Hi, folks!
I am studying the hyperpolarizability of a crystal using DFT calculations. I learned that Gaussian code is able to do that with POLAR keyword in the input file. As Gaussian is not professional at dealing with periodic structures, I abandoned the crystal box and took out the coordinates of its unit cell as well as 1*1*2，1*2*1, 2*1*1, 1*1*3 supercells for input structures of Gaussian. The calculations are performed at CAM-B3LYP/6-311++g(d,p) level. Below are some results of total beta in the unit of 10^-30 esu.
1*1*1: 33
1*1*2: 97
1*2*1: 78
2*1*1: 53
1*1*3: 199
As can been see, these values diff largely and do show any convergence. What is the correct way to simulate the crystal? How large size of crystal structure should be used?
Thank you very much in advance!
Dear Yin Li,
Your methodoly do not consider the periodic conditions necessary to describe properties and calculations of the crystalline systems. I think you could to use other softwarea like quantum espresso or Castep to do it.
Best wishes,
Bruno.
Hi, Bruno!
Thanks for your answer! I am also considering to use first-principle calculation codes include QE, VASP, Castep and CP2K as well. Do you have any suggestion that which code and settings like DFT functionals are more suitable for Second Order Hyperpolarizability?
Best,
Yin
Dear Yin Li,
The functional choices depends of your system. If you are working with a complex system, hybrid functional will have expensive computational cost. But, if is possible to work with it, you can use.
You can perform a geometry optimization using pure (GGA or LDA) functionals or hybrid (HSE06/PBE0) and calculate NLO properties. CASTEP have a specific keyword for this. You can you on the fly pseudopotentials, because it is more precise than conventional pseudopotentials.
You can contact me trhough my email [email protected] and we can discuss in more detailed way
- Badges
- Science topic
- Similar topics
- Social Science
- Media
- Journalism