2
$\begingroup$

(Note: This third method continues from this post.)

There are level-$7$ pi formulas based on the McKay-Thompson series $T_{7A}$ and Cooper's $s_7$ sequence in this paper. This third method, among other things, will enable us to use $T_{7B}$.


I. Method 3

Given the binomial coefficient $\binom{n}{k}$, some free parameters $p, r,$ and a sequence $s_1(n)$. Define a second sequence as,

$$s_2(m) = \sum_{n=0}^m r^{m-pn}\binom{m}{pn} s_1(n)$$

Then we have the transformation,

$$\sum_{n=0}^{\infty} s_1(n)\,\frac{An+B}{C^n}=\left(\frac{C^{1/p}}{C^{1/p}+r}\right)^2\,\sum_{m=0}^{\infty} s_2(m)\,\frac{A/p\,m+ B-D_3}{(C^{1/p}+r)^m}$$

where,

$$D_3 = \frac{r\,(A/p-B)}{C^{1/p}}$$


II. Examples

Given the Dedekind eta function $\eta(\tau)$. First define the functions,

\begin{align} j_{7A}(\tau) &= \left(\sqrt{j_{7B}(\tau)}+\frac{7}{\sqrt{j_{7B}(\tau)}}\right)^2\\ j_{7B}(\tau) &= \left(\frac{\eta(\tau)}{\eta(7\tau)}\right)^4 \end{align}

Let $\tau = \frac{7+\sqrt{-427}}{14},$ note that $427 = 7\times61$, and we get,

\begin{align} j_{7A}(\tau) &= -22^3+1 = -(39\sqrt7)^2\\ j_{7B}(\tau) &= -7\left(\frac{39+5\sqrt{61}}{2}\right)^2 \end{align}

where the latter involves the fundamental unit $U_{61}$. We have Cooper's formula,

$$\frac{1}{\pi} = \frac{\sqrt7}{22^3}\sum_{j=0}^\infty s_7(j)\, \frac{11895j+1286}{(-22^3)^j}$$

However, we wish to find a sequence that uses the whole $j_{7A}(\tau) = -22^3+1$ as this will lead to a second sequence that uses $j_{7B}(\tau)$. Thus $r=1$, and applying Method 3, we get,

$$\frac{1}{\pi} = \frac{\sqrt7}{(-22^3+1)^2}\sum_{k=0}^\infty t_{7A}(k)\, \frac{22^3(11895k+1286)-(-22^3+39)}{(-22^3+1)^{k}}$$

Then using Method 1, we get,

$$\frac{1}{\pi} = \frac{1}{(-22^3+1)\sqrt{-\,j_{7B}}}\sum_{h=0}^\infty t_{7B}(h)\, \frac{1272437 - 207636\sqrt{61}(1+2h)}{(j_{7B})^{h}}$$

where $j_{7B} = -7\left(\frac{39+5\sqrt{61}}{2}\right)^2$ as above.


III. Sequences

Starting with Cooper's sequence,

\begin{align}s_7(j) &= \sum_{m=0}^j \binom{j}{m}^2\binom{2m}{j}\binom{j+m}{m}\\ &= 1, 4, 48, 760, 13840, 273504\dots \end{align}

we derive,

\begin{align}t_{7A}(k) &= \sum_{j=0}^k\binom{k}{j}\sum_{m=0}^j \binom{j}{m}^2\binom{2m}{j}\binom{j+m}{m}\quad\\ &= 1, 5, 57, 917, 17185, 350805\dots\quad \end{align}

\begin{align} t_{7B}(h) &= \sum_{k=0}^h(-7)^{h-k}\binom{h+k}{h-k}\sum_{j=0}^k\binom{k}{j}\sum_{m=0}^j \binom{j}{m}^2\binom{2m}{j}\binom{j+m}{m}\\ &= 1, -2, 1, 49, -602, 5257, -39095\dots \end{align}

The advantage of Cooper's sequence $s_{7}$ is it only has a 3-term recurrence relation. The recurrence status of $t_{7A}$ and $t_{7B}$ is unknown. However, we recover the nice relation,

$$\sum_{n=0}^\infty t_{7A}(n)\,\frac{1}{\;\big(j_{7A}\big)^{n+1/2}} = \sum_{n=0}^\infty t_{7B}(n)\,\frac{1}{\;\big(j_{7B}\big)^{n+1/2}}$$

with closed-forms for the sequences, so it is now found in levels $L = 1,2,3,4,6,7,8,10,$ (but not yet in $L=5,9$).


