Topological Qubits

A topological qubit stores a logical bit not in the state of any single particle but in a global, nonlocal property of a many-body ground state — the way a knot lives in a rope rather than at any point along it. The signature of topological order is a ground-state degeneracy that no local operator can read, distinguish, or perturb: the states differ only by topologically nontrivial global configurations, so the local environment that dephases every other qubit modality acts trivially on the encoded information, up to corrections exponentially small in system size. This is error protection built into the hardware, composable with software correction. The resource that makes such ground spaces compute is anyons: pointlike excitations of two-dimensional matter whose exchange statistics interpolate beyond the boson–fermion dichotomy. For non-Abelian anyons the Hilbert space of a fixed set of particles is multidimensional, and braiding them adiabatically enacts a unitary on that space. Because the unitary depends only on the homotopy class of the braid — the topology of the worldlines in $2+1$ dimensions — geometric sloppiness in the path produces no error at all. The gates are exact, not merely accurate — the central promise of the topological road to fault tolerance. Two anyon theories anchor the discussion. Ising anyons, realized as Majorana zero modes on defects in topological superconductors, give exact but non-universal gates — braiding yields only the Clifford group, which Gottesman–Knill makes classically simulable. Fibonacci anyons are universal by braiding alone but lack a proven host, and Kitaev's toric code supplies the cleanest lattice model of $\mathbb{Z}_2$ topological order. Against the DiVincenzo criteria the modality is exceptionally coherent in principle yet, as of writing, still a frontier rather than a working machine.

Key takeaways

A topological qubit encodes information in the degenerate ground space of a topologically ordered medium, where the logical states are indistinguishable by any local operator and thus immune to local noise up to $e^{-L/\xi}$ corrections.
The carriers are non-Abelian anyons — quasiparticles whose worldline braids implement unitary gates depending only on the braid's topology, so gates are exact, not approximate.
Ising anyons (Majorana zero modes) store one qubit per four modes and braid to the Clifford group — exact but non-universal, needing magic-state distillation for the missing $T$ gate.
Fibonacci anyons are universal by braiding alone, with quantum dimension $d_\tau=\varphi=(1+\sqrt5)/2\approx1.618$ from the fusion rule $\tau\times\tau=1+\tau$ .
Kitaev's toric code encodes $2g$ logical qubits on a genus- $g$ surface ( $2$ qubits on the torus) with protection exponential in the code distance $d=L$ ; yet no topological qubit has been unambiguously demonstrated, unlike mature superconducting and trapped-ion platforms.
Key idea: information lives in a fusion-space degeneracy that no local operator can read or disturb, and is processed by braiding anyon worldlines so that gates depend only on the topology of the braid — yielding intrinsic hardware-level protection plus, for the right anyon model, exact unitary gates.

Where this lives today

The topological qubit remains the field's boldest bet and its least-settled experiment. The most prominent programme is Microsoft's Majorana effort: after a 2018 Nature report of quantized zero-bias conductance was retracted in 2021, the group reframed its case around a "topological gap protocol" and in February 2025 announced the "Majorana 1" chip, claiming a measurement-based topological qubit from indium-arsenide/aluminium tetron devices — promising but not yet independently confirmed. A separate condensed-matter route, fractional-quantum-Hall interferometry, has meanwhile confirmed Abelian anyon braiding statistics (2020), with the non-Abelian $\nu=5/2$ state the remaining prize. Meanwhile the toric code's idea is already in production: the surface code — topological order engineered by repeated stabilizer measurement rather than a Hamiltonian gap — is the leading error-correcting code on superconducting and trapped-ion processors, with the first below-threshold demonstrations in 2024. The modality thus lives a double life: its mathematics quietly runs today's best quantum memories, while its native hardware — braided, gapped, protected anyons — is still being coaxed into being.

