Athanor Chemical Lab

One number predicts how strongly atoms bond. We assign each element a single thermodynamic parameter — Pe (Péclet number) — derived from its effective nuclear charge, electron affinity, and shielding. Then we use Langevin dynamics to simulate how atoms move, collide, and form bonds. The result: Pe alone predicts bond dissociation energies across 56 known bonds with ρ = 0.881.

Why not just use electron shell theory?

Electron shell theory is the gold standard for why bonds form (orbital overlap, electron sharing, quantum mechanics). We don't replace it.

What we show is that a single transport parameter (Pe) — computed from three atomic properties that electron shell theory already predicts — recovers bond strength rankings with ρ = 0.881. That's a convergence result: the full quantum picture compresses into one number for predicting how much energy it takes to break a bond.

Think of it like body mass index (BMI): it doesn't replace cardiology, but it captures a surprising amount of health variance with one number. Pe does the same for bond stability. And because it's a transport parameter, it naturally connects to reaction kinetics, thermal stability, and material behavior under stress — things that are harder to get from orbital diagrams alone.

Pe is not a new theory of chemistry. It's a thermodynamic lens on data that electron shell theory already explains at a deeper level. The interesting part is that the lens works — and generalizes to domains beyond chemistry.

0.881
Rank correlation (ρ)
p = 3.1e-19
0.785
Cross-validated ρ
10-fold, mean
129
RMSE (kJ/mol)
training error
56
Known bonds
calibrated against
331
Novel predictions
falsifiable

Known/Predicted Bonds

Bond Predicted (kJ/mol) Actual (kJ/mol) Error % Type
+ Drag to orbit. Scroll to zoom.
Click Simulate to visualize Langevin bond formation
Theta Trajectory
Energy Landscape V(theta)
25 C Pe=0.14
Simulation (64x64 grid · LJ 12-6 · valence bonds)
Physics
Bond History
Energy
Element Inventory
Selected: 0 / 12
200 C
Crucible Simulation (server-authoritative · 500ms tick)
Crucible State
No active crucible
Genesis Ladder
0 Atoms
1 Bonds
2 Compounds
3 Autocatalytic
4 Self-Maintaining
5 Genesis
My Crucibles
My Compounds
Ancient Recipe
1000 C Pe=5.71
Forge Simulation (64x64 · LJ 12-6 · valence bonds)
Alloy Physics
Bond History
Theta Stability
Stability Pairing Matrix — elements with similar Pe (ΔPe < 0.15) stabilize each other in alloys

10-Fold Cross-Validation

Leave-30%-out splits (ATOM-04, N=56). Kill condition: CV ρ > 0.5. Mean = 0.785 — SURVIVES.

Split number
SplitTrain NTest NrhoRMSEp-value

Fitted Parameters

Mapping: atomic (Z_eff, EA, sigma) to Langevin (alpha, gamma, c) with 8 free coefficients.

Beam A (left)
Beam B (right)
50 25 C
Material Collider (64x64 · LJ 12-6 · Langevin dynamics)
Beam A Physics
Beam B Physics
Collision Data
Cross-bond events / tick
Pe Differential Matrix — Beam A vs Beam B (higher = stronger interaction at collision)
True Light Engine — 12 Spokes (live)
Connecting...
Chemistry Modulations
Onsager Feedback (Chemistry → GM)
KAUZMANN THERMODYNAMICS
Universal entropy barrier · Sconf → 0 = ideal glass
Start sim + open GM tab to see live Kauzmann metrics
GM SHARK — HEX EDITOR
Addr Hex Dec Bits Decoded Poke
Codes (click to apply)
Cosmic Ray (manual trigger)
Sim Memory (read-only)
Iron Ward Effect (Fe · Z=26 · Pe=0.049)
Binary State Log (8 bytes/sample, XOR delta, popcount)
Format: state_hex Δdelta_hex pop=bits_changed — 64-bit packed: [cascade:2|gate:2|lattice:2|drift:2][flags:4|rsv:4][fdr:8][var:8][crooks:8][cooper:8][hyst:8][pe:8]
On-Chain Snapshots (CrucibleRegistry)