– layout: default title: Chemistry Roadmap nav_order: 7 nav_exclude: true —

Chemistry Roadmap

How the ISA framework approaches chemistry — from first principles to reaction mechanisms, without fitting parameters.

---

The problem with the standard picture

Computational chemistry hands you a menu: Hartree-Fock, DFT, CCSD, CASSCF, FCI. Each method makes different approximations. Choosing the right one requires specialist expertise, and for the hardest systems — transition-metal complexes, bond-breaking, catalytic transition states, biological electron transfer — every commonly used method either fails silently or is prohibitively expensive.

The ISA framework does not replace these methods. It tells you which stratum each electron pair belongs to before you compute anything, so you use the right method on the right pairs automatically.

---

The four layers

Layer 0 — Topology: the sCeleTon

What it is. Given a molecule (atoms and connectivity only), draw the C/T graph:

• A C-box for every frozen electron pair — ionic bonds, lone pairs, core electrons. These pairs are closed to correlation. DFT or Hartree-Fock is exact here.
• A T-arrow for every active correlated pair — covalent bonds with partial character, resonance, multi-reference systems. These pairs are open and must be treated with correlated methods.

What it costs. Nothing beyond the Lewis structure. No basis set, no SCF convergence, no functional.

What it tells you. How many T-arrows = the minimum active space size for CASSCF. All C-boxes = H⁰, DFT is fine. A few T-arrows = H¹, CCSD will handle it. Many T-arrows, or T-arrows with G₂ structure = H², CASSCF essential.

The tool. alchemi — builds the sCeleTon from a molecular geometry and recommends the active space.

Papers. doi:10.5281/zenodo.21300667 (Topological Active Space, Paper 588).

---

Layer 1 — Geometry: θ_G and the Weyl coordinate

What it is. Run a cheap MP2 calculation. From the Natural Orbital Occupation Numbers (NOONs), compute two quantities for each T-arrow pair:

• θ_G — the Grassmannian angle. Measures how far the ground-state wavefunction has moved from the single-determinant (Hartree-Fock) reference on the manifold Gr(n_e, n_orb). Small θ_G = ionic / weakly correlated. Large θ_G = covalent / strongly correlated / multi-reference.

  θ_G = 2 arccos(√(n_g / 2))

  where n_g is the NOON of the bonding orbital in the pair.

• Weyl c₂ — the second Weyl chamber coordinate. Predicts DFT error with correlation r = 0.990 across all tested systems. If c₂ < δ (threshold), DFT or CCSD is reliable. If c₂ ≥ δ, CASSCF is required.

What θ_G replaces. Pauling electronegativity is a fitted table with no geometric foundation. θ_G is a coordinate on a Riemannian manifold (the Fubini-Study metric on the Grassmannian), computable from first principles, that recovers the polarity ordering — LiF more ionic than HF, H₂ more covalent than LiF — without any fitting.

The β* snap. At θ_G ≈ 20°, the MGE routing switches from H¹ to H². This is the ionic → covalent crossover geometry: the point at which NaCl transitions from Na⁺Cl⁻ to Na·Cl under bond stretching, or where an Fe spin-crossover complex switches spin state under temperature. The snap position is predicted, not fitted.

Papers. doi:10.5281/zenodo.21277821 (Universal Bonding Theory, Paper 570).

---

Layer 2 — Energetics: automatic method routing

The routing table. Once Layer 1 is done, each electron pair has a stratum assignment. The router sends it to the right method:

Stratum	θ_G	Weyl c₂	Method	Why
H⁰ (C-box)	< 5°	—	Freeze / DFT	Single-determinant exact
H¹	5°–20°	< δ	CCSD	Single-reference correlation sufficient
H²	> 20°	≥ δ	CASSCF	Derivative discontinuity; DFT fails

What this is not. This is not “run CASSCF on everything” — that is expensive and unnecessary for H⁰ and H¹ pairs. It is not “use DFT on everything” — that silently fails for H² pairs. It is the right method for each pair, with the minimum active space identified by flats.

Why DFT fails at H². The DFT derivative discontinuity — the kink in the exact energy at integer orbital occupancy — is exactly the β* snap of the MGE at the Weyl c₂ threshold. DFT approximates this kink with a smooth functional, which is why it systematically errors on spin-state assignments, bond-breaking, and transition-metal complexes. The ISA framework avoids approximating the kink by working in the algebraic (tropical) regime where it is exact.

Papers. doi:10.5281/zenodo.21224113 (Galois Chemistry as Tropical DFT, Paper 491).

---

The routing is soft, not a decision tree

The routing table above shows hard thresholds (H⁰/H¹/H²). That is the β → ∞ limit — useful as a practical summary, but not the fundamental mechanism. The actual router is the Maslov-Gibbs Einsum (MGE) at finite inverse temperature β_eff:

β_eff = Δ_HL / (corr_scale · (2S+1))

where Δ_HL is the HOMO-LUMO gap and (2S+1) is the spin multiplicity.

At finite β_eff, the MGE assigns a continuous Boltzmann weight over methods — mostly DFT with a small CASSCF correction for a borderline pair, rather than a binary choice. The hard threshold table is recovered only at β_eff → ∞ (large gap, singlet). This is maximum-entropy inference in the sense of Jaynes: given the evidence (c₂, spin, gap), the MGE posterior over methods is a Boltzmann distribution, not a decision tree.

