Native Gate Sets

The small collections of physically executable gates that define what each quantum hardware platform can do directly.

What native means

A native gate is one the hardware performs directly with a calibrated control pulse, without decomposition. Every other gate must be compiled into these. The native set is dictated by the physics of the platform, so different technologies expose very different basis gates.

Platform examples

• Superconducting (transmon): virtual-Z, √X (SX), RZ single-qubit; CZ, CNOT, or a tunable-coupler entangler
• Trapped ion: single-qubit rotations plus the Mølmer-Sørensen / RXX entangler, sometimes global multi-qubit MS
• Neutral atom: single-qubit rotations plus a Rydberg-blockade controlled-phase (CZ)
• Semiconductor spin: single-qubit ESR/EDSR rotations plus √SWAP exchange

Universality requirement

A native set must be universal: single-qubit rotations covering the Bloch sphere plus at least one entangling two-qubit gate. Any such set can approximate any circuit, though the efficiency of compilation varies. Choosing which entangler to build is a hardware design decision with large downstream compilation consequences.

Virtual-Z and the SX basis

Superconducting machines exploit that Z-rotations can be applied for free as phase-frame shifts. Combined with a single fixed √X pulse, this yields the RZ-SX-RZ-SX-RZ decomposition of any single-qubit gate — only two physical pulses per qubit operation, with all the rotation angle carried by free virtual-Z gates.

# any single-qubit U as RZ-SX-RZ-SX-RZ (superconducting native)
# angles come from an Euler ZXZXZ factorization
def native_1q_pulse_count():
    return 2  # two SX pulses; the three RZ are virtual

Cross-platform compilation

Because entanglers differ, a circuit written in CNOT is retargeted to MS on ions or CZ on atoms by KAK-based resynthesis, which finds the minimal native-entangler count for each two-qubit block. See KAK decomposition and transpilation.