1

I am trying to find the generating function $A(x)$

$[x^n]A(x) = \sum_{a+b+c=n \\ a,b,c \geq 0 \\ a,b,c \in \mathbb{Z}}\frac{a^2b}{c}$

My initial thoughts are that the 3 variables can be their own series and can have their own generating functions, but I am struggling to see how the constraints on the variables would apply to their generating functions.

  • Welcome to mse! What have you tried? Do you have any ideas of your own? Once we have a better idea of exactly where you're struggling, we can help you better ^_^ – HallaSurvivor Apr 19 '23 at 03:13
  • Hi, I'm thinking that each of the variables can probably be in their own series and so each would have their own generating function. These separate generating functions can be multiplied together to get a single function. Im struggling with understanding how to account for the constraints – bananaboy Apr 19 '23 at 03:20
  • That's a great start! You should edit your question to include this, any other ideas, and some context for the problem. As it stands a lot of people are voting to close this question, and your added context and attempts will help keep this question open – HallaSurvivor Apr 19 '23 at 03:22
  • Got it, thank you! – bananaboy Apr 19 '23 at 03:27
  • Your thoughts are very correct and on the correct path! Have you tried to push forward and write out your ideas? As in, writing out the power series and expanding for the coefficients? Check out my answer, it should help you in future problems as well. – Gareth Ma Apr 19 '23 at 03:37
  • 1
    Please correct the misprint: $c$ must be $>0$ to avoid division by $0.$ – Anne Bauval Apr 19 '23 at 07:42
  • 1
    Again: please edit your post to correct this. – Anne Bauval Apr 20 '23 at 05:42

3 Answers3

3

Welcome to MSE. Observe that \begin{align}\left(\sum_{\alpha\ge0}x_\alpha\,t^\alpha\right)\left(\sum_{\beta\ge0}y_\beta\,t^\beta\right)\left(\sum_{\gamma\ge0}z_\gamma\,t^\gamma\right) &=\sum_{\alpha,\beta,\gamma\ge0}x_\alpha\,y_\beta\,z_\gamma\,t^{\alpha+\beta+\gamma}\\[.4em] &=\sum_{n=0}^\infty\left(\sum_{\substack{\alpha,\beta,\gamma\ge0\\\alpha+\beta+\gamma=n}}x_\alpha\,y_\beta\,z_\gamma\right)t^n. \end{align}

nejimban
  • 3,663
1

Here's how I would approach the problem. We have three variables $a, b, c$ which are all conveniently non-negative and sum to $n$, so it is clear that we want something of the form

$$ G(x) = (a_0 + a_1x + a_2x^2 + \cdots)(b_0 + b_1x + b_2x^2 + \cdots)(c_0 + c_1x + c_2x^2 + \cdots) $$

This way, we have $[x^n]G(x) = \sum_{i + j + k = n \\ i, j, k \in \mathbb{Z}_{\geq 0}} a_ib_jc_k$.

Hopefully now it becomes simple. If you still don't see it, note that $a_i$ just means a function in $i$, so you can write $a(i)$ if it's clearer. Then look at your summand $\frac{a^2b}{c}$ and see how it corresponds to $a_ib_jc_k$.

Also, if you don't know how to find the coefficient from this power series, I have written an answer on solving generating functions literally a week ago. You can find it here.

Hope this helps!

Gareth Ma
  • 3,725
  • Okay so I see that the series for the series with the i,j,k, its equivalent to $(\frac{x}{(1-x)^2})^3$ correct? – bananaboy Apr 19 '23 at 04:15
  • @bananaboy I don’t think so. You seem to have gotten $b_j$ correct but not the other two. You want $a_i = i^2$ and $c_k = \frac{1}{k}$. Now substitute these back into the generating functions. What do you get? I can help you when I get back on keyboard – Gareth Ma Apr 19 '23 at 18:17
0

After correction (see last comment above), $$A(x)=\sum_{a,b\ge0, c>0}\frac{a^2b}cx^{a+b+c}=F(x)G(x)H(x)$$ where $$F(x)=\sum_{k=0}^\infty k^2x^k,\quad G(x)=\sum_{k=0}^\infty kx^k,\quad H(x)=\sum_{k=1}^\infty\frac{x^k}k.$$ There are myriads of posts on this site to calculate formal power series of the form $\sum k^px^k$ where $p\in\Bbb Z,$ so let me recall only those useful here.

  • The particular case $p=0$ corresponds to the geometric power series: $\sum_{k=0}^\infty x^k=\frac1{1-x}.$
  • Differentiating it twice, we derive $\sum_{k=0}^\infty kx^{k-1}=\frac1{(1-x)^2}$ and $\sum_{k=0}^\infty k(k-1)x^{k-2}=\frac2{(1-x)^3},$ hence $$G(x)=\frac x{(1-x)^2}\quad\text{and}\quad F(x)=\frac{2x^2}{(1-x)^3}+G(x)=\frac{x^2+x}{(1-x)^3}.$$
  • Integrating it once, we derive $$H(x)=-\ln(1-x).$$

Therefore $$A(x)=-\frac{x^2(1+x)}{(1-x)^5}\ln(1-x).$$

Anne Bauval
  • 34,650