3
$\begingroup$

Consider the following family of functions $$f_n(w):=\sum_{k=0}^{\infty}\frac{(-1)^{k-1}}{k!}(k+n)^{k-1}w^k.$$

QUESTION 1. Does the following hold? $$f_n(w)=-\frac1{n(f_{-1}(w))^n}.$$

Deeper look:

QUESTION 2. Is there a conceptual reason why the linear shift, in $n$ of $f_n$, translates into reciprocal power?


UPDATE. Thanks to Stanley's answer below, I explored the web and have found very useful resources on "polynomials of binomial type" which I post for the interested reader.

A paper by Gian-Carlo Rota et al

Slides by Richard P. Stanley

$\endgroup$
0

3 Answers 3

6
$\begingroup$

It is well-known that the Abel polynomials $p_k(x)=x(x-ak)^{k-1}$ are a sequence of polynomials of binomial type, i.e., $$ \sum_{k\geq 0} p_k(x)\frac{w^k}{k!} = \left( \sum_{k\geq 0} p_k(1)\frac{w^k}{k!}\right)^x, $$ which explains your formula.

$\endgroup$
1
  • $\begingroup$ Very nice! Thanks. That covers many polynomials, and in the present question the identity becomes valid not just for integral values of $n$ but also for an indeterminate $x$ (at least, formally). $\endgroup$ Commented Dec 9, 2016 at 0:47
2
$\begingroup$

It is also worth observing that $W(z):=-zf_1(z)$ is the Lambert function, the inverse function at $z=0$ of $ze^z$, whose power series expansion can be deduced by means of the Lagrange inversion theorem; and the expansion of any power of it as well, from which your formula follows.

$\endgroup$
2
  • 1
    $\begingroup$ Typo: it should be $W(z)=-zf_1(z)$ $\endgroup$
    – esg
    Commented Dec 11, 2016 at 16:16
  • $\begingroup$ Eqn. 2.36 on p. 338 of "On the Lambert function" by Corless et al. $\endgroup$ Commented Dec 12, 2016 at 13:44
0
$\begingroup$

This is unusual but I've also found an answer inspired by the above solutions.

Abel's formula $$\sum_{k=0}^n\binom{n}kx(x-kz)^{k-1}(y+nz)(y+kz)^{n-k-1}=(x+y+nz)(x+y)^{n-1}.$$ Lemma. Denote $c=f_{-1}(w)$ and let $i\in\Bbb{Z}$. Except for $f_0(w)$, we have the identities $$\frac{if_i(w)}{(i+1)f_{i+1}(w)}=c \qquad\text{or} \qquad f_i(w)=-\frac1{ic^i}.$$ Proof. For $(i+1)cf_{i+1}=if_i$, apply Cauchy's product formula: $$\begin{align} (i+1)\,c\, f_{i+1}&=(i+1)\sum_k\frac{(-1)^{k-1}}{k!}(k-1)^{k-1}w^k\sum_k\frac{(-1)^{k-1}}{k!}(k+i+1)^{k-1}w^k \\ &=(i+1)\sum_n\frac{(-1)^nw^n}{n!}\sum_{k=0}^n\binom{n}k(k-1)^{k-1}(n+i+1-k)^{n-k-1} \\ &=i\sum_n\frac{(-1)^{n-1}w^n}{n!}(n+i)^{n-1}=i\cdot f_i; \end{align}$$ where Abel's formula has been used with $x=-1, y=n+i+1, z=-1$. For the second part, it suffices to verify $cf_1(w)=-1$ and iterate applying the part we just proved. But, $cf_1(w)=-1$ is immediate from Abel's formula (choose $x=-1, y=n+1, z=-1$). The proof is complete.

$\endgroup$

You must log in to answer this question.

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