4
$\begingroup$

I'm hoping to find "exact" an solution to the following simple recursion:

$q_m(j+1) = m \cdot q_m(j) + j(j+m)\cdot q_m(j-1)$

with initial data $q_m(0) = 1$, $q_m(1)=m$, where $m \geq 0$ is an integer parameter.


Experimentally, the behavior seems to be slightly different based on the parity of $m$. For example, for even $m$ it seems as if $q_m(2j+1) = (2j+1)!! (2j+1+m)!! F_m(j)$ where $F_m$ is a polynomial in $j$ whose degree may depend on $m$. I can even write a recursive equation for $F_m$ that looks like

$(2j+1)(2j+1+m) F_m(j+1) =(m^2 + 2j(2j+m) + (2j-1)(2j-1+m))F_m(j) - ((2 j - 2) (2 j - 2 + m) F_m(j-1)$

with initial conditions $F(1) = m$ and $F(2) = (m^3+3m^2+5m)/(3(m+3))$, but it seems like a miracle that the solution ends up being a polynomial in the first place.

I would be also interested in relating solutions for different $m$'s, if an explicit description is out of reach.

$\endgroup$
2
  • 2
    $\begingroup$ The differential function $f(t):=\sum_j q_m(j)t^j$ satisfies some second order ODE, have you tried this way? $\endgroup$ – Fedor Petrov Feb 4 '19 at 12:18
  • $\begingroup$ Maple solves $q_0$ in closed form, but not $q_1$ or $q_2$. $\endgroup$ – Gerald Edgar Feb 4 '19 at 13:03
7
$\begingroup$

For exponential generating function, $$ f_m(z) = \sum_{j=0}^\infty \frac{q_m(j)}{j!}\;z^j $$ I get $$ f_m(z) = \frac{1}{(1-z)^{m+1/2}(1+z)^{1/2}} $$

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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