Fibonacci and Ising Anyons

A non-Abelian anyon theory is a piece of bookkeeping so rigid that it doubles as a quantum computer. Specify a finite set of charges, the fusion rules saying which charges combine into which, and the associativity ( $F$ ) and braiding ( $R$ ) data that make exchange a consistent unitary action on the degenerate fusion space — and the theory is complete. Two such theories dominate every course and lab proposal because they are the two simplest with a single self-fusing non-Abelian charge: the Fibonacci and Ising models. They sit at opposite ends of one axis, computational power, and between them draw the sharpest line topology has to offer. The Fibonacci model is the minimal universal theory. Its one nontrivial charge $\tau$ obeys $\tau\times\tau=1+\tau$ , its fusion spaces have Fibonacci-number dimensions, and the Freedman–Larsen–Wang theorem proves braiding alone generates a dense subgroup of the unitaries on each. By Solovay–Kitaev any gate is then a braid word of length polylogarithmic in the target accuracy — full quantum computation from nothing but moving particles around one another. The Ising model is the canonical non-universal theory. Its charges $\{1,\sigma,\psi\}$ are the anyon content of the Majorana / $p+ip$ superconductor and the Moore–Read $\nu=5/2$ state, and braiding $\sigma$ anyons produces exactly the Clifford gates — genuinely non-Abelian, exponentially protected, yet by Gottesman–Knill classically simulable. The organizing lesson: universality belongs to the modular tensor category, not to non-Abelian statistics as such, and Ising's one missing ingredient — a non-Clifford magic gate — connects topology to the whole fault-tolerant gate-synthesis program.

Key takeaways

Fibonacci anyons have a single nontrivial charge $\tau$ with $\tau\times\tau=1+\tau$ ; braiding them is dense in $\mathrm{PU}(2)$ on each fusion space, so braids alone are universal for quantum computation.
Ising anyons have charges $\{1,\sigma,\psi\}$ with $\sigma\times\sigma=1+\psi$ , $\sigma\times\psi=\sigma$ , $\psi\times\psi=1$ ; their braids generate only the Clifford group, which Gottesman–Knill makes classically simulable — non-universal.
Both theories are fixed by their $F$ -matrices (fusion associativity) and $R$ -matrices (braiding phases), constrained by the pentagon and hexagon equations up to gauge.
Quantum dimensions read off the fusion rules: $d_\tau=\varphi=(1+\sqrt5)/2\approx1.618$ , $d_\sigma=\sqrt2$ ; total quantum dimensions $\mathcal D_{\mathrm{Fib}}=\sqrt{1+\varphi^2}\approx1.902$ and $\mathcal D_{\mathrm{Ising}}=2$ .
Ising's missing $T=\mathrm{diag}(1,e^{i\pi/4})$ gate is supplied by magic-state distillation, restoring universality while the Clifford part stays topologically protected; Ising arises from Majorana modes and the $\nu=5/2$ state, Fibonacci is sought in $\nu=12/5$ .
Key idea: an anyon theory's computational power is decided by whether its braid-group image in $\mathrm{PU}(\text{fusion space})$ is dense (Fibonacci ⇒ universal) or finite (Ising ⇒ Clifford-only); the $F$ and $R$ data satisfying the pentagon and hexagon equations are all you need to compute either.

The core insight

An anyon theory's computational power is fixed by whether the braid-group image in $\mathrm{PU}(\text{fusion space})$ is dense (Fibonacci ⇒ universal) or finite (Ising ⇒ Clifford-only); the $F$ and $R$ matrices satisfying the pentagon and hexagon equations are all you need to compute either.

