25

This post makes a focus on a very specific part of that long post. Consider the following map:
$$f: n \mapsto \left\{ \begin{array}{ll} \left \lfloor{n/\sqrt{2}} \right \rfloor & \text{ if } n \text{ even,} \\ \left \lfloor{n\sqrt{2}} \right \rfloor & \text{ if } n \text{ odd.} \end{array} \right.$$

Let $f^{\circ (r+1)}:=f \circ f^{\circ r}$, consider the orbit of $n=73$ for iterations of $f$, i.e. the sequence $f^{\circ r}(73)$: $$73, 103, 145, 205, 289, 408, 288, 203, 287, 405, 572, 404, 285, 403, 569, 804, 568, 401, \dots$$

It seems that this sequence diverges to infinity exponentially, and in particular, never reaches a cycle. Let illustrate that with the following picture of $(f^{\circ r}(73))^{1/r}$, with $200<r<20000$:

enter image description here

According to the above picture, it seems that $f^{\circ r}(73) \sim \delta^r$ with $\delta \sim 1.02$.

Now consider the probability of the $m$ first terms of the sequence $f^{\circ r}(73)$ to be even: $$p_{0}(m):= \frac{|\{ r<m \ | \ f^{\circ r}(73) \text{ is even}\}|}{m}.$$ Then $p_1(m):=1-p_0(m)$ is the probability of the $m$ first terms of $f^{\circ r}(73)$ to be odd.

If we compute the values of $p_i(m)$ for $m=10^{\ell}$, $\ell=1,\dots, 5$, we get something unexpected: $$\scriptsize{ \begin{array}{c|c} \ell & p_0(10^{\ell}) &p_1(10^{\ell}) \newline \hline 1 &0.2&0.8 \newline \hline 2 &0.45&0.55 \newline \hline 3 &0.467&0.533 \newline \hline 4 &0.4700&0.5300 \newline \hline 5 &0.46410&0.53590 \newline \hline 6 & 0.465476& 0.534524 \end{array} }$$ (the line for $\ell = 6$ was computed by Gottfried Helms, see the comments)

It is unexpected because it seems that $p_0(m)$ does not converge to $1/2$, but to $\alpha \sim 0.465$.
It matches with the above observation because $$ \delta \sim 1.02 \sim \sqrt{2}^{(0.535-0.465)} = \sqrt{2}^{(1-2 \times 0.465)} \sim \sqrt{2}^{(1-2\alpha)}.$$

Question: Is it true that $f^{\circ r}(73)$ never reach a cycle, that $(f^{\circ r}(73))^{1/r}$ converges to $\delta \sim 1.02$, that $p_0(m)$ converges to $\alpha \sim 0.465$, and that $\delta^24^{\alpha} = 2$? What are the exact values of $\delta$ and $\alpha$? (or better approximations?)

The following picture provides the values of $p_0(m)$ for $100 < m < 20000$: enter image description here

Note that this phenomenon is not specific to $n=73$, but seems to happen as frequently as $n$ is big, and then, the analogous probability seems to converge to the same $\alpha$. If $n <100$, then it happens for $n=73$ only, but for $n<200$, it happens for $n=73, 103, 104, 105, 107, 141, 145, 146, 147, $ $ 148, 149, 151, 152, 153, 155, 161, 175, 199$; and for $10000 \le n < 11000$, to exactly $954$ ones.

Below is the picture as above but for $n=123456789$:
enter image description here

Alternative question: Is it true that the set of $n$ for which the above phenomenon happens has natural density one? Is it cofinite? When it happens, does it involves the same constant $\alpha$?

There are exactly $1535$ numbers $n<10000$ for which the above phenomenon does not happen. The next picture displays for such $n$ the minimal $m$ (in blue) such that $f^{\circ m}(n) = f^{\circ (m+r)}(n)$ for some $r>0$, together with the miniman such $r$ (in red):

enter image description here

In fact all these numbers (as first terms) reach the following cycle of length $33$:

$$(15,21,29,41,57,80,56,39,55,77,108,76,53,74,52,36,25,35,49,69,97,137,193,272,192,135,190,134,94,66,46,32,22)$$

except the following ones: $$7, 8, 9, 10, 12, 13, 14, 18, 19, 20, 26, 27, 28, 38, 40, 54,$$ which reach $(5,7,9,12,8)$, and that ones $1, 2, 3, 4, 6$ which reach $(1)$, and $f(0)=0$.

