Define $f(n)=\dfrac{n^2}{48}+\dfrac{n}{8}(1+x_2(n))+X(n),$ where $$X(n)=-\dfrac{7}{48}+\dfrac{9}{16}x_2(n)+\dfrac{1}{4}x_2(\lfloor \dfrac{n+1}{2} \rfloor)+\dfrac{1}{3}x_3(n),$$

and $\begin{eqnarray} x_2(n)= \begin{cases} 1 & 2\mid n \\ 0 & 2\nmid n \end{cases} \end{eqnarray},$ $\begin{eqnarray} x_3(n)= \begin{cases} 1 & 3\mid n \\ 0 & 3\nmid n \end{cases} \end{eqnarray}.$

  • Prove that the equation $2x+3y+4z=n,(x,y,z\in \mathbb N)$ has exactly $f(n)$ solutions.

  • The equation $x+2y+3z=n,(x,y,z\in \mathbb N)$ has exactly $\lfloor \dfrac{n^2}{12}+\dfrac{n}{2}+1 \rfloor$ solutions. Can you simplify $f(n)$ in this form?

  • $\begingroup$ I would start thinking of a generating function and then enumerate, this should be simpler considering the domain. $\endgroup$ – Torsten Hĕrculĕ Cärlemän Jul 25 '13 at 17:45
  • 1
    $\begingroup$ You could try the geometric approach. In fact, Sinai Robins has several papers about counting integer points in polygons, and they apply almost directly to this case. $\endgroup$ – Eric Naslund Jul 27 '13 at 15:07
  • $\begingroup$ How do you know that $x+2y+3z=n,(x,y,z\in \mathbb N)$ has exactly $\lfloor \dfrac{n^2}{12}+\dfrac{n}{2}+1 \rfloor$? $\endgroup$ – Bumblebee Nov 30 '16 at 21:12

Being motivated by Torsten Hĕrculĕ Cärlemän's comment I can tell you that this could be a way to have the solution.

Notice that, $$F(q)=\displaystyle\sum_{n=1}^\infty f(n)q^n=\left(\displaystyle\sum_{n=1}^\infty q^{2n}\right)\left(\displaystyle\sum_{n=1}^\infty q^{3n}\right)\left(\displaystyle\sum_{n=1}^\infty q^{4n}\right)=\frac{q^9}{(1-q^2)(1-q^3)(1-q^4)}$$ where $q=\large e^{2\pi iz}$.

So, $$f(n)=\frac{1}{2\pi i}\displaystyle\int_0^1F(q)\large e^{-2in\pi z} dz=\displaystyle\int_\gamma\frac{q^{8-n}dq}{(1-q^2)(1-q^3)(1-q^4)}$$

where $\gamma$ is unit circle centered at $0$.

So, $$f(n)=-\displaystyle\int_\gamma\frac{q^{8-n}dq}{(q-1)^3(q+1)^2(q-i)(q+i)(q-\omega)(q-\omega^2)}$$

Hence, $$f(n)=-2\pi i\left(Res(z=1)+Res(z=-1)+Res(z=i)+Res(z=-i)+Res(z=\omega)+Res(z=\omega^2)\right)$$.

$Res(z=\pm i)=-\frac{(\mp i)^n(1 \pm i)}{16}$.................. (Here your $x_2(n)$ should come and $x_3(n)$ comes from residues at $\omega$ and $\omega^2$ terms).

You may proceed like this way and I believe that result would be obtained in this way, but I am leaving the rest of the solution to you as I am not relying on accuracy of my calculation.

  • $\begingroup$ Vous lisez dans mes pensées, mon ami! $\endgroup$ – Torsten Hĕrculĕ Cärlemän Jul 25 '13 at 18:51
  • $\begingroup$ Oui, je suis un lecteur d'esprit ... (I don't know french, used google translator) :D $\endgroup$ – Kunnysan Jul 25 '13 at 18:55
  • $\begingroup$ No problem, mon ami, a very good mind reader :D $\endgroup$ – Torsten Hĕrculĕ Cärlemän Jul 25 '13 at 19:26

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.