Where this lives today

Both theories have moved from blackboard toward cryostat, at different speeds. Ising-type statistics are the largest hardware bet: Microsoft's topological-qubit program encodes qubits in Majorana zero modes at the ends of proximitized nanowires, and its 2023–2025 devices (the "topological gap protocol" and the Majorana 1 processor) braid Ising $\sigma$ excitations for protected Clifford gates, with distillation planned to close the universality gap. In fractional quantum Hall systems, interferometry has shown anyonic braiding directly (Nakamura et al., 2020), and the $\nu=5/2$ plateau — home of Ising anyons — is probed by thermal-Hall measurements. Fibonacci anyons are the harder prize: none has been produced, the leading $\nu=12/5$ candidate is fragile and contested, and proposals to engineer Fibonacci statistics from coupled Majorana or parafermion arrays are an active frontier. The framing open problem is stark: every demonstrated non-Abelian platform is Ising (Clifford-only) type, so the first Fibonacci anyon would be the first intrinsically universal topological medium ever observed.

The physical picture

Picture an anyon model as a self-consistent accounting system for fusion. Bring two anyons together and ask their total charge; the fusion rules list the allowed answers, and when a pair can fuse more than one way, the which-way label is a quantum degree of freedom no local measurement can read. Stack many anyons and the number of consistent fusion histories is the dimension of a protected Hilbert space — protected because the information lives in the global fusion pattern, not any one particle. For Fibonacci, fusing $n$ copies of $\tau$ to charge $\tau$ gives dimension the $n$ -th Fibonacci number, growing like $\varphi^{\,n}$ : the quantum dimension $\varphi\approx1.618$ is the degeneracy-growth rate per anyon, the golden ratio solving $\varphi^2=\varphi+1$ .

Figure

The protected Hilbert space of n Fibonacci τ-anyons has dimension Fₙ — the Fibonacci numbers 1, 2, 3, 5, 8, 13, 21, 34 — growing as φⁿ divided by √5 with quantum dimension φ = 1.618 obeying φ² = φ + 1

Braiding is then a unitary rotating within this protected space, and the decisive question is geometric. Think of each exchange as a rotation of a sphere: do the available rotations fill the whole rotation group densely, or close into a finite, crystalline set of points? Fibonacci's two generators are incommensurate rotations about tilted axes — their closure is dense, and dense means universal. Ising's braids are rigidly the $\pi/2$ rotations and phase gates that snap shut into the 24-element Clifford group — finite, and finite means simulable. The same braid that gives Ising its noise protection also caps its power; the second figure sets the two Bloch spheres side by side.

Figure

Two Bloch spheres of the encoded qubit: Fibonacci braids fill PU(2) densely so the theory is universal with dₜ = φ ≈ 1.618, while Ising braids land only on the finite Clifford set of 24 rotations so the theory is non-universal with dₛ = √2

Neither model is an exotic curiosity. They are the two simplest members of the $SU(2)_k$ family: $k=2$ is the Ising-like non-universal theory, $k=3$ is Fibonacci, and the Freedman–Larsen–Wang theorem shows every $k\ge3$ except $k=4$ is universal — simplicity and universality coinciding almost exactly, which is why so much of the field is written in the language of these two examples.

The core equations

The starting point of any anyon model is its fusion rules — the multiplication table saying which total charges two anyons may carry. For Fibonacci there is one nontrivial charge, and the entire theory grows from a single self-fusing rule.

\tau\times\tau=1+\tau \qquad (1)

where: $\tau$ — the unique nontrivial anyon charge; $1$ — the vacuum charge (fusion identity); $+$ — direct sum of distinct fusion channels, here multiplicity-free. Reading it: two $\tau$ 's fuse either to vacuum or back to $\tau$ , and which outcome occurs stores a qubit of protected information. Ising has three charges but the same shape: one non-Abelian $\sigma$ that self-fuses two ways, plus an Abelian fermion $\psi$ .

\sigma\times\sigma=1+\psi,\qquad \sigma\times\psi=\sigma,\qquad \psi\times\psi=1 \qquad (2)

where: $\sigma$ — the non-Abelian anyon (Majorana / Ising spin field), $d_\sigma=\sqrt2$ ; $\psi$ — an Abelian fermion, $d_\psi=1$ , its own antiparticle; $1$ — vacuum. Reading it: $\sigma\times\sigma$ has two channels so $\sigma$ is non-Abelian; $\psi\times\psi$ has one outcome so $\psi$ is Abelian. The number controlling how fast the Hilbert space grows is the quantum dimension, and it is not free: it is the largest eigenvalue of the fusion matrix $N_a$ , read off the fusion rule.

