Topological Qubits — padho.wiki

In short

A topological qubit stores quantum information in a non-local property of a many-body quantum system — so non-local that a noise process localised anywhere in the system literally cannot see it. The idea, due to Alexei Kitaev (1997, 2003), is to exploit exotic two-dimensional quasiparticles called non-abelian anyons whose exchange — "braiding" — acts as a unitary operation on a degenerate ground-state manifold that depends only on the topology of the braid, not on its microscopic details. Since local noise produces no braid, it produces no unitary — and the quantum information is protected by topology itself. The most-pursued physical realisation uses Majorana zero modes (MZMs): localised, zero-energy excitations at the ends of a one-dimensional topological superconductor. A pair of MZMs constitutes one qubit; four MZMs give one qubit whose gates are generated by exchanging (braiding) the MZMs. Microsoft has pursued this program for two decades; its Majorana 1 prototype was announced in February 2025. Crucial caveats: Majorana braiding alone is not universal for quantum computing — it produces Clifford gates only — so a non-topological "magic-state" injection step is needed for a full gate set. Experimental claims of MZM observation have been persistently contested: a 2018 Nature paper from Delft was retracted in 2021 after analysis errors; subsequent 2022–2024 Microsoft claims rest on a "topological gap protocol" that parts of the condensed-matter community continue to regard as insufficient evidence of bona fide MZMs. The Majorana 1 announcement is a significant engineering milestone — an eight-qubit device in the architecture Microsoft needs — but it is not yet a scaled, fault-tolerant topological quantum computer. Why care anyway: if topological qubits work, a single physical qubit could have an error rate below the surface-code threshold without any explicit error-correction circuitry. In India, IISc Bangalore (Shankar–Chakraborti, condensed-matter theory) and TIFR Mumbai (topological-phase theory) host active topological-matter programs. The practical hardware remains a distant goal; the theory is settled, beautiful, and will continue to shape the field regardless of whether Majorana-1 scales.

Every quantum-computing hardware platform in this series — superconducting transmons, trapped ions, neutral atoms, photons, spin qubits — fights the same enemy: local noise. A stray magnetic field flips a spin; a thermal photon ejects an atom; a trapped charge shifts a qubit's frequency; a cosmic ray ionises a patch of substrate. The response, developed over three decades, is quantum error correction: redundantly encode each logical qubit in many physical ones, constantly measure joint parities, and apply corrections. The surface code that IBM, Google, and QuEra are racing to implement achieves this — but it needs, per logical qubit, something like a thousand physical qubits at the ~1% per-gate error rate of 2024 hardware.

What if, instead, you designed a qubit that was intrinsically blind to local noise — a qubit where information simply did not live at any single physical location, so nothing localised could see it?

That is the dream of topological quantum computing. The quantum information is stored in a global, topological degree of freedom of a many-body system. No local perturbation — not a stray field, not a trapped charge, not a phonon, not a cosmic ray — can couple to a global degree of freedom, because "local" and "global" are opposites. The only way to change the quantum state is to perform a discrete, topological operation: typically, exchanging two specific quasiparticles around each other, which is called braiding.

This is not a trick. It is a theorem. If you can realise a two-dimensional system with the right kind of exotic quasiparticles — non-abelian anyons — then a qubit built from them has exponentially suppressed sensitivity to local noise, and its gates are performed by moving quasiparticles around each other rather than by fine-tuned laser pulses. No calibration, no resonant frequencies, no drift. The theoretical case is airtight.

The experimental case is the problem. Twenty-eight years after the idea was proposed, it has not yet been unambiguously realised in a laboratory. Microsoft announced its first prototype — Majorana 1 — in February 2025, drawing both enthusiasm and continued skepticism from the community. This chapter builds the idea, traces its history, and tries to give you an honest picture of where the field stands in 2026.

Step 1 — anyons: particles that are neither bosons nor fermions

The standard classification of identical particles in three-dimensional space is binary: they are either bosons (exchanging two of them leaves the wavefunction unchanged) or fermions (exchange multiplies by -1). There is no third option in 3D, because the exchange operation is a rotation, and applying it twice brings the particles back to where they started — so the phase it introduces must square to +1, giving \pm 1 as the only possibilities.

In two dimensions, this argument fails. Exchanging two particles twice does not return them to their original configuration in a topologically trivial way; instead, the two paths wind around each other and cannot be contracted to a point. The space of particle configurations is no longer simply connected, and the wavefunction can pick up any phase on exchange — or, more dramatically, the wavefunction can transform as a matrix acting on a degenerate space.

Particles with these exotic exchange statistics are called anyons, a name coined by Frank Wilczek in 1982 (anyons because their exchange phase can be anything, not just \pm 1).

