1
$\begingroup$

Remark : I've found a rather trivial answer for this question and so very likely the premise of paralleling it with the Zsigmondy-theorem is wrong, so this question might better be retracted. I'll give it a certain time loking at reactions on the discussion in the MO-meta site.
Following a hint in META I added a comment where you can vote for deletion.


I'm (recreationally) studying properties of the generalized Collatz-problem ("gCp" for shortness) for $3x+\rho$ and look for help in the question of the existence of cycles in gCp's with composite $\rho$ as parameter in relation to that in gCp's with the primefactors of $\rho$ as parameters.

To explain the problem formally, I'd like to use here the compact "Syracuse"-notation of the gCp in the following way: $$ T_\rho(a;[A]) := {3a+\rho \over 2^A} \qquad \qquad \text{where } A=v_2(3a+\rho) $$ with odd positive integer in the parameter $\rho$. By that definition the transformation $b=T(a;[A])$ works on odd numbers $a,b$ only.
For an iterated transformation of $N$ steps I write simply $a_{N+1}=T_\rho(a_1;[A_1,A_2,...,A_N]) $. I denote $S$ as the sum of exponents $A_k$ and to make the following formulae short, I define $E_{N,S} = [A_1,A_2,...,A_N]$ as the vector for a fixed set of exponents $A_k$ having length $N$ and sum $S$.

I'm interested in the question on existence of cycles in $T_\rho()$ for some composite $\rho$ in relation to that in $T_\varphi()$ where $\varphi$ is a primefactor of $\rho$ .

For example let $\varphi, \sigma \in \Bbb P$ being prime and $\rho=\varphi \cdot \sigma$ being composite. Let some vector $E_1 = E_{N_1,S_1}$ be such that $a_{N+1} = T_\varphi(a_1;E_1)=a_1$ is a cycle.
It is easy to derive that then also $ b_1 = T_\rho(b_1;E_1)$ with $b_1= \rho/\varphi \cdot a_1$ is a cycle in $T_\rho()$.
Of course the analogue is true for a cycle in $T_\sigma()$ in relation to $T_\rho()$ with another set of exponents $E_2 := E_{N_2,S_2}$

  • a) So we have first -by some simple analysis-, that a generalized Collatz-problem gCp with $T_\rho()$ with a composite parameter $\rho$ defines the same cycles as the gCp's with the primefactors of $\rho$ as parameters (where only its elements $a_k$ are rescaled by some factor as said above).

The two interesting parts are the following which I'd like to understand/to prove.

  • b) The gCp of a composite parameter $\rho$ seems to have in general always additional cycles besides that of the gCp's defined by $\rho$'s primefactors.
    This reminds me of the theorem of Zsigmondy about the primefactorization of Mersenne-numbers (where he proves the existence of the then so called "primitive primefactors"). Possibly it can be proven the same way, but I could not follow his proof. Let's call that additional cycles analoguously "primitive cycles".

  • c) Different from the "general" case: when we have the primefactor $\sigma=3$ involved, the composition $\rho=\varphi \cdot 3$ with that primefactor $T_{\varphi \cdot 3}()$ seems to not to lead to such additional "primitive cycles" - again in analogy to Zsigmondy's observation on the Mersenne-number $M_6 = M_{2 \cdot 3}$ which has no "primitive" factors.
    Of course I do such a statement only based on heuristics and the basic assumption that the Collatz-conjecture for $T_1()$ is true and that it has only the "trivial cycle" $1=T_1(1;[2,2,2,...,2])$.

So my question asks for help in proving the observations in b) and c) (under assumption of truth of the Collatz-conjecture). Perhaps it would be sufficient to only get Zsigmondy's derivations transparent enough.


*Remark: for some examples and possibly for a better explanation of the elementary analysis of this you might look at the short treatize [at my webspace][3] (instead of the term "primitive cycles" I've used "unexplained cycles" there)*
$\endgroup$
2
  • $\begingroup$ Poll for opinions: vote this comment if I should delete that question. $\endgroup$ Sep 7, 2017 at 12:44
  • $\begingroup$ On a note related to $3x+p$ you might want to look into convergence of the function $3x+p\cdot2^{v_2(x)}$, when viewed projected down to $\mathbb{N}_2^{\times}$ in the 2-adic space. You will see that the numbers which converge to a cycle of order $1$, in this alternative view of the problem, they are convergent in the 2-adic space while those converging on a cycle of order greater than one, they contain multiple convergent trajectories. Those which rise indefinitely are divergent, and their image in $\mathbb{N}_2^{\times}$ must be divergent too. $\endgroup$ Sep 20, 2017 at 17:49

1 Answer 1

1
$\begingroup$

The answers for b) and c) came out to be trivial and have likely nothing to to with Zsigmondy, so possibly I should retract my question.


For the definition of a cycle for some $a_1=T_\rho(a_1;E_{N,S})$ we have the formula (6.b) in the referred article: $$ a_1 = \rho \cdot { Q(E_{N,S}) \over 2^S - 3^N} \tag 1 $$ Let $\frac uv$ with $v>1$ denote the fully cancelled fraction in (1) then by $$ a_1 = \rho \cdot \frac uv \tag 2 $$ it is obvious that we can have a cycle at $a_1$ only if $v | \rho$. Now let $v$ be prime, then setting $\rho=v$ makes this having an integral solution and thus having a cycle at $a_1$. But of course any composite $\rho = v\cdot t$ makes this having an integral solution (and having a cycle at $t \cdot a_1$) as well. So a gCp with composite $\rho$ has cycles at the same exponentsvector $E_{N,S}$ as all gCp's with parameters being primefactor of $\rho$.

At question b): the "unexplained" (or "primitive") cycles at some $E_{N,S}$ occur if in (2) the denominator is composite and not prime - because in the gCp's with prime parameter the denominator $v$ is not completely cancelled, and thus that gCp's cannot have a cycle at that same $E_{N,S}$
This explains question b).

The question c) is simply explained by the additional fact, that the denominator in (2) cannot have the primefactor $3$ - so for any composite $\rho $ the possession of primefactor $3$ does not introduce any new ("unexplained","primitive") cycle.

So this seems to have nothing to do with Zsigmondy's theorem on mersenne-numbers.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.