First explain what you are doing! "Divergence" is a vector valued differentiation of a scalar function. It is not, in general, defined on matrices. You could, of course, define it as what you get by applying the divergence to each component of the matrix. The result would be a matrix having vectors as components. How that would be interpreted depends on your original matrix- as a linear operator over a vector space.
And, of course, you haven't said anything about dependence on x, y, z or even how many variables there are. If the matrix were a constant, its diveregence would be "0" although I'm not entirely sure what kind of object that 0 would be!
Could you give more detail about the situation in which you are dealing with the "divergence of a matrix"?