22
$\begingroup$

Are there good benchmarks available for testing homemade Ewald summation code?

Optimally it would be very nice if there was several types of systems.

NIST has a very nice benchmark for energies of SPC/E water configurations NIST SPCE made up of 100, 200, 300, 750 water molecules and breaks the energy down into LJ, LJ long range correction, Ewald real, Ewald recipricol, Ewald self, Ewald intra.

However, it would be nice to also test Forces and pressure as well. It would also be nice to do this for flexible molecules.

While it is possible to generate a configuration in an open source package like Gromacs for instance, I am hesitant because there are so many unseen variables. The NIST data is very clear on what their parameters were and it takes no digging to ensure you use the same protocol. It doesn't help that it is well known that different packages, when given the same input parameters return different results, even if slightly different. A simple, purpose made testing code/database would be nice.

Improve this question
$\endgroup$
10
  • 2
    $\begingroup$ There is an professor at Jilin University (a true expert at Ewald sums) who was a postdoc at Columbia University with Bruce Berne. I will try to get in touch with him to answer your question. $\endgroup$
    – Nike Dattani
    May 4 '20 at 19:47
  • 4
    $\begingroup$ I got a response: "Dear Nike, Thanks for contacting me. Glad to hear that you are hosting a useful computational site, which is getting popular now. I will post a benchmarking data for SPC/E water before the weekend and will try my best to clarify some misleading arguments in the post." $\endgroup$
    – Nike Dattani
    Jun 11 '20 at 6:02
  • 1
    $\begingroup$ Good stuff, I look forwards to it. $\endgroup$
    – B. Kelly
    Jun 11 '20 at 19:48
  • 2
    $\begingroup$ I might be able to get the benchmarking input data we did with CASTEP when we implemented Ewald. Let me know whether this would still be useful. You might also be interested in this simple alternative to Ewald: arxiv.org/abs/1801.01070 $\endgroup$
    – Phil Hasnip
    Jun 28 '20 at 23:22
  • 1
    $\begingroup$ Please could you link to the benchmarking data for SPC/E water, when it is live? $\endgroup$
    – Phil Hasnip
    Jun 28 '20 at 23:23
10
$\begingroup$

I have looked for a similar set of benchmarks and haven't found one, unfortunately. Hopefully the data referred to in the comments comes through!

In the meantime, I have an MIT-licensed implementation of the Wolf-like alternative to Ewald that @PhilHasnip linked to. Here is the repository. Forces and stresses are available. You could try benchmarking against that, and I'd be happy to participate in generating a database.

Improve this answer
$\endgroup$
0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.