Simple Algorithm
Quantum Computing · Rydberg State · 32 min

Topological Qubits & Majorana Modes

Encoding information non-locally in exotic quasiparticles.

What if a qubit could be stored so that no local disturbance could read it, and therefore none could damage it? Majorana zero modes promise exactly that: a qubit held jointly by the two ends of a wire. The theory is beautiful and exactly solvable. The experiments have become one of the most contested stories in modern condensed-matter physics.

What you’ll learn

  • Solve the Kitaev chain at its sweet spot and identify the unpaired Majorana zero modes.
  • Explain non-local parity encoding, fermion-parity superselection and the protection argument, including quasiparticle poisoning.
  • Prove that braiding Majorana zero modes implements only Clifford operations, and explain how universality is restored.
  • State the nanowire proposals and their topological criterion, and assess the experimental record carefully.

The Kitaev chain

Kitaev (2001) introduced a minimal model of a one-dimensional topological superconductor: spinless fermions on NN sites with chemical potential μ\mu, hopping tt and pp-wave pairing Δ\Delta (both taken real and positive),

H=−μ∑j=1Ncj†cj+∑j=1N−1[−t(cj†cj+1+cj+1†cj)+Δ(cjcj+1+cj+1†cj†)]H = -\mu\sum_{j=1}^{N}c_j^\dagger c_j + \sum_{j=1}^{N-1}\Big[-t\big(c_j^\dagger c_{j+1} + c_{j+1}^\dagger c_j\big) + \Delta\big(c_jc_{j+1} + c_{j+1}^\dagger c_j^\dagger\big)\Big]
DerivationThe sweet spot μ = 0, t = Δ
  1. cj†cj=12(1+i ajbj)c_j^\dagger c_j = \tfrac12\big(1 + i\,a_jb_j\big)

    The on-site term pairs the two Majoranas of the same site.

  2. cj†cj+1+cj+1†cj=i2(ajbj+1−bjaj+1),cjcj+1+cj+1†cj†=i2(ajbj+1+bjaj+1)c_j^\dagger c_{j+1} + c_{j+1}^\dagger c_j = \tfrac{i}{2}\big(a_jb_{j+1} - b_ja_{j+1}\big),\qquad c_jc_{j+1} + c_{j+1}^\dagger c_j^\dagger = \tfrac{i}{2}\big(a_jb_{j+1} + b_ja_{j+1}\big)

    Hopping and pairing between neighbors, simplified with the anticommutation relations.

  3. μ=0, t=Δ:H=it∑j=1N−1bj aj+1\mu = 0,\ t = \Delta:\qquad H = i t\sum_{j=1}^{N-1} b_j\,a_{j+1}

    The ajbj+1a_jb_{j+1} terms cancel and the bjaj+1b_ja_{j+1} terms add: each Majorana bjb_j pairs with the Majorana aj+1a_{j+1} of the next site.

  4. dj=12(bj+i aj+1):i bjaj+1=2dj†dj−1⇒H=2t∑j=1N−1(dj†dj−12)d_j = \tfrac12\big(b_j + i\,a_{j+1}\big):\qquad i\,b_ja_{j+1} = 2d_j^\dagger d_j - 1\quad\Rightarrow\quad H = 2t\sum_{j=1}^{N-1}\Big(d_j^\dagger d_j - \tfrac12\Big)

    Each bond hosts a new fermion that costs energy 2t2t: the bulk is gapped.

  5. f=12(a1+i bN),[H,f]=0f = \tfrac12\big(a_1 + i\,b_N\big),\qquad \comm{H}{f} = 0

    The Majoranas a1a_1 and bNb_N, one at each end, do not appear in HH: they are unpaired. Together they form a fermion mode of exactly zero energy, delocalized over the whole chain. The ground state is two-fold degenerate, with fermion parity −ia1bN=1−2f†f=±1-ia_1b_N = 1 - 2f^\dagger f = \pm1.

