### University of Heidelberg

## Large-Scale Green’s Function Calculations: The Parallel ADC(2) Method and Its Application for Weakly Bound and Metastable Anions

##### Wednesday, June 29, 2:00 PM

##### 307 Eberly Hall

**Overview:** The Cholesky Decomposition (CD) techniques has became one the key components to the computational efficiency of the P−RICDΣ program. The achieved efficiency comes, of course, at the the cost of accuracy. The advantage of CD techniques is that the introduced (absolute) error to the computed quantities (energies, properties) is rigorously controlled only by one parameter which can be choosen as small as needed. We verify the tolerance of ADC(2) to the Cholesky decomposition threshold and demonstrate the power of the P−RICDΣ program for a wide set of anions. This set covers all known kinds of bound anionic states; from weakly (dipole-, quadruple-, polarization-, and correlation-bound) to strongly (valence) bound. We found that computed electron affinities (EAs) are very robust with respect to the CD threshold and converge rapidly to numerically exact ones. Even moderate thresholds (≥ 10−5) is more than enough to achieve sub-milli-electron-volt accuracy and there is no need to use smaller ones. All CD schemes addressed in the literature are investigated. The accuracy of ADC(2) has been assessed by comparing the computed EAs with corresponding coupled-cluster singles and doubles (CCSD) ones. The key result is that the computationally much cheaper ADC(2) method performs extremely well in comparison with CCSD, a finding which is of practical relevance [1,2].

The talk is organized as follows:

- An error and efficiency analysis of CD schemes to ADC(2);
- Comparison between electron affinities computed byusing the ADC(2) and CCSD methods.

**References**:

- Thomas Sommerfeld, Bijay Bhattarai, Victor P. Vysotskiy, and Lorenz S. Cederbaum Journal of Chemical Physics 133, 114301 (2010)
- To be published

Sponsored by: Center for Molecular and Materials Simulations at the University of

Pittsburgh.