Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Dissociation energy for the NO molecule

  1. Nov 24, 2011 #1

    Does anyone know where I can find post-hartree-fock calculation benchmarks for dissociation energy of the NO molecule. Or even just the equilibrium energy of the NO molecule?

    I am looking for cc-pVTZ basis set results but I can only seem to find augmented basis set results.

  2. jcsd
  3. Nov 24, 2011 #2
    I did a geometry optimisation in qchem at CCSD(T)/cc-pVTZ level. Hope this helps.
    Z-matrix Print:
    1 N
    2 O 1 1.154418
    Final energy is -129.70273042654063
  4. Nov 28, 2011 #3
    That was fast! Thanks. How long did this calculation take? How many cores? It seems much faster than the method I was trying.
  5. Nov 28, 2011 #4


    User Avatar
    Science Advisor

    This calculation should be a matter of seconds on a single core.

    Note the horrific scaling of CCSD(T) (7th order). That means it gets very slow quickly for larger systems. That also means that it is very fast for small systems, like NO.

    As a side note: The performance of different programs differs widely for different methods and basis sets. For example, many programs cannot deal well with generally contracted basis sets (like cc-pVnZ), and then it can easily happen that one program is spending two days calculating integrals where another program does the same job in five minutes...
    So you generally want to use a program which is known to handle a certain kind of task well (e.g., Molpro is a good choice for coupled cluster calculations).
  6. Nov 28, 2011 #5
    I think this computer has 4 cores, but as cgk said this calculation is fast for small systems. I also recommend cfour for cupled cluster calculations, its free of charge for non-commercial users :) http://www.cfour.de/
  7. Nov 29, 2011 #6
    Interesting. Would either of you know how efficient a CCSDT, as opposed to CCSD(T) calculation would be? I ask because I have found a set of benchmarks, but I think they are using slightly tweaked basis sets (They optimise the polarisation exponent).


    Their result is half an eV off mine, while the result you reported is almost identical to mine (I used a CI based method).
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook