The substrate has been made rigid, then I impose a normol force Fn and a lateral force Fx to the block quasistatically, I also minimize the system before I impose force. But during the minimization, the block will deform a lot like this:

In conclusion, my problem is how to choose parameters in lj potential to make the system stable. I have done some searching and I knew the best way is using parameters from the computational result of first-principles calculation. But I have no idea about irst-principles calculation so far, so is there a easier way for me to search the suitable parameters?

No, I think your problem are that you have unrealistic expectations of how objects at the atomic or nano scale behave. It is not unexpected that you get reconstructions and you probably have a small mismatch of the lattice constants between the two parts. Nothing stops you from keeping both parts rigid, but then - of course - you have to ask yourself: is this still research or already computer animation?

Intuitively, the LJ potential is purely radial, while your desired lattice structure has unequal distances between neighbours, so your desired lattice structure cannot be a stable arrangement of just one kind of LJ particle.

Of course there is a whole zoo of three-body potentials out there – you could start with Stillinger-Weber – but I would advise you to make sure you’re solving a real scientific problem to start with, and then to make sure you have literature review that gives you some confidence that your chosen method will work.