Magic Has a Kind, Not Just an Amount

Plain-language explainer for doi:10.5281/zenodo.20541583

The central idea in one sentence

The 3-qubit phase space has a hidden seven-fold structure — the Fano plane — and a state’s kind of magic (which of the seven orbits carries its Wigner negativity) determines which computations it enables, independently of how much magic it has.

Why do we care?

Quantum computers can do things classical computers cannot. But not all quantum gates are created equal. The Clifford gates — H, CNOT, S — are powerful but ultimately simulable classically. To get genuine quantum advantage you need a non-Clifford gate: T, CS, CCZ. The resource that makes non-Clifford gates possible is called magic.

The standard picture: magic is a single number, the Wigner negativity $\mathcal{N}(\rho)$. More negativity = more magic = better. Factory protocols accept states above a negativity threshold and discard the rest.

This paper shows that picture is incomplete. Magic has a kind — a valence — not just an amount. Two states with identical negativity can be completely useless for each other’s intended computations. Worse: using the wrong kind of magic actively makes things worse than using no magic at all.

The geometry: a 63-point universe

The 3-qubit Pauli group (ignoring phases) has 63 non-identity elements. Think of these as 63 points in a discrete universe called the symplectic polar space $W(5,2)$.

This universe has a natural geometry: two points commute if their symplectic inner product is zero, anticommute otherwise. The geometry determines which observables can be measured simultaneously.

Hidden inside this 63-point universe is the Fano plane $PG(2,2)$: a 7-point subgeometry where every pair of points commutes, and every triple of points forms a “line” (a maximal commuting set of size 3). These 7 points are exactly the GHZ stabilisers:

ZZI,  ZIZ,  IZZ,
XXX,  YYX,  YXY,  XYY
The GHZ state $ 000\rangle + 111\rangle$ is supported exactly on these 7 points — its Wigner function is $+1/8$ on the Fano plane and zero everywhere else. The Fano plane is the classical, zero-magic sector of 3-qubit phase space.

The 7 orbits: where magic lives

The remaining $63 - 7 = 56$ points are where Wigner negativity can live. This paper proves they are not a featureless sea — they split into exactly 7 orbits of 8 points each, one orbit per Fano line:

Orbit O₀:  commutes with line 0  {ZZI, ZIZ, IZZ},  anticommutes with all others
Orbit O₁:  commutes with line 1  {ZZI, XXX, YYX},  anticommutes with all others
  ...
Orbit O₆:  commutes with line 6  {IZZ, YXY, XYY},  anticommutes with all others

Each orbit has a signature: it commutes with exactly one Fano line and anticommutes with all six others. No other signatures exist — the 56 non-Fano points are completely partitioned by these 7 patterns.

Theorem (Fano Orbit Decomposition). The 56 non-classical points of $W(5,2)$ decompose as $$W(5,2) \setminus PG(2,2) = \mathcal{O}_0 \sqcup \mathcal{O}_1 \sqcup \cdots \sqcup \mathcal{O}_6$$ where each orbit $\mathcal{O}_L$ has exactly 8 points, determined by commuting with Fano line $L$.

The magic valence label

Definition. For a magic state $\rho$, the orbit negativity is \(\mathcal{N}_L(\rho) = \sum_{u \in \mathcal{O}_L,\, W_\rho(u) < 0} \lvert W_\rho(u)\rvert\) and the magic valence label is ${p_L(\rho)} = {\mathcal{N}L / \mathcal{N}}{L=0}^6$ — a probability distribution over the 7 orbits.

This is a strictly finer invariant than the total negativity $\mathcal{N}$. Two states can agree on $\mathcal{N}$ but differ on every $p_L$.

Examples:

State $\mathcal{N}$ Dominant orbit Valence
$\mathrm{CS}_{01}|{+++}\rangle$ $1/8$ $\mathcal{O}_1$ only $(0,1,0,0,0,0,0)$
$\mathrm{CS}_{02}|{+++}\rangle$ $1/8$ $\mathcal{O}_2$ only $(0,0,1,0,0,0,0)$
$\mathrm{CS}_{12}|{+++}\rangle$ $1/8$ $\mathcal{O}_3$ only $(0,0,0,1,0,0,0)$
$\mathrm{CCZ}|{+++}\rangle$ higher 6 of 7 orbits spread

The three CS states have identical total negativity. A negativity-only filter cannot tell them apart.

Three levels of nested Fano structure

The orbit decomposition reveals three nested levels of Fano geometry:

Level 0 — The Fano plane itself: the 7 GHZ stabilisers, classical, zero magic.

Level 1 — The orbit labels form a second Fano plane: for any two orbits $\mathcal{O}L$ and $\mathcal{O}_M$, all cross-XOR products $u \oplus v$ land in a unique third orbit $\mathcal{O}{L \oplus M}$. The XOR triples \(\{0,1,4\},\ \{0,2,5\},\ \{0,3,6\},\ \{1,2,3\},\ \{1,5,6\},\ \{2,4,6\},\ \{3,4,5\}\) form a second copy of $PG(2,2)$ on the orbit labels. This is the XOR-Fano composition rule — it governs how magic states combine under two-qubit gates.

Level 2 — Each orbit contains two Lagrangian 4-cliques: each $\mathcal{O}_L$ splits into two groups of 4 ($\mathcal{O}_L^A$, $\mathcal{O}_L^B$) where all 4 points within each group mutually commute, and all 16 cross-pairs anticommute. Each 4-clique is a maximal abelian subgroup — the same type of structure as the Fano plane itself.

The broken-line connection

Paper 325 showed that the FMO biological light-harvesting complex implements a 6-7 architecture: 6 intact Fano lines plus 1 weakened line. This produces directed energy transfer with efficiency $\eta > 0$. The intact 7-line Fano gives $\eta = 0$.

This paper shows these are the same phenomenon in phase space:

Physical picture (6-7 architecture) Phase-space picture (this paper)
Intact Fano: $\eta = 0$, equilibrium GHZ: Wigner on $PG(2,2)$, zero magic
Break line $L$: $\eta > 0$, directed Magic: Wigner activates $\mathcal{O}_L$
$PSL(2,7)$ symmetry intact Acts transitively on all 7 orbits
One broken line $p_L = 1$ — pure single-orbit magic

The orbit $\mathcal{O}_L$ is the phase-space incarnation of the broken Fano line $L$. Magic is broken-line activation.

The spontaneous symmetry breaking analogy

The 6-7 Fano symmetry breaking $PSL(2,7) \to S_4$ is the discrete finite-group analogue of electroweak symmetry breaking $SU(2) \times U(1) \to U(1)$:

  • 6 intact Fano lines ↔ massless photon (unbroken generator)
  • 1 broken Fano line ↔ massive W and Z bosons (broken generators)
  • activated orbit $\mathcal{O}_L$ ↔ Goldstone mode
  • $\eta > 0$ ↔ Higgs VEV (order parameter)
  • 7 choices of broken line ↔ 7 discrete vacua

The mechanism differs — Fano breaking is architectural or evolutionary rather than potential-driven — but the group-theoretic structure is the same.

For the full technical treatment, see doi:10.5281/zenodo.20541583.