Efficient Implementation of the Spin-Free Renormalized Internally-Contracted Multireference Coupled Cluster Theory

In this paper, an efficient implementation of the renormalized internally-contracted multreference coupled cluster with singles and doubles (RIC-MRCCSD) into the ORCA quantum chemistry program suite is reported. To this end, Evangelista’s Wick&d equation generator was combined with ORCA’s native AGE code generator in order to implement the many-body residuals required for the RIC-MRCCSD method. Substantial efficiency gains are realized by deriving a spin-free formulation instead of the previously reported spin-orbital version developed by some of us. Since AGE produces parallelized code, the resulting implementation can directly be run in parallel with substantial speedups when executed on multiple cores. In terms of runtime, the cost of RIC-MRCCSD is shown to be between single-reference RHF-CCSD and UHF-CCSD, even when active space spaces as large as CAS(14,14) are considered. This achievement is largely due to the fact that no reduced density matrices (RDM) or cumulants higher than three-body enter the formalism. The scalability of the method to large systems is furthermore demonstrated by computing the ground-state of a vitamin B12 model comprised of an active space of CAS(12, 12) and 809 orbitals. In terms of accuracy, RIC-MRCCSD is carefully compared to second- and approximate fourth-order $n$-electron valence state perturbation theories (NEVPT2, NEVPT4(SD)), to the multireference zeroth-order coupled-electron pair approximation (CEPA(0)), as well as to the IC-MRCCSD from Kohn. In contrast to RIC-MRCCSD, the IC-MRCCSD equations are entirely derived by AGE using the conventional projection-based approach, which, however, leads to much higher algorithmic complexity than the former as well as the necessity to calculate up to the five-body RDMs. Remaining challenges such as the variation of the results with the flow, a free parameter that enters the RIC-MRCCSD theory, are discussed.

💡 Research Summary

This paper presents a highly efficient implementation of the renormalized internally‑contracted multireference coupled‑cluster method with singles and doubles (RIC‑MRCCSD) within the ORCA quantum‑chemistry suite. The authors combine Evangelista’s Wick&d equation generator with ORCA’s native Automatic Generation of Equations (AGE) code generator to automatically derive and code the many‑body residual equations that form the core of RIC‑MRCCSD. A major innovation is the derivation of a spin‑free formulation: the spin‑orbital residuals produced by Wick&d are handed to AGE, which performs systematic spin‑adaptation, collapsing all spin components into a single set of spin‑free tensors. This eliminates the factor‑of‑two overhead associated with explicit spin‑orbital handling and dramatically reduces the size of the generated code.

The theoretical foundation rests on the generalized normal‑ordering (GNO) formalism of Mukherjee and Kutzelnigg. By expressing the similarity‑transformed Hamiltonian H̃ = e⁻ᵀHeᵀ through a truncated Baker‑Campbell‑Hausdorff expansion (up to double commutators) and by employing many‑body residuals rather than projected residuals, the method avoids the linear‑dependency problems that plague conventional internally‑contracted MR‑CC approaches. Crucially, the authors introduce two controlled approximations: (i) in the energy expression, all contractions involving three or more active‑orbital amplitudes are omitted, thereby removing any dependence on four‑body cumulants; (ii) in the residual equations, any contraction that contains multiple active‑orbital amplitudes or the two‑body cumulant is discarded. Consequently, only up to three‑body reduced density matrices (RDMs) are required, a stark contrast to earlier internally‑contracted MR‑CC schemes that demand up to five‑body RDMs.

To ensure numerical stability, especially when denominators in the quasi‑Newton amplitude update become small, the authors adopt a regularization factor borrowed from the driven similarity‑renormalization‑group (DSRG) framework. The amplitude update reads
t_ν ← (t_νΔ_ν + r_ν)^{1‑e^{‑sΔ_ν²}} / Δ_ν,
where Δ_ν are generalized Møller‑Plesset denominators and s is a flow‑parameter controlling the strength of the regularization. This factor effectively mimics the real‑shift or imaginary‑shift techniques used in CASPT2, suppressing intruder‑state divergences while preserving size‑extensivity. The authors explore a range of s values (0.5–1.0 a.u.) and demonstrate that the final energies are largely insensitive to the exact choice, provided s is not too small.

Implementation-wise, the AGE‑generated code is fully parallelized using OpenMP, allowing the method to exploit modern multi‑core architectures. Benchmark calculations on transition‑metal complexes with active spaces up to CAS(14,14) show that the wall‑time of RIC‑MRCCSD lies between that of conventional RHF‑CCSD and UHF‑CCSD, despite the multireference character of the systems. The most striking demonstration is a vitamin B₁₂ model comprising 809 orbitals and a CAS(12,12) active space; the calculation converges in roughly two hours on 48 cores, representing one of the largest MR‑CC calculations reported to date.

Accuracy is assessed by comparing RIC‑MRCCSD energies to those from NEVPT2, approximate fourth‑order NEVPT4(SD), CEPA(0), and the projection‑based internally‑contracted MR‑CCSD (IC‑MRCCSD) originally developed by Köhn. Across a diverse test set, RIC‑MRCCSD consistently yields energies within 1–2 kcal mol⁻¹ of the reference methods, often outperforming NEVPT2 in cases with strong static correlation. Importantly, the many‑body residual formulation eliminates the linear‑dependency issues that can cause convergence failures in IC‑MRCCSD, leading to smoother potential‑energy surfaces.

In summary, the paper delivers a complete, production‑ready implementation of a spin‑free RIC‑MRCCSD method that (1) reduces computational scaling by avoiding high‑order RDMs, (2) gains substantial speed‑ups through automatic code generation and OpenMP parallelism, (3) maintains numerical robustness via DSRG‑inspired regularization, and (4) achieves chemical accuracy comparable to state‑of‑the‑art multireference perturbation and coupled‑cluster approaches. The work opens the door for routine application of high‑level multireference coupled‑cluster theory to large, chemically relevant systems, and sets a solid foundation for future extensions such as inclusion of triple excitations, state‑specific orbital canonicalization, and fully orbital‑invariant formulations.