IV. Questions

  1. Like the previous ones, why does Method 3 work, and how free are its parameters $p,r$?
  2. Can the closed-forms of sequences $t_{7A}$ and $t_{7B}$ be simplified?
  3. Lastly, what are their recurrence relations? (I've tested them, got nowhere, and I think it is an $m$-term relation with coefficients as polynomials of deg-$n$ where $m,n>4$.)
$\endgroup$
0

2 Answers 2

2
$\begingroup$

For Question $3$ about the recurrence relations, using my code from MMA question 285008 for $a_n := T_{7A}(n)$ I used findseqrecur[4, 4, Array[t7A, 33, 1], 1, "a", k, -1] to get

$$ 0 = 14(n+1)(n+2)(2n+3) a_n \\ -3(n+2)(19n^2+76n+80) a_{n+1} \\ + 5(2n+5)(3n^2+15n+19) a_{n+2} \\ - (n+3)^3 a_{n+3}. $$

For $b_n := T_{7B}(n)$ there are several recurrences.

For degree $4$ polynomials I used findseqrecur[6, 5, Array[t7B, 34, 1], 1, "b", k, -4] to get

$$ 0 = -7^5(k-17)(k-2)^3 b_{k-3}\quad \\ -7^3(19k^4-678k+2218k^2-2640k+1113)b_{k-2}\;\; \\ -7(85k^4-8707k^3+9978k^2-7072k+2090)b_{k-1} \\ +(85k^4+8707k^3+9978k^2+7072k+2090)b_{k}\, \\ +(19k^4+678k+2218k^2+2640k+1113)b_{k+1} \\ \quad +(k+17)(k+2)^3 b_{k+2}. $$

For degree $3$ polynomials I used findseqrecur[8, 4, Array[t7B, 38, 1], 1, "b", k, -5] to get

$$ 0 = 7^7(k-3)^3b_{k-4}\; \\ + 7^5(47k^3-300k^2+646k-470)b_{k-3} \;\\ + 2\cdot 7^3(480k^3-1830k^2+2483k-1206)b_{k-2}\quad \\ + 7^2 (1578k^3-2001k^2+1513k-433)b_{k-1} \\ + 7 (1578k^3+2001k^2+1513k+433)b_{k}\;\; \\ \;\; + 2 (480k^3+1830k^2+2483k+1206)b_{k+1} \\ \; + (47k^3+300k^2+646k+470)b_{k+2} \\ \; + (k+3)^3 b_{k+3}. $$

Notice the symmetry of these two recursions.

$\endgroup$
5
  • 2
    $\begingroup$ Wait, this is getting interesting. It is possible for a sequence to have more than one recurrence relation? Let $R(k,n)$ be a $k$-term recurrence with polynomial coefficients of deg-$n$. So the second sequence has at least three, namely $R(6,4)$, $R(8,3)$ by Somos and $R(5,6)$ by G.Edgar? $\endgroup$ Commented May 17, 2023 at 14:54
  • $\begingroup$ My interest in looking at the recurrence relations was motivated by the Apery-like sequences found by Zagier. Of a certain form, they depend only on three parameters $a,b,c$. Cooper found a more general form with additional parameter $d$, and this enabled him to find the level-$7$ and level-$18$ pi formulas. I believe they are now checking for patterns in 4-term recurrence relations. $\endgroup$ Commented May 17, 2023 at 15:34
  • 2
    $\begingroup$ @TitoPiezasIII It is very possible to have more than one recurrence relation. I have encountered this for several OEIS sequences where the recurrence relation is a homogeneous polynomial in the sequence values with constant coefficients. A good example is Somos sequences. The Somos-4 sequence has many homogeneous recurrence relations that it satisfies. $\endgroup$
    – Somos
    Commented May 17, 2023 at 17:13
  • $\begingroup$ I added the alternative forms which exhibit some symmetry, and I hope you don't mind. I noticed it first in the level-10 recurrences here, then checked if it was also in the level-7 recurrences you found. Do you know what causes such symmetry? $\endgroup$ Commented Jun 8, 2023 at 11:03
  • 1
    $\begingroup$ @TitoPiezasIII Thanks for the edit! I had noticed the symmetry myself, but went with the computer generated recurrences. I think the cause is similar to the symmetry for Somos sequences. $\endgroup$
    – Somos
    Commented Jun 8, 2023 at 11:51
2
$\begingroup$

For question 3.