N_a\,\vec v=d_a\,\vec v,\qquad d_\tau^2=d_\tau+1\Rightarrow d_\tau=\varphi,\qquad d_\sigma^2=1+1\Rightarrow d_\sigma=\sqrt2 \qquad (3)

where: $N_a$ — the matrix $(N_a)_{bc}=N_{ab}^{\,c}$ counting fusion channels; $d_a$ — quantum dimension (Perron–Frobenius eigenvalue); $\varphi=(1+\sqrt5)/2$ . Reading it: $d_\tau^2=d_\tau+1$ is the fusion rule (1) with charges replaced by dimensions — why the golden ratio is baked into Fibonacci from the start. Associativity of fusion is not automatic — reassociating three anyons requires a basis change between fusion trees, the $F$ -matrix, which for three $\tau$ 's of total charge $\tau$ is a $2\times2$ orthogonal matrix fixed (up to gauge) by the pentagon equation.

F^{\tau\tau\tau}_{\tau}=\begin{pmatrix}\varphi^{-1}&\varphi^{-1/2}\\[2pt] \varphi^{-1/2}&-\varphi^{-1}\end{pmatrix},\qquad \big(F^{\tau\tau\tau}_{\tau}\big)^2=\mathbb 1 \qquad (4)

where: $F^{\tau\tau\tau}_{\tau}$ — change of basis between the left-associated tree $((\tau\tau)\tau)$ and the right-associated $(\tau(\tau\tau))$ on the space of three $\tau$ 's with total charge $\tau$ ; $\varphi^{-1}\approx0.618$ , $\varphi^{-1/2}\approx0.786$ . Reading it: real, symmetric, and squaring to the identity, it is its own inverse — the fact used to build the second braid generator below. Braiding — exchanging two anyons — is diagonal in the fusion channel, one phase per channel; these are the $R$ -matrices, differing decisively between the two theories.

R^{\tau\tau}=\mathrm{diag}\!\left(e^{-4\pi i/5},\,e^{3\pi i/5}\right),\qquad R^{\sigma\sigma}=\mathrm{diag}\!\left(e^{-i\pi/8},\,e^{i3\pi/8}\right) \qquad (5)

where: the first entry is the phase for the pair fusing to vacuum, the second for the nontrivial channel ( $\tau$ for Fibonacci, $\psi$ for Ising). Reading it: the ratio of the Ising phases is $e^{i3\pi/8}/e^{-i\pi/8}=e^{i\pi/2}=i$ , an $S$ -gate relative phase, whereas the Fibonacci ratio $e^{i7\pi/5}$ never closes into a finite group. Finally, how big is the protected space? For Fibonacci it follows the Fibonacci numbers; for Ising it doubles every two anyons — one encoded qubit per added pair.

\dim\mathcal H_n^{\mathrm{Fib}}=F_n\sim\frac{\varphi^{\,n}}{\sqrt5},\qquad \dim\mathcal H_{2n}^{\mathrm{Ising}}=2^{\,n-1} \qquad (6)

where: $\mathcal H_n$ — space of $n$ $\tau$ -anyons at fixed total charge; $F_n$ — the $n$ -th Fibonacci number; for Ising, $2n$ $\sigma$ 's encode $n-1$ qubits, so four $\sigma$ 's give one qubit. Reading it: the Fibonacci dimension is not a power of two — the fingerprint of a genuinely non-Abelian, universal medium.

Derivation AThe Fibonacci F-matrix from the pentagon equation

The goal is to show that the $2\times2$ Fibonacci $F$ -matrix of equation (4) is not an assumption but a consequence: once you demand that the two ways of reassociating a chain of $\tau$ anyons agree, the pentagon equation together with unitarity forces every entry, and the golden ratio drops out as the only consistent value. We work on the two-dimensional space of three $\tau$ 's of total charge $\tau$ and impose the constraints until the matrix is fixed up to a gauge choice.

