8
$\begingroup$

Let $f\in \mathbb{R}[x]$ be a non-constant polynomial and $a,b\in\mathbb{R}$ be s.t. $b\not= 0$.

Then —I think so— (*) The equation, in $x\in\mathbb{C}$, $f(x)-a-ib=0$ admits at least one simple root.

For example, one can prove (using the Gröbner basis theory) that $(x-a_1-ib_1)^2(x-a_2-ib_2)^2\cdots (x-a_6-ib_6)^2$, where $a_i,b_i\in\mathbb{R}$, $b_i\ne 0$, is never in the form $f(x)-a-ib$.

Yet $f(x)-a-ib$ may have multiple roots:

$$x^3/3+x^2/2+x-(-5/12+i\sqrt{3}/4)=1/24(2x+1+2i\sqrt{3})(i\sqrt{3}-2x-1)^2.$$

Question. Is the above statement (*) true?

$\endgroup$
7
  • 2
    $\begingroup$ What is the question? $\endgroup$
    – YCor
    Commented Jun 25 at 14:29
  • 1
    $\begingroup$ ($a$ is unnecessary since you can replace $f$ with $f-a$) $\endgroup$
    – YCor
    Commented Jun 25 at 14:30
  • $\begingroup$ ($b$ is unnecessary since you can replace $f-a$ by $(f-a)/b$, hence $b=1$ is enough) $\endgroup$ Commented Jun 25 at 14:32
  • 1
    $\begingroup$ @YCor Jochen does not say that you can replace $b$ by $0$, but that you can replace $b$ by $1$. $\endgroup$ Commented Jun 25 at 14:33
  • $\begingroup$ @AlekseiKulikov I know, I erased it before you replied. Sorry. $\endgroup$
    – YCor
    Commented Jun 25 at 14:34

1 Answer 1

15
$\begingroup$

Yes. Let $g$ be a polynomial. If $g$ has no simple roots then the degree of $\gcd(g, g')$ is at least half the degree of $g$. (Each root of $g$ with multiplicity $m$ is a root of $\gcd(g, g')$ with multiplicity $m-1$ and $m-1 \geq m/2$ for $m\geq 2$.)

If $f(x)-a-bi$ has no simple roots then so does $f(x)-a+bi$. Both these polynomials have derivative $f'$ so $\gcd ( f-a -bi, f') \gcd(f-a+bi,f')$ has degree at least $\deg f$.

But $f-a-bi$ and $f-a+bi$ are coprime since their difference $2bi$ is invertible. It follows that $\gcd ( f-a -bi, f') \gcd(f-a+bi,f')=\gcd ( (f-a -bi)(f-a+bi),f') $ divides $f'$, hence (because $f$ is nonconstant and thus $f'\neq 0$) $f'$ has degree at least $\deg f$, a contradiction.

$\endgroup$
4
  • 2
    $\begingroup$ This is unexpectedly elegant! $\endgroup$ Commented Jun 25 at 14:42
  • $\begingroup$ Nice (note that the last gcd argument is incorrect if $f'=0$). $\endgroup$
    – YCor
    Commented Jun 25 at 14:43
  • 2
    $\begingroup$ Biblical simplicity. Bravissimo. $\endgroup$
    – loup blanc
    Commented Jun 25 at 15:08
  • $\begingroup$ The question was asked and answered before: mathoverflow.net/q/81732 $\endgroup$
    – GH from MO
    Commented Jun 26 at 0:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.