If the pattern continues like above up to infinity, they must have infinity many such $n$.

Bonus question: Are there infinitely many $n$ reaching a cycle? Do they all reach the above cycle of length $33$ (except the few ones mentioned above)? What is the formula of these numbers $n$?

Below is their counting function (it looks logarithmic):
enter image description here

  • Fascinating! The pattern for the $n$ with periodicity is of course close to self-similar with a ratio of $\sqrt2$, as might be expected. Just out of curiosity: could you add a graphic which displays for each of those $n$ the minimal period $r$? – Wolfgang Feb 25 '20 at 14:09
  • We can rule out convergence to anything greater than 1/2, since if $f^j(n),\ldots,f^{k-1}(n)$ are mostly even, then $f^k(n)<f^j(n)$ –  Feb 25 '20 at 14:48
  • Out of curiosity: the mapping $f$ looks like it would be expected 'on average' to be slightly contracting, since there are floor operations in the definitions of both the even and odd branches, and without those floor operations one would expect $f()$ to behave like a random process with (geometric) mean equal to its initial value. Have you looked for cycles in the iterates of $f()$ starting from 73, and is it possible that most of the numbers are falling into the same cycle? That would explain the commonality in the limit... – Steven Stadnicki Feb 25 '20 at 17:45
  • @StevenStadnicki No cycle found starting from $73$. There is a cycle of length $33$ starting from $15$. There are additional details in the long post cited at the beginning. You are right in the sense that it would be more precise to first ask whether the process never reach a cycle from $73$. Next for those never reaching a cycle, whether that probability converges to that common $\alpha$. What I have in mind when I write ‘phenomenon’ is precisely ‘never reaching a cycle’. If it reaches a cycle then the limit can also be not $1/2$ (and should) but must be closer to $1/2$ than when no cycle. – Sebastien Palcoux Feb 25 '20 at 18:50
  • I see; thank you for the pointer. I hadn't looked at the previous post, apologies. You don't really talk about process in either post, though — do you have your source available for these explorations? How many iterations did you go on $n=73$, and what was the maximum magnitude of $f^{\circ r}(n)$ you found over those iterations? Do you have any plots of, e.g. $\log(f^{\circ r}(73))$ for $r=1..10^k$ for some $k$? It would be good to see what the iterate trajectory looks like. – Steven Stadnicki Feb 25 '20 at 19:39
  • 2
    Perhaps I made a mistake, but I get a different value of $p_0(10^4)$ (0.4912 instead of 0.4700). For larger $m$ the value seems even closer to 1/2. As a control the computation for $m=10^4$ has a maximum of $f^{\circ k}(73)$ equal to $1341801280048839911857201274496$ (for $k=2584$), so roundoff errors when computing $n\cdot \sqrt{2}^{\pm 1}$ could have occured in your computation. Could some confirm the above? – Kasper Andersen Feb 25 '20 at 21:10
  • 3
    @Kasper Andersen The problem seems to be the floating point arithmetic you use. I calculated (with python and the module mpmath) $p_0(10^4)$ and $p_0(10^5)$ with a precision of 100,1000,10000 decimal places and got for 10000 decimal places the same values as in the table. But for fewer decimal places the values were different, f.i. with double precision the values seemed to converge to $0.5$. – Dieter Kadelka Feb 25 '20 at 23:33
  • @Sebastien Palcoux Can you tell which arithmetic you use for your calculations? I suspect that your results are not artefacts resulting from imprecise arithmetic, the results are "too good". – Dieter Kadelka Feb 25 '20 at 23:37
  • @KasperAndersen For f[n_]:=Floor[2^Mod[n,2] n/Sqrt[2]] on Mathematica I get for $f^{\circ2584}(73)$ the value$$670900640024419955928600637257.$$This is almost half of your value but not quite:$$1341801280048839911857201274496/2=670900640024419955928600637248$$ – მამუკა ჯიბლაძე Feb 26 '20 at 08:43
  • 6
    @მამუკა ჯიბლაძე : Sorry for the confusion. I though I was using Magma for the computation with quite large precision, unfortunately the way to compute $\sqrt{2}$ is not RealField(d)!Sqrt(2) but Sqrt(RealField(d)!2) (d is the precision). So now my results are in happy agreement with the table and also with your result for $f^{\circ 2584}(73)$. – Kasper Andersen Feb 26 '20 at 10:54
  • @Wolfgang: Done! – Sebastien Palcoux Feb 26 '20 at 16:26
