some questions about "pair style gay-berne"

The mixing rule for Gay-Berne is explained here:
http://lammps.sandia.gov/doc/pair_gayberne.html

and in the papers cited therein.

Giacomo

i read it for many times and i know that the default rule is geometric.

it is not the quantity of reading that matters, but the quality. don't
expect that the manual has the direct answer to your question right
where you look for it and exactly how you would want it to be
formulated. you have to "read around it" to understand the general
principles (e.g. read the pair_coeff, pair_modify, and pair_style
pages very carefully as well).

Actually ,i am confuse about this : for type i j and i!=j the
pair_coeff sigma—if i specify its value like D. Do lammps use the value D
or use the value which caculated by mixing rule(if the two values are
different)? It says default rule is geometric.Does it means whatever the
"pair_coeff sigma" is you specify for type i!=j,the lammps will caculate it
again by "geometric" and use it even though you do not use pair_modify?

again, this is all answered in the manual. please try again in the way
as suggested above.

you can easily verify what explicit settings LAMMPS is using through
writing out the entire force field parameters using the "write_coeff"
command.

axel.