Step 1. Label the two basis states of three $\tau$ 's (total charge $\tau$ ) by the intermediate charge $a\in\{1,\tau\}$ of the first pair. The $F$ -move rewrites the same state in the other pairing, a linear map on this two-dimensional space; unitarity forces $F$ unitary, and a standard gauge makes it real and symmetric. Why: only the intermediate charge is physical, so recoupling is a rotation among these two labels.

Step 2. Impose the pentagon equation — the two distinct sequences of $F$ -moves relating the five pairings of four $\tau$ 's must give the same unitary. For Fibonacci this collapses to scalar relations in the entries, the key one being $F_{11}(1-F_{11})=F_{1\tau}F_{\tau1}$ . Why: the pentagon is the associativity axiom, the algebraic content of "reassociation is unambiguous".

Step 3. Combine with unitarity $F^2=\mathbb 1$ , giving $F_{11}^2+F_{1\tau}^2=1$ and $F_{11}=-F_{\tau\tau}$ . With $F_{11}=x$ the pentagon relation becomes $x(1-x)=1-x^2$ , so $x=\varphi^{-1}$ (using $\varphi^{-1}+\varphi^{-2}=1$ ) and $F_{1\tau}=\sqrt{1-\varphi^{-2}}=\varphi^{-1/2}$ . Why: the golden ratio is forced — the unique solution consistent with both pentagon and unitarity, mirroring $d_\tau=\varphi$ .

Step 4. Assemble: $F^{\tau\tau\tau}_{\tau}=\begin{pmatrix}\varphi^{-1}&\varphi^{-1/2}\\ \varphi^{-1/2}&-\varphi^{-1}\end{pmatrix}$ , and squaring gives $F^2=\mathbb 1$ since the diagonal is $\varphi^{-2}+\varphi^{-1}=1$ and the off-diagonal cancels. Why: $F$ is its own inverse — exactly the property that lets us conjugate one braid generator into a second, independent one below.

Go deeper · Derivation B · 2 minWhy Fibonacci braiding is universal

Derivation BWhy Fibonacci braiding is universal

The goal is to show that the two elementary braids on three $\tau$ -anyons generate a dense subgroup of $\mathrm{PU}(2)$ , so finite braid words approximate any single-qubit gate to arbitrary accuracy, and then that a braid across two such triples supplies the entangler completing universality for $\mathsf{BQP}$ . We encode the qubit in three $\tau$ 's of total charge $\tau$ (two-dimensional since $F_3=2$ ), with basis labelled by the intermediate charge $1$ or $\tau$ .

Step 1. The first generator braids anyons 1 and 2, diagonal in this basis: $\sigma_1=R^{\tau\tau}=\mathrm{diag}(e^{-4\pi i/5},e^{3\pi i/5})$ , a $z$ -axis Bloch rotation by the phase difference $7\pi/5$ . Why: braiding the first pair acts directly on the intermediate charge, so the gate is simply the $R$ -matrix.

Step 2. The second generator braids anyons 2 and 3, so we change to the other fusion tree with $F$ , braid, and change back: $\sigma_2=F\,R^{\tau\tau}F^{-1}=F\,\sigma_1\,F$ (using $F^{-1}=F$ ). Since $F$ rotates by an angle with cosine $\varphi^{-1}$ , $\sigma_2$ is a rotation about a tilted axis. Why: recoupling tilts the axis, turning one diagonal gate into two rotations about non-parallel axes — the raw material for generating everything.

Step 3. Two rotations about distinct axes generate a dense subgroup of $\mathrm{PU}(2)$ unless the group is finite; here the angles are rational multiples of $\pi$ but the axis-separation has irrational cosine, so the group is infinite and, being non-solvable in a compact simple group, dense — the Freedman–Larsen–Wang theorem for $SU(2)_3$ . Why: density is exactly "reach any gate arbitrarily closely" — single-qubit universality by braids.