  • Can one somehow count to how many orbits do numbers up to $n$ belong? For example, you say that you only found four orbits that end in a cycle (or loop). Also $73,103,104,105,141,145,146,147,148,149,199$ are all on the same orbit; and $107,151,152,153,155$ are on the same orbit too. Is it clear that these are not all on a single orbit? Is it even clear whether there are infinitely many orbits? – მამუკა ჯიბლაძე Feb 26 '20 at 16:38
  • @მამუკაჯიბლაძე: I did not say that I found only four orbits reaching a cycle, I said that I found four cycles, and that there may have no other. Now this depends on what you mean by orbit... anyway, it seems relevant to ask whether the partition associated to the equivalence relation generated by $f(n)=m$, admits finitely many components. – Sebastien Palcoux Feb 26 '20 at 16:59
  • Yes thanks that's what I mean. Clearly any distinct cycles belong to separate equivalence classes in your sense; thus so far you established that there are at least four distinct equivalence classes. But if I understand correctly, so far it is not even clear whether there are any other equivalence classes? – მამუკა ჯიბლაძე Feb 26 '20 at 17:03
  • 2
    @მამუკაჯიბლაძე: Four distinct equivalence classes reaching cycles, and at least one which does no. The question is : are all the numbers non-reaching a cycle, in the same class? Very interesting! – Sebastien Palcoux Feb 26 '20 at 17:08
  • Thanks. The "miniman" 5-cycle should probably be 12,8,5,7,9. – Wolfgang Feb 26 '20 at 17:34
  • @Wolfgang Yes, fixed! – Sebastien Palcoux Feb 26 '20 at 17:52
  • Actually $73$ and $107$ are in the same equivalence class: $f^{\circ57}(107)=f(f(73))$ – მამუკა ჯიბლაძე Feb 26 '20 at 19:20
  • 2
    @მამუკაჯიბლაძე: I tried $73$ and $123456789$, but up to the 10000th iteration they share no point. If we assume that for a subset of numbers $n$ of natural density one, the sequence $f^{\circ r}(n)$ has exponential grows (expected $\sim 1.02^r$), then I think that an argument of density can prove that there must have infinitely many equivalence classes. – Sebastien Palcoux Feb 27 '20 at 02:56
  • Going in a different direction: If you take disjoint intervals, say $I_r:=[5\sqrt2^{,r},5\sqrt2^{,r+1}]$, then the "point clouds" in your diagrams of the $n$'s reaching a cycle should be entirely inside each $I_r$. If you count those points in each $I_r$, how do their counts grow? I guess like some $\alpha^r$, but with $\alpha$ much smaller than $\sqrt2$. – Wolfgang Feb 27 '20 at 07:00
  • @Wolfgang: This $5$ comes from where? Should it be exact or an approximation? Here is the counting your asked: [6, 1, 2, 3, 5, 7, 11, 15, 22, 28, 36, 42, 50, 61, 74, 90, 105, 122, 137, 153, 168, 182, 193] together with the corresponding approximation of your $\alpha$ (different from what is called $\alpha$ in the post):[6.00, 1.00, 1.26, 1.32, 1.38, 1.38, 1.41, 1.40, 1.41, 1.40, 1.39, 1.37, 1.35, 1.34, 1.33, 1.32, 1.31, 1.31, 1.30, 1.29, 1.28, 1.27, 1.26]. – Sebastien Palcoux Feb 27 '20 at 08:34
  • Thank you! $5$ is just an approximation. Oh yes, "$\alpha$" was a bit unfortunate. Say $\beta$ then, and rather $\beta^{r+3}$ than $\beta^{r}$, as $5\approx\sqrt2^{,3}$ (always expecting such things to start "at the origin"...). But that should make hardly a difference, of course. The choice of my intervals (with "$5$") was just to exclude "boundary effects", as those points come already in clouds, so why not try to capture those clouds inside the intervals.
    Now I see that this $\beta$ is after all not that much smaller than $\sqrt2$, but it does seem to slowly decrease. Till where??
    – Wolfgang Feb 27 '20 at 10:13
  • BTW I guess I thought of those intervals because of the "self similar sequences" that came to my mind when seeing your images. – Wolfgang Feb 27 '20 at 10:26
  • 2
    Using Pari/GP and setting internal precision to 15000 decimal digits I could get $p_1(10^6) = 0.534524$ The value of $f°^{1e6}(73)$ is about 3.89439e10394, (thus one needs such a big precision) and the $\log_{73}()$ of it is 5578.52. – Gottfried Helms Feb 28 '20 at 13:37
  • 1
    @GottfriedHelms: Thanks! My laptop was not able to get $ℓ=6$. The computation for general $\ell$ should require a precision of about $1.5 \times 10^{\ell-2}$ decimal digits. – Sebastien Palcoux Feb 28 '20 at 14:01
  • Are there cases, which have no precedessor? For instance, $a_1=300$ traverses to $73$ by $6$ steps - but I didn't find a number $a_0$ which goes to $a_1=300$ . If there is indeed no such number - can we characterize the type of numbers which have no precedessor? – Gottfried Helms Feb 28 '20 at 14:55
  • 1
    @მამუკაჯიბლაძე: you may be interested in the answer I just posted. – Sebastien Palcoux Dec 18 '20 at 04:42
  • 1
    @GottfriedHelms: you may be interested in the answer I just posted. – Sebastien Palcoux Dec 18 '20 at 04:42