The physical picture

Write a bit not on any single page of a book but in how many times a loop of string winds around the spine. Smudge one page and the winding number is untouched; the information is a topological invariant of the whole configuration. The degenerate ground space of a topologically ordered system stores qubits this way — the logical states are identical to every local probe and differ only in a global, noncontractible property. To read or flip the bit you must act coherently along an entire loop, and the environment, touching the system one atom at a time, essentially never does. A non-Abelian anyon behaves like a particle carrying a hidden internal compass entangled with its neighbours. Fuse two and the outcome is not determined: several distinct charges can result, and that multiplicity of fusion channels is the Hilbert space. Braid one around another and the hidden compasses rotate by an amount fixed entirely by the topology of the loop, so a wobbly path and a clean one enact the identical unitary. The first figure shows a single exchange of two Ising anyons.

Figure

A single exchange σ of two anyons drawn as worldlines in 2+1 dimensions: the enacted unitary ρ(σ) depends only on the braid's homotopy class, so the wobbly grey path and the clean coloured path give the identical Ising gate B = (1 + γ_iγ_j)/√2 and timing jitter contributes exactly zero error

Why is the protection "free"? Because the logical operators are nonlocal. To cause a logical error the environment must act coherently along an entire noncontractible loop, or between two well-separated defects, and the probability of such a correlated event is suppressed as $e^{-L/\xi}$ — not the linear-in-time dephasing of a local qubit. The same nonlocality also constrains: computing means moving the anyons, and at finite temperature spontaneous anyon pairs can silently braid and corrupt the code, so a finite gap $\Delta$ remains essential. The second figure shows the payoff of non-Abelian statistics — the fusion space grows exponentially in the number of anyons, faster for Fibonacci than Ising.

Figure

Fusion-space dimension grows as (d_a)^n with the number of anyons n: Fibonacci anyons (d_τ = φ ≈ 1.618) outrun Ising anyons (d_σ = √2 ≈ 1.414), reaching an 89-dimensional space at n = 10 against 32 for Ising — non-Abelian d_a > 1 is exactly what lets braiding fill Hilbert space and compute

The core equations

The starting point is that a braid is a gate. Adiabatically dragging anyons through one another realizes a homomorphism from the braid group $B_n$ into the unitaries on the fusion space, so a braiding pattern is an element of $B_n$ and its image is the gate.

|\psi_{\text{final}}\rangle = \rho(\sigma)\,|\psi_{\text{initial}}\rangle, \qquad \rho:\;B_n \to U(\mathcal{H}_{\text{fusion}}) \qquad (1)

where: $\sigma$ — an element of the braid group $B_n$ on $n$ anyons (a braiding pattern); $\rho$ — a unitary representation of $B_n$ on the fusion Hilbert space; $\mathcal{H}_{\text{fusion}}$ — the space of fusion outcomes. Reading it: the pattern of crossings, not the path's geometry, selects the unitary, since equivalent braids are the same group element. The dimension of that fusion space is fixed by the fusion rules. The integers $N_{ab}^c$ count the ways charges $a$ and $b$ fuse to $c$ , and their consistency forces each type to carry a quantum dimension $d_a$ setting the space's asymptotic growth.

a \times b = \sum_c N_{ab}^{\,c}\, c, \qquad d_a d_b = \sum_c N_{ab}^{c}\, d_c, \qquad \dim\mathcal{H}_n \sim d_a^{\,n} \qquad (2)

where: $a,b,c$ — anyon types or topological charges; $N_{ab}^{c}$ — fusion multiplicities; $d_a$ — quantum dimension of type $a$ . Reading it: $d_a=1$ for Abelian anyons (unique fusion outcome, no room for a qubit), while $d_a>1$ is the non-Abelian case where $n$ anyons span an exponentially large space $\sim d_a^{\,n}$ . The smallest universal example is the Fibonacci anyon, a single charge $\tau$ that fuses with itself either to the vacuum or back to $\tau$ . That one rule fixes its quantum dimension to the golden ratio.

