Now, the second case, where the second movement of the largest disk is back to its
original peg, can be treated in a similar fashion. Without loss of generality, assume that
there is a shortest path from 0n to v moving the largest disk at least twice—first to peg
1 and then back to peg 0. Then we can write the path as a sequence of steps the way
we did above by switching peg 2 to peg 0 in steps (3) and (4). Then the same sequence
of moves without steps (2) and (4) gives a new path two legal moves shorter, hence
giving a contradiction to our assumption. Therefore, there is no shortest path between
a corner vertex and another vertex in which the second movement of the largest disk is
back to its original peg.