optimization to TS of THF

 Hi all,
 I am trying to find transition strucutre of tetrahydrofuran (THF) at
 BLYP/6-31G* and B3LYP/6-31G* levels. I started computations from O-exo
 geometry, that is expected for TS, but computations do not converge.
 I tried optimization in internal, redundant, cartesian coordinates,
 QST3 method.I switched off testing of the number of eigenvalues, as the
 program (gaussian 94) complained about wrong number of negative
 eigenvalues. But nothing works. Can sombody give me a hint how to
 overcome these problems.
 Thanks, Marek
 Marek Strajbl
 Institute of Physics, Charles University
 e-mail: <strajbl %! at !% silicon.karlov.mff.cuni.cz>