Using Maple's gfun library, I get this $4$-term recurrence for $t_{7A}$: \begin{align} 0 &=14\, \left( 2\,n+3 \right) \left( n+2 \right) \left( n+1 \right) u \left( n \right) \\ & -3\, \left( n+2 \right) \left( 19\,{n}^{2}+76\,n+80 \right) u \left( n+1 \right) \\ & +5\, \left( 2\,n+5 \right) \left( 3\,{n }^{2}+15\,n+19 \right) u \left( n+2 \right) \\ & - \left( n+3 \right) ^{3}u \left( n+3 \right) . \end{align}

and this $5$-term recurrence for $t_{7B}$: \begin{align} 0 &= 7^4\, \left( n+3 \right) \left( {n}^{2}+6\,n+35 \right) \left( n+1 \right) ^{3}u \left( n \right) \\ &+ \big( 1274\,{n}^{6}+17199\,{n}^{5}+ 129311\,{n}^{4}+551299\,{n}^{3} \\ &\qquad\qquad+1264592\,{n}^{2}+1459808\,n+668850 \big) u \left( n+1 \right) \\ &+ \big( 267\,{n}^{6}+4005\,{n}^{5}+ 32065\,{n}^{4}+153775\,{n}^{3} \\ &\qquad\qquad+421280\,{n}^{2}+601400\,n+346290 \big) u \left( n+2 \right) \\ &+ \left( 26\,{n}^{6}+429\,{n}^{5}+3614\, {n}^{4}+18779\,{n}^{3}+57893\,{n}^{2}+94588\,n+62265 \right) u \left( n+3 \right) \\ &+ \left( n+2 \right) \left( {n}^{2}+4\,n+30 \right) \left( n+4 \right) ^{3}u \left( n+4 \right) \end{align}

It seems this is different from the Somos recurrence. So these should be examined more carefully.

Machine-readable:
2401*(n+3)*(n^2+6*n+35)*(n+1)^3*u(n)+(1274*n^6+17199*n^5+129311*n^4+551299*n^3+ 1264592*n^2+1459808*n+668850)*u(n+1)+(267*n^6+4005*n^5+32065*n^4+153775*n^3+ 421280*n^2+601400*n+346290)*u(n+2)+(26*n^6+429*n^5+3614*n^4+18779*n^3+57893*n^2 +94588*n+62265)*u(n+3)+(n+2)*(n^2+4*n+30)*(n+4)^3*u(n+4)

$\endgroup$
5
  • $\begingroup$ Somos remarked that an integer sequence may have more than one recurrence. Let $R(k,n)$ be an $k$-term recurrence with polynomial coefficients of deg-$n$. So the second sequence has at least three, namely $R(6,4)$, $R(8,3)$ by Somos and $R(5,6)$ by G.Edgar? I always assumed that, like the Fibonacci sequence, there can only be one recurrence relation. $\endgroup$ Commented May 17, 2023 at 14:57
  • 2
    $\begingroup$ @TitoPiezasIII The Fibonacci sequence is a good example. There are very many recurrence relations that it satisfies. The defining relation $F_{n+2}=F_{n+1}+F_n$ is one but there is also $F_{n+3}=2F_{n+1}+F_n.$ $\endgroup$
    – Somos
    Commented May 17, 2023 at 17:19
  • 2
    $\begingroup$ Interesting. If we take $k$ large enough, can we get $n=0$? (Recurrence with constant coefficients has a more complete theory.) Or, on the other hand, if we take $n$ large enough can we get $k=3$? (Three-term recurrences are associated with continued fractions.) $\endgroup$ Commented May 17, 2023 at 17:49
  • $\begingroup$ That is a very good MO question. Some of this is known in theory. For example, the solutions of linear with constant coefficient recursions are known to be sums of simple exponentials with overall exponential growth rates. $\endgroup$
    – Somos
    Commented May 17, 2023 at 18:28
  • $\begingroup$ I used Zagier's six 3-term recurrences to find a new cfrac for $\zeta(3)$ (I hope). Kindly see this MO post. $\endgroup$ Commented May 19, 2023 at 9:29

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .