MATERIALS ENGINE

Materials Discovery
Platform

An independent research platform for computational materials science, powered by the SOST ecosystem. Uses dedicated ML models and optimization algorithms — not ConvergenceX mining.

RESEARCH | ML models · Optimization algorithms · SOST escrow access
01 — OVERVIEW

Materials Discovery Platform

An Independent Research Platform
RESEARCH

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.

02 — COMPUTATION

Computational Approach

ML + Optimization + DFT Integration
METHODOLOGY

The platform combines multiple computational approaches for materials discovery, each using purpose-built algorithms independent of the ConvergenceX mining engine.

ML ModelsGraph neural networks for property prediction // CGCNN, MEGNet, ALIGNN
OptimizationEvolutionary algorithms and numerical optimization // composition search, structure relaxation
DFT IntegrationAPI access to VASP, Quantum ESPRESSO, GPAW // ab initio validation
Compute DonationMiners donate idle CPU cycles between blocks // voluntary, opt-in community compute
Community Compute Marketplace
FUTURE

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.

03 — ARCHITECTURE

Five-Layer Architecture

Layer 1
Data
~150K known materials
from public databases
Layer 2
Inverse Search
Specify properties
get ranked candidates
Layer 3
Prediction
ML models predict
hypothetical properties
Layer 4
Discovery
Evolutionary algorithm
generates candidates
Layer 5
Learning
Continuous improvement
from feedback loops
Layer 1 — Data Foundation
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 ProjectInorganic crystal structures + computed properties
JARVISDFT-computed properties, 2D materials, heterostructures
AFLOWAutonomous high-throughput materials discovery library
Layer 2 — Inverse Search
SEARCH

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.

Layer 3 — Property Prediction
ML MODELS

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.

Layer 4 — Generative Discovery
GENERATIVE

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.

Layer 5 — Continuous Learning
FEEDBACK

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.

04 — ACCESS

Access via SOST Escrow

SOST Deposit Tiers
ESCROW

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.

Tier 1: Search + Predict
$10 USD in SOST, 1 month lock.
Query the database, run inverse search, predict properties of known and hypothetical compositions.
Tier 2: + Generative
$20 USD in SOST, 2 month lock.
Everything in Tier 1 plus evolutionary candidate generation and optimization runs.
Tier 3: Full Access
$50 USD in SOST, 6 month lock.
All capabilities including DFT validation queue, data export, and priority compute allocation.
05 — DISCLAIMER

Research Disclaimer

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.