Step 4. Efficiency follows from Solovay–Kitaev: any target gate is reached to accuracy $\varepsilon$ by a braid word of length $L=O(\log^{c}(1/\varepsilon))$ , $c\approx3.97$ (and $c\to1$ in improved variants). Why: density guarantees existence; Solovay–Kitaev upgrades it to efficient compilation, overhead only polylogarithmic.

Step 5. Complete universality: braiding an anyon of one triple around one of another produces a two-qubit controlled-rotation. Dense single-qubit gates plus one entangler is universal, so Fibonacci braiding computes any polynomial-time quantum algorithm. Why: arbitrary local unitaries plus one entangler generate all of $\mathrm{U}(2^n)$ densely.

Go deeper · Derivation C · 1 minWhy Ising braiding gives only Clifford gates

Derivation CWhy Ising braiding gives only Clifford gates

The goal is to compute the gates $\sigma$ -braids produce, identify the generated group as the Clifford group, and invoke Gottesman–Knill to conclude that Ising braiding — despite being genuinely non-Abelian — cannot be universal, then show which ingredient repairs it. We encode one qubit in four $\sigma$ anyons of total charge $1$ , basis the fusion channel ( $1$ or $\psi$ ) of the first pair — two-dimensional, since $\dim\mathcal H_4^{\mathrm{Ising}}=2^{2-1}=2$ .

Step 1. Braiding the two anyons inside a pair multiplies the channels by $R^{\sigma\sigma}_1=e^{-i\pi/8}$ and $R^{\sigma\sigma}_\psi=e^{i3\pi/8}$ ; the ratio is $e^{i\pi/2}=i$ , so up to global phase the gate is $\mathrm{diag}(1,i)=S$ . Why: only the relative phase is physical, and here it is exactly the $\pi/2$ of the $\sqrt Z=S$ gate.

Step 2. Braiding across the two pairs is the $F$ -conjugated exchange $\sigma_2=F^{\sigma}R^{\sigma\sigma}(F^{\sigma})^{-1}$ , and with $F^\sigma=\tfrac1{\sqrt2}\begin{pmatrix}1&1\\1&-1\end{pmatrix}=H$ this is (up to phase) the Hadamard $H$ . Why: the Ising recoupling matrix is literally Hadamard, so the tilted braid is the second Clifford generator.

Step 3. The group generated by $\{S,H\}$ is the 24-element single-qubit Clifford group $\mathrm{C}_1$ — the octahedral rotation group of the Bloch sphere, finite and not dense, so single-qubit universality already fails. Why: $S$ and $H$ have orders 4 and 2 and close after 24 elements; no braid escapes the octahedron.

Step 4. Braiding between two four- $\sigma$ qubits gives an entangler equivalent to $\mathrm{CNOT}$ up to Clifford, so the braid group maps onto $\langle S,H,\mathrm{CNOT}\rangle$ — the full Clifford group. By Gottesman–Knill, Clifford circuits on stabilizer inputs are classically simulable in polynomial time. Why: a classically simulable gate set cannot be universal, so Ising braiding alone is provably weak.

Step 5. Restore universality by injecting a non-Clifford resource: the magic state $|A\rangle=(|0\rangle+e^{i\pi/4}|1\rangle)/\sqrt2$ gate-teleports $T=\mathrm{diag}(1,e^{i\pi/4})$ using only Clifford operations and measurement, and distilling noisy copies into clean ones promotes the protected Clifford set to a universal one. Why: $\{\text{Clifford},T\}$ is universal, so the one missing $\pi/8$ rotation — paid for non-topologically — closes the gap.

Limiting cases

A handful of invariants cleanly separate the two theories, each computable directly from the fusion rules and braid data.

