Hardware Smoke Test - Setup & Methods¶

Experiment ID: SMOKE-HW Workstream: S (Shadows) Executable: C:\Users\User\Desktop\Projects\QuartumSE\experiments\shadows\S_T01_ghz_baseline.py

Circuit Description¶

3-Qubit GHZ State¶

Identical to SMOKE-SIM for direct comparison:

|GHZ(3)⟩ = (|000⟩ + |111⟩) / √2

Circuit:

     ┌───┐
q_0: ┤ H ├──■────■──
     └───┘┌─┴─┐  │
q_1: ─────┤ X ├──┼──
          └───┘┌─┴─┐
q_2: ──────────┤ X ├
               └───┘

Depth: 3 Gates: 1 H + 2 CNOT Qubit Requirement: 3 connected qubits (linear or triangle topology)

Observable Set¶

Same as SMOKE-SIM: - Z Observables: ZII, IZI, IIZ (expected: 0.0) - ZZ Observables: ZZI, ZIZ (expected: 1.0)

Total: 5 Pauli strings

Backend Configuration¶

IBM Quantum Hardware¶

Selected Backend: ibm_fez (Nov 3, 2025) - Qubits: 156 (superconducting transmon) - Topology: Heavy-hex lattice - Queue Status: 77 pending jobs (low) - Calibration: 2025-11-03T13:17:32Z (fresh, < 1 hour)

Alternative Backends: - ibm_torino (133q, 485 pending) - used in Oct 22 smoke test - ibm_marrakesh (156q, 298 pending) - ibm_brisbane (127q, 3175 pending) - avoid due to queue

Qubit Selection¶

Use first 3 connected qubits (qubits 0-1-2 if linear connectivity available):

Quality Metrics (ibm_fez qubits 0-2): | Qubit | T1 (μs) | T2 (μs) | Readout Error | Gate Error (SX) | |-------|---------|---------|---------------|-----------------| | 0 | 63.6 | 49.7 | 0.98% | ~0.04% | | 1 | 174.8 | 199.1 | 2.22% | ~0.03% | | 2 | 208.9 | 178.7 | 0.77% | ~0.04% |

Two-Qubit Gates: - CZ(0,1): 1.08% - CZ(1,2): 1.08% (typical)

Assessment: Excellent qubit quality, suitable for smoke test.

Classical Shadows Configuration¶

v0 Baseline (No Mitigation)¶

shadow_config = ShadowConfig(
    shadow_size=100,              # Small budget for smoke test
    random_seed=42,               # Match SMOKE-SIM
    confidence_level=0.95,
    version=ShadowVersion.V0_BASELINE,
    apply_inverse_channel=False   # No noise correction
)

Rationale for Small Shadow Budget: - 100 snapshots minimizes queue time (< 10 min target) - Sufficient for qualitative validation (not precision) - Reduces cost if issues arise

Comparison to SMOKE-SIM¶

Parameter	SMOKE-SIM	SMOKE-HW
Backend	aer_simulator	ibm:ibm_fez
Shadow Size	500	100
Baseline Shots	1000	1000 (computed for SSR)
Mitigation	None	None
Random Seed	42	42

Key Difference: Reduced shadow budget (500 → 100) to minimize hardware runtime.

Mitigation Strategy¶

None for This Experiment

Smoke test uses v0 baseline to characterize raw hardware noise. Mitigation introduced in S-T02: - MEM (Measurement Error Mitigation): Confusion matrix calibration - ZNE (Zero-Noise Extrapolation): Planned for Phase 2 - Inverse Channel: v1 noise-aware shadows in S-T02

Hardware Execution Workflow¶

Step 1: Backend Selection¶

# Check backend status
quartumse runtime-status --json \
  --backend ibm:ibm_fez \
  --instance ibm-q/open/main

Output includes: - Queue depth (pending jobs) - Runtime quota remaining - Calibration timestamp

Decision: Proceed if queue < 200 jobs and calibration < 24 hours old.

