In short

Fibonacci anyons are a model of non-abelian anyons with exactly two particle types — the vacuum \mathbf{1} and a non-trivial charge \tau — and one non-trivial fusion rule: \tau \times \tau = \mathbf{1} + \tau. That single rule encodes the golden ratio \phi = (1+\sqrt{5})/2: the quantum dimension of \tau is d_\tau = \phi, and the number of fusion channels for n copies of \tau grows as the Fibonacci sequence 1, 1, 2, 3, 5, 8, 13, \ldots — which is how the model gets its name. The payoff is spectacular. Freedman, Larsen, and Wang (2002) proved that the unitaries you get from braiding Fibonacci anyons form a dense subgroup of SU(2) on the two-dimensional qubit subspace (and of the corresponding special-unitary groups on larger subspaces). By the Solovay–Kitaev theorem, a dense subgroup can approximate any unitary to any precision using only O(\log^c(1/\epsilon)) braid generators. In plain language: braiding alone is universal for Fibonacci anyons. No magic-state injection, no external T gate, no Clifford-plus-magic hybrid — the full set of quantum gates falls out of the topology for free. This is the maximum dream of topological quantum computing. Compare with the Ising anyons / Majorana zero modes you met in chapter 185: braids there generate only the finite Clifford group, and magic states must be supplied from outside. Fibonacci anyons are stronger. They are also harder to find. The leading physical candidate is the \nu = 12/5 state of the fractional quantum Hall effect — the second plateau in the second Landau level — predicted by Read and Rezayi (1999) to host a k=3 parafermion state whose quasiparticles are Fibonacci anyons. Weak signatures at \nu = 12/5 have been measured since 2008 (Xia et al.) but the non-abelian nature has never been confirmed. No other laboratory system has been established as a Fibonacci host either. This is, in 2026, the single biggest prize in topological quantum computing: a platform where braid-universality is a theorem, a physical candidate where the experiments are at the edge of feasibility, and a wait of unknown length for the decisive result.

You have just seen, in chapter 185, a topological quantum computer that almost works. Majorana zero modes in a topological superconductor host non-abelian anyons. Braiding those anyons applies unitaries to the encoded qubit. The unitaries are topologically protected — they do not care about the speed or wobble of the exchange, only about which strand crossed which. The gates are, in a meaningful sense, exact: no calibration, no drift, no pulse shaping. It is the cleanest picture of a quantum computer anyone has drawn.

The catch was one sentence, and it was load-bearing. The braid group of Majorana anyons generates only the Clifford subgroup of quantum gates — the group of Hadamards, phase gates S, and CNOTs. And the Clifford subgroup, by the Gottesman–Knill theorem, is not universal; it is classically simulable. For a speedup you need at least one non-Clifford gate — the T gate — and the T gate has to be injected from outside the topological platform via magic-state distillation. The Majorana computer is 90% of a fault-tolerant quantum computer. The last 10% is conventional magic-state machinery, with all of its overhead.

Now ask the obvious question. Is there a different kind of non-abelian anyon — one whose braid group is bigger, so big that it generates not just the Clifford gates but every gate? An anyon whose braiding alone is universal for quantum computation, with no external magic states, no distillation, no concession to the classical world?

The answer is yes. The anyon is called the Fibonacci anyon, after the number sequence that appears in its fusion rules. The universality theorem was proved in 2002 by Michael Freedman, Michael Larsen, and Zhenghan Wang. And if a laboratory system were ever confirmed to host Fibonacci anyons, topological quantum computing would reach its theoretical ceiling: a platform where every single gate is a topologically protected braid, and the hardware itself is the error correction.

The catch, this time, is that no such system has been cleanly observed. The Fibonacci model is the dream. The Majorana model is what Microsoft actually fabricates. This chapter explains the gap.

Step 1 — fusion rules and why the golden ratio appears

Every non-abelian anyon model is defined by two pieces of data: a list of particle types, and a fusion rule that tells you what you get when two particles are brought together. Think of fusion as a kind of multiplication — if you fuse particle a with particle b, the result is a sum over possible product particles c, weighted by integers N_{ab}^c that count how many distinct ways the fusion can go.

For Fibonacci anyons, the list of types is short. There are only two:

There is exactly one non-trivial fusion rule:

\tau \times \tau \;=\; \mathbf{1} \;+\; \tau.

