Surface Code

The leading quantum error correction code, using a 2D grid of qubits with local measurements.


The surface code is the most promising error correction code for near-term fault-tolerant quantum computers. It has high error thresholds and requires only nearest-neighbor interactions.

Structure

Qubits arranged in a 2D grid:

    ●───○───●───○───●
    │   │   │   │   │
    ○───●───○───●───○
    │   │   │   │   │
    ●───○───●───○───●
    │   │   │   │   │
    ○───●───○───●───○

● = Data qubit
○ = Ancilla (measurement) qubit

Two types of measurements:

  • X stabilizers: Detect Z errors (measure X on 4 neighbors)
  • Z stabilizers: Detect X errors (measure Z on 4 neighbors)

Logical Qubit

A surface code “patch” encodes one logical qubit:

  • Size: data qubits ( = code distance)
  • Logical , : Defined by boundary conditions
  • Logical X, Z: Chains of operators across the patch

Error Correction

Errors create pairs of “defects” in syndrome measurements:

    0   0   0   0       0   0   0   0
      0   0   0   0  →    1   0   0   0   (error here)
    0   0   0   0       0   1   0   0

Decoder matches defects to identify errors. Most likely error pattern is corrected.

Code Distance

Distance = minimum number of errors to cause logical error.

DistanceData QubitsTotal Qubits
3917
52549
74997
d~~

Higher distance → better protection (but more qubits).

Error Threshold

The surface code has threshold .

If physical error rate :

Logical error rate decreases exponentially with distance!

Logical Operations

OperationImplementation
Logical XChain of X across patch
Logical ZChain of Z across patch
Logical HCode deformation
Logical SMagic state injection
Logical TMagic state distillation
Logical CNOTLattice surgery

Non-Clifford gates (T) are expensive and require “magic states.”

Lattice Surgery

Merge and split surface code patches to perform logical operations:

Patch A ──[merge]── Patch B
        ──[measure]──
        ──[split]──

This implements logical CNOT between patches.

Advantages

AdvantageDetails
High threshold~1%, achievable with current technology
Local operationsOnly nearest-neighbor interactions
2D layoutMatches chip architecture
FlexibleCan change code size dynamically

Challenges

ChallengeIssue
Qubit overhead~1000 physical per logical
Classical decodingMust be fast enough
Magic statesT gates require distillation
Physical footprintLarge chips needed

Current Status

  • Google, IBM, Quantinuum demonstrating surface code
  • Error rates approaching threshold
  • Full fault-tolerant computation still requires significant scaling

See also: Quantum Error Correction, Logical Qubit, Fault Tolerance