Can you please explain the reason behind taking the upper limit for binary search n (and not n-1)?

In my case taking upper limit to n-1 gives segmentation fault and for n works fine.