An abelian anyon picks up a phase e^{i\theta} on exchange, for some \theta other than 0 (boson) or \pi (fermion). Such particles exist in the fractional quantum Hall effect — the Laughlin quasiparticles at filling factor \nu = 1/3 pick up e^{i\pi/3} on exchange, observed in transport measurements in 2020.
A non-abelian anyon does something stronger: if n non-abelian anyons sit in the system, there are multiple degenerate ground states (say, d^n of them for some d > 1), and exchanging two of the anyons acts on this degenerate space as a unitary matrix — not a single phase. Different exchange orderings give different matrices, and they generally do not commute. Hence "non-abelian."

For quantum computing, non-abelian is the magic. The degenerate ground-state space is the Hilbert space of a topological qubit (or a topological register of qubits), and braiding operations — physically, moving the anyons around each other in 2D space — act as non-trivial unitary gates on that space.

Braiding two anyons in 2+1 dimensions (two spatial + one time dimension). Left: their world lines form a braid — one anyon winds around the other on its way past. Right: the many-body wavefunction acquires a unitary $U$ determined by the topological class of the braid; small wobbles in the trajectory, or slow-down and speed-up, do not change $U$. This is the source of the topological protection: $U$ depends only on *which* anyons went around *which*, not on how they got there.

Step 2 — what a topological qubit is

Here is the key picture. Suppose a physical system has four non-abelian anyons at fixed positions. Because of their non-abelian character, the many-body ground state is two-fold degenerate: there are two linearly independent states, call them |0\rangle_L and |1\rangle_L (the subscript L stands for "logical"), both at exactly the same energy.

A local perturbation — say, a small potential well at some location in the system — couples the ground states only to excited states a gap away (the topological gap \Delta), shifting the ground-state manifold by an amount that vanishes as e^{-L/\xi} where L is the distance between anyons and \xi is the coherence length. Crucially, the perturbation cannot mix |0\rangle_L with |1\rangle_L because the two differ by a topological property — a property that is not expressible as a local operator on the system.

That is the protection. Local noise cannot cause bit flips or phase flips of the logical qubit, because no local operator connects |0\rangle_L and |1\rangle_L. The coherence time of the qubit is, in principle, exponentially long in the separation of the anyons.

Gates on the logical qubit are implemented by braiding. Exchange anyon 1 with anyon 2 while 3 and 4 remain fixed: the unitary U_{12} acts on the logical qubit in a specific way determined by the topology of the exchange. A different exchange — anyons 2 and 3, say — gives a different U_{23}, and U_{12} and U_{23} do not commute. Concatenating braids is quantum circuit composition; the set of unitaries you can implement by braiding is called the braid-group representation of the anyon model.

Measurement is performed by fusing anyons: bring two of them together and see what they combine into. In a non-abelian model, the fusion has multiple possible outcomes, and which one you get encodes the measurement result.

A topological qubit from four non-abelian anyons. The many-body ground state is two-fold degenerate; the two states $|0\rangle_L$ and $|1\rangle_L$ are distinguished by a *topological* invariant of the anyon configuration — not by any local observable. A local perturbation (red X) shifts both logical states by a tiny, equal amount (exponentially small in the inter-anyon spacing) but cannot mix them, because no local operator can change a topological invariant.

Step 3 — Kitaev 2003 and the toy model

Alexei Kitaev proposed the first concrete lattice model of a topological qubit in a 2003 paper titled Fault-tolerant quantum computation by anyons. Consider a 1D chain of N spinless fermions (for example, electrons with fixed spin orientation) with a specific kind of "p-wave" superconducting pairing. The Hamiltonian is

H_{\text{Kitaev}} = -\mu \sum_{j=1}^{N} c_j^\dagger c_j - t \sum_{j=1}^{N-1}(c_j^\dagger c_{j+1} + c_{j+1}^\dagger c_j) + \sum_{j=1}^{N-1}(\Delta\, c_j c_{j+1} + \Delta^* c_{j+1}^\dagger c_j^\dagger),

where c_j annihilates a fermion at site j, \mu is the chemical potential, t is the hopping amplitude, and \Delta is the p-wave pairing. Rewrite each fermion operator c_j as a combination of two new operators \gamma_{2j-1} and \gamma_{2j} called Majorana operators, defined so that \gamma_k^\dagger = \gamma_k (Hermitian) and \gamma_k^2 = 1. There are 2N Majorana operators; the chain has 2N "Majorana sites."

In a specific topological regime (|\mu| < 2t and \Delta \ne 0), the Hamiltonian, when rewritten in Majorana language, pairs up every Majorana with its neighbour except for the Majorana operators \gamma_1 (at the leftmost site) and \gamma_{2N} (at the rightmost site). Those two are left unpaired: they have zero energy and are spatially separated by the entire length of the chain. They are called Majorana zero modes (MZMs).

The ground state is two-fold degenerate. The two ground states differ only by the occupation of the non-local fermion mode f = (\gamma_1 + i\gamma_{2N})/2 — a fermion composed of two Majoranas at opposite ends of the chain. Because the Majoranas are far apart, no local perturbation can change this occupation: that is the topological protection.

