Diagrammatic Chemistry

Where the algebraic skeleton ends and the Hamiltonian begins — and which diagram calculi already do the work.

The dividing line
1. Where the line actually bites
The diagram calculi, and who already uses them
1. The pattern worth naming
What is genuinely open
And what is not open
Further reading

The dividing line

Chemistry splits cleanly into two halves, and confusing them is the commonest error in applying mathematics to it.

The skeleton is combinatorial. Which states exist, how they are labelled, what can couple to what, which transitions are forbidden — all of this is fixed by symmetry and combinatorics before any energy is computed. It is exact, it is cheap, and it is often the same in every molecule of a given shape.

The flesh needs a Hamiltonian. Which of the allowed states is lowest, how big the gaps are, how fast a reaction goes — none of this follows from symmetry. It requires solving the Schrödinger equation with real integrals over real distances.

Question	Settled by	Needs a Hamiltonian?
Which orbitals exist for a given n?	SO(4) → SO(3) branching	no
Which terms arise from a d³ configuration?	Young tableaux, character tables	no
Which transitions are allowed?	selection rules, triangle conditions	no
How many CSFs in an active space?	Weyl–Paldus dimension formula	no
Which point group does a molecule have?	its geometry	no
Which orbitals belong in the active space?	judgement — see below	partly
Which term is the ground state?	Hund’s rules, then computation	yes
What is the bond length?	energy minimisation	yes
What is the barrier height?	transition-state calculation	yes
Is this molecule multireference?	occupation numbers, not symmetry	yes

The skeleton is why a chemist can tell you the term symbols of an ion in seconds and cannot tell you its excitation energy without a computer.

Where the line actually bites

Take FeMoco, the iron–molybdenum cofactor of nitrogenase. Five things make it hard, and only one is a symmetry problem:

Difficulty	Symmetry problem?
Coupling 7 Fe spins to a total S = 3/2	yes — this is angular-momentum recoupling
Active space of ~54 electrons in ~54 orbitals	no — combinatorial size, not structure
Dynamic correlation through the sulfur bridges	no — needs real integrals
Which Fe is 2+ and which 3+ in each E-state	partly — broken-symmetry DFT is the tool
Where N₂ binds; the E₄ Janus intermediate	no — structural and experimental

One in five. That ratio is worth remembering before proposing that a new diagrammatic method will crack a catalysis problem.

The diagram calculi, and who already uses them

Mathematics has built a dozen graphical calculi. Chemistry uses several of them, often under different names, and the ones it does not use are mostly the ones that answer questions chemistry does not ask.

Calculus	What it draws	Used in chemistry?
Goldstone / Hugenholtz diagrams	many-body perturbation terms; antisymmetrised vertices handle exchange in one diagram	yes — the standard language of MBPT and coupled cluster since the 1950s
Brandow diagrams	folded diagrams for effective Hamiltonians	yes — open-shell and quasi-degenerate PT
Wick contraction diagrams	operator contraction bookkeeping	yes — every CC derivation
GUGA (Graphical Unitary Group Approach)	the Gel’fand–Tsetlin lattice as a walk on a graph	yes — in production MCSCF codes since the 1970s
Yutsis / JLV diagrams	3nj recoupling of angular momenta	yes — molecular magnetism; MAGPACK computes polynuclear cluster spectra with 6j and 9j symbols
Young tableaux	irrep labels, branching rules, CSF counts	yes — spin eigenfunctions, term symbols
Character tables	point-group irreps and selection rules	yes — undergraduate spectroscopy onward
Tensor-network diagrams	MPS and PEPS contraction	yes — DMRG
Weight and root diagrams	Lie algebra structure	rarely — implicit in SO(4,2) treatments of the periodic table
Birdtracks (Cvitanović, Penrose)	Casimirs, irrep dimensions, invariant tensors	no — computes things chemists get from tables
Crystal bases (Kashiwara)	representation theory as coloured graphs at q → 0	no — but GUGA is the same poset walked differently
ZX / ZW / ZH	qubit and fermionic processes	not yet — ZW has a fermionic variant with a completeness theorem
Dynkin diagrams	the classification of simple Lie algebras	not applicable — classifies algebras, not states

The pattern worth naming

Three times while auditing this corpus, a claim of the form “chemists do not have this tool” turned out to be false:

Recoupling. Yutsis diagrams draw the 6j/9j algebra that molecular magnetism has used since at least 2001.
Crystal bases. GUGA is the graph-walk formulation of the same poset, in production codes since the 1970s.
Diagrammatics for exchange. Antisymmetrised Goldstone diagrams have done precisely this job since the 1950s.

The fields whose tools appear absent from chemistry have usually been imported decades ago under a different name. Anyone proposing a new diagrammatic method should search for its chemical alias first.

What is genuinely open

Not the existence of a calculus — the completeness of one.

Categorical quantum mechanics has completeness theorems: ZX-calculus is complete in the sense that every true equation between the processes it describes is derivable from its rewrite rules. That is what makes it a calculus rather than a notation, and the theorem came years after the rules.

Goldstone diagrams have no such theorem. Diagram equivalence in MBPT is handled by symmetry factors and topological equivalence — conventions, not a stated rule set. Whether the calculus is complete, and whether it is a fragment of the fermionic ZW-calculus (whose W generator encodes antisymmetry natively), appear to be open questions.

Two cautions, both real:

The linked-cluster theorem already does much of what a completeness result would do. It states exactly which diagrams contribute to a size-extensive energy. A categorical theorem here risks restating a 1950s result.
Goldstone diagrams are perturbative. They assume a single dominant configuration, which fails precisely in the strongly correlated regime — stretched bonds, transition-metal clusters — that makes chemistry hard. A completeness theorem would be a result about the well-behaved case.

And what is not open

The non-perturbative problem is solved, several times over. Coupled cluster’s exponential ansatz e^T resums infinite classes of diagrams in closed form and has been the workhorse since the 1960s; DMRG, Green’s-function methods and quantum Monte Carlo are further mature answers. What remains hard is strong correlation, where DMRG is currently the best available tool.

There is no amplituhedron shortcut. The amplituhedron is not a general technique for summing diagrams; it is a statement about planar N = 4 super-Yang–Mills, whose dual conformal and Yangian symmetries are large enough to fix the answer geometrically. Chemistry has no conformal symmetry (Coulomb plus fixed nuclei break it), no supersymmetry, no planarity, massive non-relativistic electrons, and computes an energy rather than an S-matrix element.

Two-electron integral sparsity is ordinary angular momentum. Structural zeros come from SO(3) triangle conditions and parity, which every code already exploits. Tested directly: SO(4) accounts for about 0.2% of the structural zeros, and the zero pattern is a property of the basis rather than of the element — hydrogen and helium in the same basis have identical patterns.