There are multiple problems of technical and conceptual nature here
your syntax is incorrect. both your pair_coeff statements assign interactions to both atom types
where is your Oxygen atom type?
pair style reax/c does not support being used as a hybrid sub-style multiple times
you need ReaxFF potentials that are suitable for your problem
using a hybrid pair style configuration is conceptually a problem for your system, since you will do two independent calculations and have the interaction between them only approximated with a simple LJ potential
it is generally strongly recommended for ReaxFF calculations to use only ReaxFF for all atom types in the system and have a parameterization that is applicable. ReaxFF potentials are generally not very transferable. So it can be challenging to find something.
Thanks for the answers and useful instructions. At the first stage, I tried to deposit only Ni without Oxygen. If hybrid pair styles work well, I will try to simulate with Oxygen. However, I did not get what I expect.
Unfortunately, I did not find any ReaxFF potentials that included Ni,O and Si atoms. Thereofore I used the hybrid pair styles.
Please keep in mind that having a syntactically correct LAMMPS input does not automatically mean you get meaningful results. That is particularly true for using hybrid potential setups. Many pair styles in LAMMPS have many-body components and thus may not be computed correctly with a hybrid pair style.
When using a hybrid pair style in particular you have to be certain of what it is that you want to study. So a partitioning of Si and NiO (or just Ni) with a LJ or Morse interaction to combine those can be acceptable, provided you are mainly interested in either the Si structural properties of the NiO/Ni structural properties and less so in the interactions between Si and NiO/Ni.
For that you can set up the Si component with a Tersoff pair style and use, e.g. EAM for Ni or some pair style with a parameterization suitable for studying NiO, for the other component.
If you have to do charge equilibration (which is likely required for the oxide), then you already have some limitations since you will have to treat that part of the system as an isolated chunk and thus cannot have any polarization and polar interactions across the hybrid substyles. This is when using ReaxFF (or something equivalent like COMB/COMB3) for everything may be required.
Given the difficulty of representing especially the oxide structure, you may also consider looking into some semi-empirical or tight-binding models, which may offer more flexibility at the expense of computational effort but will be much more accessible than a full quantum treatment.