Clifford + T Universality

Clifford gates alone are efficiently simulable on a classical computer; adding the T gate makes the set universal -- the basis of nearly every fault-tolerant gate construction.

Two tiers of gates

The Clifford group is generated by the Hadamard, phase (S), and CNOT gates. Cliffords normalize the Pauli group -- they map Pauli operators to Pauli operators -- which gives them a clean stabilizer description.

Gottesman-Knill: Cliffords are not enough

The Gottesman-Knill theorem shows that a circuit built only from Clifford gates, stabilizer-state inputs, and computational measurements can be simulated efficiently on a classical computer. So Cliffords alone give no quantum advantage; something more is required.

Adding the T gate

The T gate (a pi/8 Z rotation) is the standard non-Clifford ingredient. Clifford + T is universal: any unitary can be approximated to arbitrary precision, with the Solovay-Kitaev theorem bounding the overhead. Almost all fault-tolerant hardware targets exactly this gate set.

Why fault tolerance cares

In most error-correcting codes the Clifford gates are cheap (often transversal), while the T gate is expensive -- it typically requires magic-state distillation. As a result the T-count of an algorithm is the dominant cost metric for fault-tolerant resource estimates. Kronos's KQROSS estimator counts T-gates for exactly this reason when it computes when a quantum computer could beat a classical method for a fusion kernel.