0

enter image description here

This book says it's easy, but to me, it's not. :(

As for 'at most k summands', in terms of Combinatorics, by using MSET(), $$ MSET_{\le k}(Positive Integer) = P^{1,2,3,...k}(z) = \prod_{m=1}^{k} \frac{1}{1-z^m} $$

because the number of summands and the largest of summands are interchangable.

I understood that by rotating enter image description here

So I think 'at most $k$ summands each at most $l$' means applying $\prod \frac{1}{1-z^m}$ 'TWICE', compositing $k$ and $l$.

And I tried many times but failed. I don't exactly know what 'the recurrence' in the book means.

How to get the Ordinary Generating Function above?

David Lee
  • 205

1 Answers1

2

Let $P(z;k,\ell)$ denote the ordinary generating function of the partitions $\mathcal P$ with at most $k$ parts each at most $\ell$.

Hint. By distinguishing the cases $i_\ell=0$ and $i_\ell\ge1$, show that $$P(z;k,\ell)=z^\ell P(z;k-1,\ell)+P(z;k,\ell-1).\tag{$\star$}$$ Then proceed by induction on $k+\ell\ge0$.


Solution. Such a partition $\mathcal P$ is uniquely determined by a $\ell$-tuple $(i_1,\ldots,i_\ell)$ where:

  1. for each $j\in\{1,\ldots,\ell\}$, $i_j\ge0$ indicates how many times part $j$ is repeated;
  2. $1i_1+2i_2+\cdots+\ell i_\ell=n$;
  3. $i_1+\cdots+i_\ell\le k$.

Therefore \begin{align*} P(z;k,\ell)\, \quad &=\sum_{i_1+\cdots+i_\ell\le k}z^{1i_1+2i_2+\cdots+\ell i_\ell}\\[.4em] &=\quad z^\ell\sum_{\substack{i_1+\cdots+i_\ell\le k\\[.2em]i_\ell\ge1}}z^{1i_1+2i_2+\cdots +\ell(i_\ell-1)} \quad+\sum_{\substack{i_1+\cdots+i_{\ell-1}\le k\\[.2em](i_\ell=0)}}z^{1i_1+2i_2+\cdots+(\ell-1)i_{\ell-1}}\\[.4em] &=\quad z^\ell\sum_{i_1+\cdots+i_\ell\le k-1}z^{1i_1+2i_2+\cdots+\ell i_\ell} \quad+\sum_{i_1+\cdots+i_{\ell-1}\le k}z^{1i_1+2i_2+\cdots+(\ell-1)i_{\ell-1}}\\[.4em] &=z^\ell\,P(z;k-1,\ell)+P(z;k,\ell-1).\tag{$\star$} \end{align*}

We now prove $$k+\ell=s\implies P(z;k,\ell)=\binom s{k,\ell}_z:=\dfrac{\prod\limits_{m=1}^s(1-z^m)}{\prod\limits_{i=1}^k(1-z^i)\cdot\prod\limits_{j=1}^\ell(1-z^j)}\tag{$\mathcal H_s$}$$ by induction on $s\ge0$.

We have $(\mathcal H_0)$ since $P(z;0,0)=1$ counts the empty partition. Let $s\ge1$ and suppose $(\mathcal H_{s-1})$. Let $k,\ell\ge0$ with $k+\ell=s$. Then \begin{align*} P(z;k,\ell)&\stackrel{(\star)}=z^\ell\,P(z;k-1,\ell)+P(z;k,\ell-1)\\[.4em] &\!{\smash{\stackrel{(\mathcal H_{s-1})}=}}z^\ell\binom{s-1}{k-1,\ell}_z+\binom{s-1}{k,\ell-1}_z\\[.4em] &\,=\left(\frac{1-z^s}{1-z^{s-\ell}}-\frac{1-z^\ell}{1-z^{s-\ell}}\right)\binom{s-1}{k-1,\ell}_z+\binom{s-1}{k,\ell-1}_z\\[.4em] &\,=\frac{1-z^s}{1-z^k}\binom{s-1}{k-1,\ell}_z+\underbrace{\binom{s-1}{k,\ell-1}_z-\frac{1-z^\ell}{1-z^k}\binom{s-1}{k-1,\ell}_z}_{=\,0}\\[.4em] &\,=\binom s{k,\ell}_z. \end{align*} Hence $(\mathcal H_s)$ is proved for all $s\ge0$.

Note. $(\star)$ is a $q$-analogue of Pascal's triangle formula. See Gaussian binomial coefficient.

nejimban
  • 4,107
  • Thank you very much. Your proof is perfect! WOW. I never thought of using induction like that. I really appreciate your answer. – David Lee Jan 22 '24 at 02:39
  • Is there any other way by direct combinatorial reasoning (deductively) rather than the induction? – David Lee Jan 24 '24 at 06:43
  • 1
    @DavidLee I asked myself the same question a while ago (and to date, it is actually the only question I posted on math.SE). See here (with a bit different notation). – nejimban Jan 24 '24 at 12:30