CCL:G: Problem with ONIOM calculations

 Sent to CCL by: "Lesley Rae Rutledge" [lesley.rutledge|*|uleth.ca]
 Hello,
 I am trying to run ONIOM calculations (M06-2x/6-31G(d):AMBER) in Gaussian 09
 Revision A.02 and I am having a few problems. My system has 4162 atoms, where
 149 are in the QM region (1316 basis functions). I am running my calculations on
 8 processors with 8GB of memory and 160 GB of disk space.
 In order to build my model, there are multiple stages of calculations, where I
 came to the stage where I wish to fix all of the low level region and optimize
 just the hydrogen atoms of the QM region. This calculation finished. However,
 the next stage was to use this system, and optimize the whole QM region (with
 the MM region fixed). These calculations die just before it starts the
 high-level calculation on the model system with the following error:
  Cartesian Forces:  Max     0.059474500 RMS     0.000788502
  GradGradGradGradGradGradGradGradGradGradGradGradGradGradGradGradGradGrad
  Berny optimization.
  Internal  Forces:  Max     0.047983376 RMS     0.006291829
  Search for a local minimum.
  Step number   1 out of a maximum of  894
  All quantities printed in internal units (Hartrees-Bohrs-Radians)
  NRF ne Abs NRFX
  Error termination via Lnk1e in /opt/g09/l103.exe
 If I reduce the size of the MM region, I get the same error. Also, if I delete
 all of the MM region and just run an M06-2X/6-31G(d) calculation (on the QM
 part), I have no problems (goes about 35 steps in the optimization in 24 hours).
 We are aware that you can split up ONIOM calculations into their parts (to run
 separately), which I have tried to generate, yet these don't finish either. I
 get the following error:
  GradGradGradGradGradGradGradGradGradGradGradGradGradGradGradGradGradGrad
  Berny optimization.
  Operation on file out of range.
 However, this calculation does generate the three inputs for the single-point
 calculations on the system before it dies. The high-level calculation on the
 model system has all of the MM atoms present, but they appear as:
  Bq-#1-0.442200(Iso=1.00782504,Spin=1,GFac=2.792846)  -1    -25.988729094663
 -0.265707167253    -16.971277852321
   Bq-#1--0.631800(Iso=15.99491464)          -1    -26.778340617607
 -0.824800703903    -17.146345953700
   Bq-#1--0.006900(Iso=12)                   -1    -26.646213612701
 -2.047053577460    -16.427388903608
 ..
 If I take this partial output as input for the high-level optimization
 calculation on the model system, this calculation will run (but very slowly -
 only 2 optimization steps in 24 hours, even though the basis functions are
 correct for only the QM atoms being present).
 Therefore, my questions are:
 Why are my ONIOM calculations dying? Is this a memory or disk problem?
 How do I properly calculate the required memory (and disk) for an ONIOM
 calculation in Gaussian?
 Any help would be greatly appreciated.
 Thanks very much for your time,
 Lesley Rutledge