This is the simplest concrete model of Majorana zero modes, and it maps — remarkably — onto a one-dimensional topological superconductor. In the real world, the Kitaev chain is approximated by a semiconductor nanowire (InAs or InSb) with strong spin-orbit coupling, covered with a thin superconducting layer (Al), placed in a magnetic field. The resulting system is predicted to host Majorana zero modes at its ends. This is the central proposal of the Microsoft program.

Braiding Majoranas — a CNOT-like gate

Take four Majorana zero modes: \gamma_1, \gamma_2, \gamma_3, \gamma_4. These form one qubit: the computational basis is the two fermion-parity states of the non-local fermion modes (\gamma_1 + i\gamma_2)/2 and (\gamma_3 + i\gamma_4)/2, restricted to the sector where the total parity is fixed (e.g. even). The two basis states are |0\rangle_L and |1\rangle_L.

Exchanging \gamma_1 with \gamma_2 in 2D space (move one MZM around the other in a half-loop) implements the unitary

U_{12} = \frac{1}{\sqrt{2}}(1 + \gamma_1 \gamma_2) = e^{i\pi\gamma_1\gamma_2/4}.

On the logical qubit, this acts as e^{i\pi Z_L / 4} — a single-qubit T-like phase gate (actually, in Majorana conventions, an R_z(\pi/2) gate). Exchanging \gamma_2 with \gamma_3 gives e^{i\pi X_L / 4} — an R_x(\pi/2)-like gate. All braids combining these are Clifford operations — they generate the Clifford group on the logical qubit.

What braids cannot do — and why magic states are needed

Clifford gates alone are not universal. The Clifford group — generated by H, S, and CNOT — can be efficiently simulated on a classical computer (the Gottesman–Knill theorem). For universal quantum computing you additionally need a non-Clifford gate, usually the T gate (e^{i\pi Z/8}).

Majorana braiding naturally gives e^{i\pi\gamma_1\gamma_2/4} \sim R_z(\pi/2), which is a Clifford. It does not give you T. To complete the gate set, topological quantum computing with Majoranas must fall back on magic state injection: prepare an ancilla in a specific non-stabilizer state (the "magic state" |T\rangle = (|0\rangle + e^{i\pi/4}|1\rangle)/\sqrt 2), and use it as fuel for a T gate via a gadget that uses only Clifford operations plus measurement. The magic state itself is prepared by a non-topological process and then "distilled" to high fidelity using a protocol that costs many copies.

This is the famous caveat of Majorana-based topological quantum computing: its braiding is only half of what you need. The other half — magic state distillation — is not topologically protected and must rely on standard error-correction techniques. Topological protection gets you 85–90% of the way to a low-overhead logical qubit; the last 10–15% still requires conventional machinery. An alternative is to use Fibonacci anyons (a different non-abelian anyon model) whose braiding is already universal on its own — but no physical system is predicted to host Fibonacci anyons yet.

Example 1: Braiding two Majoranas — getting a NOT-like gate

Walk through what happens when you exchange two Majorana zero modes.

Step 1. Identify the physical setup. Four Majoranas \gamma_1, \gamma_2, \gamma_3, \gamma_4 sit at four spatial positions in a 2D system — imagine the four ends of two separate topological superconducting wires arranged in a "T" or "H" shape. The ground-state manifold has dimension 2 (even total parity); the computational basis states are |0\rangle_L (both non-local fermion modes unoccupied) and |1\rangle_L (both occupied). Why dimension 2: four Majoranas fuse into two fermion modes; the even-parity sector of Fock space on two fermion modes is spanned by two states (|00\rangle and |11\rangle in occupation-number basis). These are the computational basis of the topological qubit.

Step 2. Perform the braid. Physically, drag \gamma_1 through 2D space in a half-loop that swaps its position with \gamma_2, keeping \gamma_3 and \gamma_4 undisturbed. On the many-body wavefunction, the effect is

B_{12}\,\gamma_1\,B_{12}^{-1} = \gamma_2, \qquad B_{12}\,\gamma_2\,B_{12}^{-1} = -\gamma_1,

(the sign is the non-abelian signature — one Majorana turns into the other, the other turns into minus the first) and the unitary implementing this is B_{12} = e^{i\pi\gamma_1\gamma_2/4} = (1 + \gamma_1\gamma_2)/\sqrt 2. Why the signs: in two dimensions, a half-exchange of two identical particles is a \pi rotation in the relative-coordinate plane; for Majorana fermions this rotation acts as a Bogoliubov rotation on the operators themselves, producing the specific sign pattern above. Derivation details are in Kitaev's 2003 paper.

Step 3. Write it as a logical gate. Define Z_L = i\gamma_1\gamma_2 (this is a Hermitian operator satisfying Z_L^2 = 1 — a Pauli operator on the logical qubit). Then B_{12} = e^{-i\pi Z_L/4} \cdot e^{i\pi/4} (up to a global phase), which is R_z(-\pi/2) — an eighth-of-a-turn phase gate. Why Z_L and not X_L: the specific bilinear i\gamma_1\gamma_2 that governs the braid, when translated through the Majorana-to-fermion dictionary, turns out to be the Pauli Z on the logical qubit. A different choice of Majorana-to-logical-basis dictionary would call it X, but the physics is invariant.

