You could try writing the mergesort() to handle three input arrays, storing to a fourth bigarray, but that's an awkward way. We generally want to solve problems in terms of things that are already solved.
You know how to merge two arrays into one. There is no requirement that the two source arrays be of the same size, only that the destination be at least big enough to hold the total number of elements.
What if you have, in main, another array of at least size 100, which will be the temporary holding place for the result of merging x and y. Then, call mergesort( ) again, passing z and the temp array, with big array as the destination. This will require that mergesort() also be told the size of the source arrays. Side benefit is that now your mergesort will handle any size problem, not the just specific case of inputs of size 50. This more general approach means never having to revise the function when the problem scale changes. So, your prototype for mergesort ()might look like:
void mergesort ( int a[], int b[], int dest[], int size_a, int size_b );
use the size parameters as limits on index_a and index_b in the loops.
Marked 
