Real-Time Coupled Cluster Theory with Approximate Triples

Area of Science:

• Computational chemistry
• Quantum chemistry
• Theoretical chemistry

Background:

• Coupled cluster (CC) methods are essential for accurate electronic structure calculations.
• Real-time (RT) methods capture dynamic properties beyond static approximations.
• High levels of electron correlation pose challenges for computational efficiency.

Purpose of the Study:

• To develop and validate a time-dependent CC3 (RT-CC3) method.
• To investigate the impact of electron correlation on real-time coupled cluster formalism.
• To assess the computational performance and accuracy of the new method.

Main Methods:

• Introduced a time-dependent implementation of the CC3 singles, doubles, and approximate triples method.
• Incorporated triples into existing CCSD equations, resulting in N^7 scaling.
• Utilized graphics processing unit (GPU) acceleration for computational speedup.
• Compared single-precision with double-precision arithmetic.

Main Results:

• RT-CC3 demonstrates validity for frequency-dependent properties.
• GPU acceleration achieved speedups up to a factor of 13 for water clusters.
• Single-precision arithmetic showed minimal impact on polarizabilities but increased error for hyperpolarizabilities.
• RT-CC3 results showed low percentage errors (<0.1% for polarizabilities, <1% for hyperpolarizabilities) compared to linear response CC3.

Conclusions:

• The RT-CC3 method provides an accurate and efficient approach for studying electron correlation.
• GPU acceleration significantly reduces computational cost.
• The method is suitable for investigating dynamic properties and orbital-optimization effects in electronic systems.