Step 4. Braid two pairs — get a NOT gate. A full exchange (braid twice) gives B_{12}^2 = \gamma_1\gamma_2 = -iZ_L — a pure Z_L gate, up to a phase. A different braid — exchanging \gamma_2 with \gamma_3 — gives B_{23} = e^{i\pi\gamma_2\gamma_3/4}, which on the logical qubit is R_x(\pi/2) up to a phase. Combining six Majorana braids gives a full single-qubit Clifford: for instance, B_{23}B_{12}B_{23} realises a 90° rotation about the y-axis of the logical Bloch sphere — a Hadamard-like operation. Why Cliffords only: each braid is a \pi/4 rotation about a Pauli axis on the logical qubit; the group generated by such rotations is precisely the Clifford group (up to global phases), which is provably not universal without additional resources.

Step 5. Make a CNOT between two logical qubits. Two qubits means eight Majoranas \gamma_1, \ldots, \gamma_8. The braid B_{45}\,B_{56}\,B_{23}\,B_{34} (a specific four-crossing braid) implements a CNOT on the two logical qubits. This is the entangling operation and it is also Clifford. Why this specific braid: the Majorana-to-Pauli dictionary maps the product \gamma_2\gamma_3\gamma_4\gamma_5 to X_L \otimes X_L; the four-crossing braid produces a unitary proportional to \exp(i\pi X_L X_L / 4) which, combined with single-qubit Cliffords, gives CNOT. The explicit construction is worked out in Nayak et al. (2008).

Result. The full Clifford group on n logical qubits is implemented by braids of 4n Majoranas. A universal gate set requires one more ingredient — a non-Clifford T gate — which is added by non-topological magic-state distillation, breaking the elegance but completing the computation.

The world-line braid for exchanging $\gamma_1$ and $\gamma_2$. In 2+1D, the paths cross — a topologically non-trivial event in the anyon's configuration space. The resulting unitary on the logical qubit is $B_{12} = (1+\gamma_1\gamma_2)/\sqrt 2 = e^{-i\pi Z_L/4}$, a $\pi/4$ phase rotation. Unlike a microwave-pulse gate, this gate's angle is set by pure topology — no calibration, no drift, no pulse-shape error. It can go wrong only if the braid is *topologically* wrong (the wrong particles are exchanged, or the exchange is interrupted) — which is a discrete, detectable event, not an analogue drift.

What this shows: a logical gate is performed not by tuning the amplitude and duration of a microwave pulse to 0.1% precision, but by moving particles around each other. The gate's fidelity is bounded not by analogue pulse-shape errors but by the topological coherence of the anyons. If the theoretical framework holds in the laboratory, topological gates are intrinsically perfect to the extent that the anyons themselves are well-defined.

Step 4 — why topology protects

Now make the protection argument precise. Suppose the many-body ground state has energy E_0 and the first excited state has energy E_0 + \Delta (the topological gap). Consider a local Hermitian perturbation V = V(\vec r_0) acting near some position \vec r_0 far from any Majorana. To first order in perturbation theory, the shift in any ground state is

\delta E_n = \langle n | V | n \rangle + \sum_{m \ne n} \frac{|\langle m | V | n \rangle|^2}{E_n - E_m} + \ldots

The key question: can the matrix element \langle 0 | V | 1 \rangle_L (between the two logical states) be nonzero?

The answer is exponentially small in L/\xi, where L is the separation of the Majoranas and \xi is the coherence length of the topological superconductor. Proof sketch: the two logical states differ by the occupation of the non-local fermion mode f = (\gamma_1 + i\gamma_{2N})/2. The operator V at position \vec r_0 couples only to local degrees of freedom near \vec r_0, which have exponentially small overlap with \gamma_1 (at one end) or \gamma_{2N} (at the other) — each Majorana wavefunction decays as e^{-x/\xi} away from its localisation point. The matrix element therefore scales as e^{-L/\xi}, where L is the distance from \vec r_0 to the nearest Majorana, and an even smaller amount from the far one.

In a macroscopic system with L \gg \xi, the error rate per unit time for a single Pauli error on a topological qubit is exponentially suppressed in L/\xi. For realistic parameters (\xi \sim 100 nm, L \sim 1 μm), that suppression is \sim e^{-10} \approx 4\times 10^{-5}. This is the exponential error suppression that topological protection promises.

Example 2: Local noise on a topological qubit vs a regular one

Work through what a local Pauli-like error does to a topological qubit compared with a standard one.

