A CPD-enabled low-scaling environment solver in a coupled cluster based static quantum embedding theory

We incorporate a canonical polyadic decomposition (CPD) based low-level solver as a means to accelerate the environment-level solver for the recently developed MPCC embedding framework. Using CPD, we both factorize the three dominant order-three density-fitting two-electron integral (DF TEI) tensors and develop a novel formulation that reduces the storage complexity of the low-level solver from ${O}(N^3)$ to $O(NR)$, where $R$ is the CPD rank, and the computational scaling of the most time-consuming contractions from ${O}(N^4)$ to ${O}(NR^2)$. We provide benchmarks on representative chemical environments, namely water clusters $\mathrm{(H_2O)n}$ with $n = 1$ to $6$ and linear alkane chains $\mathrm{C_nH{2n+2}}$ with $n = 1$ to $6$. For both test sets, using the CPD-compressed DF TEI tensors reproduces the DF reference convergence behavior of the low-level solver, the subsequent high-level step, and the fully self-consistent MPCC iterations, while introducing only small, rank-controlled shifts in absolute energies. At a fixed tolerance in the absolute MPCC energy, the CP ranks required for these tensor approximations increase linearly with system size. Chemically relevant energy differences are likewise preserved, as demonstrated for water-cluster dissociation energies and in a proof-of-concept embedding calculation of methane in a four-water cluster.

💡 Research Summary

This paper presents a novel acceleration strategy for the low‑level (environment) solver within the recently introduced MPCC (Multi‑Level Coupled Cluster) quantum embedding framework. MPCC partitions a molecular system into a chemically important fragment treated with a high‑level coupled‑cluster method (typically CCSD) and an environment treated with a cheaper perturbative approach, coupling the two via a self‑consistent downfolded Hamiltonian. While prior work incorporated the density‑fitting (DF) approximation to reduce the storage of the four‑index two‑electron integral tensor from O(N⁴) to O(N³), the remaining three‑index DF tensors and their repeated contractions still dominate memory usage and computational cost, scaling as O(N³) and O(N⁴) respectively.

The authors address this bottleneck by applying the Canonical Polyadic Decomposition (CPD) to the three‑index DF tensors. CPD expresses an order‑N tensor as a sum of R rank‑1 outer products, effectively factorizing the DF tensor J_Q^{a i b j} into three factor matrices A, K, and L:
J_Q^{a i b j} ≈ Σ_{s=1}^{R} A_{a s} K_{i s} L_{b s}.
Here R is the CP rank, which can be chosen to meet a desired accuracy. This factorization reduces the storage requirement from O(N³) to O(N R) (approximately O(N²) for modest R) and lowers the dominant tensor contraction cost from O(N⁴) to O(N R²) (approximately O(N³)). Importantly, the required CP rank grows only linearly with system size for a fixed energy tolerance, ensuring that the overall scaling remains favorable.

Algorithmically, the low‑level MPCC equations are rewritten to replace the DF tensors with their CPD approximations, eliminating the need to form intermediate three‑index objects. This redesign cuts both memory overhead and data movement, while the CPD accuracy can be systematically improved using standard alternating‑least‑squares (ALS) optimization. The method thus provides a single tunable parameter (the rank R) that controls the trade‑off between computational cost and precision.

Benchmark calculations were performed on two representative sets: water clusters (H₂O)ₙ with n = 1–6 and linear alkane chains CₙH₂ₙ₊₂ with n = 1–6. For both test families, the CPD‑compressed environment solver reproduces the convergence behavior of the original DF‑based solver, the subsequent high‑level fragment step, and the fully self‑consistent MPCC iterations. Energy differences of chemical relevance—such as water‑cluster dissociation energies and a proof‑of‑concept embedding of methane in a four‑water cluster—are preserved within a few kcal·mol⁻¹, demonstrating that the rank‑controlled approximation does not compromise chemical accuracy. The required CP ranks increase linearly with the number of monomers, confirming the anticipated scaling.

In summary, the integration of CPD into the MPCC environment solver dramatically reduces both storage and computational scaling while maintaining high fidelity of the embedded quantum chemical results. This advancement makes MPCC embedding viable for larger molecular systems (tens to hundreds of atoms) where the environment would otherwise be prohibitive. The authors suggest future extensions, including combination with tensor‑hypercontraction techniques, application to time‑dependent or higher‑order coupled‑cluster embeddings (e.g., CCSDT), and exploration of adaptive rank selection strategies to further enhance efficiency.