Read this as: when two \tau particles are fused together, the result is either the vacuum \mathbf{1} or another \tau — two possible outcomes, each occurring once. The plus sign is not an arithmetic sum; it is a direct sum of fusion channels. The meaning is that the two-\tau system has a two-dimensional fusion space: one basis state where the pair fuses to \mathbf{1}, one basis state where it fuses to \tau. This two-dimensional space is where the qubit will live.

That single fusion rule is the entire model. Everything else — the golden ratio, the Fibonacci numbers, the universality theorem, the \nu = 12/5 candidate — flows from it.

The golden ratio falls out of d_\tau^2 = 1 + d_\tau

Each particle type has a quantum dimension d_a — a positive real number (not necessarily an integer) that measures how fast the fusion space grows when you add more copies of that particle. For abelian anyons, all quantum dimensions are 1, and the fusion space stays one-dimensional no matter how many anyons you add. For non-abelian anyons, at least one d_a > 1, and the fusion space grows exponentially.

Quantum dimensions obey the fusion rules as a simple algebraic identity: if a \times b = \sum_c N_{ab}^c \, c, then d_a \, d_b = \sum_c N_{ab}^c \, d_c. Applied to the Fibonacci rule \tau \times \tau = \mathbf{1} + \tau:

d_\tau^2 \;=\; d_{\mathbf{1}} \;+\; d_\tau \;=\; 1 \;+\; d_\tau.

Why: d_{\mathbf{1}} = 1 always — the vacuum has quantum dimension 1 by definition. The rest follows by substituting the fusion rule into the multiplicativity identity.

Solve the quadratic d_\tau^2 - d_\tau - 1 = 0, keeping the positive root:

d_\tau \;=\; \frac{1 + \sqrt{5}}{2} \;=\; \phi \;\approx\; 1.618.

That is the golden ratio. It is the same \phi that appears in pentagon geometry, Bhaskara II's calculations on continued fractions, Fibonacci's rabbit problem, and a thousand other places in mathematics where two things fuse into one-plus-itself. It appears here for the same reason it appears everywhere else: a self-referential relation of the form next = previous + current uniquely determines \phi once you demand a positive real solution.

The quantum dimension of \tau being irrational has a direct consequence: the fusion space dimensions for n copies of \tau do not grow as any neat power of 2 or 3. They grow as the Fibonacci numbers.

Fibonacci numbers are the dimensions

Let F_n be the dimension of the Hilbert space of n \tau-anyons that fuse to the vacuum overall. Then F_n obeys the recursion

F_n \;=\; F_{n-1} \;+\; F_{n-2}, \qquad F_0 = 1,\; F_1 = 0,\; F_2 = 1.

Why: pick any fusion tree. The last two \tau anyons fuse to either \mathbf{1} (and the remaining n-2 must also fuse to \mathbf{1} overall — F_{n-2} ways) or to \tau (and the remaining n-2 together with this intermediate \tau must fuse to \mathbf{1} — F_{n-1} ways, because this is the same as asking n-1 \tau's to fuse to \mathbf{1}).

The sequence F_0, F_1, F_2, \ldots starts 1, 0, 1, 1, 2, 3, 5, 8, 13, 21, \ldots — shift by one, and it is the classical Fibonacci sequence familiar to Indian readers from Pingala's binary-pattern work (~300 BCE), which listed the same numbers in a prosody context centuries before they were rediscovered in Pisa. The Hilbert-space dimension grows as F_n \sim \phi^n / \sqrt{5} — exponential in the number of anyons, with base \phi.

This is already the heart of the computational power. With n Fibonacci anyons, you get roughly \phi^n orthogonal states to compute with — far more than you would get from n independent qubits, which gives only 2^n. But the real payoff is what you can do with those states by braiding them.