Step 2: Execute Smoke Test¶

cd C:\Users\User\Desktop\Projects\QuartumSE

python experiments/shadows/S_T01_ghz_baseline.py \
  --backend ibm:ibm_fez \
  --variant st01 \
  --shadow-size 100 \
  --seed 42 \
  --data-dir ./data

Expected Execution: 1. Job Submission: Circuit transpiled and submitted to ibm_fez 2. Queue Wait: 0-30 minutes depending on queue depth 3. Execution: ~7-15 seconds for 100 shadow snapshots 4. Retrieval: Results fetched and processed 5. Manifest Save: Provenance captured with IBM calibration data

Step 3: Monitor Execution¶

# In separate terminal, tail logs if available
# Or use IBM Quantum dashboard to monitor job status

Backend Provenance Capture¶

IBM Calibration Snapshot¶

Manifest includes: - Calibration Timestamp: When backend was last calibrated - Properties Hash: SHA-256 of calibration data for versioning - T1/T2 Times: Per-qubit coherence times - Gate Errors: Single-qubit (SX, X, RZ) and two-qubit (CZ) error rates - Readout Errors: Per-qubit measurement fidelity - Topology: Qubit connectivity graph

Example manifest field:

{
  "backend_snapshot": {
    "name": "ibm_fez",
    "num_qubits": 156,
    "calibration_timestamp": "2025-11-03T13:17:32Z",
    "properties_hash": "a1b2c3...",
    "qubits_used": [0, 1, 2],
    "t1_times": [63.6, 174.8, 208.9],
    "t2_times": [49.7, 199.1, 178.7],
    "readout_errors": [0.0098, 0.0222, 0.0077],
    "gate_errors": {
      "sx_0": 0.000364,
      "cx_0_1": 0.01083,
      ...
    }
  }
}

Runtime Tracking¶

Manifest captures: - Submission Time: When job entered IBM queue - Execution Start: When quantum processor began executing - Execution Duration: Wall-clock time on hardware - Retrieval Time: When results returned to client

Smoke Test Metrics (Nov 3, 2025, ibm_fez): - Total execution: 7.82 seconds - Queue wait: < 5 minutes (low queue depth)

Data Storage¶

Same structure as SMOKE-SIM: - Manifest: data/manifests/{experiment_id}.json - Shot Data: data/shots/{experiment_id}.parquet - Console Logs: Captured in terminal output

Validation Checks¶

Automated validation in script: 1. Execution Success: Job completes without IBM errors 2. Result Retrieval: Counts dict populated with 2^3 = 8 bitstrings 3. Observable Estimation: All 5 observables have expectation values and CIs 4. Manifest Completeness: Backend snapshot non-null

Expected Warnings: - Observables may fall outside CIs due to hardware noise - SSR may be < 1.0× if noise dominates - Runtime may exceed estimate if queue saturated

Comparison to Simulator¶

Expected Degradation¶

Metric	SMOKE-SIM	SMOKE-HW (Expected)
ZZI Estimate	1.0000	0.85-0.95 (noise)
ZIZ Estimate	1.0000	0.85-0.95 (noise)
CI Coverage	100%	60-80% (wider CIs)
SSR	17.37×	1.1-2.0× (realistic)
Execution Time	8s	8-15s (overhead)

Noise Sources: - Gate Errors: ~1% for CNOTs, ~0.04% for single-qubit - Readout Errors: 0.77-2.22% across qubits 0-2 - Decoherence: T1=63-209 μs, T2=49-199 μs

Impact Estimate: - ZZ expectation reduced by 5-15% due to CNOT errors and T2 dephasing - Single-qubit Z observables less affected (no entanglement degradation)

Key Code Differences from SMOKE-SIM¶

Backend Descriptor¶

# SMOKE-SIM
backend_descriptor = "aer_simulator"

# SMOKE-HW
backend_descriptor = "ibm:ibm_fez"

Shadow Size¶

# SMOKE-SIM
shadow_size = 500

# SMOKE-HW
shadow_size = 100  # Reduced for speed

Expected Execution Time¶

SMOKE-SIM: < 1 second (local) SMOKE-HW: 5-30 minutes (queue + execution + retrieval)

Next Experiments¶

Upon successful completion: 1. S-T01: Increase shadow_size to 200-500, run ≥10 trials 2. S-T02: Add MEM, compare v0 vs. v1 on same ibm_fez backend 3. C-T01: Apply validated hardware access to H₂ chemistry experiment

Link to Executable Notebook¶

Interactive notebook: notebooks/experiments/shadows/smoke_test_hardware.ipynb [TBD]

For now, use standalone script: experiments/shadows/S_T01_ghz_baseline.py --backend ibm:ibm_fez