4
$\begingroup$

Let

  • $f(n,m)$ be a function such that $$ f(n,m) = mf(n-1,m) + 1, \\ f(0,m) = 1. $$
  • $T(n,m)$ be a coefficients such that $$ T(n,m) = n! \sum\limits_{k=1}^{n} \frac{m^{k-1} n^{n-k-1}}{(n-k)!}. $$ See related sequences A001865 ($m=1$) and A133297 ($m=-1$).
  • $R(n,k,m)$ be a coefficients such that $$ R(n,k,m) = nR(n-1,k,m) + (k+1)R(n-1,k+1,m) + R(n,k-1,m), \\ R(n,0,m) = nR(n-1,0,m) + R(n-1,1,m), \\ R(0,k,m) = f(k,m). $$

I conjecture that for $n>0$ and arbitrary $m$ we have $T(n,m) = R(n-1,0,m)$.

Here is the PARI/GP program to check it numerically:

upto1(n,m) = my(v1); v1 = vector(n, i, i!*sum(k=1, i, m^(k-1)*i^(i-k-1)/(i-k)!)) upto2(n,m) = {my(v1, v2); v1 = vector(n, i, 0); v1[1] = 1; for(i=1, n-1, v1[i+1] = m*v1[i] + 1); v2 = vector(n, i, 0); v2[1] = 1; for(i=1, n-1, v1[1] = i*v1[1] + v1[2]; for(j=2, n-i, v1[j] = i*v1[j] + j*v1[j+1] + v1[j-1]); v2[i+1] = v1[1]); v2} test(n,m) = upto1(n,m) == upto2(n,m) 

Note that on my very old PC for $10^3$ terms it takes about $2$ minutes for upto1 and about $2$ seconds for upto2.

Is there a way to prove it?

New contributor
Notamathematician is a new contributor to this site. Take care in asking for clarification, commenting, and answering. Check out our Code of Conduct.
$\endgroup$

1 Answer 1

6
$\begingroup$

Take a fixed $m$ and first solve the recursion relation for $f(k,m)$, $$ f(k,m)= \frac{m^{k+1}-1}{m-1},\;\;m\neq 1, $$ with $\lim_{m\rightarrow 1} f(k,m)=k+1$.

Then define the generating function $$ G_n(z)=\sum_{k\ge 0} R(n,k,m)z^k. $$ The recursion relation and initial conditions give \begin{eqnarray*} &(1-z)G_n(z)=nG_{n-1}(z)+G_{n-1}'(z),\\ &G_0(z)=\sum_{k\ge 0} f(k,m)z^k=\frac{1}{(1-z)(1-mz)}. \end{eqnarray*} We introduce the exponential generating function $$H(x,z)=\sum_{n\geq 0}G_n(z)\frac{x^n}{n!},$$ which satisfies the differential equation $$(1-x-z)\frac{\partial H}{\partial x}-\frac{\partial H}{\partial z}=H(x,z).$$ The initial condition is $H(0,z)=G_0(z)$.

Mathematica solves the differential equation for $H(x,z)$ in terms of the (principal branch) Lambert W-function, which gives $$H(x,0)=\frac{W(-x)}{-x[W(-x)+1] [m W(-x)+1]},\;\;x\leq 1.$$ Since $$H(x,0)=\sum_{n\geq 0}R(n,0,m)\frac{x^n}{n!}, $$ we need to extract the term of order $x^n$ in the expansion of $H(x,0)$ in powers of $x$. This is done below. The result is \begin{eqnarray*} H(x,0)={}&\sum_{n\geq 0}\left(\sum_{r=1}^{n+1} \frac{m^{r-1}n!(n+1)^{n+1-r}}{(n+1-r)!}\right)\frac{x^n}{n!}. \end{eqnarray*} The sum over $r$ is $T(n+1,m)$, so we have proven the desired result $$R(n,0,m)=T(n+1,m).$$


Calculation of the series expansion of $H(x,0)$.

The Lambert W-function produces the expansions \begin{eqnarray*} &W(-x)^r=(-1)^r r\sum_{n=r}^\infty \frac{n^{n-r-1}}{(n-r)!}x^n,\\ &\frac{1}{W(-x)+1}=\sum_{n=0}^\infty\frac{n^n}{n!}x^n, \end{eqnarray*} which we substitute into \begin{eqnarray*} H(x,0)={}&\frac{1}{-x}\frac{1}{W(-x)+1}\sum_{r=1}^\infty (-m)^{r-1} W(-x)^{r}\\ ={}&\frac{1}{x}\left(\sum_{n=0}^\infty\frac{n^n}{n!}x^n\right)\sum_{r=1}^\infty m^{r-1}\left( r\sum_{n=r}^\infty \frac{n^{n-r-1}}{(n-r)!}x^n\right). \end{eqnarray*} Collecting powers of $x$, one arrives at \begin{eqnarray*} H(x,0)={}&\sum_{n=0}^{\infty}\left(\sum_{r=1}^{n+1}m^{r-1} s_{n,r}\right)\frac{x^n}{n!},\\ s_{n,r}={}&rn!\sum_{k=0}^{n+1-r}\frac{k^k}{k!}\frac{(n+1-k)^{n-k-r}}{(n+1-k-r)!}. \end{eqnarray*} It remains to evaluate $s_{n,r}$, which can be done with the help of Abel's binomial theorem. (Thanks to ChatGPT for pointing me to that classic result.) First rewrite the sum identically in terms of binomial coefficients. Define $N_r=n+1-r$,
$$s_{n,r}=\frac{(N_r+r-1)!}{N_r!}r\sum_{j=0}^{N_r}{{N_r}\choose j}(N_r-j)^{N_r-j}(r+j)^{j-1}. $$ Abel's formula gives $$r\sum_{j=0}^{N_r}{{N_r}\choose j}(N_r-j)^{N_r-j}(r+j)^{j-1}=(N_r+r)^{N_r}.$$ We thus arrive at $$s_{n,r}=\frac{n!}{(n+1-r)!}(n+1)^{n+1-r}.$$

$\endgroup$
5
  • $\begingroup$ Thank you for answer! It looks like question is quite simple. Can you figure out where is the problem? $\endgroup$ Commented 23 hours ago
  • 1
    $\begingroup$ I need more work to fix it, hopefully later tonight. $\endgroup$ Commented 21 hours ago
  • $\begingroup$ Nice! +1 for correct EGF. Can you finish it? $\endgroup$ Commented 8 hours ago
  • 1
    $\begingroup$ done; quite a bit more involved than anticipated. $\endgroup$ Commented 5 hours ago
  • $\begingroup$ Thank you very much! $\endgroup$ Commented 5 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.