Fibonacci fusion rule and fusion-space dimensionsTwo panels. Left panel shows two tau anyons fusing and branching into two outcomes: the vacuum or another tau. Right panel shows a table of Fibonacci numbers for the dimension of the fusion Hilbert space with n tau anyons.τ × τ = 1 + τ → golden-ratio dimensions, Fibonacci growthfusion of two τ anyonsττ1τchannel 1channel τtwo-dimensional fusion spacedim(n τ → 1) = Fn ≈ φn / √5n23456Fn11235n7891011Fn813213455Pingala (~300 BCE) listed these first
Left: the Fibonacci fusion rule $\tau \times \tau = \mathbf{1} + \tau$ read as a branching — two $\tau$ particles can fuse into either the vacuum $\mathbf{1}$ or another $\tau$, giving a two-dimensional fusion space. Right: the dimensions of the "$n$ $\tau$'s fuse to the vacuum" Hilbert spaces are exactly the Fibonacci numbers. The growth rate is $\phi^n$ where $\phi$ is the golden ratio — the quantum dimension of the $\tau$ anyon.

Step 2 — a qubit in three Fibonacci anyons

To build a qubit, you need a two-dimensional Hilbert space that braiding can act on. The smallest Fibonacci configuration that gives exactly that is three \tau anyons fused to total charge \tau.

Here is why. Fuse two of the three \tau's first. The result is either \mathbf{1} or \tau. Now fuse that intermediate with the third \tau. The total must equal \tau by assumption:

That is two basis states, and they are distinguished by what the first two \tau's fused to. Call them

|0\rangle_L \;=\; |\,(\tau\tau)_\mathbf{1}\, \tau\rangle, \qquad |1\rangle_L \;=\; |\,(\tau\tau)_\tau\, \tau\rangle.

Why this labelling is a qubit: the two states are orthogonal — they lie in different fusion channels of the first pair, which are physically distinguishable by a measurement that asks "what is the fusion outcome of anyons 1 and 2?". They span a two-dimensional subspace of the total three-\tau Hilbert space (of total dimension F_4 + F_3 = 3 when the total charge is allowed to vary; restricted to total charge \tau, the dimension is exactly 2).

The qubit sits in the non-local correlation between the three anyons. No local measurement of any single \tau can tell you whether the qubit is in state |0\rangle_L or |1\rangle_L — only a joint measurement of fusion outcomes can.

What braids do to this qubit

Now move the three anyons around each other. Each elementary braid \sigma_i — exchanging anyons i and i+1 — acts as a 2 \times 2 unitary matrix on the \{|0\rangle_L, |1\rangle_L\} subspace.

For Fibonacci anyons, these matrices are explicit. In the basis above (with a convention choice for the F-symbols, see the Going Deeper section for the formal construction), the generators are:

\sigma_1 \;=\; \begin{pmatrix} e^{-4\pi i/5} & 0 \\ 0 & e^{3\pi i/5} \end{pmatrix}, \qquad \sigma_2 \;=\; \begin{pmatrix} -\phi^{-1} e^{-\pi i/5} & -\phi^{-1/2} e^{-3\pi i/5} \\ -\phi^{-1/2} e^{3\pi i/5} & -\phi^{-1} \end{pmatrix}.

These matrices look exotic because they are: the entries are complex numbers whose phases are multiples of \pi/5 (a signature of the underlying pentagon symmetry — recall that \phi is the diagonal-to-side ratio of a regular pentagon), and the real parts involve \phi^{-1} and \phi^{-1/2}.

What matters for quantum computing is not the explicit form but a structural property of the group they generate.

Step 3 — the Freedman–Larsen–Wang theorem: braiding is universal

Take the two matrices \sigma_1, \sigma_2 above. Multiply them together in all possible orders. Multiply products by more products. You generate an infinite family of 2 \times 2 unitary matrices. What group do they form?

Three options, in decreasing order of strength:

  1. They generate a discrete finite group (e.g. the Clifford group).
  2. They generate a discrete infinite group (e.g. a countable dense subgroup of some proper subgroup of SU(2)).
  3. They generate a dense subgroup of SU(2) — that is, the products get within distance \epsilon of every SU(2) element for every \epsilon > 0.

For Majorana / Ising anyons, the answer was option (1): a finite group (the single-qubit Clifford subgroup, eight elements up to phase). That is why Majorana braids are not universal.

For Fibonacci anyons, the answer is option (3), and it is a theorem:

Freedman–Larsen–Wang (2002). The image of the braid group B_n on the Fibonacci fusion Hilbert space is a dense subgroup of SU(F_n) on the relevant total-charge subspaces, for every n \geq 3. In particular, on the three-anyon qubit subspace, the image is dense in SU(2).

