Author Topic: Riper displays non-positive definite matrix in cholesky decomposition  (Read 6353 times)


  • Jr. Member
  • **
  • Posts: 12
  • Karma: +0/-0

I was trying to run a periodic single point calculation for a 4*4*4 Ru(0001) surface with an adsorbed moiety using riper module. I used SVP basis set for adsorbed moiety atoms and ecp fr Ru, I also included the auxiliary basis set as the manual suggested. The following specifies my cell information in the control file-

  10.87272 9.41606 21.50000 90.0 90.0 90.0
$periodic 3
$kpoints nkpoints 4 4 1

Since the manual didn't suggest anything different, I have the coord file as usual.

I'm also using a b-p functional and m3 grid size. I tried with SMP parallelization and ran the calculation on 24 cores, unfortunately after reading the MO's, here's where the calculation stops-

              |              NUMERICAL INTEGRATION               |
       Will use smaller grid during SCF iterations and a grid
       of size 3 for the last iteration.
       Integration of exchange-correlation term is performed with
       Stratmann/Scuseria renormalization.
       Specified sharpness parameter is .64000D+00.
       Maximum number of grid points per atom:             9914
       Total number of grid points of the system:          362947
   Periodic boundary conditions - crystal field construction:
     extent of the nuclei:             10.7500
     extent of dipole correction:      10.7500
     extent of electron density:       43.0250
     extent of auxiliary density:      25.0926
     final crystal field radius:      135.0440
     number of lattice vectors:       4689
   Continuous Fast Multipole Method for the Coulomb term
     order of multipole expansions:       20
     well-separateness criterion:         3.0000
     local multipole expansions switched on
     local expansions incremented by:      2
     number of octree levels:              9
     number of boxes at all levels:   147028
     max. number of NF boxes per box:  32486
     length of the lowest level box:      0.7742
     length of the highest level box:   198.1962
 ERROR: non-positive definite matrix in Cholesky decomposition (routine: usad)
    of MOs in cho_ortho.f
 MODTRACE: no modules on stack

 negative root in usad, see output
  ended abnormally

Can someone help me with this? Thank you so much!



  • Administrator
  • Full Member
  • *****
  • Posts: 63
  • Karma: +0/-0
You probably used ECPs, Turbomole 7.0.1 neither supports ECPs nor metallic systems for periodic calculations. The new 7.1 version does both.