Away from the sweet spot the end modes survive as long as the bulk gap stays open: the chain is in its topological phase for ∣μ∣<2t\abs\mu < 2t (with Δ≠0\Delta\neq0). The zero modes are then exponentially localized near the ends, with decay length ξ\xi, and their energy splitting in a chain of length LL is δE∼Δ e−L/ξ\delta E \sim \Delta\,e^{-L/\xi}, up to a prefactor.

A qubit stored non-locally

Fermion parity is a superselection rule, so the even and odd ground states of a single chain cannot be superposed. A qubit therefore uses four Majoranas γ1,…,γ4\gamma_1,\dots,\gamma_4, for example the ends of two wires, at fixed total parity −γ1γ2γ3γ4=+1-\gamma_1\gamma_2\gamma_3\gamma_4 = +1. The logical Pauli operators are Majorana bilinears:

ZL=−iγ1γ2,XL=−iγ2γ3,YL=iγ1γ3Z_L = -i\gamma_1\gamma_2,\qquad X_L = -i\gamma_2\gamma_3,\qquad Y_L = i\gamma_1\gamma_3
All three commute with the total parity and obey the Pauli algebra (XLYL=iZLX_LY_L = iZ_L) on the fixed-parity subspace. Reading the qubit means measuring the joint parity of two distant Majoranas.
Predict before you look

The zero-mode splitting scales as δE∝e−L/ξ\delta E\propto e^{-L/\xi}. By how many orders of magnitude does it shrink when a wire is doubled from L=10ξL = 10\xi to L=20ξL = 20\xi?

Show the answer

4.3 orders

The ratio is e−10≈4.5×10−5e^{-10}\approx4.5\times10^{-5}, i.e. 10/ln⁡10≈4.310/\ln10\approx4.3 orders of magnitude. For scale: with Δ∼0.2\Delta\sim0.2 meV, L=10ξL = 10\xi still leaves δE∼10\delta E\sim10 neV and a dephasing time of only about 0.10.1 μs. The protection is exponential, but the exponent has to be large.

Braiding is Clifford

DerivationThe braid unitary
  1. Bij=exp⁡ ⁣(π4 γiγj)=12(1+γiγj)B_{ij} = \exp\!\Big(\frac\pi4\,\gamma_i\gamma_j\Big) = \frac{1}{\sqrt2}\big(1+\gamma_i\gamma_j\big)

    Adiabatically exchanging two Majoranas γi,γj\gamma_i, \gamma_j applies this unitary to the ground space (Ivanov, 2001); the closed form uses (γiγj)2=−1(\gamma_i\gamma_j)^2 = -1.

  2. Bij γi Bij†=−γj,Bij γj Bij†=γi,Bij γk Bij†=γk  (k≠i,j)B_{ij}\,\gamma_i\,B_{ij}^\dagger = -\gamma_j,\qquad B_{ij}\,\gamma_j\,B_{ij}^\dagger = \gamma_i,\qquad B_{ij}\,\gamma_k\,B_{ij}^\dagger = \gamma_k\ \ (k\neq i,j)

    Direct computation with the anticommutation relations: an exchange maps Majoranas to Majoranas, with a sign. Exchanging twice sends γi→−γi\gamma_i\to-\gamma_i and γj→−γj\gamma_j\to-\gamma_j, which is not the identity: the signature of non-Abelian (Ising) anyons.

  3. B12=exp⁡ ⁣(iπ4ZL)≃S†,B23=exp⁡ ⁣(iπ4XL)≃(X)†B_{12} = \exp\!\Big(i\frac\pi4 Z_L\Big)\simeq S^\dagger,\qquad B_{23} = \exp\!\Big(i\frac\pi4 X_L\Big)\simeq\big(\sqrt X\big)^\dagger

    On the four-Majorana qubit, using γ1γ2=iZL\gamma_1\gamma_2 = iZ_L and γ2γ3=iXL\gamma_2\gamma_3 = iX_L. Both are quarter-turn Clifford rotations; the orientation of the exchange decides between a gate and its inverse.

