Why Clifford Circuits Are Easy to Simulate

Why Clifford Circuits Are Easy to Simulate

Clifford gates can entangle qubits while a stabilizer simulator tracks compact Pauli structure instead of every statevector amplitude.

A classical computer usually needs 2^n complex amplitudes to store an arbitrary pure state of n qubits. Yet some circuits with many qubits can be simulated efficiently because their structure is unusually regular. The important clue is that certain gates preserve the relationships among Pauli operators.

The Clifford rule

A Clifford gate maps Pauli operators to Pauli operators when they pass through the gate. Hadamard H, phase S, and controlled-NOT CX generate the multi-qubit Clifford gates. Stabilizer circuits start in a stabilizer state such as |0> on every qubit, use Clifford gates, and measure Pauli observables. Classical choices based on earlier measurement results can be included too.

The Gottesman-Knill theorem says this restricted circuit family can be simulated efficiently on a classical computer. That does not mean every quantum circuit is easy to simulate. It means that this particular combination of inputs, gates, and measurements has a compact description.

Track symmetries, not every amplitude

A stabilizer state is the +1 eigenstate of a set of commuting Pauli operators. For |00>, for example, the two stabilizers are Z on qubit 0 and Z on qubit 1. A simulator stores these generators and updates them as gates act. Hadamard swaps X and Z, while S maps X to Y. CX can turn a Pauli on one qubit into a product involving both qubits.

This compact bookkeeping can represent superposition and entanglement without listing all 2^n amplitudes. That is why a circuit can look unmistakably quantum yet remain classically tractable. Qiskit Aer provides a stabilizer method for this class of circuits, as described in the IBM Quantum guide to stabilizer simulation.

A small Qiskit example

The circuit below uses H, S, CX, and computational-basis measurement, all within the Clifford family. It prepares a state with equal measurement probabilities for 00 and 11. The exact counts vary from run to run because the simulator samples measurement outcomes.

from qiskit import QuantumCircuit, transpile
from qiskit_aer import AerSimulator

qc = QuantumCircuit(2, 2)
qc.h(0)
qc.s(0)
qc.cx(0, 1)
qc.measure([0, 1], [0, 1])

simulator = AerSimulator(method="stabilizer")
compiled = transpile(qc, simulator)
result = simulator.run(compiled, shots=1000).result()
print(result.get_counts())

Where the shortcut stops

The T gate, defined by diag(1, e^(iπ/4)), is not Clifford. It does not preserve the Pauli description, so a circuit with T gates is outside the exact stabilizer method. Clifford gates together with T are universal for quantum computing, which is why non-Clifford gates matter. Other simulation methods can handle them, but the simple stabilizer guarantee no longer applies and the cost can rise as more T gates are added.

The practical lesson is useful when choosing test circuits. A large stabilizer circuit can be a strong check of hardware and software even though it is classically simulable. To test behavior beyond that boundary, include non-Clifford operations and use a simulator or device that supports them.