Combine this with the Solovay–Kitaev theorem, which states that any dense subgroup of SU(2) can approximate any target unitary U to precision \epsilon using at most O(\log^c(1/\epsilon)) generator words (with c \approx 3.97). The conclusion is explicit and strong:

To apply any single-qubit gate U to precision \epsilon, you need only a braid of length O(\log^4(1/\epsilon)) on three Fibonacci anyons. To apply any two-qubit gate, you need a braid of comparable length on six Fibonacci anyons.

No magic states. No T-gate distillation. No Clifford-plus-magic hybrid. Every gate — Hadamard, phase, T, CNOT, Toffoli, whatever you need — is a specific topologically protected braid, and the braid can be written down once and executed many times with exponentially small error.

The word "dense" is doing a lot of work

Dense-in-SU(2) is strictly stronger than "generates infinitely many matrices." A discrete infinite subgroup (like the integer rotations by any irrational angle) is infinite but not dense. You could rotate by \pi/\sqrt 2 as many times as you like and miss many points of the circle entirely.

Density means every point is reachable to within any chosen tolerance. In concrete terms: for any target angle \theta and any error budget \epsilon, there exists a word \sigma_{i_1}\sigma_{i_2}\ldots\sigma_{i_k} in the Fibonacci generators whose SU(2) image lands within \epsilon of R_z(\theta) (or any other target). The length k grows only polylogarithmically in 1/\epsilon, so hitting \epsilon = 10^{-15} needs a braid of roughly 60^4 \sim 10^7 generators — long but finite and explicitly constructible. [The Solovay–Kitaev algorithm produces the sequence.]

Why dense, physically

The density comes from an arithmetic accident in the \sigma matrices. Their phases are multiples of \pi/5, which is irrational with respect to \pi/2 and \pi/4 and every other angle that gives rise to finite groups. The \pi/5 (pentagonal) structure is what makes the Fibonacci braid group infinite and dense, while the \pi/4 (square) structure of Majorana anyons makes their braid group finite and Clifford-limited.

Behind the arithmetic lies representation theory: SU(2) has a one-parameter family of finite subgroups (cyclic, dihedral, tetrahedral, octahedral, icosahedral) that exhaust its closed subgroups of finite order. The Fibonacci braid group maps into SU(2) in a way that avoids all of these finite subgroups — and the only closed groups left to land in are SU(2) itself or dense subgroups of it. Freedman–Larsen–Wang show the latter holds.

Majorana braids versus Fibonacci braids on the Bloch sphereTwo Bloch sphere cartoons side by side. Left sphere is labelled Majorana-Ising and has six accent dots at the cardinal points and a large red X over the T-gate location. Right sphere is labelled Fibonacci and is peppered with many small accent dots covering the surface densely, including one near the T-gate location.Reachable set of single-qubit gates, by braid groupMajorana / Ising — finite Clifford× T unreachableonly 6 points + Clifford closureFibonacci — dense in SU(2)T reachablebraids get within ε of any point
Contrast of the gate sets. Left: Majorana braids (the Ising representation of the braid group) hit only a finite set of Bloch-sphere points — the six cardinals plus short Clifford compositions. The $T$ gate, at a $\pi/4$ rotation angle, is unreachable by any braid. Right: Fibonacci braids are dense — for every target unitary and every error budget $\epsilon$, some braid word lands within $\epsilon$. A $T$ gate falls out of a word of length $\sim 10$ in the generators; precision $10^{-15}$ needs length $\sim 10^7$. The topological platform is universal on its own.

Step 4 — two worked examples

Example 1: Three τ anyons have exactly three fusion states, and the qubit uses two of them

Count the states of three \tau-anyons directly, to see where the "two-dimensional qubit" comes from.

Step 1. Write all legal fusion trees. Fuse anyons 1 and 2 first, then fuse the result with anyon 3. The first fusion gives \tau \times \tau = \mathbf{1} + \tau — two outcomes. For each outcome, fuse with the third \tau:

  • First fusion = \mathbf{1}: then \mathbf{1} \times \tau = \tau. Total charge \tau. One state, which we call |(\mathbf{1}), \tau\rangle.
  • First fusion = \tau: then \tau \times \tau = \mathbf{1} + \tau. Total charge is either \mathbf{1} or \tau. Two states, |(\tau), \mathbf{1}\rangle and |(\tau), \tau\rangle.

