1

Continued question here

I encountered this question: For quadratic polynomials such as $x^2\pm5x\pm6$ or $x^3\pm5x\mp6$, can be factorised over the integers. The main problem is to find a generator which can generate every polynomial that have this identity.

What have I tried:

For $x^2\pm mx+n=(x\pm p)(x\pm q)$ and $x^2\pm mx-n=(x\pm s)(x\mp t)$, solving a simultaneous equation for $q$ and $s$, getting $$q=\frac {t(p+t)} {p-t}$$ $$s=\frac {p(p+t)} {p-t}$$

and applying into the general equation getting $$x^2+(\frac {p^2+t^2} {p-t})x+\frac {pt(p+t)} {p-t}$$ but it doesn't suit for every integer $p$ and $t$, how can we get furthur than that?

xxxx036
  • 775
  • Typo first line $x^3 + \ldots$ Also can you write in your simultaneous equation for $q$ and $s$? –  Feb 09 '21 at 13:48

1 Answers1

-1

Let $m$ and $n$ be two positive integers such that the two polynomials $$x^2+mx+n\qquad\text{ and }\qquad x^2+mx-n,$$ both factor over the integers. Note that then also the two polynomials $$x^2-mx+n\qquad\text{ and }\qquad x^2-mx-n,$$ factor over the integers. It means there exist integers $p$, $q$, $s$ and $t$ such that \begin{eqnarray*} x^2+mx+n&=&(x+p)(x+q)=x^2+(p+q)x+pq,\\ x^2+mx-n&=&(x+s)(x+t)=x^2+(s+t)x+st,\\ \end{eqnarray*} and so in particular $m^2-4n$ and $m^2+4n$ are both perfect squares, because \begin{eqnarray*} m^2-4n&=&(p+q)^2-4pq=(p-q)^2,\\ m^2+4n&=&(s+t)^2-4st=(s-t)^2,\\ \end{eqnarray*} and hence $(p-q)^2$, $m^2$ and $(s-t)^2$ are three perfect squares in arithmetic progression, with common difference $4n$. Conversely, if $a^2$, $b^2$ and $c^2$ are three perfect squares in arithmetic progression with common difference $4d$ then \begin{eqnarray*} p&=&-\frac{b-a}{2},\qquad\text{ and }\qquad q&=&-\frac{b+a}{2},\\ s&=&-\frac{b+c}{2},\qquad\text{ and }\qquad t&=&-\frac{b-c}{2}, \end{eqnarray*} are integers satisfying $$(x-p)(x-q)=x^2+bx+d,$$ $$(x-s)(x-t)=x^2+bx-d.$$ So your question is equivalent to characterizing all arithmetic progressions of perfect squares of length $3$. This was somewhat famously done by Fermat, see wikipedia, for example. The parametrization given there yields the parametrization \begin{eqnarray*} m&=&u^2+v^2,\\ n&=&uv(u^2-v^2), \end{eqnarray*} of the coefficients of such polynomials. Here $u$ and $v$ range over the integers.

Servaes
  • 67,306
  • 8
  • 82
  • 171
  • Would also notice that $m=u^2+v^2$ and $n=uv(u^2-v^2)$ is quite similar to $\frac {p^2+t^2} {p-t}$ and $\frac {pt(p+t)} {p-t}$, can you relate? – 00xxqhxx00 Feb 05 '21 at 11:20
  • I can't tell how OP derived this parametrization in the first place. The two expressions differ from mine by a factor $(p-t)$ and $(p-t)^2$, respectively, which suggests that the OP's argument might have skipped over some homogenezation somewhere. But it's just a wild guess without knowing their reasoning. – Servaes Feb 05 '21 at 11:25