Fault Tolerance

The ability to perform reliable computation despite faulty components. The goal of large-scale quantum computing.


Fault tolerance means a system can continue working correctly even when individual components fail. In quantum computing, it means performing accurate computation despite noisy qubits and gates.

The Problem

Quantum hardware is noisy:

  • Gates have ~0.1-1% error rates
  • Qubits lose coherence over time
  • Measurements can be wrong

Without fault tolerance, errors accumulate faster than computation proceeds.

The Solution: Error Correction Done Right

Quantum error correction encodes information protectively, but the correction process itself uses noisy gates!

Fault tolerance ensures errors during correction don’t spread uncontrollably:

  • Errors in one qubit don’t propagate to many
  • Syndrome extraction doesn’t amplify errors
  • The whole system is self-consistent

Fault-Tolerant Design Principles

1. Transversal Gates

Apply gates qubit-by-qubit across code blocks:

Block 1: ──U──U──U──
Block 2: ──U──U──U──

Errors stay localized.

2. Flag Qubits

Detect when errors might have spread during syndrome measurement.

3. Code Concatenation

Encode logical qubits within logical qubits:

Physical → Level 1 logical → Level 2 logical → ...

Each level suppresses errors further.

4. Careful Circuit Design

Every syndrome extraction circuit must be fault-tolerant.

Threshold Theorem

The foundation of fault tolerance:

If the physical error rate is below a threshold , logical error rates can be made arbitrarily small by using enough physical qubits.

where is related to code distance.

Threshold Values

CodeApproximate Threshold
Surface code~1%
Steane code~0.01%
Concatenated codes~0.001%

Surface code’s high threshold is why it’s favored.

Requirements for Fault Tolerance

RequirementWhy
Error rate below thresholdFoundation of everything
Fast classical processingDecode syndromes in real-time
Scalable architectureNeed millions of qubits
Mid-circuit measurementMeasure syndromes during computation
Low-latency feedbackCorrect errors before they spread

Current Status

MilestoneStatus
Demonstrate error correctionDone
Below-threshold error ratesAchieved for some gates
Logical qubit better than physicalRecent demonstrations
Useful fault-tolerant computationYears away

The Path Forward

Today:     NISQ (noisy, no error correction)
           ↓
Next:      Early fault tolerance (limited logical qubits)
           ↓
Goal:      Full fault tolerance (arbitrary computation)

See also: Quantum Error Correction, Logical Qubit, Fault-Tolerant Quantum Computing