Step 2. Classify by total charge.

  • Total charge \mathbf{1}: one state, |(\tau), \mathbf{1}\rangle. This is F_3 = 1.
  • Total charge \tau: two states, |(\mathbf{1}), \tau\rangle and |(\tau), \tau\rangle. This is a 2-dimensional subspace — the qubit.

Why the F_3 = 1 and "two states of total charge \tau" match the Fibonacci counting: the recursion says F_3 = F_2 + F_1 = 1 + 0 = 1 for total-charge-\mathbf{1}, and by the same argument the dimension of the total-charge-\tau subspace is F_4 - F_3... actually by direct enumeration above, two. The counting works.

Step 3. Identify the qubit. In the total-charge-\tau subspace, define

|0\rangle_L = |(\mathbf{1}), \tau\rangle, \qquad |1\rangle_L = |(\tau), \tau\rangle.

These are orthogonal — a measurement of "what do anyons 1 and 2 fuse to?" distinguishes them perfectly.

Step 4. Interpret the encoding. The logical information is stored in a global fusion channel — the fusion outcome of a pair of anyons — not in any single anyon. No local disturbance of one anyon can reveal or change the qubit value. This is the topological protection, baked into the encoding.

Result. Three Fibonacci anyons with total charge \tau encode one qubit in their fusion channel. The logical basis states correspond to the two ways the first pair of anyons can fuse. What this shows: the minimal Fibonacci quantum computer uses three anyons per qubit, not two — because with two anyons the fusion space has a fixed total charge and gives only one fusion channel (trivial for computation). The third anyon provides the non-trivial branching.

Three tau anyons encoding one qubitA fusion tree diagram showing three tau anyons on top, fusing pairwise into the two basis states of the logical qubit. The two fusion trees are drawn side by side with labels indicating intermediate and total charges.Three τ anyons with total charge τ → qubit|0⟩_L = |(τ τ)_1, τ⟩τττ1τintermediate 1 → total τ|1⟩_L = |(τ τ)_τ, τ⟩τττττintermediate τ → total τ
Two fusion trees for three Fibonacci anyons constrained to total charge $\tau$. Left: the first two anyons fuse to the vacuum $\mathbf{1}$, forcing the final fusion with the third anyon to be $\tau$ — this is $|0\rangle_L$. Right: the first two anyons fuse to $\tau$, and combining with the third $\tau$ the final fusion picks out the $\tau$ outcome (out of the two possibilities $\mathbf{1} + \tau$) — this is $|1\rangle_L$. The two trees span the qubit subspace.

Example 2: A toy Fibonacci braid that approximates a π/10 rotation

Work out what one specific short braid does to the qubit, and see a universal-rotation building block appear.

Step 1. Use the generator matrix for \sigma_1 in the three-anyon qubit basis:

\sigma_1 \;=\; \begin{pmatrix} e^{-4\pi i/5} & 0 \\ 0 & e^{3\pi i/5} \end{pmatrix}.

This is a diagonal matrix, so in the Bloch-sphere picture it is a rotation about the z-axis. The rotation angle is the difference of the two phase angles: \theta_z = (3\pi/5) - (-4\pi/5) = 7\pi/5, or equivalently -3\pi/5 \pmod{2\pi}.

Why just the difference: a diagonal \text{diag}(e^{i\alpha}, e^{i\beta}) acts on |\psi\rangle = c_0 |0\rangle + c_1 |1\rangle by multiplying c_0 by e^{i\alpha} and c_1 by e^{i\beta}. Factoring out the global phase e^{i(\alpha+\beta)/2} leaves \text{diag}(e^{i(\alpha-\beta)/2}, e^{-i(\alpha-\beta)/2}) = R_z(\alpha - \beta). Global phases are physically irrelevant, so only \alpha - \beta matters.

Step 2. Compute what \sigma_1 does. It is R_z(-3\pi/5) = R_z(-108°) — a rotation of the Bloch sphere by 108° about the z-axis (with a sign depending on convention). This is already a non-Clifford rotation: 3\pi/5 is not a multiple of \pi/4, so this angle cannot be realised by any Majorana braid.

Step 3. Approximate R_z(\pi/10) = R_z(18°) using powers of \sigma_1.

