-1
$\begingroup$

(this was discovered analyzing Collatz empirically.)

a key aspect of resolving Collatz involves looking at the number of iterations for trajectories to "fall below" the initial value.

consider a trajectory starting with the initial value $a \cdot 2^n + m$. it always "falls below" this initial value for all $a$ and "large enough" $n$, depending on $m$. more specifically $n$ is approximately the number of iterations in the full trajectory starting with $m$.

the proof is not so complicated.

does this property follow from something in the published literature on Collatz?

(plz cite ref if the answer is in the affirmative.)

$\endgroup$
7
  • $\begingroup$ I think you are claiming that for every $m$ there exists $n$ such that for every $a$ Collatz is true for $a\cdot2^n+m$. Is that right? $\endgroup$ Aug 21, 2014 at 0:13
  • $\begingroup$ the "fall below" property is interrelated but not exactly the same as the "terminates at 1" property. the conjecture is also true iff all trajectories "fall below". but a "fall below" property of a trajectory does not nec(?) guarantee it terminates at 1. (by induction) a "fall below" trajectory does terminate at 1 if all trajectories starting below it also have the "fall below" property. here "full trajectory" means a trajectory ending at 1. $\endgroup$
    – vzn
    Aug 21, 2014 at 0:40
  • $\begingroup$ OK, so what you are asserting is that for every $m$ there exists $n$ such that for every $a$ $T^k(a\cdot2^n+m)<a\cdot2^n+m$ for some $k$, where $T$ is the Collatz iteration, right? $\endgroup$ Aug 21, 2014 at 3:55
  • 2
    $\begingroup$ I believe the statement the OP is aiming for is: for all $m$ and $a$, there exists $N$ such that for all $n>N$ there exists $k$ with $T^k(a\cdot 2^n+m) < a\cdot 2^n + m$. $\endgroup$
    – Aeryk
    Aug 21, 2014 at 5:10
  • 1
    $\begingroup$ The proof to this is simple, so I always assumed it's in Lagarias's book. But I've never bothered to check because if not, the proof is simple. $\endgroup$
    – Aeryk
    Aug 21, 2014 at 5:13

1 Answer 1

1
$\begingroup$

Your claim is surely unsubstantiated.

For a clearer derivation of your suggested process, you should formalize the odd numbers, to be taken under Collatz-transformation $$ a_{n,j}=j \cdot 2 \cdot 2^n + m_n \qquad \qquad j,a_{n,j} \text{ are odd }$$ then $m_n$ is a function of $n \gt 0$: $$ m_n = \cases { { 1 \cdot 2^n-1 \over 4-1} =\lbrace 1,5,21,85,...\rbrace \qquad n \text{ is even } \\ \\ {5 \cdot 2^n-1 \over 4-1} =\lbrace 3,13,53,213,...\rbrace\qquad n \text{ is odd } } $$ Now by one transformation $$ b_{n,j}= {3 a_{n,j}+1 \over 2^n} = {3(j \cdot 2 \cdot 2^n + m_n )+1 \over 2^n} \\ = 6j +{ 3 m_n +1 \over 2^n} \\ = 6j + \cases{1 \qquad \small \text{ if n even}\\ 5 \qquad \small \text{ if n odd}} $$ So after one transformation we have some $b_{n,j} = 6j +3 \pm 2 $ which is again of the form $a_{m,k}$ for some other $m,k$. Of course this can then be iterated.
Because the indexes $m,k$ of $a_{m,k}$ are independent from $n,j$ it might be that $a_{m,k} \gt a_{n,j}$ (namely if $n=1$, $a_{n,j}=4j+3$ , $b_{n,j}=6j+5$) and thus there is no guarantee, that iterations fall below the initial value.

(Remark: If you had such a guarantee, you'd solved the Collatz-problem.)

$\endgroup$
1
  • $\begingroup$ thx for further attn however afaik collatz implies the "fall below" property highlighted in the question for an infinite # of cases but not vice versa as your answer seems to assert. but the strict formulation of the property is more found in the followup comments. suggest/ encourage further discussion/ clarification in number theory chat $\endgroup$
    – vzn
    Mar 15, 2015 at 18:30

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.