PNNL's AI Screens 1.5 Million Glass Formulas for Hanford Nuclear Waste Vitrification

Pacific Northwest National Laboratory (PNNL) deploys AI models to screen 1.5 million glass compositions for Hanford Site's 56 million gallons of legacy nuclear waste. The U.S. Department of Energy (DOE) funds the effort. John Vienna, PNNL glass scientist, says models train on 12,000+ datasets (PNNL release).

Vitrification immobilizes radionuclides in stable glass logs for geologic storage. Hanford stores waste in 177 underground tanks.

AI Training Targets Key Glass Properties

PNNL trains machine learning (ML) models on oxide compositions like silica (SiO2, 40-60 wt%) and alumina (Al2O3, 10-20 wt%). Outputs include glass transition temperature (Tg, 500-600°C), normalized leach rate (<10^{-7} g/m²/day), and viscosity (10-100 Poise at 1,000°C).

Albert Kruger, PNNL materials engineer, states top candidates pass IEC 12113 chemical durability tests. AI explores over 10^6 compositional spaces. Thermodynamic models flag crystallization risks in 1,000°C melts. This cuts physical experiments by 50% versus trial-and-error.

Bechtel National integrates results into Hanford's Waste Treatment Plant (WTP).

Hanford Waste Demands Precise AI Optimization

Hanford waste features variable radionuclides and salts (up to 30 wt% Na2O). Phase separation risks arise from phosphates. Donghee Park, PNNL researcher, notes AI detects unstable blends early for 1,000+ year repository life.

Industrial melters produce 300 gallons daily. PNNL models simulate 90-day corrosion tests, capping solubility at <0.1 wt%. Park adds simulations ensure melt homogeneity under WTP conditions.

$19 Billion WTP Relies on AI Throughput Gains

WTP processes low-activity waste into 10-foot glass logs by 2030s (DOE Hanford site). Each log endures 10^5-year isolation.

Vienna reports AI boosts throughput 20-30%, offsetting $19 billion overruns. Reliable waste handling maintains U.S. nuclear at 20% grid capacity. This pairs with lithium-ion batteries for baseload power.

AI Parallels Boost Energy Storage Materials

PNNL applies AI to solid-state battery electrolytes targeting 500 Wh/kg density and 1,000 cycles at 80% retention. Glass ceramics match with 300 Wh/kg volumetric density and 2,000 cycles.

U.S. silica supply chains avoid China lithium risks. Optimized glasses cut costs 15% to $80/kWh LCOE. Flow batteries hit $150/kWh; lithium-ion packs $100/kWh.

Manufacturing reaches MRL 6 for prototypes. Pilot lines mirror Hanford melters.

• Metric: Energy Density (Wh/kg) · Vitrification Glass: N/A · Solid-State Electrolyte: 500 · Glass Ceramic Storage: 300 (volumetric)
• Metric: Cycles · Vitrification Glass: 1,000+ yr stability · Solid-State Electrolyte: 1,000 @ 80% · Glass Ceramic Storage: 2,000
• Metric: Cost ($/kWh) · Vitrification Glass: N/A · Solid-State Electrolyte: Target $80 · Glass Ceramic Storage: Benchmark $100

Grid Benefits from Accelerated Cleanup

AI reduces cycles from 2 years to 3 months. Hanford deploys glasses by 2026. NRC fast-tracks proven formulas.

Operators gain firm nuclear capacity in 4-hour battery hybrids. PNNL advances DOE AI for waste (DOE page). Pilots confirm gains for TWh-scale storage.

Faster vitrification unlocks nuclear for renewables integration, stabilizing grids with long-duration storage.