\tau \times \tau = 1 + \tau, \qquad d_\tau = \varphi = \frac{1+\sqrt{5}}{2} \approx 1.618 \qquad (3)

where: $\tau$ — the nontrivial Fibonacci anyon; $1$ — the vacuum (trivial charge); $\varphi$ — the golden ratio. Reading it: $d_\tau^2=1+d_\tau$ has positive root $\varphi$ ; the space of $n$ $\tau$ 's is a Fibonacci number growing as $\varphi^n$ , and braids on it are dense in $SU(2)$ — hence universal. The concrete carrier of Ising anyons is the Majorana zero mode: a Hermitian, self-conjugate half of an ordinary fermion. Its algebra makes each mode square to the identity and anticommute with the others.

\gamma_i = \gamma_i^\dagger, \qquad \{\gamma_i,\gamma_j\} = 2\delta_{ij}, \qquad c = \tfrac12(\gamma_1 + i\gamma_2) \qquad (4)

where: $\gamma_i$ — Majorana (real-fermion) operator at defect $i$ ; $\{\cdot,\cdot\}$ — anticommutator; $c$ — a Dirac fermion built from two Majoranas. Reading it: four Majoranas make two fermion modes; fixing total parity leaves a two-dimensional logical space — one qubit stored nonlocally, in the joint parity of separated defects. Exchanging two Majoranas is the elementary Ising braid. Because $(\gamma_i\gamma_j)^2=-1$ , the algebra collapses the exponential of $\gamma_i\gamma_j$ to a simple closed form.

B = e^{\frac{\pi}{4}\gamma_i\gamma_j} = \frac{1}{\sqrt2}\left(1 + \gamma_i\gamma_j\right) \qquad (5)

where: $B$ — the unitary exchanging Majoranas $i$ and $j$ ; $\gamma_i\gamma_j$ — the product being braided, behaving like an imaginary unit. Reading it: conjugation by $B$ sends $\gamma_i\to-\gamma_j$ , $\gamma_j\to\gamma_i$ — a $\pi/2$ rotation realizing Clifford gates ( $S$ , Hadamard, and across pairs CNOT), but nothing more. Finally, Kitaev's toric code realizes the same physics on a spin lattice with a gapped Hamiltonian. Commuting star and plaquette stabilizers pin down a ground space whose degeneracy carries the protected qubits.

H = -J_v \sum_v A_v - J_p \sum_p B_p, \qquad A_v = \prod_{i\in v}\sigma^x_i,\;\; B_p = \prod_{i\in p}\sigma^z_i, \qquad \dim(\text{ground}) = 4^{\,g} \qquad (6)

where: $A_v$ — vertex (star) stabilizer; $B_p$ — plaquette stabilizer; $J_v,J_p>0$ — couplings setting the gap; $g$ — genus of the surface. Reading it: all $A_v,B_p$ commute and square to $1$ , the ground space is their joint $+1$ eigenspace, $4^g$ -fold degenerate ( $2$ qubits on the torus), with logical operators the noncontractible Wilson loops.

Derivation AThe Majorana qubit and its parity-protected logical space

The goal is to build one fully protected logical qubit from four Majorana zero modes and prove the claim that names topological hardware: that no operator acting near a single Majorana can read out or flip the stored bit. We pair the modes into two fermions, split the Fock space by total parity, identify the logical Paulis as nonlocal products, and show their matrix elements between the logical states vanish up to corrections exponentially small in $L$ . The result, $\dim\mathcal{H}_{\text{logical}}=2^{n-1}=2$ for $2n=4$ Majoranas, is the smallest topological qubit.

Step 1. Pair the four Majoranas into two Dirac fermions, $c_1=\tfrac12(\gamma_1+i\gamma_2)$ and $c_2=\tfrac12(\gamma_3+i\gamma_4)$ . Using $\gamma_i=\gamma_i^\dagger$ and $\{\gamma_i,\gamma_j\}=2\delta_{ij}$ one checks $\{c_a,c_b^\dagger\}=\delta_{ab}$ and $c_a^2=0$ . Why: two real Majoranas carry the content of one complex fermion, so four defects give a four-dimensional Fock space.

