G98 and SCRF

 I'm trying to optimize the structure of a simple silica compund
 (H3SiOH) in THF solvent by means of SCRF calculation in
 Gaussian 98. When I run the calculation with opt=(loose) and
 SCRF=(cpcm,solvent=THF) the calculation converges in a couple
 of steps. However, when I run the calculation with the standard
 convergence criteria (or tighter) the calculation does never
 converge. The energy remains oscilating (typically in the second
 decimal), as does the force and displacements, and the calculation
 typically exits with a L9999 error. Increasing the maximum number
 of steps only causes the calculation to stop after more steps and
 not the job to converge.
 In both job files G98 warnes that sometimes some elements are
 discarded. However, only in the tighter criteria run there are warning
 that a certain electronic charge (in this case charge 144) has
 positive or a negative charge ("Electronic charge 144 has a
 negative value:  -0.349460"). Could this screw up convergence or
 are there other important details wich are important to insure
 convergence in such SCRF calculations.
 Martijn Zwijnenburg
 Lab. of Applied Organic Chemistry and Catalysis
 Delft University of Technology
 Julianalaan 136
 2628 BL Delft
 The Netherlands
 Tel: 0031-(0)152782691
 Fax: 0031-(0)152784700
 e-mail: M.A.Zwijnenburg (+ at +) tnw.tudelft.nl
 web page: http://come.to/tock