MP2: Counterpoise corrected geometry optimization fails


Just Got Here
Hi,

I'm trying to do counterpoise corrected geometry optimizations at the MP2/6-311G(d,p) level. However, every single one fails with an error like this:

me = 11 MAXITS exceeded in inv_it 

c1= 363 cn= 364 b1= 363 bn = 364 
nrm = 1.0783 stpcrt = 1.11803 
xj = 1.598721e-14 onenrm = 3.554916e-15 
d[363] = -2.409250e-15 e[363] = -9.746273e-16 
d[364] = -2.331243e-15 e[364] = 1.145666e-15 

me = 11 Eigenvector 364 failed to convergve in inv_it3


It seems like this happens sometimes after some steps of optimization, since it was cycling through the different fragments several times, and sometimes it fails right at the first step.

So what's happening here and is there a way to fix it?

Example of Input:

start BCN_pAz_TS_2_saddle

title "BCN_pAz_TS_2_saddle"

scratch_dir /scratch

memory 3000 mb

geometry units angstroms print xyz autosym
	C         -0.14435        0.97154       -0.46844
	C          1.08312        0.88478       -0.23927
	C         -0.64826       -1.65774        0.72275
	C          2.28507        0.08618        0.05335
	C          0.84474       -1.76973        0.98256
	C          1.91937       -1.40965       -0.02245
	H          3.08792        0.32068       -0.65531
	H         -1.22331       -1.32667        1.58723
	H          2.66619        0.33004        1.05383
	H          1.13189       -1.51122        2.00156
	C         -1.45246        0.31249       -0.63810
	H         -1.91560        0.60932       -1.58738
	H         -2.14308        0.61861        0.16111
	C         -1.25997       -1.21559       -0.58906
	H         -2.22577       -1.71993       -0.71783
	H         -0.63043       -1.51063       -1.43424
	H          1.59559       -1.62678       -1.04454
	H          2.81607       -2.01480        0.17179
	C         -0.01099       -3.02177        0.88990
	H         -0.22009       -3.52704        1.82947
	C          0.07697       -3.96154       -0.27763
	H          0.75387       -4.79091       -0.02623
	H          0.48168       -3.45369       -1.16078
	O         -1.24265       -4.44544       -0.53304
	H         -1.15813       -5.10800       -1.22405
	N          1.78437        2.93097       -0.38270
	N          0.71293        3.39967       -0.58762
	N         -0.47834        3.05701       -0.84134
	C         -1.52092        3.62991        0.03390
	H         -1.64748        4.68942       -0.21361
	H         -2.43723        3.11209       -0.25962
	C         -1.22757        3.43827        1.51640
	H         -0.30512        3.95512        1.79555
	H         -2.04620        3.84084        2.12032
	H         -1.10313        2.37682        1.74442
	end

basis spherical
    * library 6-311G**
	bqO library O 6-311G**
	bqN library N 6-311G**
	bqC library C 6-311G**
	bqH library H 6-311G**
end

mp2
	freeze atomic
end

bsse
 mon bla1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
 mon bla2 26 27 28 29 30 31 32 33 34 35
end


task mp2 saddle


I can provide the whole output file if it's needed.

Thanks a lot

Forum Vet
Yes, the output file would be useful since I am not able to reproduce your failure

Just Got Here
Here are two examples. After some hours the problem occured

[[1]]
[[2]]

The system I'm running those on is NWChem 6.6 on a CentOS 7 cluster, one node has two Intel Xeon E5-2650v2, 2.6 GHz, 8 cores processors and 64GB RAM. Those calculations run on 6 or 8 nodes using MPI.

I've now tried using the full direct MP2 method and for now it's running fine, but will still take a while to finish so I don't know yet if that solves the problem.

Forum Vet
I have just realized that your NWChem is not linked with SCalapack, since your error message shows that the diagonalization calls are handled by the built-in Peigs library.
Could you try to link your executable with Scalapack and see if the error goes away (that's what I have seen in my NWChem 6.6 installation)?
http://nwchemgit.github.io/index.php/Compiling_NWChem#Optimized_math_libraries

Just Got Here
Thanks. I will look into that.


Forum >> NWChem's corner >> General Topics