Write 18° as a combination of 108°s modulo 360°. Seek integers k so that -108° \cdot k \equiv 18° \pmod{360°}. Try k = 3: -108 \times 3 = -324 \equiv 36°. Try k = 2: -108 \times 2 = -216 \equiv 144°. Try k = 7: -108 \times 7 = -756 \equiv -36°. Try k = 8: -864 \equiv -144°. Exact closure does not happen in few steps — the angle 3\pi/5 is incommensurate with 2\pi/(\pi/10) = 20, so no finite power of \sigma_1 alone hits \pi/10 exactly.

But by density, a mixed word \sigma_1^a \sigma_2^b \sigma_1^c \sigma_2^d \ldots gets arbitrarily close. The Solovay–Kitaev algorithm would explicitly find such a word of length O(\log^4(1/\epsilon)).

Step 4. Interpret. The single generator \sigma_1 provides a non-Clifford z-rotation; \sigma_2 provides a non-axial rotation whose composition with \sigma_1 can reach every point of the Bloch sphere. Because the rotation angles of \sigma_1, \sigma_2 are irrational multiples of \pi (involving \pi/5), products avoid landing on any finite-group orbit and instead fill SU(2) densely.

Result. A short Fibonacci braid of two generator types already gives rotations about two non-parallel axes at \pi/5-related angles. Combining these fills SU(2) densely; any target gate is approximable in poly-log braid length. What this shows: the T gate, which costs a magic state in the Majorana setup, comes free from a Fibonacci braid of modest length. The universality of Fibonacci braiding is the reason this platform is the theoretical gold standard.

Fibonacci sigma 1 as a z-rotation on the Bloch sphereA Bloch sphere showing a state starting at plus x, rotating around the z-axis by 108 degrees to a new position at an accent-coloured dot. Beside it a small circuit showing sigma-1 applied to the state with the rotation angle labelled.σ₁ on |+⟩ → R_z(−3π/5) |+⟩|0⟩|1⟩|+⟩σ₁|+⟩−108° rotation about zσ₁ = diag(e^{−4πi/5}, e^{3πi/5})= R_z(−3π/5) · (global)non-Clifford: angle = −108°no Majorana braid can produce this
The first Fibonacci braid generator $\sigma_1$ is a $z$-rotation on the Bloch sphere by $-3\pi/5 = -108°$. This alone is already a non-Clifford rotation — Majorana braiding can only produce multiples of $\pi/2$ ($90°$), so the Fibonacci braid explores angles that Majorana braids cannot reach. Combining $\sigma_1$ with the second generator $\sigma_2$ (whose rotation axis is tilted relative to $z$) produces, through the density theorem, every $SU(2)$ rotation up to any required precision.

Step 5 — the physical candidate: ν = 12/5 fractional quantum Hall

Theory is one thing. Does any physical system actually host Fibonacci anyons?

The short honest answer, as of 2026, is that no laboratory has cleanly demonstrated Fibonacci anyons. The leading candidate is a specific state in the fractional quantum Hall effect (FQHE).

Recall what FQHE is: a 2D electron gas confined at a semiconductor interface (typically GaAs/AlGaAs) and subjected to an intense perpendicular magnetic field (about 10 T) at millikelvin temperatures. Under these conditions, the electrons organise into highly correlated incompressible liquids characterised by the filling factor \nu, the ratio of electron density to magnetic-flux-quantum density. Different values of \nu give different exotic phases.

Why 12/5? The Read–Rezayi family is a sequence of trial wavefunctions labelled by an integer k — at k = 1 you get Laughlin (abelian), at k = 2 you get Moore–Read (Ising), at k = 3 you get parafermions whose non-trivial quasiparticle has the Fibonacci fusion rule. The filling factor for k = 3 in the second Landau level works out to \nu = 2 + 2/5 = 12/5 — a fraction that pops out of the theory, not one chosen by us.

What has been measured at ν = 12/5

A weak incompressible quantum Hall plateau at \nu = 12/5 was first clearly resolved by Xia and collaborators in 2008 using ultra-high-mobility GaAs samples at temperatures below 10 mK. The plateau is narrow, fragile, and disappears if the sample quality drops or the temperature rises — consistent with a gapped topological phase but with a very small gap (on the order of 50 microkelvin in energy units, compared with ~10 mK sample temperatures; the ratio is barely favourable).