Case	Result	Meaning
Fibonacci quantum dimension	$d_\tau=\varphi=\tfrac{1+\sqrt5}{2}\approx1.618$	solves $d^2=1+d$ , the fusion rule $\tau\times\tau=1+\tau$ ; degeneracy grows as $\varphi^{\,n}$
Ising quantum dimensions	$d_\sigma=\sqrt2,\ d_\psi=1$	$d_\sigma^2=1+1$ from $\sigma\times\sigma=1+\psi$ ; $2n$ $\sigma$ 's span a $2^{\,n-1}$ -dimensional space
Total quantum dimension	$\mathcal D=\sqrt{\textstyle\sum_a d_a^2}$	Fibonacci $\mathcal D=\sqrt{1+\varphi^2}\approx1.902$ ; Ising $\mathcal D=\sqrt{1+2+1}=2$
Topological spins	$\theta_\tau=e^{4\pi i/5},\ \theta_\sigma=e^{i\pi/8},\ \theta_\psi=-1$	$\theta_\psi=-1$ marks $\psi$ as a fermion; Ising carries chiral central charge $c=1/2$
Braid-group image	$\overline{\rho(B_n)}=\mathrm{PU}(2)$ (Fib); $\lvert\rho(B_n)\rvert<\infty$ (Ising)	dense ⇒ universal (Solovay–Kitaev); finite ⇒ Clifford, simulable (Gottesman–Knill)
$SU(2)_k$ placement	$k=3$ (Fib), $k=2$ (Ising)	universal for all $k\ge3$ except $k=4$ ; the two simplest points bracket the family

Worked examples

1 · The topological qubit and its exponential growth (confidence builder)

How many qubits can $n$ Fibonacci $\tau$ -anyons store, and at what rate does that capacity grow per anyon? Take the total charge fixed to $\tau$ , so the protected dimension is $F_n$ .

Step 1. Smallest case: for $n=3$ , equation (6) gives $\dim\mathcal H_3=F_3=2$ — one qubit, its basis labelled by whether the first two $\tau$ 's fused to $1$ or $\tau$ . This is the elementary Fibonacci qubit of every braiding-gate construction. Three τ-anyons encode exactly one qubit.

Step 2. Scale up: for $n=20$ , $\dim\mathcal H_{20}=F_{20}=6765$ , so the qubit count is $\log_2 6765=\mathbf{12.7}$ — twenty anyons hold more than twelve qubits, and the count is not a power of two, the signature of a non-Abelian medium.

Step 3. Growth rate: by Binet's formula $F_n\sim\varphi^{\,n}/\sqrt5$ , successive dimensions ratio to $\varphi$ , so each anyon multiplies the space by $\varphi\approx1.618$ — a capacity of $\log_2\varphi=\mathbf{0.694}$ qubits per anyon. Check: $\varphi^{8}/\sqrt5=21.01$ versus $F_8=21$ , agreeing to $0.05\%$ . The space grows by φ ≈ 1.618 — about 0.694 qubits — per anyon.

2 · An Ising σ-braid is the S gate

Encode a qubit in the fusion channel of a $\sigma$ pair and compute the single-qubit gate produced by one elementary exchange. Then say precisely what is still missing for universality.

Step 1. The two channels acquire the braid phases of equation (5): the vacuum channel gets $R^{\sigma\sigma}_1=e^{-i\pi/8}$ and the $\psi$ channel gets $R^{\sigma\sigma}_\psi=e^{i3\pi/8}$ . Only the relative phase is physical, so factor out the global $e^{-i\pi/8}$ . The braid is $e^{-i\pi/8}\,\mathrm{diag}(1,\,e^{i\pi/2})$ .

Step 2. Evaluate the relative phase: $e^{i3\pi/8}/e^{-i\pi/8}=e^{i(3\pi/8+\pi/8)}=e^{i\pi/2}=i$ . The gate is therefore $\mathrm{diag}(1,i)=S$ up to global phase — the Clifford phase gate, satisfying $S^2=Z$ . One in-pair σ-braid realizes the S gate exactly.

Step 3. The inter-pair braid ( $F^\sigma=H$ ) gives Hadamard, so $\{S,H\}$ generates the 24-element Clifford group and an inter-qubit braid adds a $\mathrm{CNOT}$ -equivalent entangler. No braid ever produces $T=\mathrm{diag}(1,e^{i\pi/4})$ , the $\pi/8$ rotation — the square root of $S$ that braids cannot reach. The missing gate is $T=\mathrm{diag}(1,e^{iπ/4})$ , supplied by magic-state distillation.