Universality needs a non-Clifford ingredient. Bravyi and Kitaev (2005) showed that ideal Clifford operations plus noisy copies of a magic state suffice, via magic-state distillation, so a Majorana computer would look like other fault-tolerant designs, with non-Clifford gates prepared outside the topological protection. In practice braiding may be replaced altogether by sequences of joint parity measurements (measurement-only topological computing; Bonderson, Freedman & Nayak, 2008), the basis of most current device designs. Other anyon types escape the theorem: braiding Fibonacci anyons is universal (Freedman, Larsen & Wang, 2002), but no system has been shown to host them in a controllable form.

Check your understanding

A Majorana-based processor implements all Clifford gates topologically. Which additional resource makes it universal?

  1. More Majorana zero modes
  2. Longer braids
  3. A non-Clifford resource, such as distilled magic states for TT gates
  4. Faster parity measurements
Show the answer

C. A non-Clifford resource, such as distilled magic states for TT gates

Braids and parity measurements generate only Clifford operations, however many and however long (the braid group's image is finite). A TT gate, for example implemented by injecting distilled magic states, completes a universal set.

From nanowires to the lab

Proposals. A semiconductor nanowire with strong spin–orbit coupling (InAs or InSb), proximity-coupled to an ss-wave superconductor (typically aluminum) and placed in a magnetic field along the wire, realizes an effective Kitaev chain (Lutchyn, Sau & Das Sarma, 2010; Oreg, Refael & von Oppen, 2010). In the simplest single-subband model the wire is topological when the Zeeman energy exceeds a critical value:

EZ>Δ2+μ2E_Z > \sqrt{\Delta^2 + \mu^2}
Δ\Delta is the induced superconducting gap and μ\mu the chemical potential measured from the band bottom. At the transition the bulk gap closes and reopens.
SignatureIdeal MZM predictionWhy it is not conclusive alone
Zero-bias conductance peakA peak at zero voltage when tunneling into a wire end, quantized at 2e2/h2e^2/h at zero temperatureTrivial Andreev bound states, from smooth confinement or disorder, produce similar near-zero peaks
Gap closing and reopeningThe bulk gap closes and reopens at the transition, visible in nonlocal conductanceDisorder and multiple subbands can mimic or obscure the pattern
Exponential splittingPeak-spacing oscillations in Coulomb-blockaded islands that shrink as e−L/ξe^{-L/\xi}Needs clean, long wires and careful modeling
Fractional Josephson effectA 4π4\pi-periodic supercurrentMasked by quasiparticle poisoning and Landau–Zener processes
Fusion and braidingNon-deterministic fusion outcomes; non-Abelian exchange statisticsThe decisive tests, not yet achieved convincingly

Meanwhile, related physics is being built from the bottom up. Minimal Kitaev chains of two quantum dots coupled through a superconductor show "poor man's Majorana" signatures at fine-tuned sweet spots (Dvir et al., 2023). Non-Abelian anyons have also been prepared and braided as entangled states on superconducting and trapped-ion processors (2023–2024). Those are digital simulations of topological order: valuable physics, but not hardware-level protection.

Key takeaways

  • The Kitaev chain at μ=0\mu = 0, t=Δt = \Delta pairs Majoranas across bonds and leaves a1a_1 and bNb_N unpaired: a zero-energy fermion stored non-locally, with splitting ∝e−L/ξ\propto e^{-L/\xi} away from the sweet spot.
  • Qubits use four Majoranas at fixed total parity; local parity-conserving noise acts trivially up to e−L/ξe^{-L/\xi}, but quasiparticle poisoning is not protected.
  • Braids act as exp⁡(π4γiγj)\exp(\tfrac\pi4\gamma_i\gamma_j), mapping Majoranas to Majoranas: only Clifford gates, so magic states are still required for universality.
  • Nanowire proposals require EZ>Δ2+μ2E_Z > \sqrt{\Delta^2+\mu^2}. As of the mid-2020s, Majorana zero modes and non-Abelian braiding have not been convincingly demonstrated, and several prominent claims have been retracted or disputed.

