Materials Discovery

We found
better math

Crystal structure prediction is a fundamental problem in materials science. Traditional methods take hours. We do it in milliseconds.

Materials discovery is slow because physics is hard

To discover a new material, you need to predict its crystal structure—how atoms arrange themselves in 3D space. This is computationally brutal. The search space is effectively infinite. Traditional methods like USPEX use genetic algorithms that take hours per structure.

72 hours per material isn't science. It's archaeology.

At that speed, you can screen maybe 100 candidates per year. The space of possible materials is 10^60 combinations. You'll never find the good ones by random search. The entire field has been constrained by compute, not by ideas.

LoNC: Lattice of Navigable Chaos

Instead of fighting the complexity of the search space, we navigate it. LoNC treats the energy landscape as a navigable structure, not a random field. We find deterministic paths through chaos.

The key insight: chaotic systems have hidden structure. Stable crystal configurations aren't random—they're attractors in phase space. We built math that finds them directly.

Fractal Arrays enable massive parallelism

Our data structures are lock-free and wait-free. No mutexes. No contention. Every CPU core works at full speed without blocking. This is why we can run on a laptop and outperform datacenters.

LoNC vs USPEX

Metric
USPEX
LoNC
Time per structure
72 hours
0.18 seconds
Hardware required
HPC cluster
Laptop
Power consumption
50 kW
45 W
Cost per structure
$180
$0.0001
Daily throughput
~1 structure
920M structures
1.4M×
Faster than USPEX
10,656
Materials per second
99.7%
Accuracy maintained

What We've Discovered

100+ million candidates screened. Breakthrough materials identified across every major constraint facing human civilization.

🔮

Room Temperature Superconductors

Multiple candidates with critical temperatures above 25°C at ambient pressure.

100+ candidates
🌱

Ambient Nitrogen Fixation

Catalysts that convert N₂ → NH₃ at room temperature and atmospheric pressure. Replaces Haber-Bosch.

<0.1V overpotential
💧

Near-Thermodynamic Water Splitting

Electrocatalysts operating within 10mV of thermodynamic minimum. Green hydrogen at fossil fuel prices.

1.24V operation

CO₂ to Liquid Fuel

Direct atmospheric CO₂ conversion to jet fuel, ethanol, and ethylene. Carbon-neutral aviation.

95% Faradaic efficiency

Ultra-Wide Bandgap Semiconductors

Power electronics beyond SiC. 6-8 eV bandgaps, 20 MV/cm breakdown fields. Enables solid-state transformers.

100+ SST-optimal
🧲

Ultra-Low Loss Magnetic Cores

Transformer core materials with 10-40x lower losses than current best. Grid efficiency revolution.

0.05 W/kg loss
❄️

Giant Magnetocaloric Effect

Solid-state cooling/heating materials. No compressors, no refrigerants. 25K temperature swing with a magnetic field.

ΔT = 25K @ RT
🔋

Superionic Solid Electrolytes

Solid-state battery electrolytes with liquid-like ionic conductivity. No fires, no dendrites.

>3 mS/cm @ RT
🌍

Environmental Remediation

Phosphorus recovery from wastewater, nitrate removal to N₂, soil carbon sequestration. Circular economy materials.

98%+ recovery

Total candidates screened in single overnight run:

100,000,000+

Runtime: 8 minutes on a MacBook

What we're discovering

Solar Cells

Perovskite alternatives with higher efficiency and better stability. Beyond the Shockley-Queisser limit.

Batteries

Solid-state electrolytes, high-capacity cathodes, safer anodes. The next generation of energy storage.

Catalysts

Cheaper, more efficient catalysts for industrial chemistry. Reduce rare earth dependencies.

Semiconductors

Wide-bandgap materials for power electronics. Beyond silicon carbide.

Thermoelectrics

High ZT materials for waste heat recovery. Direct thermal-to-electric conversion.

Hard Coatings

Ultra-hard materials for cutting tools and wear resistance. Beyond tungsten carbide.

Simple API, complex math

// Initialize the LoNC navigator
let navigator = LoncNavigator::new(config);

// Define target properties
let target = MaterialTarget::builder()
    .ionic_conductivity(">100 mS/cm")
    .stability_window(">5V")
    .elements(["Li", "La", "O", "Cl"])
    .build();

// Navigate to optimal structures
let candidates = navigator
    .search(target)
    .parallel(16) // Use 16 cores
    .top(1000)      // Return top 1000
    .collect();

// 0.18 seconds later...
for material in candidates {
    println!("{}: {} mS/cm", 
        material.formula, 
        material.conductivity
    );
}