Experiments to date have measured:

The community consensus, echoed in reviews by Feldman, Halperin, and others, is that the \nu = 12/5 plateau is consistent with the Read–Rezayi k = 3 state, but other candidate wavefunctions (anti-Pfaffian analogues, other competing phases) have not been ruled out. The decisive non-abelian statistics demonstration remains unachieved.

Why it is so hard

Fibonacci anyons are fragile. The energy gap separating the Read–Rezayi ground state from excited states at \nu = 12/5 is exceptionally small — the factor 5 in the denominator of the filling factor means there are many low-lying competing states. Disorder, finite temperature, and Landau-level mixing all tend to destabilise the predicted state and give rise to alternative phases that superficially look similar but have abelian rather than non-abelian quasiparticles.

Experimentally, you need:

None of this exists in any Indian laboratory today. The handful of groups making the running — Pan at Purdue, Manfra at Purdue, Xia at Rice, small groups at Weizmann and Max-Planck — work with collaborations spanning continents and years. A confirmed Fibonacci observation would be one of the most important experimental results in condensed-matter physics of the decade, and it has not yet happened.

Other candidate platforms

Beyond FQHE, several alternative routes to Fibonacci anyons have been proposed:

The engineered-simulation approach is intriguing — it decouples the question "does nature host Fibonacci anyons?" from the question "can you do Fibonacci-braid quantum computing?" — but it has not yet matured.

Ising versus Fibonacci: the full contrast

The two non-abelian anyon models that this chapter and chapter 185 have built are worth placing side by side.

Ising / Majorana Fibonacci
Particle types \mathbf{1}, \sigma, \psi \mathbf{1}, \tau
Non-trivial fusion \sigma \times \sigma = \mathbf{1} + \psi \tau \times \tau = \mathbf{1} + \tau
Quantum dimension d_\sigma = \sqrt{2} d_\tau = \phi = 1.618\ldots
Fusion-space dimension 2^{n/2} for n \sigma's F_n (Fibonacci)
Braid group image finite — Clifford subgroup dense — universal
T gate by braiding? No — needs magic state Yes — by braid of length O(\log^4(1/\epsilon))
Universal by braiding alone? No Yes
Physical candidate Majorana nanowires (Microsoft 2025) \nu = 12/5 FQHE state (Read–Rezayi)
Experimental status braiding not yet demonstrated; devices claimed plateau observed; anyons not confirmed
Effort to realise very hard harder still

Both models share one virtue — topological protection of all gates they can produce. The decisive difference is what they can produce. Ising gives you the Clifford subgroup; Fibonacci gives you everything.

From a hardware roadmap point of view, the field's implicit strategy is: try Ising first (Majorana), get the Clifford piece protected topologically, import T gates from outside; if and when Fibonacci becomes feasible, switch. Whether "when" is ten years or a century is unknown.

Common confusions

Going deeper

If you understand that Fibonacci anyons have two types and fusion rule \tau \times \tau = \mathbf{1} + \tau, that the quantum dimension is d_\tau = \phi and the fusion-space dimensions are Fibonacci numbers, that three \tau anyons encode one qubit whose basis states are distinguished by the first-pair fusion channel, that the Freedman–Larsen–Wang theorem proves braiding alone generates a dense subgroup of SU(2) making the model universal for quantum computing without magic states, and that the leading physical candidate is the Read–Rezayi k = 3 state at \nu = 12/5 whose non-abelian nature is not yet experimentally confirmed — you have chapter 186. What follows sketches the F and R matrices that define the model, states the Solovay–Kitaev theorem precisely, and maps out the Read–Rezayi wavefunctions.

The F and R matrices

The full algebraic data of a non-abelian anyon model are two sets of matrices: the F matrices (which recouple fusion trees) and the R matrices (which describe exchanges).

For Fibonacci anyons, there is a single non-trivial F matrix — the one for three \tau's fusing to \tau:

F^{\tau\tau\tau}_\tau \;=\; \begin{pmatrix} \phi^{-1} & \phi^{-1/2} \\ \phi^{-1/2} & -\phi^{-1} \end{pmatrix}.

This matrix describes the change of basis between two different fusion-tree orderings: fusing anyons 1, 2 first versus fusing anyons 2, 3 first. Its determinant is -1; its square is the identity (as required for a recoupling matrix).

