Implementing binary search trees on disk presents challenges such as pointer reassignment and poor locality, leading to inefficiencies. Transitioning to b trees allows for higher fan-out and better data locality, as they utilize multiple keys per node and maintain a relationship between keys and child nodes. This structure significantly enhances search efficiency by enabling quicker key location.
