qiskit.ignis.verification.calculate_2q_epg

calculate_2q_epg(gate_per_cliff, epc_2q, qubit_pair, list_epgs_1q=None, two_qubit_name='cx')

Convert error per Clifford (EPC) into error per gate (EPG) of two qubit cx gates.

Given that a standard 2Q RB sequences consist of u1, u2, u3, and cx gates, the EPG of cx gate can be roughly approximated by $EPG_{CX} = EPC/N_{CX}$, where $N_{CX}$ is number of cx gates per Clifford which is designed to be 1.5. Because an error from two qubit gates are usually dominant and the contribution of single qubit gates in 2Q RB experiments is thus able to be ignored. If list_epgs_1q is not provided, the function returns the EPG calculated based upon this assumption.

When we know the EPG of every single qubit gates used in the 2Q RB experiment, we can isolate the EPC of the two qubit gate, ie $EPG_{CX} = EPC_{CX}/N_{CX}$ [1]. This will give you more accurate estimation of EPG, especially when the cx gate fidelity is close to that of single qubit gate. To evaluate EPGs of single qubit gates, you first need to run standard 1Q RB experiments separately and feed the fit result and gate counts to calculate_1q_epg().

import qiskit.ignis.verification.randomized_benchmarking as rb
 
# assuming we ran 1Q RB experiment for qubit 0 and qubit 1
gpc = {0: {'cx': 0, 'u1': 0.13, 'u2': 0.31, 'u3': 0.51},
       1: {'cx': 0, 'u1': 0.10, 'u2': 0.33, 'u3': 0.51}}
epc_q0 = 1.5e-3
epc_q1 = 5.8e-4
 
# calculate 1Q EPGs
epgs_q0 = rb.rb_utils.calculate_1q_epg(gate_per_cliff=gpc, epc_1q=epc_q0, qubit=0)
epgs_q1 = rb.rb_utils.calculate_1q_epg(gate_per_cliff=gpc, epc_1q=epc_q1, qubit=1)
 
# assuming we ran 2Q RB experiment for qubit 0 and qubit 1
gpc = {0: {'cx': 1.49, 'u1': 0.25, 'u2': 0.95, 'u3': 0.56},
       1: {'cx': 1.49, 'u1': 0.24, 'u2': 0.98, 'u3': 0.49}}
epc = 2.4e-2
 
# calculate 2Q EPG
epg_no_comp = rb.rb_utils.calculate_2q_epg(
    gate_per_cliff=gpc,
    epc_2q=epc,
    qubit_pair=[0, 1])
 
epg_comp = rb.rb_utils.calculate_2q_epg(
    gate_per_cliff=gpc,
    epc_2q=epc,
    qubit_pair=[0, 1],
    list_epgs_1q=[epgs_q0, epgs_q1])
 
print('EPG without `list_epgs_1q`: %f, with `list_epgs_1q`: %f' % (epg_no_comp, epg_comp))

EPG without `list_epgs_1q`: 0.016107, with `list_epgs_1q`: 0.013622

References

[1] D. C. McKay, S. Sheldon, J. A. Smolin, J. M. Chow, and J. M. Gambetta, “Three-Qubit Randomized Benchmarking,” Phys. Rev. Lett., vol. 122, no. 20, 2019 (arxiv:1712.06550).

Parameters

• gate_per_cliff (Dict[int, Dict[str, float]]) – dictionary of gate per Clifford. see gates_per_clifford().
• epc_2q (float) – EPC fit from 2Q RB experiment data.
• qubit_pair (List[int]) – index of two qubits to calculate EPG.
• list_epgs_1q (Optional[List[Dict[str, float]]]) – list of single qubit EPGs of qubit listed in qubit_pair.
• two_qubit_name (Optional[str]) – name of two qubit gate in basis gates.

Return type

float

Returns

EPG of 2Q gate.

Raises

QiskitError – when cx is not found, specified qubit_pair is not included in the gate count dictionary, or length of qubit_pair is not 2.