MergeOrbitalRotations¶

class MergeOrbitalRotations¶

Bases: GenericPass[DAGCircuit, DAGCircuit]

A transpilation pass merging runs of consecutive orbital rotations into a single one.

Two OrbitalRotation gates applied back-to-back on the same modes compose into a single orbital rotation whose rotation_unitary is the matrix product of the two. This pass detects maximal runs of such consecutive rotations in a FermionicDAGCircuit and rewrites each into one OrbitalRotation, so the synthesis stage lowers a single (still square) decomposition instead of one per gate in the run.

A “run” is a maximal chain of OrbitalRotation nodes acting on the same mode set with nothing else intervening on those wires. Following the OrbitalRotation convention \(a^\dagger_i \mapsto \sum_j U_{ji} a^\dagger_j\), applying a rotation \(U_1\) and then a rotation \(U_2\) maps \(a^\dagger_i \mapsto \sum_j (U_2 U_1)_{ji} a^\dagger_j\), so the run’s rotations are multiplied in circuit order (later rotation on the left) to form the merged rotation_unitary. Because the merged gate implements exactly that composed basis change, the rewrite leaves the simulated state vector unchanged.

Only rotations on the identical mode set are merged: two rotations on different (even overlapping) mode sets, or rotations separated by any other operation on a shared wire, break the run and are left untouched. This is the shape produced, e.g., by two adjacent per-spin rotations on the same spin half, or by consecutive UCJ layers whose trailing and leading orbital rotations meet on the same modes.

Caution

This is an early development prototype. Beware of changes to its interface without warning during the pre-release development of this package.