escf module: "Switching to fine grid"

[Matteo Guglielmi]

When I compute excited states (rpa) usually I have the following
most relevant settings:

Code Select Expand


...
$dft
   functional b3-lyp
   gridsize   m4   \*My system has more than 70 atoms*\
$scfconv   7
$denconv 1d-7
$soes
a           5
$rpacor  15000
...


dscf > dscf.out && escf > escf.out &

My questions are:

1) do I have to worry about the "NOTE" message on the mgrid
option I always see in the escf.out file?

2) how does the code work?  Does it try to converge the requested
roots twice?

Code Select Expand


...
<get> : LAST BUFFER REACHED --> switching to 'direct' mode !
read in  (=     64) + newbuffer  >  max (=     64)
   13       a        5        9.766471718504769D-06


converged!


Switching to fine grid

preparing numerical integration ....
          Overall gridpoints after grid construction =        742954

Calculating ground state density on molecular grid

Integral of ground state density: N =   364.0000318735833



Iteration IRREP Converged      Max. Euclidean
                 roots          residual norm

<get> : LAST BUFFER REACHED --> switching to 'direct' mode !
read in  (=     64) + newbuffer  >  max (=     64)
   14       a        0        5.207066891872824D-05


NOTE: Residual norms are larger than 1.00D-05 due to mgrid option!

Eigenpairs written on logfile sing_a
...


Thanks,
M.

[christof.haettig]

With the m grids (multiple grids) the SCF and response equations are always solved in a coarse grid and after convergence improved energies are evaluated in a finer grid.

Christof