Step 1. The regular qubit. Consider a transmon sitting at some location, in superposition |\psi\rangle = \alpha|0\rangle + \beta|1\rangle. A stray magnetic field fluctuation \delta B(t) near the qubit shifts its Zeeman energy and therefore adds a phase \phi(t) = g\mu_B\int \delta B(t')\,dt'/\hbar to the |1\rangle state. Accumulated over time, this is dephasing: \beta \to \beta e^{i\phi}, and averaging over the noise distribution shrinks the coherence. Why: the qubit's quantum information is stored in a local degree of freedom — the electron-spin state of a specific location in the chip — and any local perturbation at that location couples directly to the information.

Step 2. The topological qubit. Consider a pair of Majoranas \gamma_1, \gamma_2 at opposite ends of a nanowire of length L \gg \xi, with logical state |\psi\rangle_L = \alpha|0\rangle_L + \beta|1\rangle_L. The same stray magnetic field fluctuation \delta B(\vec r_0, t) near some position \vec r_0 couples to the local density of states at \vec r_0 but has essentially zero matrix element connecting |0\rangle_L and |1\rangle_L: this matrix element scales as e^{-|\vec r_0 - \vec r_1|/\xi} \cdot e^{-|\vec r_0 - \vec r_2|/\xi}, where \vec r_1 and \vec r_2 are the positions of \gamma_1 and \gamma_2. Why the double exponential: the matrix element is a two-point function that must "reach" both Majorana ends, so it picks up one exponential suppression factor from each end. If \vec r_0 is near neither, the product is vanishingly small.

Step 3. Compare the error rates. For the regular qubit, the error rate from local magnetic-field noise is of order \gamma = (g\mu_B)^2 \langle \delta B^2\rangle \tau_c / \hbar^2 where \tau_c is a correlation time — a typical number is 10^3 Hz (a T_2 of a millisecond). For the topological qubit, the same noise multiplied by the e^{-2L/\xi} matrix-element suppression gives \gamma_{\text{top}} \sim \gamma \cdot e^{-2L/\xi}. At L = 10\xi, this is a factor of e^{-20} \sim 2\times 10^{-9} — a billion-fold suppression. The topological T_2 is correspondingly billion-fold longer than the bare T_2. Why this is the punchline: the topological qubit is not just a bit better than a regular qubit; it is exponentially better, and the exponent grows with the physical separation of the Majoranas. A 10-micron wire with \xi = 100 nm gives L/\xi = 100, and the error rate is suppressed by e^{-200} — astronomically.

Step 4. Where the argument breaks. Two effects spoil this paradise. (a) Non-zero temperature: thermal photons or phonons with energy above the topological gap \Delta can create excited quasiparticles ("poisoning" the topological qubit by filling the non-local fermion with a thermal quasiparticle). The error rate from this is \sim e^{-\Delta/k_B T}, which is exponentially small in \Delta/T — still very good at 20 mK where \Delta \approx 10\,k_B T, but not zero. (b) Finite wire length: L is finite, and at short enough L the exponential suppression is small. Real Majorana nanowires have L \sim 1 μm and \xi \sim 100 nm, giving L/\xi \sim 10 — good but not astronomical. Why these are the bottlenecks: in every real experiment, the protection is exponential but limited by the weakest exponential — usually the topological gap or the Majorana overlap. The Microsoft "topological gap protocol" is precisely about measuring and optimising \Delta in candidate materials.

Step 5. The honest summary. Topological protection is exponential, but the exponent is finite. The scale of the exponent — set by \Delta, \xi, L, T — determines how protected a real device is. For near-future devices, the protection is a few orders of magnitude beyond unprotected qubits, not infinite. That is already enormous: a topological qubit with a physical error rate of 10^{-6} would mean logical error rates approaching fault-tolerant thresholds without explicit error correction. But "topological = perfect" is a slogan, not a derived result. Why: in condensed-matter and quantum-information discussions, the phrase "topologically protected" is sometimes used as shorthand for "exponentially protected in the gap-to-temperature ratio and the distance-to-coherence-length ratio." These are large numbers in good devices, but they are not infinite.

Result. A topological qubit is exponentially less sensitive to local noise than a regular qubit — a factor that can be 10^3 to 10^{10} depending on material parameters. If realised in a clean enough system, this makes quantum error correction dramatically cheaper (fewer physical qubits per logical one) or even unnecessary for moderate circuit depths.

Left: local noise (red arrows) hits a regular qubit directly at its localisation point, giving a dephasing rate $\gamma$. Right: the same noise hits somewhere in the middle of a topological-superconductor wire that has Majorana zero modes at its two ends. The noise has exponentially small matrix element connecting the two logical states, because the logical degree of freedom (the parity of the non-local fermion composed of $\gamma_1$ and $\gamma_2$) requires the perturbation to "reach both ends" simultaneously — suppressed by $e^{-2L/\xi}$. The practical suppression is typically a factor of $10^4$ to $10^9$, depending on wire length and coherence length.