Check yourself

In the interactive lesson these are scored with your confidence, and what you miss comes back for review.

Question 1

At the sweet spot μ=0\mu = 0, t=Δt = \Delta of an NN-site Kitaev chain, how many zero-energy fermion modes are there, and where do they live?

  1. None: the chain is fully gapped
  2. Two, one localized at each end
  3. NN, one per site
  4. One, built from the unpaired Majoranas a1a_1 and bNb_N at opposite ends
Show the answer

D. One, built from the unpaired Majoranas a1a_1 and bNb_N at opposite ends

H=2t∑j(dj†dj−12)H = 2t\sum_j(d_j^\dagger d_j - \tfrac12) involves only bond fermions. The leftover Majoranas a1a_1 and bNb_N form a single fermion f=12(a1+ibN)f = \tfrac12(a_1 + ib_N) at zero energy.

Question 2

Why does a Majorana qubit use four zero modes rather than two?

  1. Fermion-parity superselection forbids superposing the even and odd ground states of one pair, so the qubit lives in two pairs at fixed total parity
  2. Two zero modes cannot be separated far enough
  3. Each Majorana stores half a qubit
  4. Braiding needs at least four modes to do anything
Show the answer

A. Fermion-parity superselection forbids superposing the even and odd ground states of one pair, so the qubit lives in two pairs at fixed total parity

One pair's two states differ in parity and cannot be superposed. With fixed total parity, 2n2n Majoranas encode n−1n-1 qubits.

Question 3

Why can't braiding alone make a Majorana processor universal?

  1. Braids are too slow
  2. Braiding destroys the topological protection
  3. Each braid maps Majorana operators to ± Majorana operators, hence Paulis to Paulis: every braid is Clifford, and the generated group is finite
  4. Braids commute with each other, so they generate only diagonal gates
Show the answer

C. Each braid maps Majorana operators to ± Majorana operators, hence Paulis to Paulis: every braid is Clifford, and the generated group is finite

Clifford unitaries are exactly those that map Paulis to Paulis under conjugation, and the Clifford group modulo phases is finite, so it cannot be dense in the unitary group.

Question 4

Why is a zero-bias conductance peak, even one close to 2e2/h2e^2/h, not proof of a Majorana zero mode?

  1. Conductance cannot be measured at zero bias
  2. Trivial Andreev bound states, for example from smooth confinement or disorder, can produce similar near-zero-energy peaks
  3. Majorana modes never produce conductance peaks
  4. Quantized conductance occurs only in quantum Hall systems
Show the answer

B. Trivial Andreev bound states, for example from smooth confinement or disorder, can produce similar near-zero-energy peaks

An ideal MZM implies a quantized zero-bias peak, but the converse fails: non-topological states can mimic it. That is why nonlocal, fusion and braiding signatures are needed.

Further reading

  • A. Yu. Kitaev, "Unpaired Majorana fermions in quantum wires", Physics-Uspekhi 44, 131 (2001).
  • D. A. Ivanov, "Non-Abelian statistics of half-quantum vortices in p-wave superconductors", Physical Review Letters 86, 268 (2001).
  • C. Nayak, S. H. Simon, A. Stern, M. Freedman & S. Das Sarma, "Non-Abelian anyons and topological quantum computation", Reviews of Modern Physics 80, 1083 (2008).
  • J. Alicea, "New directions in the pursuit of Majorana fermions in solid state systems", Reports on Progress in Physics 75, 076501 (2012).
  • S. Bravyi & A. Kitaev, "Universal quantum computation with ideal Clifford gates and noisy ancillas", Physical Review A 71, 022316 (2005).