Hi,

I'm having difficulty with performing TLS refinement on a highly symmetric system. The difference between the R_work and R_free increases with TLS refinement (16.5/21.4 changes to 14.4/22.2). The TLS tensors for some of the group are not sensible (if I understand the literature correctly) -there are some negative numbers for the T and L tensors (see below). The TLS groups were identified with phenix.find_tls_groups. Any suggestions would be appreciated. 

REMARK   3    SELECTION: chain 'A' and (resid   38  through   68 )

REMARK   3    ORIGIN FOR THE GROUP (A):  56.6459 -23.5648 -59.0695

REMARK   3    T TENSOR

REMARK   3      T11:   0.2105 T22:   0.2895

REMARK   3      T33:   0.0962 T12:   0.0829

REMARK   3      T13:   0.0473 T23:   0.0007

REMARK   3    L TENSOR

REMARK   3      L11:   0.0446 L22:   0.0120

REMARK   3      L33:   0.0096 L12:   0.0246

REMARK   3      L13:   0.0209 L23:   0.0106

REMARK   3    S TENSOR

REMARK   3      S11:   0.0228 S12:  -0.0268 S13:  -0.0023

REMARK   3      S21:  -0.0507 S22:   0.0526 S23:  -0.0287

REMARK   3      S31:  -0.0296 S32:  -0.0137 S33:   0.0534​

The data set is complete at 2.3Å resolution and there are no strange crystal phenotypes. 

Best wishes,
Reza