The non-trivial R matrices — the phases picked up when two \tau's are exchanged and fuse to \mathbf{1} or \tau — are:

R^{\tau\tau}_\mathbf{1} \;=\; e^{-4\pi i / 5}, \qquad R^{\tau\tau}_\tau \;=\; e^{3\pi i / 5}.

Combining F and R using the standard formula \sigma_i = F \cdot \text{diag}(R^{\tau\tau}_\mathbf{1}, R^{\tau\tau}_\tau) \cdot F reproduces the generator matrices \sigma_1, \sigma_2 from Step 3. The appearance of \pi/5 phases is not a choice — it follows from the consistency equations (the pentagon and hexagon identities that F and R must satisfy for a well-defined anyon model).

The Solovay–Kitaev theorem, stated

Theorem (Solovay–Kitaev, 1995–2002). Let G be a finite subset of SU(d) that generates a dense subgroup. Then for every target unitary U \in SU(d) and every \epsilon > 0, there exists a word g_{i_1} g_{i_2} \cdots g_{i_L} in the elements of G and their inverses such that \|g_{i_1} \cdots g_{i_L} - U\| < \epsilon, with length L \leq C \log^c(1/\epsilon), where C depends on G and c < 4.

For Fibonacci braids, G = \{\sigma_1, \sigma_2, \sigma_1^{-1}, \sigma_2^{-1}\} on the three-anyon qubit subspace, and the theorem guarantees polylogarithmic braid words. In practice, the Solovay–Kitaev algorithm is constructive: given a target U, it returns an explicit braid word that approximates U to precision \epsilon.

The Read–Rezayi wavefunction at k = 3

The Read–Rezayi trial wavefunction for k = 3 at filling \nu = 3/5 (which sits inside the second Landau level as \nu = 2 + 3/5... up to complications, the standard quoted filling is 12/5) is an explicit polynomial in the electron coordinates z_j, a generalisation of the Moore–Read Pfaffian:

\Psi_{k=3}(z_1, \ldots, z_N) \;=\; \mathcal{S} \prod_{\text{triples}} \frac{1}{z_a - z_b} \cdot \prod_{i < j} (z_i - z_j)^M \cdot e^{-\sum_k |z_k|^2/4},

where \mathcal{S} is a symmetrisation over triples and M is fixed by the filling. The topological order of this state is precisely SU(2)_3 Chern–Simons, and its quasiparticles are exactly Fibonacci anyons.

What a Fibonacci quantum computer would look like

Imagine the best case: a stable \nu = 12/5 FQHE sample with well-controlled Fibonacci quasiparticles.

Overhead: roughly 3n quasiparticles per logical qubit, versus \sim 1000 physical qubits per logical qubit in the surface-code approach. If built, Fibonacci TQC would reduce the hardware-count for a fault-tolerant machine by a factor of roughly 300.

Comparison with Majorana in one sentence

Majorana is likely to be built but not sufficient; Fibonacci is sufficient but unlikely to be built. The field's current bet is that the hybrid — Majorana-hardware plus magic-state distillation — is the pragmatic near-term route, and Fibonacci is the theoretical ceiling that may or may not be reached.

Where this leads next

References

  1. Michael Freedman, Michael Larsen, Zhenghan Wang, A modular functor which is universal for quantum computation (2002), Commun. Math. Phys. 227 — arXiv:quant-ph/0001108.
  2. Chetan Nayak, Steven Simon, Ady Stern, Michael Freedman, Sankar Das Sarma, Non-abelian anyons and topological quantum computation (2008), Rev. Mod. Phys. 80, 1083 — arXiv:0707.1889.
  3. Nick Read, Edward Rezayi, Beyond paired quantum Hall states: Parafermions and incompressible states in the first excited Landau level (1999), Phys. Rev. B 59, 8084 — arXiv:cond-mat/9809384.
  4. J. S. Xia et al., Electron correlation in the second Landau level: a competition between many nearly degenerate quantum phases (2004), Phys. Rev. Lett. 93, 176809 — arXiv:cond-mat/0406724.
  5. Simon Trebst, Matthias Troyer, Zhenghan Wang, Andreas Ludwig, A short introduction to Fibonacci anyon models (2008) — arXiv:0902.3275.
  6. Wikipedia, Fibonacci anyons.