4
$\begingroup$

I define the sequence $\{a_n\}_{n \ge 0}$ by the generating function $$ G(g, \xi) = \sum_{n \ge 0} a_n g^n = \frac{1}{g} \, \frac{4\bar g \, (2\bar g^2 - 3\bar g + \xi/2+1/2)} {(\xi-1) \, (2\bar g - \xi)}, $$ where $\bar g$ satisfies the cubic algebraic equation $$ \frac{\bar g \, (2\bar g - \xi/2 -1/2 )^2} {(2\bar g - 1)(2\bar g - \xi)} = g \, . $$

The first few coefficients (if we expand $G(g, \xi)$ in series and looking at the numerator ) are $ 1,\; 1,\; 4,\; 1,\; 1,\; 10,\; 22,\; 10,\; 1,\; 1,\; 18,\; 88,\; 150,\; 88,\; 18,\; 1,\; \dots $

Question I would like to find a linear recurrence (or linear differencial recurrence) satisfied by the coefficients $a_n$, for example of the general form $$ c_0(n)\,a_n + c_1(n)\,a_{n-1} + \cdots + c_k(n)\,a_{n-k} = 0, $$ where each $c_i(n)$ is a polynomial (or constant) in $n$.

And generation functional meets next functional equation $$G[ g, \xi]= 8- G[ g/\xi , 1/\xi]$$ and differential equation of third order: \begin{eqnarray} g^3 \frac{\partial^3 G}{\partial g^3} +3 g^2 (\xi -1) \frac{\partial^3 G}{ \partial g^2 \partial \xi} +3 g (\xi-1) \xi \frac{\partial^3 G}{ \partial g\partial^2 \xi } +(\xi^2-1) \xi \frac{\partial^3 G}{\partial \xi^3}+ \nonumber \\ +6 \left( g^2 \frac{\partial^2 G}{\partial g^2} +g (2 \xi-1) \frac{\partial^2 G}{ \partial \xi \partial g} + \xi^2 \frac{\partial^2 G}{\partial \xi^2} + g \frac{ \partial G}{\partial g}+ \xi\frac{\partial G}{\partial \xi}\right) =0 \, . \end{eqnarray}

I have derived that $G(g, \xi)$ satisfies a third–order algebraic differential equation, but I was unable to extract from it a closed–form recurrence for $\{a_n\}$. Standard tools for guessing recurrences did not yield a stable result.

$\endgroup$
8
  • 2
    $\begingroup$ Please check your question, $\xi$ is not used. For which values of $\kappa,m$ do you find the given coefficients? $\endgroup$ Commented Oct 21 at 18:19
  • 2
    $\begingroup$ I verified your third–order ADE with $ \frac{1}{8}\frac{(1+\xi)}{1-\xi}G\left(\frac{1}{8} g (1+\xi)^3,\xi \right) =\frac{1}{(1-\xi)^2} - \frac{\xi^2}{(1-\xi)^2}g + \xi^2 g^2 - \xi^2(1-4\xi+\xi^2)g^3 + O(g^4). $ Do you agree? $\endgroup$ Commented Oct 21 at 23:53
  • 1
    $\begingroup$ @Fred Hucht first terms of expansion $G = -\frac{8}{\xi ^2-1}+\frac{64 g \xi ^2}{(\xi -1) (\xi +1)^4}-\frac{512 g^2 \left((\xi -1) \xi ^2\right)}{(\xi +1)^7}+\frac{4096 g^3 (\xi -1) \xi ^2 \left(\xi ^2-4 \xi +1\right)}{(\xi +1)^{10}}-\frac{32768 g^4 \left((\xi -1) \xi ^2 \left(\xi ^4-10 \xi ^3+22 \xi ^2-10 \xi +1\right)\right)}{(\xi +1)^{13}}+O\left(g^{9/2}\right)$ $\endgroup$ Commented Oct 22 at 6:53
  • 1
    $\begingroup$ SergiiVoloshyn OK, i got it. The rewritten series by @Somos can be summed, for $\xi=-1$, to be equal to $\frac 1 {24g}\left[{}_2F_1\left(-\frac23,-\frac13;\frac12;\frac{27g}2\right)-1 \right].$ However, I don't know if this helps. $\endgroup$ Commented 2 days ago
  • 1
    $\begingroup$ It would be very good to know where you got $G(g,\xi)$ from. That is, why are you so interested in this particular function and its power series coefficients? $\endgroup$ Commented yesterday

