Spectral amplification for ground-state energy estimation of electronic structure in first quantization

Technical Summary: Spectral Amplification for Ground-State Energy Estimation in First Quantization

Problem Statement
Simulating the electronic structure of chemical and material systems is a primary target for quantum algorithms. While most work focuses on second-quantized representations, first-quantized approaches offer an exponential space advantage by scaling memory as $O(\eta \log N)$ (where $\eta$ is the number of electrons and $N$ is the number of basis functions), compared to $O(N)$ for second quantization. This makes first quantization particularly suitable for targeting the large-basis-set limit and materials simulations requiring convergence to the thermodynamic limit.

However, despite favorable memory scaling, the gate complexity for ground-state energy estimation in first quantization remains prohibitively high when using large plane-wave basis sets. The primary bottleneck is the block encoding normalization constant ($\lambda$) required for Quantum Phase Estimation (QPE). Recent work utilizing compact basis sets has attempted to minimize this factor, but high gate costs persist. The challenge lies in reducing the normalization constant without introducing additional physical approximations or sacrificing the space advantage of first quantization.

Methodology
The authors propose employing the Sum-of-Squares Spectral Amplification (SOSSA) framework to improve the asymptotic gate complexity of ground-state energy estimation. The core of the methodology involves constructing an analytical Sum-of-Squares (SOS) representation of the first-quantized electronic structure Hamiltonian in a plane-wave basis.

1. SOS Representation: The Hamiltonian $H$ is shifted by a constant $\beta$ to form a non-negative operator $H + \beta I = \sum_\alpha O_\alpha^\dagger O_\alpha$. The authors identify a specific set of operators $O_\alpha$ generated by the total charge density operator.
   • The kinetic energy term is reproduced by operators $T_{j,w}$.
   • The potential energy (electron-electron and electron-nuclear) is reproduced by operators $\tilde{V}_\nu$, which correspond to the Fourier transform of the total charge density (electronic density minus nuclear density).
   • A self-interaction term arising from squaring $\tilde{V}_\nu$ is converted into an identity term (a constant shift) using an additional SOS generator $\tilde{X}_{j,\nu}$.
2. Block Encoding Construction: The authors compile a block encoding for the operator $H_{SOSSA} = \sum_\alpha O_\alpha$. This involves a nested structure of "outer" and "inner" prepare and select oracles.
   • To manage the differing summation structures of the SOS terms (where electron indices appear on the outside for kinetic energy but on the inside for potential terms), the authors introduce a direct-sum Hilbert space with a branch selection register.
   • They utilize oblivious amplitude amplification and specific overflow-flag qubits to handle boundary conditions (in-bounds vs. out-of-bounds momentum shifts) efficiently, avoiding complex conditional arithmetic.
3. Spectral Amplification: By using the square root of the SOS representation, the protocol constructs a walk operator with eigenphases $\phi_k = \cos^{-1}((2E_k + \beta)/\lambda_{SOS} - 1)$. The cost of QPE scales with an effective normalization factor $\lambda_{eff} = \sqrt{E_{gap}(\lambda_{SOS} - E_{gap})}$, where $E_{gap}$ is the gap between the ground state and the shifted bound.

Key Contributions

• Analytical SOS Construction: The paper provides the first analytical SOS representation for a first-quantized electronic structure Hamiltonian in a plane-wave basis that is implemented with costs comparable to prior Linear Combination of Unitaries (LCU) methods.
• Asymptotic Improvement: The authors derive a new scaling for the block encoding normalization. The effective normalization $\lambda_{eff}$ scales as $O(\eta \Delta^{-1.5} + \eta^{1.5} \Delta^{-1})$, where $\Delta$ is the grid spacing. This improves upon the prior state-of-the-art LCU scaling of $O(\eta \Delta^{-2} + \eta^2 \Delta^{-1})$.
• Resource Compilation: The paper details the compilation of the SOSSA block encoding, including specific optimizations for state preparation, selection oracles, and reflections, demonstrating that the overhead of the SOS framework is minimal compared to the LCU method.

Results
The authors provide detailed logical resource estimates for various benchmark systems, including molecular systems (ethylene carbonate, LiPF6), lithium metal supercells, and uniform electron gas (UEG) targets.

• Speedup Magnitude: The SOSSA method demonstrates a speedup over the prior LCU method ranging from 2$\times$ to 44$\times$ in total Toffoli gate counts.
• System Dependence: The speedup is most pronounced for larger, more realistic systems (e.g., the deuterium target shows a 44$\times$ improvement).
• Scaling Verification: Numerical analysis confirms the theoretical asymptotic scaling. In the continuum limit (kinetic energy dominated), the complexity reduces from $O(\eta \Delta^{-2})$ to $O(\eta \Delta^{-1.5})$. In the large-system limit (Coulomb interaction dominated), it reduces from $O(\eta^2 \Delta^{-1})$ to $O(\eta^{1.5} \Delta^{-1})$.
• Dominant Costs: The analysis identifies that the dominant cost in the algorithm remains the data movement (swapping electron registers to workspace registers), consistent with other first-quantized algorithms. However, the reduction in the effective normalization constant directly translates to fewer calls to the walk operator, yielding the observed speedup.

Significance
The paper claims that this work resolves the difficulty of high gate costs in first-quantized simulations without introducing new physical approximations. By leveraging the total charge density operator to generate a low-frustration SOS representation, the authors achieve a provable asymptotic advantage. This makes first-quantized algorithms a more viable choice for ab initio materials simulation, particularly for systems requiring large basis sets and convergence to the thermodynamic limit. The authors note that while the current implementation focuses on the Born-Oppenheimer approximation, the framework is generalizable to non-Born-Oppenheimer cases and future work will extend it to include pseudopotentials, a necessary step for practical applications.