Quantum runtime for peaked circuit 49q x 5072 CZ's on heavy-hex

[PetarJurcevic]

Name

Quantum runtime for peaked circuit 40q x 5072 CZ's on heavy-hex

Circuit

peaked_circuit_heavy_hex_49x5072

Value

100

Method

Quantum

Method proof

Method proof

We run a classically verifiable peaked circuit (S. Aaronson arXiv:2404.14493) of size 49qubits and 5072 two-qubit gates. Circuits are constructed by BlueQubit as described in arXiv:2510.25838v1. The circuit consist of 3 unique 2Q-gate batched layers, A-B-C.

Circuit is executed on ibm_boston
with a quantum runtime of 11min 37s

Circuit is transpiled to qubits (see device graph and highlighted qubits)
initial_layout = [87, 88, 89, 90, 91, 92, 93, 94, 95, 97, 98, 99, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 117, 118, 119, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 137, 138, 139, 147, 148, 149, 150, 151, 152, 153, 154, 155]

Transpiler Settings:
transpile(
circuit,
backend=backend,
basis_gates= ['rx', 'rz', 'cz'],
initial_layout = best_layout,
optimization_level=3,
approximation_degree=1,
scheduling_method='alap')

Circuit Overhead:
'rz': 20257,
'rx': 10193,
'cz': 5072,
'delay': 4058,
'measure': 49,
'barrier': 10144,

Circuit duration:
3.049e-05 s

Execution settings:
sampler = Sampler()
sampler.options.dynamical_decoupling.enable = True
sampler.options.dynamical_decoupling.sequence_type = 'XpXm'
sampler.options.twirling.enable_measure = False
sampler.options.twirling.enable_gates = False
sampler.options.execution.rep_delay = 0.0001

Total number of shots
5'000'000

Custom 2Q-gate calibration scheme
We leverage the circuit's ABC-layer structure for gate calibration. The circuit consists of unique 2Q-gate layers, i.e. each layer A, B and C have a unique composition of which set of 2Q gates that run in parallel:

A: 87_88, 90_91, 92_93, 97_107, 98_111, 105_106, 109_110, 112_113, 118_129, 119_133, 127_128, 131_132, 138_151, 152_153
B: 87_97, 89_90, 91_98, 93_94, 95_99, 105_117, 107_108, 109_118, 111_112, 113_119, 114_115, 125_126, 127_137, 129_130, 131_138, 133_134, 135_139, 147_148, 149_150, 151_152, 154_155
C: 88_89, 91_92, 94_95, 99_115, 106_107, 108_109, 110_111, 113_114, 117_125, 126_127, 128_129, 130_131, 132_133, 134_135, 137_147, 139_155, 148_149, 150_151, 153_154

We use these unique sets of 2Q-gates to batch the calibration routine accordingly, and run the calibration for those gates in a single set in parallel. Below, we show the EPLG comparison that is run exactly in the A,B,C batching for default calibrations (orange) and the ABC-batched calibration (blue).
These improvements in the median 2Q gate fidelities on the same layout were crucial for peak-finding on these deeper circuits, compared to previous experimental results on the same device.

The first 20 entries in the sorted (by highest occurance) count ditionary are shared:
'1001100011100000001010100001011111111011110010110': 3,
'1000110001101010000010101001110001000101011010010': 1,
'0011110100110111110111110000110010001101100001001': 1,
'0101100101110101111010001100000101110011100010101': 1,
'1111111100010100011101110000110001011011111001110': 1,
'1010011001011000001111100001100111011101111100100': 1,
'0101010010111101100010100010010101000110011000110': 1,
'1011100111110101101110011010001100101100101011111': 1,
'0011101000010010001111110010001100111101001101011': 1,
'1101101101101011010100000111000100011010000010001': 1,
'1101000101000011001111000001000000101110111101110': 1,
'1100011110100110000010100000001011000011001111000': 1,
'0001000011011100000100000011101101001001011110110': 1,
'0011011001111111001101010111000011111001100100111': 1,
'1011110100100111010110101011001111101111011011110': 1,
'0010010111100100100101110101101101100101101101000': 1,
'0100000101100111001010101100011100110001101001001': 1,
'0010111011101001011100101000111111110111111101111': 1,
'1001111000101100110001101101101111010110000000001': 1,
'0011101011100100110000000101101101001001010110110': 1,

Authors

Petar Jurcevic

Institutions

IBM Quantum

Quantum runtime (seconds)

697

Classical runtime (seconds)

0

Compute resources (quantum)

ibm_boston Heron R3 + custom calibrations

Compute resources (classical)

N/A

Notes

No response

[PetarJurcevic]

in Case the method description is to large to fit, it could be moved to notes/comments and we could summerize the method as Quantum Sampling

[abhinavkandala]

The title and the name here say 40q. Could we correct that to 49q?

[n-e-sherman]

Overall, the submission is sufficiently detailed, and provides adequate information to allow for reproducing the results. I see one particular bitstring with a count of 3 while all others are 1, indicating strong evidence for properly identifying the peak. I accept the submission if the authors address, or adequately respond to, the following questions/comments:

• How many shots were used in the results? On a related note, was does the “Value = 100" mean? Is that the number of shots?
• Could you include more information about peaked circuits? Or possibly include a link like this one suggested by David Kremer: https://github.com/quantum-advantage-tracker/quantum-advantage-tracker.github.io/tree/main/data/classically-verifiable-problems/circuit-models/peaked_circuit
• Could you include the qasm, or a link to the qasm for the actual circuit used? Also, what about including the actual target bitstring for the circuit as well? I’m open to keeping that private, but would like to hear your thoughts.
• Are there any estimates for the classical runtime, or at least estimates, that can be used as a comparison against the quantum runtime?

[PetarJurcevic]

Eddit's made:

• added total number of shots
• added references to peaked circuits (S. Aaranson) and Circuit creation (BQ paper)
• added additional info about circuit structure

I believe value means how many correct bistrings were found. I think the circuit submitter should decide if they want the true solution to be public or not.

I believe my submission should be intrinsically linked to the qasm as I filed the submission under that particular circuit instance. Let me know if that is incorrect

[n-e-sherman]

The edits look great, and I accept this submission.

Good point about the true solution, we will handle that elsewhere.

[cinciolukasz]

The authors addressed all the issues. I accept the submission.