Step 2. Split that Fock space by total parity $P=(-1)^{n_1+n_2}$ , $n_a=c_a^\dagger c_a$ . The even sector $\{|00\rangle,|11\rangle\}$ and odd sector $\{|01\rangle,|10\rangle\}$ are each two-dimensional, and a superconductor conserves parity, pinning the state to one sector. Why: a fixed-parity two-dimensional subspace is one qubit, and superselection forbids mixing the sectors.

Step 3. Write the logical operators in Majorana form: $\overline{Z}=i\gamma_1\gamma_2=2n_1-1$ and $\overline{X}=i\gamma_2\gamma_3$ . These anticommute and square to $1$ — a legitimate logical Pauli pair — yet each is a product of Majoranas that may sit micrometres apart. Why: the qubit lives in the joint parity of a nonlocal pair, so "where the information is" has no local answer.

Step 4. Bound the leakage. Any $\mathcal{O}_{\text{local}}$ near one Majorana commutes with the parity constraint and cannot connect $|0_L\rangle$ to $|1_L\rangle$ ; the residual matrix element comes only from overlap of the two modes, giving $\langle 0_L|\mathcal{O}_{\text{local}}|1_L\rangle = O(e^{-L/\xi})$ . Why: this single exponential is the whole of "topological protection" here — double the separation and the error squares.

Go deeper · Derivation B · 1 minGround-state degeneracy of the toric code on a torus

Derivation BGround-state degeneracy of the toric code on a torus

The goal is to count how many protected qubits Kitaev's toric code stores by turning the ground-state condition into linear algebra over the stabilizer group. A spin-1/2 sits on every edge, every vertex gives one $A_v$ and every plaquette one $B_p$ , all commuting and squaring to the identity, and the codespace is their joint $+1$ eigenspace. The one subtlety is that the product of all stars and of all plaquettes are each the identity — two redundancies that leave a residual degeneracy. We obtain $\dim=2^N/2^{N-2}=4$ on the torus and $4^g$ on a genus- $g$ surface.

Step 1. Fix the lattice: an $L\times L$ torus has $L^2$ vertices, $L^2$ plaquettes, and $N=2L^2$ edges (spins), consistent with the Euler characteristic $V-E+F=0$ . This gives $2L^2=N$ stabilizer generators $A_v,B_p$ . Why: matching generators to spins is the bookkeeping that pins down how much freedom is left unfixed.

Step 2. Check the algebra: any $A_v$ and $B_p$ share zero or two edges, so they commute, and each squares to $1$ . Hence they are simultaneously diagonalizable, and the ground space is their joint $+1$ eigenspace. Why: only a commuting, square-to-one set defines a well-posed stabilizer codespace.

Step 3. Find the redundancies: on a closed surface $\prod_v A_v=\mathbb{1}$ and $\prod_p B_p=\mathbb{1}$ , since every edge is shared by two vertices and two plaquettes, so each Pauli appears twice and cancels. That is two constraints, leaving $N-2$ independent stabilizers. Why: dependent generators impose no new condition — they are what the degeneracy remembers.

Step 4. Count dimensions: each independent stabilizer halves the $2^N$ -dimensional spin space, so $\dim=2^N/2^{N-2}=4=2^2$ — two logical qubits, with $\overline{X},\overline{Z}$ the Wilson loops on the two cycles. On genus $g$ there are still exactly two redundancies, but $V-E+F=2-2g$ leaves only $N-2g$ independent stabilizers, so $\dim=2^N/2^{N-2g}=4^g$ . Why: the degeneracy is a topological invariant of the surface, not the lattice size — the hallmark of topological order.