What this shows: the quantitative case for topological qubits is not a slogan. It is a calculable exponential suppression factor, given by the ratio of wire length to coherence length. The challenge is not the theory — the theory is airtight — but realising a wire with the right material properties, long enough and clean enough, to reach the large exponent regime.

Step 5 — Majorana zero modes in the lab (2012–2024)

Theoretical predictions of Majorana zero modes in specific materials appeared in 2008–2010 (Fu–Kane, Sau–Lutchyn–Tewari–Das Sarma, Oreg–Refael–von Oppen). The recipe: an InSb or InAs semiconductor nanowire, a thin Al superconducting film proximity-coupled, a magnetic field of ~300 mT. At low temperature and specific gate voltages, the system should enter the topological phase, and Majorana zero modes should localise at the wire ends.

The experimental story has been long and contested:

2012 (Mourik et al., Delft) reported a zero-bias conductance peak (ZBP) in InSb–Al nanowires — one of the predicted signatures of MZMs. The ZBP is, however, also produced by several non-topological mechanisms (Andreev bound states, disorder-induced zero-energy states), and the experiment did not uniquely identify MZMs.
2012–2017 saw a succession of increasingly sophisticated ZBP experiments by groups at Delft, Copenhagen, Weizmann, UCLA, Copenhagen, and Microsoft's Station Q labs. Each was consistent with MZMs but also with alternative explanations. The field settled into cautious optimism.
2018 (Zhang et al., Delft) published a Nature paper claiming to have observed the quantised conductance 2e^2/h predicted for a Majorana zero mode — a much stronger claim than a generic ZBP. The result was a flagship Microsoft achievement.
2021 — the 2018 paper was retracted after external analysis revealed data selection and analysis errors. An independent investigation concluded that the authors had "eliminated data that was inconsistent with the hypothesis." The retraction was a serious blow to the field and triggered broad reassessment of MZM claims.
2022–2024 (Microsoft, Aghaee et al.) introduced the topological gap protocol — a measurement suite designed to distinguish genuine topological signatures from trivial alternatives by looking at both ends of the wire simultaneously and checking that the gap closes and reopens as a function of magnetic field (the predicted topological phase transition). The 2023 Nature publication was Microsoft's argument that their devices had now passed these stricter tests.
February 2025 — Microsoft announced Majorana 1, described as an "eight-qubit topological processor" fabricated at their Station Q lab. The announcement claimed single-qubit operations in the correct Majorana-based architecture; independent verification is ongoing, and the condensed-matter community remains divided on whether the claimed MZMs are bona fide topological states or more sophisticated non-topological mimics.

The honest 2026 assessment:

Theory: airtight. Majorana zero modes in topological superconductors are real; the Kitaev chain and its realisations are well-understood.
Experiment: contested. Many experiments see signatures consistent with MZMs; no single experiment has yet produced evidence that is universally accepted as decisive. Braiding (the actual topological computation step) has not yet been demonstrated on any platform at the time of writing.
Engineering: advanced but slow. Microsoft has the most mature platform; Majorana 1 is real hardware; whether it scales to thousands of topological qubits in the next decade is an open question.

A short timeline. Kitaev's theoretical proposal (2003) is the anchor. The first zero-bias peak consistent with Majorana zero modes came in 2012, but alternative explanations could not be ruled out. The 2018 Delft paper in *Nature* claimed the sharper "quantised zero-bias peak" signature, then was retracted in 2021 after analysis errors surfaced. Microsoft's topological-gap protocol (2022, with major publications in 2023) and the Majorana 1 announcement (February 2025) are the most recent milestones; braiding — the gate operation — has not been demonstrated in any experiment as of this writing.

Who is betting on topological qubits

Microsoft has by far the largest industrial program, running two decades and funding dedicated fabrication labs (Station Q, Copenhagen) plus collaborations with Delft, Copenhagen, UCSB, and others. The company's long bet: a small number of topological qubits per error-corrected logical qubit — perhaps 10× fewer physical qubits than a surface-code machine — making a useful fault-tolerant quantum computer accessible at a far smaller scale.
Alice & Bob (Paris), while not Majorana-focused, builds cat qubits — another route to intrinsically error-biased qubits. Not quite topological, but in the same philosophical family of "design the qubit so the errors are easier."
Academic groups working on MZMs or other non-abelian anyons: Delft (Kouwenhoven, Bakkers), Copenhagen (Marcus, Nygård), UCSB (Palmstrøm), Harvard (Marcus — moved from Copenhagen), Weizmann (Heiblum), and others.
Fractional quantum Hall: independent of the Majorana route, a handful of groups pursue non-abelian anyons at filling fraction \nu = 5/2 in GaAs 2DEGs — predicted to host the "Moore–Read" non-abelian state. Progress has been slow; as of 2024 the community does not consider \nu = 5/2 definitively non-abelian.

