10 dimensional jump

Notice that the parity of a point’s coordinates does not change when it undergoes a jump, so if we take the unit $10$ dimensional cube (all coordinates at $0$ or $1$ ), we have $2^{10} = 1024$ points, and no two will ever coincide because that would imply their parities are the same. Thus the answer is at least $1025$ .

To show that $1025$ works, we use induction on the number of dimensions.

Base Case : Any $3$ points in $1$ -dimensional space can be moved so that some two coincide.

Proof : Jump an outer point over the middle point, maximum distance decreases. So, eventually some two will coincide.

Inductive step : Consider any $2^{n+1} + 1$ points in $(n+ 1)$ -dimensional space. We’re going to show that we can make some $2^{n} + 1$ of them lie in an $n$ -dimensional “plane,” so that we can apply the inductive step.

Let’s project the points down onto $1$ -dimensional space; say, by looking only at the first coordinate.

Then any jump in $(n+1)$ -space corresponds directly to a jump of the projected points (feet) in $1$ -space.

Now divide up the points into two sets $A$ and $B$ based on the parity of the first coordinate. We can assume WLOG that both $A$ and $B$ are non-empty, else we can either divide by $2$ , or shift over and divide by $2$ , and the sequence of moves remains the same. By pigeonhole, we know that one of the sets has at least $2^n + 1$ elements, let that be $A$ . Now, choose any element $P$ of $B$ . By manipulating the feet of any two elements $a_1, a_2 \in A$ and $P$ , we can get the feet of $a_1$ and $a_2$ to coincide (by the base case, and parity) and thus $a_1$ and $a_2$ lie in the same $n$ -dimensional plane. Now we repeat this algorithm, subsequently always treating $a_1$ and $a_2$ as a unit (so that they remain in the same plane whenever they jump over another point).

Thus, by repeatedly manipulating the feet of $P$ and two elements from $A$ , we can accumulate all the points in $A$ into $n$ -space, and we are done. Thus, the answer is $\boxed{1025}$ .