3 Answers3

7

A list of predecessors as mentioned in my comment.

I document pairs of $(m,n)$ for consecutive $m$ and their 1-step predecessors $n$ such that $f(n)=m$. The value $n=0$ indicates, that $m$ has no predecessor. I didn't reflect, that one $m$ can have two predecessors, but if $n/2$ is odd, then $n/2$ is a second predecessor.(This makes the table more interesting, because all odd predecessors $n$ are overwritten by the even predecessors $2n$...

Moreover, a nearly periodic structure occurs. I tried to resemble this by the arrangement of three or four columns of $(m,n)$ such that the first column contains all $m$ which have no predecessor. The basic pattern is not really periodic, but has super-patterns which again seem to be periodic but actually aren't. This pattern-superpattern-structure is also recursive. It reminds me of a similar structure when I looked at $\beta=\log_2(3)$ and found a similar style of pattern-superpattern-supersuperpattern-... and is there related to the continued fraction of $\beta$.
So I think we'll get no nice description for the cases $m$ which have no predecessor...

 m     n         m     n         m     n         m    n
------------------------------------------------------------- 
                 1     2         2     4    
 3     0         4     6         5     8    
 6     0         7    10         8    12         9    14    

10     0        11    16        12    18    
13     0        14    20        15    22        16    24    

17     0        18    26        19    28    
20     0        21    30        22    32    
23     0        24    34        25    36        26    38    

27     0        28    40        29    42    
30     0        31    44        32    46        33    48    

34     0        35    50        36    52    
37     0        38    54        39    56    
40     0        41    58        42    60        43    62    

44     0        45    64        46    66    
47     0        48    68        49    70        50    72    

51     0        52    74        53    76    
54     0        55    78        56    80        57    82    

58     0        59    84        60    86    
61     0        62    88        63    90    
64     0        65    92        66    94        67    96    

Update Some more explanation on the idea of "recursive aperiodic pattern". If we list the values $m$ which have no predecessor, we get

m_k:   3, 6,10,13, 17,20,23,27,30,... 

Writing the differences (I have prepended a zero-value to the above list of $m_k$)

    ,3,3,4  ,3,4  ,3,3,4  ,3,4  ,3,3,4  ,3,4  ,3,4  ,3,3,4 , ...      

We note, that we have a pattern of two different words: 3,3,4 and 3,4 repeating, but aperiodical. Let's denote the longer one with the capital A and the shorter one with the small a (and A means a difference of 10 and a of 7).
We get

 Aa Aa Aaa 
 Aa Aaa 
 Aa Aa Aaa 
 Aa Aaa
 Aa Aa Aaa
 Aa Aaa
 Aa ...       

Again we find only two kind of "words". Let's them shorten by Aaa=B and Aa=b. B means now a difference of 24, b of 17. Then we get

   bbB bB
   bbB bB
   bbB bB bB
   bbB bB
   bbB bB bB
   bbB bB
   bbB bB
   bbB bB bB
   ... 

