Matrix Approach
The Collatz conjecture is a well-known problem. I won't make the claim to make any new findings here. However, I have never seen the problem be approached like this (but well, I didn't search very much). So this could be of interest to you.
Let the Collatz process be defined as
Let
we then have
This is a linear equation. We can then define the two vectors
Let also define the matrix
Note that all elements of the vectors should be positive integers to represent a valid Collatz sequence. We have a cycle in the Collatz sequence if
To solve this, we can transform U in the following way:
Transformation | \(l_n\) |
---|---|
\(l_n\leftarrow u_0l_n+2l_0\) | \(l_n = \begin{pmatrix} 0 & 4&0&0&\cdots&-u_0u_n\end{pmatrix}\) |
\(l_n\leftarrow u_1l_n+4l_1\) | \(l_n = \begin{pmatrix} 0 & 0&8&0&\cdots&-u_0u_1u_n\end{pmatrix}\) |
\(\vdots\) | \(\vdots\) |
\(l_n\leftarrow u_{n-1}l_n+2^nl_{n-1}\) | \(l_n = \begin{pmatrix} 0 & 0&0&0&\cdots&2^{n+1}-\prod_{i=0}^n u_i\end{pmatrix}\) |
To keep equality, we apply the same transformation to \(V\)
Transformation | \(v_n\) |
---|---|
\(v_n\leftarrow u_0v_n+2v_0\) | \(v_n = u_0v_n+2v_0\) |
\(v_n\leftarrow u_1v_n+4v_1\) | \(v_n = u_1(u_0v_n+2v_0)+4v_1\) |
\(\vdots\) | \(\vdots\) |
\(v_n\leftarrow u_{n-1}v_n+2^nv_{n-1}\) | \(v_n = v_0\prod_{i=0}^{n-1}u_i+\sum_{i=0}^{n-1}(2^{i+1}v_i\prod_{j=i+1}^{n-1}u_j)\) |
The equality thus gives
For a valid cycle, we thus have
with \(k\), the number of odd numbers in the cycle. This gives us an upperbound on the number of odd numbers in a cycle:
I want to point out this is in now way new but I didn't see this way to get it, even if equivalent methods have probably been used.
If one could prove that
is never an positive integer other than 1 and 2, he would have proven that there is no other cycles. It would still not be sufficient to prove the Collatz conjecture as this does not eliminates the possibility that a sequence infinitely increases.