Physics of the portrait

The claim at the centre of this article — that some anyons compute anything and others almost nothing — is a theorem, due to Michael Freedman with Michael Larsen and Zhenghan Wang.

Origins · 1 minMichael Freedman and how this was discovered

Michael Freedman

b. 1951, Fields Medal 1986 · topological quantum computation

Michael Freedman won the Fields Medal in 1986 for proving the four-dimensional Poincaré conjecture, using geometric topology far from physics. In the late 1990s he made an audacious bet: that the fault tolerance a quantum computer needs could come not from active error correction but from the topology of anyon worldlines — an idea developed alongside Alexei Kitaev's 1997 proposal of fault-tolerant computation by anyons. Its technical heart is the theorem he proved in 2000–2002 with Larsen and Wang: the braid-group representations of the $SU(2)_k$ Chern–Simons theories are dense in the projective unitary groups for every level except $k=1,2,4$ — exactly why Fibonacci ( $k=3$ ) is universal while Ising ( $k=2$ ) is not. The two models here are the boundary stones of that theorem: Ising, the largest non-universal example, tied by Moore and Read to the $\nu=5/2$ state and by Kitaev to Majorana modes; and Fibonacci, the smallest universal one, still hunted in $\nu=12/5$ . A Fields-medal topologist's density theorem now setting the design target for quantum hardware is one of the cleaner cases of pure mathematics dictating engineering.

History

1977

— Leinaas and Myrheim show two space dimensions permit statistics beyond bosons and fermions

1982

— Wilczek coins the name anyon for these fractional-statistics quasiparticles

1991

— Moore and Read tie non-Abelian anyons to the ν = 5/2 fractional quantum Hall state

1997

— Kitaev proposes fault-tolerant quantum computation by braiding anyons

2002

— Freedman, Larsen and Wang prove Fibonacci-type braiding is universal for quantum computation

Going further

Texts. Preskill, Lecture Notes on Quantum Computation, ch. 9 · Nielsen & Chuang, Quantum Computation and Quantum Information, ch. 10 · Pachos, Introduction to Topological Quantum Computation (Cambridge, 2012) · Wang, Topological Quantum Computation (AMS, 2010) · Kitaev, Anyons in an exactly solved model and beyond, Ann. Phys. 321 (2006) 2.

Provenance. A verifier should confirm: (1) $\varphi=(1+\sqrt5)/2$ satisfies $\varphi^2=\varphi+1$ and the Fibonacci $F$ -matrix squares to the identity via $\varphi^{-1}+\varphi^{-2}=1$ ; (2) $d_\sigma=\sqrt2$ and $\mathcal D_{\mathrm{Fib}}=\sqrt{1+\varphi^2}\approx1.902$ , $\mathcal D_{\mathrm{Ising}}=2$ ; (3) the Ising ratio $R^{\sigma\sigma}_\psi/R^{\sigma\sigma}_1=e^{i\pi/2}=i$ giving an $S$ gate, with $\{S,H,\mathrm{CNOT}\}$ the Clifford group (Gottesman–Knill); (4) the FLW density result for $SU(2)_3$ and Solovay–Kitaev scaling $L=O(\log^c(1/\varepsilon))$ . Spins $\theta_\tau=e^{4\pi i/5}$ , $\theta_\sigma=e^{i\pi/8}$ , $\theta_\psi=-1$ and $F_3=2$ , $F_{20}=6765$ are standard. Cross-check against Preskill ch. 9, Nayak et al. Rev. Mod. Phys. 80 (2008) 1083, Kitaev Ann. Phys. 321 (2006), and Freedman–Larsen–Wang, Commun. Math. Phys. 227 (2002) 605.

Prerequisites. Anyons and Fractional Statistics · Non-Abelian Braiding and Fusion Rules · Topological Quantum Computation. Leads to. Magic State Distillation · Majorana Zero Modes · The Cost of Magic and the Resource Theory of Magic.