Next obvious step gives

   Cc Cc Ccc
   Cc Ccc
   Cc Cc Ccc
   Cc Ccc
   Cc Cc Ccc
   Cc Ccc
   ... 

with c representing a difference of 17+24=41 and C of 17+17+24=58.
And so on.
If I recall correctly, then with the mentioned case of working with $\beta = \log_2(3)$ the same style of recursive pattern reflected the convergents of the continued fractions of $\beta$.
The first few differences here match the convergents of the continued fraction of $\sqrt2$ so far:

            a    b    c                    short patterns
 -------------------------------------
[1  1  3    7    17   41  99   239   577  ...  ]  convergents of contfrac(sqrt(2))
[0  1  2    5    12   29  70   169   408  ...  ] 
 -------------------------------------...
          A/2   B/2   C/2                  long patterns              

Update 2 The above can be explained by the following:

  • a number of the form $\lfloor2k\sqrt2\rfloor$ has exactly one predecessor $4k$;
  • a number of the form $\lfloor(2k-1)\sqrt2\rfloor$ has exactly two predecessors $2k-1$ and $4k-2$;
  • a number has no predecessors iff it has form $\lfloor n(2+\sqrt2)\rfloor$.

The first two statements are easily checked, while the third follows from the Beatty theorem, as explained in another answer by @Dattier

Update 3 Using a back-step algorithm (recursive) it seems I've got the predecessing tree of $m=73$. If no bugs, then this tree would also be complete. (But my routine may still be buggy, please check the results!)

The back-steps go from top-right south-west (antidiagonal) downwards. When there are two possible predecessors, they occur in the same column, but on separate rows.
If there is a predecessor without further predecessor, a short line (---) is printed.

                                    73   <--- start
                                104   
                            148   
                        105 ---

