This post was instigated by Syracuse University Professor of Chemistry and well-known non-blogger Tim Korter concerning efforts to, I believe, generate proper Møller–Plesset Perturbation Theory Of The 4th Order (MP4, and also testing coupled cluster CCSD(T) calculations) intermolecular potentials for improving terms for Grimme dispersion-corrected density functional theory (DFT) calculations with the Gaussian09 package (a program for which many people grumble about various issues but which is, by nearly all metrics, a fantastic set of quantum chemical programs). The examples below, using water only, are just for ease-of-testing, which produce the following results based on the form of the input of the molecular coordinates. For those wondering why, z-matrices are the preferred format for performing SCAN or other automated trajectory calculations (an absolutely useless format, in my opinion, now that we have computers that can handle more than five atoms).

1. Molecular coordinates defined in the z-matrix (obviously in z-matrix format) and “z-matrix” called in the **opt** keyword…

#p scf=tight opt(tight,z-matrix) MP4/aug-cc-pVTZ
h2o test 1
0 1
O
H 1 0.96000000
H 1 0.96000000 2 109.50000032

… produces:

************************************************
** ERROR IN INITNF. NUMBER OF VARIABLES ( 0) **
** INCORRECT (SHOULD BE BETWEEN 1 AND 50) **
************************************************

2. Molecular coordinates defined in the z-matrix (obviously in z-matrix format) and “z-matrix” NOT called in the **opt** keyword…

#p scf=tight opt=tight MP4/aug-cc-pVTZ
h2o test 2
0 1
O
H 1 0.96000000
H 1 0.96000000 2 109.50000032

… produces:

************************************************
** ERROR IN INITNF. NUMBER OF VARIABLES ( 0) **
** INCORRECT (SHOULD BE BETWEEN 1 AND 50) **
************************************************

3. Molecular coordinates defined by variables in the z-matrix (obviously in z-matrix format) and “z-matrix” called in the **opt** keyword…

#p scf=tight opt(tight,z-matrix) CCS(D)/aug-cc-pVTZ
h2o test 3
0 1
O
H 1 B1
H 1 B1 2 A1
B1 0.96000000
A1 109.50000032

… produces:

Normal termination of Gaussian 09 at Fri Apr 27 11:28:33 2012.

4. Molecular coordinates defined by variables in the z-matrix (obviously in z-matrix format) and “z-matrix” NOT called in the **opt** keyword…

#p scf=tight opt=tight MP4/aug-cc-pVTZ
h2o test 4
0 1
O
H 1 B1
H 1 B1 2 A1
B1 0.96000000
A1 109.50000032

… produces:

Normal termination of Gaussian 09 at Fri Apr 27 11:28:33 2012.

So, a solution, for when the dreaded…

NEF-NEF-NEF-NEF-NEF-NEF-NEF-NEF-NEF-NEF-NEF-NEF-NEF-NEF-NEF-NEF-NEF-NEF-
NUMERICAL EIGENVECTOR FOLLOWING MINIMUM SEARCH
INITIALIZATION PASS
************************************************
** ERROR IN INITNF. NUMBER OF VARIABLES ( 0) **
** INCORRECT (SHOULD BE BETWEEN 1 AND 50) **
************************************************

… error appears, the simple solution is to re-define your system.

In the interest of wasting a few cycles, I ran a series of the same calculations with other post-Hartree-Fock methods [B3LYP, MP2, MP4, CCS(D), CCSD(T)] to see how pervasive the issue with coordinate definitions might be. It appears to be limited only to **MP4** and **CCSDT** (CCS(D) worked but took an unbelievably long time for Option 3 above), meaning people generally running significant molecules likely have never come across the issue.