What this means practically. A borderline pair with c₂ just above δ does not need full CASSCF — it needs an MGE-blended energy with β_eff chosen from the gap. This is systematically cheaper than hard-routing everything above the threshold to CASSCF, and more accurate than ignoring the correlation altogether. The optimal δ* = (π/4)√α, where α = β_eff · ε (ε = correlation energy per pair), follows from the MGE cost-accuracy tradeoff (experiment x596e).

The routing table is for orientation. The MGE at finite β_eff is the engine.

---

Layer 3 — Mechanisms: Grassmannian geodesics

For reactions, not just static molecules. A CASSCF wavefunction is a point on the Grassmannian Gr(n_e, n_orb). The Fubini-Study metric defines arc-lengths between points. For a reaction R → TS → P:

• G_R, G_TS, G_P are points on Gr(k,n)
• The reaction coordinate is the arc-length from G_R to G_P
• Pauling’s complementarity principle — that an enzyme active site is complementary to the transition state, not the substrate — is a theorem: G_TS is the Fubini-Study geodesic midpoint between G_R and G_P

Catalytic Carnot efficiency.

η_cat = 1 − arc(G_R, G_TS) / arc(G_R, G_P)

At the geodesic midpoint, η_cat = 0.500 exactly. Computed for Diels-Alder (butadiene + ethylene → cyclohexene): η_cat = 0.503, within 0.3% of the midpoint. This is not a fitting result — it follows from the geometry.

Synchronous mechanism diagnostic. When arc(G_R, G_TS) ≈ arc(G_TS, G_P), the mechanism is synchronous and concerted. When the arcs are asymmetric, the mechanism is stepwise. This is the first computable, basis-independent diagnostic for reaction mechanism type.

Papers. doi:10.5281/zenodo.21279006 (Grassmannian Bonding and Scattering, Paper 574).

---

Layer 4 — Spectra: Racah recoupling

For spectroscopy, no DFT required. Given the quantum numbers produced by Layers 0–3 (spin state S, orbital angular momentum L, symmetry group G), compute spectroscopic observables exactly using the Racah algebra:

• Term symbols and selection rules — from ORBIT labels in Rep(SU(2))
• Transition energies — from FLOP (Wigner 6j symbol) via the Wigner-Eckart theorem
• EPR spectra, optical spectra — from TWIST (ribbon element) and FLOP
• The G₂ wall — where SU(2) seniority labelling becomes incomplete; Rep(G₂) provides the missing label. Occurs at the atomic f-shell (j = 7/2) and the nuclear 1g₉/₂ shell (j = 9/2)

The tool. spectrafold — Racah coefficients, Pandya transform, G₂ wall diagnosis, exact sympy arithmetic.

Papers. doi:10.5281/zenodo.21219702 (Valence as Orbit Occupancy, Paper 487); doi:10.5281/zenodo.21219722 (Galois Chemistry, Paper 488).

---

The roadmap as a decision tree

Given a molecule:

  Step 1 — Build sCeleTon (Layer 0)
    All C-boxes? ──→ H⁰: use DFT, done.
    Has T-arrows?  ──→ continue

  Step 2 — Compute θ_G and Weyl c₂ from MP2 NOONs (Layer 1)
    c₂ < δ for all T-arrows? ──→ H¹: use CCSD on T-arrows, done.
    c₂ ≥ δ for any T-arrow?  ──→ H²: continue

  Step 3 — Run CASSCF on H² T-arrows only (Layer 2)
    Active space = H² T-arrows identified by alchemi
    CCSD on remaining H¹ T-arrows
    Freeze all C-boxes

  Step 4 — Predict spectrum from quantum numbers (Layer 4)
    Use spectrafold on ORBIT labels from Step 3

  For reactions: compute Grassmannian geodesic (Layer 3)
    Arc-length = reaction coordinate
    η_cat = catalytic efficiency

---

Three reader paths

Drug designer / biochemist. Start with Layer 3 (Pauling theorem) — this is why your enzyme works. Then Layer 0 (sCeleTon tells you which bonds are active in catalysis). Papers: 574, 490 (protein design).

Computational chemist. Start with Layer 1 (θ_G and Weyl router) — this is the DFT failure detector you have been missing. Then Layer 2 (automatic active-space selection). Tools: alchemi. Papers: 596, 588, 491.

Theoretical chemist / physicist. Start with the ISA zoo entries C01–C05 and S01–S05 — validated ISA programmes across chemistry and spectroscopy. Then the categorical foundations: Paper 591.

---

ISA zoo entries for chemistry

See the ISA Zoo filtered to Chemistry and Spectroscopy for the full list of validated entries. Key entries:

ID	System	Layer	Result
C01	FeMoco N₂ fixation	0+2	14-opcode programme; spin ladder confirmed
C02	Fe(II) spin crossover	0+1	SCO gate = β* snap; 20/20 benchmark
S01	TM d-shell spectroscopy	4	Racah exact; Tanabe-Sugano from ORBIT
S04	FMO exciton dynamics	2+4	BIND fires at 7-BChl Fano geometry