                        210   
                    149   
                212   
            300 ---

                    298   
                211 ---

                422   
            299   
        424   
    600 ---

            598   
        423 ---

        846 ---
    ---------------------------- tree seems to be complete (please check for errors!)
Gottfried Helms
  • 5,214
  • 1
  • 21
  • 36
  • 1
    Interesting! A number $m$ admits no predecessor iff the interval $[m\sqrt{2},(m+1)\sqrt{2}]$ admits no even number and the interval $[m/\sqrt{2},(m+1)/\sqrt{2}]$ admits no odd number. There are exactly $r_{\ell}$ such numbers $m<10^{\ell}$ with $\ell=1,2,3,4,5,6$ and $r_{\ell}=2,29,292,2928,29289,292893$. Strangely, for $\ell \le 6$ we observe that $r_{\ell-1} = ⌊ r_{\ell}/10⌋$, it should be true in general, I don't know why but it should be due to some pattern. – Sebastien Palcoux Feb 28 '20 at 16:51
  • 2
    Numbers without predecessors are described in the answer by @Dattier as those of the form $\lfloor n(2+\sqrt2)\rfloor$ – მამუკა ჯიბლაძე Feb 29 '20 at 07:28
  • 1
    Moreover it seems that numbers with one predecessor are those of the form $\lfloor2k\sqrt2\rfloor$ and numbers with two predecessors those of the form $\lfloor(2k-1)\sqrt2\rfloor$ – მამუკა ჯიბლაძე Feb 29 '20 at 07:56
  • 1
    In fact, predecessor of $\lfloor2k\sqrt2\rfloor$ is $4k$ while predecessors of $\lfloor(2k-1)\sqrt2\rfloor$ are $2k-1$ and $4k-2$. – მამუკა ჯიბლაძე Feb 29 '20 at 08:02
  • @მამუკაჯიბლაძე - very nice observations, indeed! For ease-of-reading: should I incorporate your results in my answer or even better won't you like to compose an own answer of yours? – Gottfried Helms Feb 29 '20 at 08:10
  • 1
    I believe it is better to extend your answer, not to have too many partial answers. Note that relevant keywords here are Sturmian sequences and Beatty sequences (the latter mentioned in another answer) – მამუკა ჯიბლაძე Feb 29 '20 at 08:12
  • @მამუკაჯიბლაძე - you might even incorporate this in my answer yourself. (I'm already a bit tired due to eyes-weakness) – Gottfried Helms Feb 29 '20 at 08:18
  • Well I added it to the end although I am not sure how to do it better – მამუკა ჯიბლაძე Feb 29 '20 at 08:26
  • @მამუკაჯიბლაძე - all well, thanks! – Gottfried Helms Feb 29 '20 at 08:28
  • 1
    @მამუკაჯიბლაძე: the formula for the numbers without predecessor you pointed out from Dattier answer explains the cardinal I computed in the first comment as $1/(2+\sqrt{2}) = 1-\sqrt{2}/2 = 0.2928932188\cdots$ – Sebastien Palcoux Feb 29 '20 at 09:54
  • Do you expect that these investigations will lead to an answer of the question? – Sebastien Palcoux Feb 29 '20 at 12:49
  • @Sebastien - primarily I try to get familiar with the whole system. My backtracer showed (immediately) that the predecessor-tree of the small cycle (at $5$) is finite - which is different compaed to the Collatz-type tree-structures. For the 3x-1 and the 5x+1 problem the predecessortrees of all cycles are always infinite. Q: has also the 33-step tree here only a finite precedessor-tree? The backtracing-algorithm might give numerical evidence if it is so. If all cycles must have finite predecessor-trees, are there more cycles? And so on... It's just exploratory so far. – Gottfried Helms Feb 29 '20 at 13:26
  • ... but, well, of course I don't want to spoil your track so far! – Gottfried Helms Feb 29 '20 at 13:33
  • No problem, I appreciate your investigation, I just wanted to know whether you have expectation. – Sebastien Palcoux Feb 29 '20 at 13:36
  • 1
    All branches going to the cycle $15,21,...,22,15$ are finite except, possibly, one, joining the cycle at $80=f(114)$. Successive $f$-inverse images of ${114}$ go like$${81,162},{115,230},{163,164,326},{231,232,462},{327,328,654},{463,464,926},{655,656,1310},{927,928,1854},{1311,1312,2622},{1856},{1313,2626},{929,1857,1858,3714},...$$ – მამუკა ჯიბლაძე Feb 29 '20 at 14:16
  • I confirm the calculation with the backtracking tree of $73$. The backtracking tree for $f(73)=103$ is finite too. However that for $f(f(73))=145$ seems to be infinite – მამუკა ჯიბლაძე Feb 29 '20 at 14:19
  • 1
    Just in case - a quick algorithm for $f^{-1}(n)$: let $m=\lceil\frac n{\sqrt2}\rceil$. Then $f^{-1}(n)=\varnothing$ if $m\sqrt2\geqslant n+1$; otherwise it is ${2m}$ if $m$ is even and ${m,2m}$ if $m$ is odd. – მამუკა ჯიბლაძე Feb 29 '20 at 15:11
  • 80is the only entry of the 33-step cycle which has not short finiteteness in backtracing. Document: 80 (part of 33-step-cycle)
    ...<-- 114 <-- 162 <-- 230 <-- 326 <-- 462 <-- 654 <-- 926 <-- 1310 <-- 927 <-- 1312 <-- 1856 <-- 1313 <-- 929 <-- 657 : split into two branches of *each infinite?* Branch 1 (657)<-- 465 <-- 329 <-- 233 <-- 165 <-- 234 <-- 332 <-- 470 <-- 333 <-- 472 <-- 668 <-- infinite ???
    Branch 2 (657)<-- 930 <-- 1316 <-- 1862 <-- 2634 <-- 3726 <-- 5270 <-- 7454 <-- 10542 <-- 7455 <-- 10544 <-- infinite ???
    – Gottfried Helms Feb 29 '20 at 15:18
  • Seems like the inverse tree for 114 has infinitely many infinite branches, as well as infinitely many finite ones – მამუკა ჯიბლაძე Feb 29 '20 at 15:33
  • 1
    Hmm having calculated more now I am not so sure about it. Actually your second branch dies out in 52 steps after 10544. However the first branch further branches at 1449 into 1025 <-- 725 <-- 513 <-- 363 <-- 514 <--... and 2050 <-- 2900 <-- 4102 <-- 5802 <-- 8206 <--...; whether either of them is infinite, I don't know – მამუკა ჯიბლაძე Feb 29 '20 at 16:56
  • 1
    @მამუკაჯიბლაძე - I began to guess, that there will be exactly 1 infinite backwards branch begining at 80 - just because I looked at the patterns of the died/finite branches. Because the steps increase by a relatively small amount the elements of a branch lie near together and this might be too dense for the existence of multiple infinite branches. But I've not yet an estimate for the according densities/distributions. The reverse Collatz-tree has a much wider stepping and so any subtree can have/has infinite subtrees. – Gottfried Helms Feb 29 '20 at 17:04
  • 1
    Difficult to say. The branch at 1025 branches still further at 195185 while the branch at 2050 branches at 16417, etc. I don't see any method to guess whether any of these branches goes on infinitely – მამუკა ჯიბლაძე Feb 29 '20 at 17:45
5

Here is a heuristic answer inspired by this comment of Lucia.