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

Derivation CWhy braiding Ising anyons yields only Clifford gates

The goal is to show precisely why braiding Ising anyons, for all its exactness, cannot by itself run a universal quantum computer. Expanding the braid generator with the Majorana algebra reveals that it conjugates Paulis into Paulis — the signature of a Clifford gate; Gottesman–Knill then certifies any Clifford circuit is efficiently classically simulable, so a non-Clifford resource must come from outside the braid. The upshot, $\langle B_{ij}\rangle=\text{Clifford}\subsetneq U(2^k)$ , is why Ising braiding still needs magic-state distillation to reach universality.

Step 1. Note $(\gamma_i\gamma_j)^2=-\gamma_i^2\gamma_j^2=-1$ , so $\gamma_i\gamma_j$ acts like an imaginary unit and $B_{ij}=e^{\frac{\pi}{4}\gamma_i\gamma_j}=\cos\tfrac{\pi}{4}+\gamma_i\gamma_j\sin\tfrac{\pi}{4}=\tfrac{1}{\sqrt2}(1+\gamma_i\gamma_j)$ . Why: the Majorana algebra makes the Euler expansion terminate exactly, giving a closed-form braid gate — no small-angle approximation.

Step 2. Conjugate the Majoranas: $B_{ij}\gamma_i B_{ij}^\dagger=-\gamma_j$ and $B_{ij}\gamma_j B_{ij}^\dagger=\gamma_i$ , leaving other $\gamma_k$ fixed. Why: braiding rotates the two exchanged Majoranas into each other with a sign — a rigid, exact $\pi/2$ rotation on the operator algebra.

Step 3. Translate to logical Paulis: since $\overline{Z}=i\gamma_1\gamma_2$ and $\overline{X}=i\gamma_2\gamma_3$ , the braids permute $\{\overline{X},\overline{Y},\overline{Z}\}$ up to signs — mapping Paulis to Paulis, the definition of the Clifford group. One obtains $S$ , Hadamard up to phase, and inter-pair braids give CNOT. Why: mapping the Pauli group to itself is the Clifford property, so the braid image sits inside the Cliffords.

Step 4. Invoke Gottesman–Knill: any Clifford circuit on stabilizer input is classically simulable in polynomial time, hence non-universal. Restore universality by distilling a magic state $|T\rangle=(|0\rangle+e^{i\pi/4}|1\rangle)/\sqrt2$ . Why: universality is strictly stronger than exactness — the braid gives perfect Cliffords, a separate non-topological resource supplies the rest.

Limiting cases

How protection and gate power vary with the physical regime:

Case	Result	Meaning
Zero temperature, infinite gap	$\Delta/k_BT \to \infty$	Ideal limit: no thermal anyon pairs, degeneracy exact up to tunnelling; protection limited only by finite size.
Finite size	$\delta E \sim \Delta\,e^{-L/\xi}$	Two defects a distance $L$ apart split the degeneracy exponentially, hybridizing the logical states and slowly dephasing them.
Finite temperature	$n_{\text{anyon}} \sim e^{-\Delta/k_BT}$	Thermal anyons braid uncontrollably; lifetime scales as $e^{\Delta/k_BT}$ , so a large gap is essential.
Abelian anyons ( $d_a=1$ )	$\dim\mathcal{H}_n = 1$	Fusion is unique, exchange a mere phase — a good memory (toric code) but no space to compute in by braiding.
Ising anyons ( $d_\sigma=\sqrt2$ )	$\langle\text{braid}\rangle=\text{Clifford}\subsetneq U(2^n)$	Exact but non-universal; classically simulable, needs magic states.
Fibonacci anyons ( $d_\tau=\varphi$ )	braids dense in $SU(2)$	Exact and universal by braiding alone — the theorist's ideal, with no proven host.

