arXiv:2610.37840
Ground-state energy estimation compiled on cat qubits: 7 164 qubits for FeMoco-54 in 23.5 hours
Larissa Kroell, Hugo Jacinto, Victor Hertz, Axel Pappalardo, Adel Ben Moussa, Manuel Rispler, Peter Shanahan, Élie Gouzien, Christophe Vuillot, Vivien Londe
Abstract
We provide a detailed resource estimate of the state-of-the-art sum-of-squares block encoding algorithm for electronic structure Hamiltonians of Low et al. on superconducting cat qubits. To achieve this, we present compilations with closed-form costs of the dominant subroutines such as skew-tree Qrom, a Swap network and a Phaseup for both a repetition-cat and an LDPC-cat architecture. Our compilation represents algorithmic qubits in the $X$ basis, taking advantage of the asymmetric cost of $X$ and $Z$ access that the concatenation of cat qubits with phase-flip classical code brings. Additionally, we provide a protocol for $\vert i \rangle$-state distillation and a two-level Toffoli-like $\vert \text{C}_X\text{C}_X X \rangle$ magic state distillation protocol, including three factory layouts. To minimize the cost of Clifford corrections induced by magic state injections we tailor injection gadgets to each subroutine. Combining our compilation work with the adder compilation from the companion work Pappalardo et al., we provide global layouts for the block encoding in both architectures which we use to obtain resource estimates for ground-state energy estimation for the benchmarks of FeMoco-54 and FeMoco-76. Assuming a cycle time of 1$μ$s, a reaction time of 10$\mathrmμ$s, a phase estimation precision of 1 mHa and circuit-level phase-flip noise with probability $p_Z = 5 \times 10^{-4}$ with negligible bit-flip error, FeMoco-54 requires fewer than $7\,500$ cat qubits with an expected runtime of under one day for a single phase estimation run using an LDPC-cat architecture. Relaxing the bit-flip requirement to $p_X = 10^{-11}$ with an additional layer of elevator codes raises this to an estimated $13\,000$ cat qubits.
AI Summary
The paper estimates the resources for electronic-structure ground-state energy estimation on repetition-cat and hybrid LDPC-cat architectures. It compiles an existing DFTHC sum-of-squares block encoding with spectral amplification, using previously determined Hamiltonian parameters rather than introducing a new chemistry algorithm. Algorithmic qubits are represented in the $X$ basis to exploit cheaper logical $X$ access. The compilation provides closed-form costs and layouts for skew-tree QROM, controlled-SWAP networks and phase correction, together with $|i\rangle$ distillation and two-level Toffoli-like magic-state factories. These primitives, combined with an adder compilation from companion work, yield end-to-end layouts and space–time trade-offs. Circuit-level simulations support the distillation protocols and LDPC logical-error models; the latter extrapolate from Monte Carlo data using a most-likely-error decoder. With $1\,\mu\mathrm{s}$ cycles, $10\,\mu\mathrm{s}$ reaction time, $1$ mHa phase-estimation precision, $p_Z=5\times10^{-4}$ and negligible bit flips, FeMoco-54 requires 7,164 cat qubits and an expected 23.5 hours per phase-estimation run, consuming approximately $2.01\times10^9$ Toffolis. FeMoco-76 requires 7,818 qubits and 3.31 days. Larger compute zones reduce runtime; LDPC-cat uses 36–38% fewer qubits than repetition-cat in the reference configurations. The estimates assume serial Toffolis, one factory and fixed state-preparation choices. Neglecting bit flips requires extremely low physical $p_X$; an additional elevator-code layer at $p_X=10^{-11}$ raises the FeMoco-54 LDPC estimate to approximately 12,700 qubits, but this is a proportional-overhead extrapolation, not a complete compilation. FeMoco is treated as a benchmark, not a quantum-advantage demonstration.
Error found: The paper gives incompatible estimates, approximately 10^-18 and 10^-16, for the physical bit-flip rate required when no active bit-flip protection is used. The spin-combined one-electron OuterPrep weight should be sqrt(2 mu), whereas the block-encoding description specifies 2 sqrt(mu), inconsistent with its own SOS normalization and formula for Lambda.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 81 pages, 54 figures, 14 tables