First, let assume that the probabilty for an integer $n$ to be odd is $\frac{1}{2}$, and that the probabilty for $f(n)$ to be odd when $n$ is even (resp. odd) is also $\frac{1}{2}$. We will observe that (surprisingly) it is no more $\frac{1}{2}$ for $f^{\circ r}(n)$ when $r \ge 2$ (in some sense, the probability does not commute with the composition of $f$ with itself).

  • if $n$ and $m=f(n)$ are even: note that $\frac{n}{\sqrt{2}} = m+\theta$ (with $0 < \theta < 1$) so that $m=\frac{n}{\sqrt{2}}- \theta$, then $$f^{\circ 2}(n) = f(m) = \left \lfloor{\frac{m}{\sqrt{2}}} \right \rfloor = \left \lfloor{\frac{\frac{n}{\sqrt{2}}- \theta}{\sqrt{2}}} \right \rfloor = \left \lfloor \frac{n}{2} - \frac{\theta}{\sqrt{2}}\right \rfloor$$ but $\frac{n}{2}$ is even with probability $\frac{1}{2}$, so in this case, $f^{\circ 2}(n)$ is odd with probability $\frac{1}{2}$.

  • if $n$ is even and $m=f(n)$ is odd: $$f^{\circ 2}(n) = f(m) = \left \lfloor\sqrt{2}m \right \rfloor = \left \lfloor \sqrt{2}(\frac{n}{\sqrt{2}} - \theta) \right \rfloor = \left \lfloor n - \sqrt{2} \theta) \right \rfloor$$ but $n$ is even and the probability for $0<\sqrt{2} \theta<1$ is $\frac{\sqrt{2}}{2}$ (because $\theta$ is assumed statistically equidistributed on the open interval $(0,1)$), so $f^{\circ 2}(n)$ is odd with probability $\frac{\sqrt{2}}{2}$.

  • if $n$ is odd and $m=f(n)$ is even:
    $$f^{\circ 2}(n) = f(m) = \left \lfloor{\frac{m}{\sqrt{2}}} \right \rfloor = \left \lfloor{\frac{\sqrt{2}n-\theta}{\sqrt{2}}} \right \rfloor = \left \lfloor n - \frac{\theta}{\sqrt{2}} \right \rfloor $$ but $n$ is odd and $0 < \frac{\theta}{\sqrt{2}}<1$, so $f^{\circ 2}(n)$ is even.

  • if $n$ is odd and $m=f(n)$ is odd:
    $$f^{\circ 2}(n) = f(m) = \left \lfloor \sqrt{2} m \right \rfloor = \left \lfloor \sqrt{2} (\sqrt{2}n-\theta) \right \rfloor = \left \lfloor 2n - \sqrt{2} \theta \right \rfloor $$ but $2n$ is even and the probability for $0<\sqrt{2} \theta<1$ is $\frac{\sqrt{2}}{2}$, so $f^{\circ 2}(n)$ is odd with probability $\frac{\sqrt{2}}{2}$.

By combining these four cases together, we deduce that the probability for $f^{\circ 2}(n)$ to be odd is $$\frac{1}{2} \times \frac{1}{2} \times (\frac{1}{2} + \frac{\sqrt{2}}{2} + 0 + \frac{\sqrt{2}}{2}) = \frac{2\sqrt{2}+1}{8}$$

By continuing in the same way, we get that the probability for $f^{\circ 3}(n)$ to be odd is:
$$ \frac{1}{4} (\frac{1}{2}\frac{1}{2} + \frac{1}{2}\frac{\sqrt{2}}{2} + \frac{\sqrt{2}}{2}\frac{\sqrt{2}}{2} + 1\frac{1}{2} + \frac{\sqrt{2}}{2}\frac{\sqrt{2}}{2}) = \frac{\sqrt{2}+7}{16}$$