Common trapTopological protection is absolute and braiding alone gives a universal computer

Neither half is true. Protection is exponential, not infinite: it scales as $e^{-L/\xi}$ in the anyon separation and $e^{\Delta/k_BT}$ in the gap-to-temperature ratio, so finite temperature, finite size (tunnelling between defects), and quasiparticle poisoning (a stray fermion flipping the parity sector) all reintroduce errors — a large gap and large separation are mandatory, and active error correction is still expected on top. And braiding Ising anyons is emphatically not universal: it generates only the Clifford group, which Gottesman–Knill makes efficiently classically simulable, so a non-topological magic-state resource is unavoidable. Only Fibonacci anyons are universal by braiding alone, with no robust host established. "Topological" buys an exceptionally quiet qubit and, for Ising anyons, exact Clifford gates — not a free universal quantum computer.

Worked examples

1 · Fibonacci fusion space — how fast does it fill? (fully worked)

Take $n$ Fibonacci anyons of type $\tau$ . Find the quantum dimension $d_\tau$ , the growth of the fusion-space dimension, and how many logical qubits a chain of $10$ anyons is worth.

Step 1. Apply $d_a d_b=\sum_c N_{ab}^c d_c$ to $\tau\times\tau=1+\tau$ with $d_1=1$ : $d_\tau^2 = 1 + d_\tau$ , i.e. $d_\tau^2-d_\tau-1=0$ . The positive root is $d_\tau=(1+\sqrt5)/2=\varphi=\mathbf{1.618}$ .

Step 2. The dimension of $n$ $\tau$ 's is a Fibonacci number, $\dim\mathcal{H}_n=F_{n+1}$ , growing as $\varphi^n/\sqrt5$ . So ten anyons give $F_{11}=\mathbf{89}$ , tracking the quantum-dimension growth $\varphi^{10}=122.99$ ; contrast Ising with $d_\sigma=\sqrt2$ : $(\sqrt2)^{10}=\mathbf{32}$ , nearly three times smaller.

Step 3. Convert to qubits via $\log_2(\dim)$ : $\log_2(89)=\mathbf{6.48}$ , so a ten-anyon chain holds roughly $6\tfrac12$ qubits' worth of space ( $2^6=64$ , $2^7=128$ ). Because $d_\tau>1$ and braids are dense in $SU(2)$ , any single-qubit gate is approximable to accuracy $\varepsilon$ by a braid of length $O(\log^c(1/\varepsilon))$ (Solovay–Kitaev). $d_\tau=\varphi\approx1.618$ , $\dim=F_{n+1}\sim\varphi^n$ , $n=10\Rightarrow89\approx6\tfrac12$ qubits — universal by braiding alone.

2 · The coherence budget of a Majorana qubit (fully worked)

A Majorana wire has induced gap $\Delta/k_B=1\ \mathrm{K}$ , coherence length $\xi=100\ \mathrm{nm}$ , its two zero modes $L=2\ \mu\mathrm{m}$ apart, at $T=20\ \mathrm{mK}$ . Estimate the finite-size splitting and its dephasing time, and the thermal suppression, and say which dominates.

Step 1. Finite-size splitting. With $L/\xi=20$ , $\delta E\sim\Delta\,e^{-L/\xi}=k_B(1\ \mathrm{K})\,e^{-20}$ . Since $e^{-20}=\mathbf{2.06\times10^{-9}}$ and $k_B(1\ \mathrm{K})=1.38\times10^{-23}\ \mathrm{J}$ , $\delta E=\mathbf{2.85\times10^{-32}\ J}$ .

Step 2. Dephasing time: $\tau\sim\hbar/\delta E=\mathbf{3.7\ ms}$ . Doubling $L$ to $4\ \mu\mathrm{m}$ squares the factor ( $e^{-40}$ ) and pushes $\tau$ to about $1.8\times10^{6}\ \mathrm{s}$ , roughly three weeks — the exponential is the whole game.

Step 3. Thermal poisoning: the anyon density is $\sim e^{-\Delta/k_BT}=e^{-50}=\mathbf{1.9\times10^{-22}}$ , negligible. The finite-size splitting ( $\sim3.7$ ms) dominates the budget; thermal excitation ( $e^{-50}$ ) is irrelevant at $20$ mK, so the engineering lever is longer wires, not colder fridges.