Ask the tutor: Ask me to braid three Fibonacci anyons into a gate, or trace a Majorana pair through the Ising $\sigma$ -braid to watch a Clifford gate appear.

What does "quantum dimension" actually count?

The quantum dimension $d_a$ is the rate at which the protected space grows per added anyon of type $a$ : stack $n$ of them and it grows like $d_a^{\,n}$ . For Fibonacci $d_\tau=\varphi\approx1.618$ , so each $\tau$ multiplies the space by the golden ratio; for Ising $d_\sigma=\sqrt2$ , so two $\sigma$ 's double it. Formally $d_a$ is the largest eigenvalue of the fusion matrix, which is why $\tau\times\tau=1+\tau$ becomes $d_\tau^2=d_\tau+1$ . A non-integer $d_a$ is the unmistakable mark of a non-Abelian theory — the many-anyon space is not a tensor product of local qubits.

If Ising anyons only give Clifford gates, why build Ising hardware?

Because those Clifford gates are essentially free and superbly protected. Braiding Majorana-based Ising anyons realizes Clifford operations whose action depends only on the topology of the worldlines, so local noise that does not move anyons around each other cannot corrupt them — exponential error suppression that ordinary qubits reach only through costly active correction. The one missing gate, the non-Clifford $T$ , is added by magic-state distillation. The bet is that a little expensive distillation on a lot of cheap, protected braiding beats error-correcting conventional qubits.

Why does the golden ratio appear in the Fibonacci model?

It is forced by the fusion rule $\tau\times\tau=1+\tau$ . Replace each charge by its quantum dimension and the rule becomes $d_\tau^2=1+d_\tau$ , whose positive root is $\varphi=(1+\sqrt5)/2$ . The same relation sizes the fusion space: fusion histories obey the recurrence $F_n=F_{n-1}+F_{n-2}$ , whose successive-term ratio converges to $\varphi$ , and the $F$ -matrix entries are powers of $\varphi$ because the pentagon equation inherits the same quadratic. The golden ratio is the one number consistent with a charge that self-fuses to either nothing or itself.

Is a single braid a quantum gate I can just apply?

A single braid is a fixed unitary — a specific rotation of the fusion space — not an adjustable gate. For Ising that fixed unitary is a Clifford gate ( $S$ , $H$ , or an entangler), directly useful. For Fibonacci a single braid is some irrational rotation of the Bloch sphere; only the closure of all braid words is dense in $\mathrm{PU}(2)$ . To realize a target gate you compile it into a braid sequence with the Solovay–Kitaev algorithm, of length polylogarithmic in the accuracy demanded. "Braiding is universal" is about long braid sequences, not one exchange equalling an arbitrary gate.

How are Ising anyons related to Majorana zero modes?

They are two descriptions of the same physics. A Majorana zero mode is a self-conjugate zero-energy bound state — at the end of a topological superconducting wire, or in a $p+ip$ vortex core — and a pair of them shares one nonlocal fermionic state that can be empty or filled. That two-way choice is exactly the $\sigma\times\sigma=1+\psi$ channel of the Ising $\sigma$ anyon, and exchanging two vortices braids the Majoranas to reproduce the Ising $R$ -matrix and its Clifford gate. This is why "Majorana-based" and "Ising anyon" hardware mean the same thing, and why the $\nu=5/2$ state (Moore–Read Pfaffian) hosts Ising anyons.

Do the two theories have the same total quantum dimension by coincidence?

Their total quantum dimensions are close — $\mathcal D_{\mathrm{Fib}}\approx1.902$ versus $\mathcal D_{\mathrm{Ising}}=2$ — but this does not govern computational power, and treating it as such is a trap. The total quantum dimension $\mathcal D=\sqrt{\sum_a d_a^2}$ measures how much protected information the theory stores and fixes the topological entanglement entropy $\gamma=\ln\mathcal D$ (0.643 versus 0.693). What decides universality is instead the braid-group image: dense for Fibonacci, finite for Ising — capacity and capability are separate axes.