Indian context

IISc Bangalore (Centre for Theoretical Physics and Department of Physics) hosts active condensed-matter theory groups working on topological phases, Majorana physics, and the Kitaev model. Subroto Mukerjee and collaborators have published on topological superconductors and lattice models of anyons.
TIFR Mumbai (Department of Theoretical Physics) has a topological-matter program with contributions to the theory of surface codes, non-abelian anyons, and fault tolerance.
IIT Madras and IIT Bombay have smaller topological-matter theory groups.
The National Quantum Mission does not — at least publicly — identify topological qubits as an experimental priority (silicon, NV, superconducting, and neutral atoms get the hardware funding). NQM-funded theoretical work on topological codes and fault tolerance is ongoing.

The Indian topological-qubit community is predominantly theoretical. This reflects the global pattern: the experiments are done at a handful of extremely well-resourced Western labs, while the theory — mathematically rich, cheap to do — has broad international participation.

Common confusions

"Topological qubits already work, Microsoft made a chip." The 2025 Majorana 1 announcement is a serious engineering milestone, but (a) the claimed Majorana zero modes remain contested by parts of the condensed-matter community, (b) braiding — the actual topological gate operation — has not been demonstrated in a Majorana device at any scale, and (c) Majorana 1 is a prototype with 8 qubits, orders of magnitude short of a useful quantum computer. It is a milestone, not a finished product.
"Topological = perfect." Topological protection is exponential in L/\xi and in \Delta/k_B T. At realistic material parameters, the suppression is 10^3 to 10^9 — large but not infinite. Topological qubits are still subject to "quasiparticle poisoning" from thermal excitations above the gap and from finite-wire end-to-end tunnelling.
"Braiding alone is universal." For Majorana zero modes, braiding generates only the Clifford group — not universal. A non-Clifford resource (magic state injection) is needed. The universality-from-braiding-alone is true for Fibonacci anyons, a different non-abelian model, but Fibonacci anyons are not predicted to exist in any physical system so far engineered.
"Anyons are free-floating exotic particles." Non-abelian anyons are localised excitations (quasiparticles or defect modes) in specific two-dimensional quantum many-body systems. In the Majorana case, the "particle" is a bound zero-energy mode at the end of a topological superconductor wire — not a thing you can isolate in vacuum.
"Braiding is physical motion." In some proposals, yes — you physically move Majorana positions by adjusting gate voltages. But in many practical architectures, braiding is implemented by measurement-based operations: sequences of parity measurements that have the same net effect as a braid without physically moving anything. Microsoft's Majorana 1 is measurement-based.
"The retraction means Majorana zero modes don't exist." The 2021 retraction concerned specific data-selection practices in one experiment. It did not disprove the theory. Subsequent experiments under the topological gap protocol remain consistent with MZMs; the debate is about whether the evidence is strong enough, not whether the phenomenon is impossible.

Going deeper

If you understand that topological qubits encode quantum information in the degenerate ground space of a non-abelian anyonic system, that Majorana zero modes are the best-understood experimental candidate, that braiding gives Clifford gates (not universal without magic states), that local noise is exponentially suppressed in L/\xi, and that Microsoft's Majorana 1 (Feb 2025) is a prototype whose claims the community is still evaluating — you have chapter 172. What follows is a closer look at Kitaev's 2003 paper, the Nayak et al. (2008) review, why Fibonacci anyons would give universality, the continued controversy around Majorana evidence, and the long-term prospects.

Kitaev's 2003 paper

Alexei Kitaev's Fault-tolerant quantum computation by anyons (Annals of Physics, 2003) is the foundational reference for topological quantum computing. The paper introduces: (a) the Kitaev chain 1D model of p-wave superconductivity with Majorana end states; (b) the general framework for non-abelian anyons on a lattice; (c) the toric code, a separate but related topological stabilizer code that became the basis of surface-code error correction; (d) the theoretical framework for braiding as quantum computation. The paper is long (50+ pages), dense, and remains the single best reference for getting the theory right. The arXiv preprint is freely accessible.

Non-abelian anyons — the Nayak et al. 2008 review

Chetan Nayak, Steven Simon, Ady Stern, Michael Freedman, and Sankar Das Sarma wrote the canonical review Non-abelian anyons and topological quantum computation (Rev. Mod. Phys. 2008). The review covers: the abstract theory of anyons, Ising vs Fibonacci vs Moore–Read anyons, candidate physical systems (quantum Hall, topological superconductors, spin liquids), proposed experiments, and universality results. It is the reference for understanding why Fibonacci anyons would be universal via braiding alone (their braiding group is dense in SU(2)) and why Majorana-based (Ising) anyons are not (the Ising braid group is finite, giving only Cliffords).

Fibonacci anyons — the other path

A Fibonacci anyon model has a single non-trivial anyon type \tau satisfying the fusion rule \tau \times \tau = 1 + \tau. The dimension of the n-anyon Hilbert space grows as the Fibonacci sequence (hence the name). Braiding Fibonacci anyons generates a dense subgroup of SU(2) — every unitary can be approximated to arbitrary precision by a sufficiently long braid. Braiding alone is universal. No magic state distillation needed.

