Hamiltonian ID
Sn2O2_CAS_6_6
Hamiltonian file
analyze_blind_sqd.zip
Hilbert space size
400
Problem category
Baseline benchmarks
Problem description
Ground-state energy of Sn2(mu-O)2 in a CAS(6,6) active space, recovered BLIND on ibm_fez via sample-based quantum diagonalization (LUCJ ansatz, parameters from classical CCSD t2). The comparison targets were cryptographically sealed before the device ran (sha256 9963553e..., E_HF/E_FCI/E_noiseless committed). Hardware E = -156.2719 Ha: beats Hartree-Fock by 4.04 mHa (P=0.999 bootstrap), sits 0.53 mHa above FCI -- a valid variational upper bound, within chemical accuracy. The recovery comes from which determinants the device sampled, not from any energy supplied to it (a random-configuration null stays above HF). Honest placement: the hardware energy lands 0.65 mHa below the noiseless SQD point, within shot noise, still above FCI -- statistical variance, not a variational violation. NOT a quantum-advantage claim: CAS(6,6) is ~400 determinants, classically tractable; the claim is blind recovery with a sealed target, not speedup. Full method + reproduction: https://github.com/sharadbachani-oss/blind-sqd-sn2o2-66 (METHOD.md). Raw counts (job d8elna1vjngc73aojnp0) available on request.
Name
Sn2(mu-O)2 CAS(6,6) blind SQD, ibm_fez (Merlin Quantum)
Qubits
12
Gates
~27 (LUCJ k=1)
Energy (Eh)
-156.2719
Low error bound (Eh)
-0.00015
High error bound (Eh)
+0.00015
Method
Blind SQD (sealed pre-registration)
Method proof
https://github.com/sharadbachani-oss/blind-sqd-sn2o2-66
Quantum runtime (seconds)
see job d8elna1vjngc73aojnp0
Classical runtime (seconds)
120
Compute resources (quantum)
IBM Heron r2 (ibm_fez), 4000 shots
Compute resources (classical)
single CPU core (PySCF CCSD + CI diagonalization)
Notes
Blind: target sealed pre-run; beats HF; above FCI
Authors
Suhail Bachani
Institutions
Merlin Quantum
Hamiltonian ID
Sn2O2_CAS_6_6
Hamiltonian file
analyze_blind_sqd.zip
Hilbert space size
400
Problem category
Baseline benchmarks
Problem description
Ground-state energy of Sn2(mu-O)2 in a CAS(6,6) active space, recovered BLIND on ibm_fez via sample-based quantum diagonalization (LUCJ ansatz, parameters from classical CCSD t2). The comparison targets were cryptographically sealed before the device ran (sha256 9963553e..., E_HF/E_FCI/E_noiseless committed). Hardware E = -156.2719 Ha: beats Hartree-Fock by 4.04 mHa (P=0.999 bootstrap), sits 0.53 mHa above FCI -- a valid variational upper bound, within chemical accuracy. The recovery comes from which determinants the device sampled, not from any energy supplied to it (a random-configuration null stays above HF). Honest placement: the hardware energy lands 0.65 mHa below the noiseless SQD point, within shot noise, still above FCI -- statistical variance, not a variational violation. NOT a quantum-advantage claim: CAS(6,6) is ~400 determinants, classically tractable; the claim is blind recovery with a sealed target, not speedup. Full method + reproduction: https://github.com/sharadbachani-oss/blind-sqd-sn2o2-66 (METHOD.md). Raw counts (job d8elna1vjngc73aojnp0) available on request.
Name
Sn2(mu-O)2 CAS(6,6) blind SQD, ibm_fez (Merlin Quantum)
Qubits
12
Gates
~27 (LUCJ k=1)
Energy (Eh)
-156.2719
Low error bound (Eh)
-0.00015
High error bound (Eh)
+0.00015
Method
Blind SQD (sealed pre-registration)
Method proof
https://github.com/sharadbachani-oss/blind-sqd-sn2o2-66
Quantum runtime (seconds)
see job d8elna1vjngc73aojnp0
Classical runtime (seconds)
120
Compute resources (quantum)
IBM Heron r2 (ibm_fez), 4000 shots
Compute resources (classical)
single CPU core (PySCF CCSD + CI diagonalization)
Notes
Blind: target sealed pre-run; beats HF; above FCI
Authors
Suhail Bachani
Institutions
Merlin Quantum