Physics of the portrait

The topological qubit is where a 1937 curiosity of particle theory, a 1970s puzzle about statistics in flatland, and a 1990s insight on fault tolerance converge.

Origins · 1 minAlexei Kitaev and how this was discovered

Alexei Kitaev

b. 1963 · topological quantum computation, anyons, and the toric code

Alexei Kitaev turned topological quantum computation from a curiosity into a research programme. In a 1997 preprint, "Fault-tolerant quantum computation by anyons," he proposed that a quantum computer could be protected not by constant active correction but by the topology of a medium's excitations — braiding non-Abelian anyons to enact gates immune to local error. The same work introduced the toric code, the simplest exactly solvable model of topological order and the direct ancestor of the surface code now central to error correction. In 2001 he wrote down the Majorana chain, showing an ordinary-looking wire could bind unpaired Majorana zero modes at its ends — the blueprint every Majorana-nanowire experiment since has chased. With Freedman, Larsen, and Wang he proved braiding certain anyons is polynomially equivalent to the circuit model, and with Bravyi he formalized magic-state distillation; his 2006 "sixteenfold way" classification of anyon models remains field-defining. At Caltech and the Kavli Institute, Kitaev works where condensed-matter physics, quantum information, and topology are one subject.

History

1937

— Majorana writes down a fermion that is its own antiparticle

1977

— Leinaas and Myrheim show two-dimensional particles admit fractional statistics

1982

— Wilczek coins the term anyon for these fractional-statistics particles

1997

— Kitaev proposes fault-tolerant quantum computation by anyons and the toric code

2003

— Freedman, Kitaev, Larsen and Wang prove braiding matches the quantum circuit model

Going further

Texts. Nielsen & Chuang, Quantum Computation and Quantum Information, 10th anniv. ed., Ch. 10 and Ch. 7 · Preskill, Physics 219 Lecture Notes, Ch. 9 (Topological Quantum Computation) · Kitaev, "Fault-tolerant quantum computation by anyons," Ann. Phys. 303, 2 (2003) · Nayak, Simon, Stern, Freedman, Das Sarma, "Non-Abelian Anyons and Topological Quantum Computation," Rev. Mod. Phys. 80, 1083 (2008) · Bravyi & Kitaev, Phys. Rev. A 71, 022316 (2005) · Watrous, The Theory of Quantum Information (CUP, 2018) · Mermin, Quantum Computer Science (CUP, 2007). Key numerical checks: $2^{n-1}=2$ for four Majoranas; the torus count gives $N=2L^2$ generators, two global constraints, and degeneracy $4=4^{g=1}$ , consistent with $V-E+F=0$ ; $d_\tau=\varphi$ follows from $\tau\times\tau=1+\tau$ , with $\dim=F_{n+1}$ giving $F_{11}=89$ at $n=10$ (quantum-dimension growth $\varphi^{10}=122.99$ ) and $(\sqrt2)^{10}=32$ for Ising; $(\gamma_i\gamma_j)^2=-1$ yields $B=(1+\gamma_i\gamma_j)/\sqrt2$ and Clifford conjugation; the coherence estimate gives $\delta E=k_B(1\,\mathrm{K})e^{-20}=2.85\times10^{-32}\,\mathrm{J}$ , $\tau=\hbar/\delta E=3.7$ ms, and $e^{-50}=1.9\times10^{-22}$ . The golden ratio and the conductance quantum $2e^2/h$ are the only constants quoted, both standard.

Prerequisites. DiVincenzo Criteria · Quantum Error Correction · Group Theory and Unitary Representations · Second Quantization and BCS Theory. Leads to. Surface Code · Magic-State Distillation · Superconducting Qubits · Trapped-Ion Qubits.

Ask the tutor: Want me to walk through how a specific braid word acts on a four-Majorana qubit, or why Fibonacci anyons reach universality that Ising anyons cannot?

How is a topological qubit different from a surface-code qubit if both trace back to the toric code?

