Calculations without density fitting

Is is possible to run calculations without using density fitting or cholesky decomposition? If possible, could you send me an input file as example?

Thanks very much
Antonio

Hi Antonio,
for SCF you can tweak the ERI algorithm with the scf_type keyword, see here
To use exact ERIs in SCF calculations, you can set the keyword to PK, OUT_OF_CORE or DIRECT. For example:

molecule NH3 {
N     -0.0000000001    -0.1040380466      0.0000000000
H     -0.9015844116     0.4818470201     -1.5615900098
H     -0.9015844116     0.4818470201      1.5615900098
H      1.8031688251     0.4818470204      0.0000000000
units bohr
}

set {
  basis STO-3G
  scf_type pk
}

You might also find this table in the manul useful.

I was benchmarking PK calculations on some recent (Intel Skylake-X on X299, and AMD Threadripper on X399) systems with different disk and memory configurations and found that SCF=PK (with Disk-Based J/K Matrices) speeds up nicely when given a faster disk; I saw almost linear scaling with sequential disk access speed. Higher RAM throughput (more memory channels) was also helpful. However, with my setup (Psi4 1.2a1 compiled with ICC and linked to Intel MKL), the scaling with threads was not nearly as impressive as for SCF=DF. Hope this helps.

Thank you all for the suggestions. It is running now.

Best wishes,
Antonio

@kalju The exact integral code has not nearly seen the attention the DF code has. There is quite a bit of room for improvement in both direct and PK unfortunately. Always happy to take PRs however!