Skip to content

v4.8.0: Frontier Quantum Control & Control Compiler v2

Choose a tag to compare

View all tags
@devinphillipdavis devinphillipdavis released this 05 Nov 05:39
· 92 commits to main since this release
v4.8.0
9cadb05

OSIRIS v4.8.0 — Sovereign Dual-Substrate Architecture: Frontier Quantum Control & Control Compiler v2

Release Tag: v4.8.0
Target Commit: 9cadb05ae57a4a6d20db18fa7019ade9f9840c92
Permanent Zenodo Archive: 10.5281/zenodo.23969941 (Record #23969941)
Sovereign Ledger Receipt: Block #15 (038bb10ea98a9fae9d6d3dc3a0be1ee8242445831dbf08c3653139366df04724)

Executive Summary

OSIRIS release v4.8.0 delivers the physical and algorithmic maturation of the quantum control substrate within the Sovereign Dual-Substrate Architecture. This milestone resolves the OSIRIS vs. Qiskit dynamical decoupling benchmark through three-engine numerical simulation, formalizes the empirical lessons of the held-out Experiment E2 campaign into OSIRIS Control Compiler v2, and confirms its prospective superiority on physical IBM Heron r2 hardware (ibm_marrakesh, Job dbm00gd8mdtc73d6vjcg).

Key Breakthroughs & Empirical Deliverables

1. Resolution of the OSIRIS vs. Qiskit Benchmark

  • Toggling-Frame Bisection: Confirmed across three independent numerical engines (analytical toggling frame, SciPy matrix exponentiation, and Qiskit Aer) that ideal ZZ spectator crosstalk refocusing is governed by signed exposure time cancellation ($A_{uv} = \int_0^T s_u(t)s_v(t),dt = 0$).
  • Parity Under Matched Staggering: Proved that when Qiskit is configured with matched sublattice staggering, both engines achieve identical ideal refocusing ($\langle XX \rangle = 1.000000$, numerical agreement within $3.93 \times 10^{-7}$ in Aer and $1.58 \times 10^{-12}$ in SciPy ODE integration).
  • Control vs. SDK: Formally established that the advantage lies in hardware-aware scheduling policies rather than native SDK limitations, correcting previous unsigned metric anomalies.

2. Forensic Post-Mortem of Prospective Experiment E2

  • Empirical Falsification: Documented the transparent falsification of Adaptive Scheduler v1.0 across 4 held-out physical couplers on ibm_marrakesh (mean difference $-0.28\text{ pp}$ relative to fixed 80 ns stagger).
  • Physical Root Cause: Identified that single-qubit dephasing thresholds caused improper fallback to synchronous control, failing to account for bilateral partner dephasing under CPMG breakdown.
  • Cluster-Robust Re-Analysis: Re-evaluated with cluster-robust standard errors ($G=4$ couplers, $df=3$), establishing that Fixed 80 ns stagger beat synchronous XY8 by $+0.446 \pm 0.113\text{ pp}$ ($p_{\text{2s}} = 0.0289$).

3. OSIRIS Control Compiler v2 (Risk-Aware Graph Optimization)

  • Surrogate Objective: Implemented $J(s) = \hat{\mathcal{F}}(s) - \lambda \sigma_{\mathcal{F}}(s) - \gamma C(s)$, balancing mean predicted fidelity against bilateral parameter uncertainty and multi-body spectator risk.
  • Open-System Lindblad Dynamics: Evaluated under 6-arm open-system Master equation simulations across all 14 ibm_marrakesh coupler topologies ($28$ state-evolution cells with finite pulse width $\tau_p = 36\text{ ns}$, detuning $\delta\nu$, $T_1/T_2$, and filter leakage):
    • OSIRIS v2 vs. Fixed STAG-80: $+1.825\text{ pp}$ ($p_{\text{2s}} = 0.0004$)
    • OSIRIS v2 vs. Synchronous XY8: $+2.356\text{ pp}$ ($p_{\text{2s}} = 0.0069$)
  • 5-Qubit Heavy-Hex Graph Optimization: On a connected 5-qubit chain ($Q_{12}\text{--}Q_{19}$), localized shifting of central vertex $Q_{15}$ outperformed uniform staggering by $+2.24\text{ pp}$.

4. Prospective Hardware Validation on IBM Heron r2 (Experiment E3 Pilot)

  • Execution Environment: ibm_marrakesh (Heron r2), Job ID dbm00gd8mdtc73d6vjcg.
  • Workload: 216 ISA-compiled circuits, 221,184 physical shots across 3 held-out physical couplers ($Q_{15}\text{--}Q_{19}$, $Q_{32}\text{--}Q_{31}$, $Q_{35}\text{--}Q_{34}$) over 3 independent acquisition blocks ($K=3$).
  • Empirical Confirmation:
    • OSIRIS Compiler v2: $78.529%$ mean Bell-state fidelity.
    • Fixed STAG-80: $77.878%$ mean Bell-state fidelity.
    • Synchronous XY8: $78.003%$ mean Bell-state fidelity.
    • Advantage over Fixed STAG-80: $+0.651 \pm 0.182\text{ pp}$ ($t = 3.585, df = 2, p_{\text{1s}} = \mathbf{0.0348}$, positive in 3 of 3 blocks).
    • Advantage over Synchronous XY8: $+0.526 \pm 0.116\text{ pp}$ ($t = 4.536, df = 2, p_{\text{2s}} = \mathbf{0.0454}$, positive in 3 of 3 blocks).

5. Standalone Compiler Performance Benchmarks

Evaluated across 5 heavy-hex circuit workloads ($n = 2\text{--}12$ qubits, depth $4\text{--}16$ cycles):

  • Compilation Speed: $6.80\times$ mean speedup vs. Qiskit Opt-1 + PadDynamicalDecoupling; $13.87\times$ speedup vs. Qiskit Opt-3 + Staggered PadDynamicalDecoupling.
  • Memory Footprint: $3.71\times$ lower peak memory allocation.
  • Unitary Equivalence: Exactly verified against standard statevectors with error $|1 - \mathcal{F}_{\text{Statevector}}| \le 3.3 \times 10^{-15}$.

Release Artifacts & Cryptographic Checksums

Algorithm: SHA-256
----------------------------------------------------------------------------------------------------
411f41c89365f6e1677b551f51b6e3a6cd130e95fd66085e221397b1142cb72b  osiris_v4.8.0_frontier_quantum_control_release.zip
55efbd9b04ab6488730fc6e7db5c5be3e7e78ee2b24c6ef9149021626a5750d4  osiris_v4.8.0_frontier_quantum_control_release.tar.gz
d7ba8de360d363b2b494b0db844098eadd141501c012c75a987d92ecc04eeb27  SOVEREIGN_DUAL_SUBSTRATE_MONOGRAPH.pdf
07f6e07c33dc6eb273f3c4dbce395d8869151c89e13df9aeeb26590c689bfe15  SOVEREIGN_DUAL_SUBSTRATE_MONOGRAPH.md
a79f04263db54e58b16fa05d8bc6028ea28cf7e7aa3f0f9076bc11b988f63567  zenodo_deposit_v4.8.0.json
32a7e78d9b152d1ea3caad4d49d977e5d8ec1d782ea255e2c56aeb632f7823f6  MANIFEST_SHA256_v4.8.0.txt
----------------------------------------------------------------------------------------------------

How to Verify & Reproduce

# 1. Clone repository and checkout v4.8.0
git clone https://github.com/osiris-dnalang/osiris-cli.git
cd osiris-cli
git checkout v4.8.0

# 2. Verify release bundle manifest
sha256sum -c dist/MANIFEST_SHA256_v4.8.0.txt

# 3. Reproduce the 3-engine benchmark
python3 experiments/experiment_e/bench_ideal_bipartite_dd.py

# 4. Reproduce the open-system Lindblad simulation
python3 experiments/experiment_e/bench_compiler_lindblad.py

# 5. Reproduce the compiler performance benchmark
python3 experiments/experiment_e/bench_compiler_performance.py
Assets 7
Loading