For $2 \le r \le 24$, we computed the probability $p_r$ for $f^{\circ r}(n)$ to be odd (see Appendix). It seems (experimentally) that $p_r$ converges to a number $\simeq 0.532288725 \simeq \frac{8+3\sqrt{2}}{23}$ by Inverse Symbolic Calculator. This leads to the following question/conjecture:
$$\lim_{r \to \infty}p_r = \frac{8+3\sqrt{2}}{23} \ \ ?$$

If so, consider the number $\alpha$ mentioned in the main post, then $$\alpha = 1-\frac{8+3\sqrt{2}}{23} = \frac{15-3\sqrt{2}}{23} \simeq 0.467711,$$ which matches with the computation in the main post. And next, we would have:
$$ \delta = \frac{\sqrt{2}}{2^{\alpha}}= 2^{\frac{1}{2}-\alpha} = 2^{\frac{6\sqrt{2}-7}{46}} \simeq 1.022633$$


Appendix

Computation

sage: for i in range(3,26):
....:     print(sq2(i))
....:
[1/4*sqrt(2) + 1/8, 0.478553390593274]
[1/16*sqrt(2) + 7/16, 0.525888347648318]
[3/32*sqrt(2) + 13/32, 0.538832521472478]
[15/64*sqrt(2) + 13/64, 0.534581303681194]
[5/128*sqrt(2) + 61/128, 0.531805217280199]
[39/256*sqrt(2) + 81/256, 0.531852847392776]
[93/512*sqrt(2) + 141/512, 0.532269260352925]
[51/1024*sqrt(2) + 473/1024, 0.532348527032254]
[377/2048*sqrt(2) + 557/2048, 0.532303961432938]
[551/4096*sqrt(2) + 1401/4096, 0.532283123258685]
[653/8192*sqrt(2) + 3437/8192, 0.532285334012406]
[3083/16384*sqrt(2) + 4361/16384, 0.532288843554459]
[3409/32768*sqrt(2) + 12621/32768, 0.532289246647030]
[7407/65536*sqrt(2) + 24409/65536, 0.532288816169701]
[22805/131072*sqrt(2) + 37517/131072, 0.532288667983386]
[24307/262144*sqrt(2) + 105161/262144, 0.532288700334941]
[72761/524288*sqrt(2) + 176173/524288, 0.532288728736551]
[159959/1048576*sqrt(2) + 331929/1048576, 0.532288729880941]
[202621/2097152*sqrt(2) + 829741/2097152, 0.532288725958633]
[639131/4194304*sqrt(2) + 1328713/4194304, 0.532288724978704]
[1114081/8388608*sqrt(2) + 2889613/8388608, 0.532288725350163]
[1825983/16777216*sqrt(2) + 6347993/16777216, 0.532288725570602]
[5183461/33554432*sqrt(2) + 10530125/33554432, 0.532288725561857]

Code

def sq2(n):
    c=0
    for i in range(2^n):
        l=list(Integer(i).digits(base=2,padto=n))
        if l[-1]==1:
            cc=1/4
            for j in range(n-2):
                ll=[l[j],l[j+1],l[j+2]]
                if ll==[0,0,0]:
                    cc*=1/2
                if ll==[0,0,1]:
                    cc*=1/2
                if ll==[0,1,0]:
                    cc*=(1-sqrt(2)/2)
                if ll==[0,1,1]:
                    cc*=sqrt(2)/2
                if ll==[1,0,0]:
                    cc*=1
                if ll==[1,0,1]:
                    cc=0
                    break
                if ll==[1,1,0]:
                    cc*=(1-sqrt(2)/2)
                if ll==[1,1,1]:
                    cc*=sqrt(2)/2
            c+=cc
    return [c.expand(),c.n()]
2

You can say with Beatty theorem : $A=\{E(n(\sqrt{2}+2)) \text{ ; } n\in\mathbb N^*\}$ and $B=\{E(n \sqrt{2});n\in\mathbb N^*\}$ is a partition of $\mathbb N^*$

And we have $E(n(\sqrt{2}+2))=2n+E(n\sqrt{2})$

with $E$ is the function integer part

Dattier
  • 3,759
  • 1
  • 15
  • 40