2

We have a polynomial $p(x)=x^5 -ax^3+b$. We need to find the relationship between $a$ and $b$ such that $p(x)$ has multiple roots.

Assume that $p(x)$ has 2 roots $c$ and $d$ with multiplicities $2$ and $3$ respectively. Then 1) $3c+2d=0$ 2) $3c^2+6cd+d^2=-a$ 3) $(c^3)*(d^2)= -b$

Thus, $a=3.75c^2$ and $b=-2.25c^5$. and we can find the relationship between $a$ and $b$ then.

Is it the right approach?

user52045
  • 1,430
  • 13
  • 25
Nika
  • 285
  • 4
    $p$ has multiple (complex) roots iff $\gcd(p,p')\ne1$. – lhf Dec 03 '13 at 13:19
  • Note that if "$p(x)$ has 2 roots $c$ and $d$ with multiplicities $2$ and $3$ respectively," then the sum of roots is $2c+3d$, not $3c+2d$. Are you assuming only multiple roots? I read the Question as asking for at least one multiple root. – hardmath Dec 03 '13 at 14:50

2 Answers2

4

There is an algorithm which allows you to find if polynomials $f$, $g$ have a common root. You have to calculate the resultant of $f$ and $g$, $R(f,g)$. If it's $0$, then there are common roots. Now polynomial $f$ has multiple roots iff $f$ and $f'$ has common root.

hardmath
  • 37,015
user52045
  • 1,430
  • 13
  • 25
  • 1
    For univariate (one-variable) polynomials, it's likely efficient just to compute the GCD of $f,g$, which is nontrivial if and only if $f,g$ have a common root. – hardmath Dec 03 '13 at 13:38
  • @hardmath Ofc you are right but i think it can lead to more complicated computations. – user52045 Dec 03 '13 at 13:51
  • Can the resultant of $p,p'$ in this particular case be found more easily than $\gcd(p,p')$? The determinant of the Sylvester matrix gives the resultant, but this is a $9\times 9$ matrix determinant here. – hardmath Dec 03 '13 at 14:10
  • I dont think its worth arguing but in this case in any row there will be only 3 or 2 nonzero elements and Laplace formula gives almost trivial computation. I guess this is more matter of taste in this case tho. – user52045 Dec 03 '13 at 14:17
  • @hardmath can I use the Euclidian algorithm and after coming to the stage when just a constant (consisting of a and b) is equal to 0? thus, I secure that a and b have GCD which is not 0? – Nika Dec 03 '13 at 14:20
  • 1
    Yes, I wasn't trying to be argumentative, just trying to compare. It seems to me the special structure of coefficients might help to do both approaches. – hardmath Dec 03 '13 at 14:20
  • @Nika: Yes, the condition for $p,p'$ to have nontrivial GCD would be setting that expression in $a,b$ to zero. – hardmath Dec 03 '13 at 14:22
2

It is not so hard to show that if a polynomial $f$ has $\alpha$ as a root with multiplicity $k$ then $f'$ has $\alpha$ as a root with multiplicity $k-1$. Since $f'(x)=x^2(5x^2-3a)$ you can easily find the root with multiplicity $3$. From this you will be able to find the conditions on $a$ and $b$.

P..
  • 14,929