The catch: no one has yet identified a physical system that cleanly hosts Fibonacci anyons. The leading candidates are some fractional quantum Hall states (particularly \nu = 12/5), but the experimental signatures are elusive. If a Fibonacci-anyon platform were discovered, topological quantum computing would become the cleanest route to universal fault-tolerant quantum computing — no magic states, no error correction, just braiding. This is the "holy grail" scenario.

Measurement-based topological computing

In modern Majorana architectures — including Microsoft's Majorana 1 — braids are not performed by physically moving Majoranas around each other. Instead, measurement-based protocols perform the same net unitary via sequences of fermion-parity measurements on pairs of Majoranas. This is more practically implementable (no moving parts) and preserves the topological protection: each measurement outcome is random, but the logical unitary implemented by the measurement sequence is deterministic up to Pauli byproducts that can be tracked in classical software. This is the topological analogue of measurement-based quantum computing on regular qubits.

Why the community remains cautious

Several concerns keep the field in "show me" mode:

Alternative explanations for zero-bias peaks. Disorder-induced "trivial" Andreev bound states can mimic MZM signatures. Distinguishing them requires measuring both ends of the wire and the topological-gap-protocol signatures.
No direct demonstration of non-abelian statistics. The decisive experiment would be to braid Majoranas and measure a non-abelian unitary on a logical qubit. No such experiment has been performed; all current evidence is indirect (transport signatures, not braiding outcomes).
Historical false positives. The 2018 Nature retraction demonstrated that experimental claims in this field can be wrong even in top-tier journals. The community has adopted a "show me braiding" standard.
Material complexity. InSb and InAs nanowires with high-quality proximity-induced superconductivity require extremely careful materials science. Reproducibility across labs has been challenging.

The field is genuinely at a crossroads: either the next few years produce unambiguous braiding experiments and topological qubits become a real platform, or the program reaches its terminal difficulty and the field pivots. Both outcomes remain possible in 2026.

Topological vs surface-code qubits

Even if Majorana-based topological qubits are realised, they are not a replacement for surface-code error correction in the large-scale limit. The argument: topological protection gives exponentially good physical qubits, but the exponent depends on hardware parameters (wire length, coherence length, temperature). For a cryptographically relevant quantum computer (~2048-bit RSA), the required logical error rate is ~10^{-12}; achieving that from physical error rates alone, via topological protection, requires wires with L/\xi \gtrsim 30, which is materials-physics-hard. Most realistic roadmaps assume topological qubits would be used alongside surface-code error correction, with the topological protection reducing the physical error rate to ~10^{-5} and the surface code boosting to 10^{-12}. The topological qubit's role is to reduce the surface-code overhead by perhaps 10×, not to eliminate error correction entirely.

Comparison to other platforms

Platform	Physical qubit count (2024)	2Q fidelity	Braiding demonstrated
Microsoft Majorana 1 (topological)	8 (claimed)	not published	no
IBM Condor/Heron (superconducting)	1121 / 133	99.7%	n/a
Quantinuum H2 (trapped ion)	56	99.91%	n/a
Atom Computing (neutral atom)	1180	99.4%	n/a

Topological qubits lag dramatically on qubit count and have not yet demonstrated the gate operation that defines the platform. The bet is not "topological is currently better" but "topological, if it works, will scale differently and require less error correction."

Where this leads next

Spin Qubits — the other "not-yet-dominant" matter-qubit platform, with a much clearer experimental track record but weaker intrinsic noise protection.
Toric / Surface Codes — the topological stabiliser code inspired by the same physics as topological qubits; surface-code error correction is the main competitor framework and, in a sense, the "software" version of what topological qubits achieve in hardware.
DiVincenzo's Criteria — the criteria against which all hardware platforms are judged; topological qubits score very well on coherence if they work, but currently score zero on "demonstrated braiding."
Superconducting Transmon — the dominant non-topological platform; much of the Majorana hardware stack borrows from superconducting-circuit fabrication.
Neutral Atom Qubits — the platform with the largest current qubit count; a reasonable comparison for what "scaling succeeds" actually looks like in 2024-2026.

References

Alexei Kitaev, Fault-tolerant quantum computation by anyons (2003), Annals of Physics 303, 2 — arXiv:quant-ph/9707021.
Chetan Nayak et al., Non-abelian anyons and topological quantum computation (2008), Rev. Mod. Phys. 80, 1083 — arXiv:0707.1889.
Jason Alicea, New directions in the pursuit of Majorana fermions in solid state systems (2012), Rep. Prog. Phys. 75, 076501 — arXiv:1202.1293.
Microsoft Azure Quantum (Aghaee et al.), InAs-Al hybrid devices passing the topological gap protocol (2023) — arXiv:2207.02472.
Wikipedia, Topological quantum computer.
John Preskill, Lecture Notes on Quantum Computation, Chapter 9 — theory.caltech.edu/~preskill/ph229.