivalent in SOST · 6 months — Layers 1–5 // full access |
The algorithm is FREE. The escrow deposit is a spam-prevention mechanism, NOT a fee. 100% of the deposit is returned to the user upon tier expiry. No portion is burned, redistributed, or retained by the protocol.
The Materials Discovery Platform is an independent research initiative within the SOST ecosystem. It uses dedicated ML models and optimization algorithms, not the ConvergenceX mining engine. Practical deployment requires significant additional research in computational materials science. SOST commits to exploring this application without making premature claims about timelines or capabilities. Any deployment will be contingent on peer-reviewed validation and demonstration of scientifically meaningful results. This is a research direction, not a product commitment.