Orbital-Optimized Unitary Coupled Cluster for Indirect Nuclear Spin-Spin Coupling Constants within a Quantum Linear Response Framework

Indirect nuclear spin-spin coupling constants are crucial for predicting and interpreting Nuclear Magnetic Resonance (NMR) spectra, and this work presents an implementation of these constants within a quantum linear response framework using unitary coupled cluster methods. The method employs the unitary coupled cluster (UCC) ansatz and its orbital-optimized variant (ooUCC), both suitable for quantum computing implementations, to evaluate spin-spin coupling constants via qLR. Test calculations on five small molecules are compared with CASCI, CASSCF, and conventional CCSD results, showing that ooUCC yields spin-spin coupling constants comparable to classical methods.

The gist

Orbital optimization is important for accurate NMR coupling predictions within quantum-computing-friendly correlated methods.

Theoretical Framework and Operators

The calculation of indirect nuclear spin-spin coupling constants relies on relating the reduced indirect nuclear spin-spin coupling constant, KAB, to the trace of a spin-spin coupling tensor. This tensor can be calculated as the derivative of the (quasi-)energy E with respect to the perturbation (nuclear magnetic moments) evaluated at zero perturbation. The first-order perturbation in the Hamiltonian consists of three contributions:

1. The paramagnetic spin-orbit operator (PSO).

2. The Fermi contact (FC) operator.

3. The spin dipolar (SD) operator.

The second-order perturbation consists of the diamagnetic spin-orbit (DSO) operator, which is given by Equation 2.5 in the text, involving the two-electron integral eˆpqrs and one-electron integrals hpq and gpqrs. The calculation extends the qLR method to a triplet spin-adapted operator manifold necessary for the FC and SD contributions.

Wavefunction Ansatz: UCC vs. ooUCC

The active space approximation partitions the wavefunction into inactive, active, and virtual spaces, where only the active space needs to be simulated on quantum hardware. The unitary coupled cluster (UCC) ansatz is defined as:

((2.21)

UCC(θ)⟩ = e Tˆ1(θ)-Tˆ†1(θ)+Tˆ2(θ)-Tˆ2(θ)†Φ0⟩

The orbital-optimized UCC (ooUCC) ansatz further parametrizes the wavefunction with an exponential orbital rotation operator:

((2.26)

ooUCC(θ,κ)⟩ = e−κˆ(κ)UCC(θ)⟩

The parameters θ and κ are found by variational minimization using a quantum device, known as orbital-optimized VQE (ooVQE).

Quantum Linear Response Implementation

The calculation of the coupling constant in Equation 2.15 requires the linear response function, which is obtained as:

((2.30)

⟨⟨Aˆ; Bˆ⟩⟩ = -V[1]A†βB

The linear response column vector βB is the solution to the linear response equation:

((2.31)

E[2]βB = V[1]B, where E[2] is the electronic Hessian. The property gradient V[1]B is defined as Equation 2.35, which involves terms like ⟨0Gˆn, Bˆ0⟩ and orbital rotation operators between spaces.

Key Findings from Molecular Calculations

Test calculations on five small molecules were compared with CASCI, CASSCF, and conventional CCSD results. Key observations include:

1. For H2, the results for methods without orbital optimization (UCCSD and CASCI) are identical because double excitations are the highest order. The orbital-optimized methods (ooUCCSD and CASSCF) show nearly invariant results with respect to active space, with a difference of only 1.94 Hz between the smallest and full space, indicating a 0.7% difference.

2. For water (H2O), comparing the two orbital-optimized methods, CASSCF and ooUCCSD, shows they yield very similar results for both 2JHH and 1JOH couplings in the (6,5) active space.

3. The impact of orbital optimization is significant: ooUCCSD results are relatively insensitive to the active space choice and qualitatively agree with full-space CCSD even for small active spaces.

4. The breakdown of coupling constants into individual contributions reveals that the Fermi contact term is the most challenging to capture, and orbital rotations improve the paramagnetic spin-orbit, spin-dipolar, and Fermi contact terms. This demonstrates that orbital rotations improve the paramagnetic spin-orbit, spin-dipolar, and Fermi contact terms.

5. For CO, the poor agreement of UCCSD with CCSD is primarily due to "drastic differences in the PSO and FC contributions.

As a fastidious researcher, I have analyzed this paper, Orbital-Optimized Unitary Coupled Cluster for Indirect Nuclear Spin-Spin Coupling Constants within a Quantum Linear Response Framework. The key scientific contribution is developing a quantum linear response (qLR) framework using orbital-optimized unitary coupled cluster (ooUCC) methods to accurately compute indirect nuclear spin-spin coupling constants, which are crucial for NMR spectroscopy predictions.

Here are the specific improvements I propose for AI systems, categorized by the type of capability they would gain:

---

) 1. Enhanced Quantum Chemistry Simulation Accuracy (System-Specific Property Prediction)

The primary improvement is enabling AI systems to perform highly accurate, quantum-chemically grounded property predictions that are currently computationally prohibitive or inaccurate on classical hardware.

• This AI system can accurately predict the indirect nuclear spin-spin coupling constants for a wide range of small to medium molecules (e.g., H2, H2O, NH3, CH4, CO) with high fidelity (comparable to CCSD benchmarks).

• It can distinguish between different computational methodologies (UCCSD vs. ooUCCSD) and quantify the exact benefit of orbital optimization in mitigating systematic errors related to active-space truncation.

• The system can provide a detailed decomposition of the coupling constant into its fundamental physical contributions (Diamagnetic Spin-Orbit, Paramagnetic Spin-Orbit, Fermi Contact, and Spin Dipolar terms), allowing researchers to pinpoint which electronic effects dominate specific molecular couplings.

) 2. Robust Active Space Selection and Convergence Prediction

The paper demonstrates that the choice of active space significantly impacts results without orbital optimization (e.g., the large variation in JHH coupling in H2O).

• This AI system can be trained to predict the optimal active space configuration (number of electrons/orbitals) required to achieve a target level of accuracy (e.g., within 1% error margin) for a given molecular property calculation.

• It can predict the convergence behavior of coupled-cluster methods based on molecular properties, preventing wasted computational resources on sub-optimal active spaces.

) 3. Model Selection and Error Quantification

The AI system can move beyond simply providing a result to providing a rigorous assessment of the underlying theoretical model's reliability.

• It can compare the performance gap between truncated methods (UCCSD/ooUCCSD) and full-space reference methods (CASCI/CASSCF/CCSD) for specific chemical classes, quantifying the exact error introduced by truncation.

• For systems where truncation is insufficient (like CO), it can identify which specific physical terms (e.g., the Fermi Contact term) are most sensitive to model approximation, guiding further theoretical development.

) 4. Transfer Learning and Generalization of Correlated Methods

The use of orbital rotation operators allows the quantum method to be more robust across different molecular environments and basis sets than standard UCCSD.

• The AI can generalize the findings from small molecules (H2, H2O, CO) to predict coupling constants for larger or structurally diverse molecules where direct high-level classical calculations are infeasible.

• It can identify transferable orbital optimization strategies that improve accuracy across different types of molecular interactions (e.g., nonbonded vs. bond couplings).

In summary, this paper enables the development of an AI system capable of performing highly accurate, quantum-mechanically informed predictions for nuclear spin properties—a capability that is essential for advanced NMR spectroscopy modeling and fundamental studies in strongly correlated quantum chemistry.