2 Answers 2

4
$\begingroup$

"Guessed", not proven, but validated for the first 80 coefficients:

$$(1 + 3n + 2n^2) a_n = -\frac{2((10 - 27n + 23n^2) + (-4 + 54n - 62n^2)\xi + (10 - 27n + 23n^2)\xi^2)}{(\xi+1)^3} a_{n-1} - \frac{48((26 - 27n + 7n^2) + (-88 + 90n - 22n^2)\xi + (26 - 27n + 7n^2)\xi^2)}{(\xi+1)^4} a_{n-2} - \frac{128(n-4)(5n-13)}{(\xi+1)^3} a_{n-3} + \frac{1024(n-5)(n-4)}{(\xi+1)^4} a_{n-4}$$

I first used resultants to eliminate $\bar g$, obtaining $$16g^2(\xi-1)^3G^3 + 8g(\xi-1)^2(-8g^2-20g\xi+\xi^2+28g+2\xi+1)G^2 + (\xi-1)(512g^3\xi+192g^2\xi^2+24g\xi^3+\xi^4-512g^3-1024g^2\xi-88g\xi^2+4\xi^3+704g^2+40g\xi+6\xi^2+152g+4\xi+1)G + 8(64g^2\xi+16g\xi^2+\xi^3-64g^2-112g\xi+3\xi^2+88g+3\xi+1) = 0$$

This allows expansion to obtain coefficients, and using an ansatz for the recurrence, the first $15$ coefficients suffice to extract the guess.

$\endgroup$
1
  • $\begingroup$ Thank you for the answer! $\endgroup$ Commented 18 hours ago
3
$\begingroup$

The question is about the power series expansion of $G(g, \xi) = \sum_{n \ge 0} a_n g^n$ where $G$ satisfies an algebraic differential equation. I examined the first few coefficients and found

$$ \frac18\frac{1\!+\!\xi}{1\!-\!\xi} G\left(\frac{1}{8}g(1\!+\!\xi)^3,\xi \right) \!=\! \frac{1}{(1\!-\!\xi)^2} \!-\! \frac{\xi^2}{(1\!-\!\xi)^2}g \!+\! \xi^2\sum_{k=2}^\infty b_k(\xi) g^k \tag1 $$

where $b_k(\xi)$ is a palindromic polynomial in $\xi$ with integer coefficients of degree $2(k-2)$.

$$ b_2 = 1,\: b_3 = -(1-4\xi+\xi^2),\: b_4 = 1-10\xi+22\xi^2-10\xi^3+\xi^4,\: \dots \tag2 $$

Based on the A.D.E., and using the code in my answer to https://mathematica.stackexchange.com/questions/285008), my calculations suggest that

$$ 0 = c_0\,b_k + c_1\,b_{k+1} + c_2\,b_{k+2} + c_3\,b_{k+3} \tag3 $$

for all $k\ge2$ with

\begin{align} c_0 &= -k(k-1)(3k+8)(1+\xi)^6, \\ c_1 &= +k(18k^2+75k+67)(1+\xi)^4, \\ c_2 &= +(3k+7)(80(1-4\xi+\xi^2)+(3k^2+13k)(5\xi^2-26\xi+5)), \\ c_3 &= +4(k+4)(2k+7)(3k+5). \tag4 \end{align}

Note carefully that the recurrence in equation $(3)$ is valid for $k=0$ or $k=1$ no matter what $b_0$ or $b_1$ is. This is because of the factor of $k$ in $c_1$ and the factor $k(k-1)$ in $c_0$.

$\endgroup$
5
  • 1
    $\begingroup$ Interestingly, this recurrence also holds for $k\in\{0,1\}$. $\endgroup$ Commented 23 hours ago
  • $\begingroup$ I think, after the edit the statement is incorrect. To fulfil the recurrence for $k\geq0$, one has to set $b_0=(1-\xi)^{-2}\xi^{-2}$ and $b_1=-(1-\xi)^{-2}$, remove the two corresponding terms in (1) and run the sum over all $k\geq 0$. $\endgroup$ Commented 22 hours ago
  • 1
    $\begingroup$ @FredHucht Thanks for that comment! I will explain further in my answer. $\endgroup$ Commented 21 hours ago
  • $\begingroup$ @Somos Thank you for the answer! This sequence is a part of my research arxiv.org/abs/2507.00689 (formula 35 and below) $\endgroup$ Commented 18 hours ago
  • $\begingroup$ @Somos I believe that differential recursion (for the terms $G_k$ with denominators from $G=\sum_kG_k$ ) may be has more simple form. $\endgroup$ Commented 16 hours ago

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.