How Much Computing Power is Required for Full CI Calculation of a Molecule?

Click For Summary

Discussion Overview

The discussion revolves around estimating the computing power and processing time required for performing exact Full Configuration Interaction (Full CI) calculations on molecules, particularly focusing on the complexities introduced by the number of nuclei and electrons involved. The scope includes theoretical considerations and practical examples, such as the H2O molecule.

Discussion Character

  • Technical explanation
  • Exploratory
  • Debate/contested

Main Points Raised

  • One participant notes that an exact Full CI calculation requires an infinite number of molecular orbitals, leading to the conclusion that infinite computing power is necessary for systems with more than one electron.
  • Another participant suggests using a finite basis set, such as VQZ, and provides an example calculation involving 100 basis functions per atom, indicating that the number of Slater determinants becomes astronomically large, making calculations impractical.
  • A participant seeks clarification on how to quantify the computing power needed for Full CI calculations, particularly for a specific example involving an H2O molecule with defined parameters.
  • One participant discusses the relationship between the number of basis functions and the number of Slater determinants, proposing a formula to estimate the number of determinants based on the number of spin orbitals and electrons.
  • Another participant inquires about typical values for the number of basis functions for simple systems like the He molecule or H- ion and questions the scaling of processing time from a simple case to a more complex one like H2O.

Areas of Agreement / Disagreement

Participants generally agree that exact Full CI calculations are computationally intensive and often impractical for systems with multiple electrons. However, there is no consensus on specific values for basis functions or how to effectively estimate computing power for different molecules.

Contextual Notes

Limitations include the dependence on the choice of basis set and the assumptions made regarding the number of Slater determinants. The discussion does not resolve the complexities involved in estimating processing time or computing power for various molecular systems.

Big-Daddy
Messages
333
Reaction score
1
For a molecule of [itex]n[/itex] nuclei and [itex]x[/itex] total electrons, how can I work out how much computing power or processing time is required for an exact Full CI calculation for that molecule?
 
Chemistry news on Phys.org
Big-Daddy said:
For a molecule of [itex]n[/itex] nuclei and [itex]x[/itex] total electrons, how can I work out how much computing power or processing time is required for an exact Full CI calculation for that molecule?

For an exact full CI you need a complete MO basis which contains an infinity of molecular orbitals and you can form an infinity of Slater determinants from it. Hence you will need infinite computing power for more than one electron in general.
So you have to be less ambitious. E.g. you can chose a certain finite basis set, e.g. VQZ. Let's say you have 100 basis functions per atom (quite a big basis already) (this will also depend on the type of atom). If you have 10 atoms with 10 electrons each, you can form 100 out of 1000 Slater determinants (roughly 10^300) so this will still take almost forever.
 
Hang on, please clarify. Ok so as you've pointed out an exact solution will take infinite time, i.e. impossible, except for one-electron systems. If you have a certain basis set with k basis functions per nucleus, and n nuclei with x total electrons ... I still don't get how you are proposing to work out the amount of computing power needed? Thanks for the guidance. Can you do an example, maybe, let there be k basis functions, but now we're talking about an H2O molecule (n=3, x=10), how much processing power/time is needed?
 
Big-Daddy said:
Hang on, please clarify. Ok so as you've pointed out an exact solution will take infinite time, i.e. impossible, except for one-electron systems. If you have a certain basis set with k basis functions per nucleus, and n nuclei with x total electrons ... I still don't get how you are proposing to work out the amount of computing power needed? Thanks for the guidance. Can you do an example, maybe, let there be k basis functions, but now we're talking about an H2O molecule (n=3, x=10), how much processing power/time is needed?

So if you have k basis functions and x electrons, you have 2k spin orbitals. As each Slater determinant has to be made up from x different spin orbitals, you can form $$m=\begin{pmatrix} 2k\\x \end{pmatrix}\approx x^{2k}$$ different Slater determinants (you could classify them by spin and symmetry to reduce the number, but we are interested in the gross picture). So basically your Hamiltonian becomes an $$m \times m$$ matrix which you have to diagonalize. The matrix will be sparse as the two electron matrix elements vanish between most of the Slater determinants but still the size increases factorially with the size of the system.
 
Ok thanks.
So what value of k would we typically use for something very simple like a He molecule or H- ion?

And with the same k, let's say processing time on a certain computer is 10 seconds for a H- ion (2 electrons); will it then be 10!/2!*10 seconds for a (10-electron) water molecule?
 

Similar threads

Replies
3
Views
2K
  • · Replies 16 ·
Replies
16
Views
3K
Replies
0
Views
2K
  • · Replies 8 ·
Replies
8
Views
2K
  • · Replies 9 ·
Replies
9
Views
2K
  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 1 ·
Replies
1
Views
2K
  • · Replies 1 ·
Replies
1
Views
2K
  • · Replies 16 ·
Replies
16
Views
4K
  • · Replies 9 ·
Replies
9
Views
3K