qp.transforms.decompositions.make_phase_gradient_distillation_decomp

make_phase_gradient_distillation_decomp(aux_wires, work_wires, mode='repeat-until-success')[source]

Create a decomposition rule for PhaseGradientStatePrep based on Jones distillation.

The rule implements the Fourier-state distillation protocol from Jones (2013) and uses only Clifford gates, SemiAdder, (Pauli) mid-circuit measurements, and classical control. The prepared state is approximate: the number of distillation rounds grows as \(\mathcal{O}(\log n)\) for an \(n\)-qubit register and targets an infidelity of at most \(\sin^2(\pi/2^n)\).

The output, auxiliary, and work registers require \(n\), \(n\), and \(n-1\) wires, respectively. All wires must be distinct and start in \(|0\rangle\). The decomposition uses the negative-phase convention of PhaseGradientStatePrep.

Parameters:
  • aux_wires (WiresLike) – auxiliary register with the same size as the output register, i.e. the target register of the PhaseGradientStatePrep operator to be decomposed.

  • work_wires (WiresLike) – at least \(n-1\) clean wires used by SemiAdder

  • mode (str) – "repeat-until-success" (default) or "postselect", see below

Returns:

the fixed decomposition rule

Return type:

pennylane.decomposition.DecompositionRule

Example

Register the rule and select it by name in qp.decompose() via fixed_decomps:

import pennylane as qp

registers = qp.registers({"grad": 3, "aux": 3, "work": 2})
rule = qp.transforms.decompositions.make_phase_gradient_distillation_decomp(
    registers["aux"], registers["work"]
)
gate_set = {
    "Hadamard", "S", "Adjoint(S)", "T", "Adjoint(T)", "CNOT", "X", "Z",
    "measure", "PauliMeasure",
}

qp.add_decomps(qp.PhaseGradientStatePrep, rule)

@qp.qjit(capture=True)
@qp.decompose(gate_set=gate_set, fixed_decomps={qp.PhaseGradientStatePrep: rule.name})
@qp.qnode(qp.device("lightning.qubit", wires=8))
def circuit():
    qp.PhaseGradientStatePrep(registers["grad"])
    return qp.state()

Only Clifford+T gates and (Pauli) mid-circuit measurements remain:

>>> sorted(qp.analyze(circuit)().resources.quantum_operations)
['Adjoint(S)', 'Adjoint(T)', 'CNOT', 'Hadamard', 'MidCircuitMeasure', 'PPM-w1', 'PauliX',
 'PauliZ', 'T']