They share mathematics but differ in where the protection comes from. A native topological qubit relies on a Hamiltonian with an energy gap: the degeneracy is a property of the material's ground state, and the gap suppresses errors even with no measurement. A surface-code qubit engineers the same structure actively, measuring stabilizers thousands of times a second on ordinary qubits and correcting in software — no protecting gap, only fast feedback. The surface code works today; native hardware would need far less active correction. Most roadmaps expect a hybrid.

If braiding Ising anyons is exact, why isn't that already a fault-tolerant quantum computer?

Because exactness and universality are different properties. Braiding Ising anyons implements Clifford gates perfectly — no timing or path error — but the Clifford group is not universal, and by Gottesman–Knill any Clifford circuit on stabilizer states can be simulated efficiently on a laptop. To compute something classically hard you need one non-Clifford gate, such as $T$ , supplied non-topologically by preparing and distilling noisy magic states. So Ising anyons give a superb, exact Clifford backbone, and universality is bolted on from outside the braid. Fibonacci anyons would be universal by braiding alone, but no material is known to host them robustly.

Where is the information physically located in a Majorana qubit?

Nowhere local — that is the point. The logical $Z$ is the joint fermion parity $i\gamma_1\gamma_2$ of two Majorana zero modes that can sit micrometres apart at opposite ends of a wire. Neither end "has" the bit; only their joint parity does. An instrument probing one end measures an operator that commutes with the logical information and learns nothing, and cannot flip it. The only way to read or corrupt the qubit is to act on both defects coherently, with amplitude $\sim e^{-L/\xi}$ . The delocalization is the protection; "which end holds the 0" is like asking which page holds a loop's winding number.

What exactly is a quantum dimension, and why must it exceed 1 to compute?

The quantum dimension $d_a$ measures how fast the fusion Hilbert space grows as you add anyons of type $a$ : $\dim\mathcal{H}_n\sim d_a^{\,n}$ . For Abelian anyons $d_a=1$ : two fuse to a unique outcome, exchange is a mere phase, and there is no room to store a qubit. When $d_a>1$ the fusion outcome is genuinely ambiguous, and that ambiguity is a Hilbert space braids can rotate — Ising has $d_\sigma=\sqrt2$ , Fibonacci $d_\tau=\varphi\approx1.618$ . A non-integer quantum dimension is the fingerprint of non-Abelian statistics — what lets a braid be a nontrivial gate rather than a phase.

What are the main ways topological protection fails in a real device?

Three effects, each with a clear scaling. First, finite size: two defects a distance $L$ apart have overlapping wavefunctions, splitting the degeneracy by $\sim\Delta\,e^{-L/\xi}$ and slowly dephasing the qubit — fixable by longer wires. Second, finite temperature: thermal anyon pairs at density $\sim e^{-\Delta/k_BT}$ braid uncontrollably, giving a lifetime $\sim e^{\Delta/k_BT}$ — fixable by a larger gap and colder operation. Third, quasiparticle poisoning: a stray electron tunnels in and flips the fermion parity, directly flipping a Majorana qubit; this is often the practical limiter, since even a tiny non-equilibrium quasiparticle population defeats the protection. All three point to one recipe: large gap, large separation, clean parity.

Has anyone actually built a topological qubit?

Not unambiguously, as of writing. Zero-bias conductance peaks near the predicted $2e^2/h$ — consistent with Majorana zero modes — have been reported many times, but such peaks can also come from trivial Andreev bound states, and a prominent 2018 claim was retracted in 2021. Microsoft's 2025 "Majorana 1" claims a measurement-based topological qubit passing its own "topological gap protocol," but the community treats it as encouraging rather than settled, pending independent reproduction of full braiding and fusion. Interferometry has convincingly shown Abelian anyon braiding in fractional quantum Hall systems (2020); the non-Abelian $\nu=5/2$ state remains under study. The honest status: the code descended from these ideas already protects real quantum memories, but a demonstrated, braided, native topological qubit is still a frontier result.