Thank for your mail and suggestions Steve. In fact I've used the command "compute myTemp all temp/asphere". When the system is equilibrium enough there is very small difference between the temperature original and the temp/asphere.
After discussed with my tutor, we found that the problem should be the energy scale parameter in pair_coeff for ellipse. We set the parameter to 2 in order to match reports in article. Even for LJ particle we also need to do that setting for matching. I did not know the reasons. With the setting the pressure reported by thermo is not reduced value then. That is to say when I set npt pressure as 2 in fact the reduce pressure is just 1. So we need to double our setting. The same problem also exist at temperature. In conclusion, the data I got now is not pressure 2 I wanted but just for reduce pressure 1. And compared with the reported density for that pressure my data is acceptable. Much more validations are going on. If convenient, could you tell me why we need doubling the energy scale parameters?
Thank again. You see Steve there is seldom open source provider can do so kindly help for users. It is lucky for all Lammps user with your help. And I hope when I familiar with Lammps I also can help some other beginner.