3

Let $f=f(t),g=g(t)\in \mathbb{C}[t]$, with $\deg(f),\deg(g) \geq 3$.

A known result about the resultant of $f$ and $g$ says the following: The resultant of $f$ and $g$ is $0$ if and only if $f$ and $g$ have a common root in $\mathbb{C}$ (notice that I assumed that the base field is algebraically closed; otherwise, the common root is in an algebraic closure of the base field).

Further assume that the greatest common divisor of $f$ and $g$, $\gcd(f,g)$, is $(t-a)(t-b)$ for some $a \neq b \in \mathbb{C}$.

Obviously, the resultant of such $f$ and $g$ is zero, since they have a common root $a$ (and $b$).

(1) Is there a 'generalized resultant' which tells how many distinct common roots $f$ and $g$ have (and counts multiplicities of common roots)?

Remarks: (i) If $a=b$ then $a$ is a common root of $f$ and $f'$ (and of $g$ and $g'$), hence the discriminant is zero. But the discriminant does not help in answering my question which is concerned with $f$ and $g$, one is not assumed to be the derivative of the other.

(ii) Perhaps subresultants are relevant to my question.

(iii) This answer is relevant if we assume that $k(f,g)=k(t)$. However, it only tells that in that case there exist $\lambda,\mu \in k$ with $\gcd(f-\lambda,g-\mu)=x-\nu$ for some $\nu \in k$, but it does not exclude the possibility of $\deg(\gcd(f-\tilde{\lambda}, g-\tilde{\mu}) \geq 2$ for some $\tilde{\lambda},\tilde{\mu} \in k$. For example: $f=t^3-4t$ and $g=t^2+1$. Taking $\lambda=0$ and $\mu=5$ yields $f-\lambda=f-0=t^3-4t=t(t^2-4)$ and $g-\mu=g-5=t^2+1-5=t^2-4$, therefore $\gcd(f-\lambda,g-\mu)=t^2-4$ which is of degree $2$. Notice that indeed $k(f,g)=k(t)$ since $t=\frac{f}{g-5}$.

(2) Same question as (1), with the additional assumption that $k(f,g)=k(t)$.

Thank you very much!

user237522
  • 7,267
  • 1
    Perhaps subresultants are relevant Maybe this answers your question: "the GCD of P and Q has a degree d if and only if $,{\displaystyle s_{0}(P,Q)=\cdots =s_{d-1}(P,Q)=0\ ,s_{d}(P,Q)\neq 0},$.". – dxiv Jul 30 '18 at 21:36
  • Thanks! Nice (I have seen that wikipedia page before; this is why I have remarked "Perhaps subresultants are relevant to my question"). Anyway, it is good that you reminded me, and any explanation concerning subresultants is welcome, for example, how to apply the quoted claim to my above example $P=t^3-4t, Q=t^2+1$. (You can write your comment as an answer, if you like). – user237522 Jul 30 '18 at 21:53
  • The problem is that I do not know how to apply the quoted claim in a 'general case', namely: $f=\sum_{i=0}^{n}a_it^i$ and $g=\sum_{j=0}^{m}b_jt^j$. But truly I am not yet enough familiar with subresultants, so perhaps something can be done. – user237522 Jul 30 '18 at 22:01
  • Too long for a comment, posted as an answer. – dxiv Jul 30 '18 at 22:14

1 Answers1

3

Perhaps subresultants are relevant to my question.

From the wikipedia page on resultants:

      ...the GCD of P and Q has a degree d if and only if

      ${\quad\quad\displaystyle s_{0}(P,Q)=\cdots =s_{d-1}(P,Q)=0\ ,s_{d}(P,Q)\neq 0}$.

      In this case, $S_d(P ,Q)$ is a GCD of $P$ and $Q$ and

      $\quad\quad{\displaystyle S_{0}(P,Q)=\cdots =S_{d-1}(P,Q)=0.}$.

For example: $f=t^3-4t$ and $g=t^2+1$

Avoiding tedious hand calculations, the subresultants of $x^3-4x-\lambda$ and $x^2+1-\mu$ courtesy WA subresultants[ x^3 - 4x - \lambda , x^2 + 1 - \mu , x ] are:

$$ \begin{align} s_0 &= λ^2 - μ^3 + 11 μ^2 - 35 μ + 25 \\ s_1 &= μ - 5 \\ s_2 &= 1 \end{align} $$

Therefore the necessary and sufficient condition for $f - \lambda$ and $g - \mu$ to have (at least) a common root is $\,s_0=λ^2 - μ^3 + 11 μ^2 - 35 μ + 25=0\,$, and for a second common root the additional condition $\,s_1=\mu-5=0\,$. The latter gives $\,\mu = 5\,$, which substituted in the former gives $\,\lambda = 0\,$.


[ EDIT ] In the simple case above, it is of course straightforward to verify the result by hand. Direct euclidean division gives $\,x^3 - 4x - \lambda = \big(x^2 + 1 - \mu\big) \cdot x + \big((\mu - 5) x - λ\big)\,$, so the condition for two common roots is $\,(\mu - 5) x - λ \equiv 0 \;\iff\; \mu-5=\lambda=0\,$.
dxiv
  • 78,115
  • 2
    Great! Thank you very much. (1) I did not know that Wolfram Alpha http://www.wolframalpha.com/ can compute subresultants. (2) Applying the quoted claim to the example I requested also helped. – user237522 Jul 30 '18 at 22:36
  • 2
    Nice edit, thanks. – user237522 Jul 31 '18 at 01:06
  • Please, if you can take a look at the following question of mine https://math.stackexchange.com/questions/3056189/computing-subresultants-in-wolframalpha – user237522 Dec 29 '18 at 19:46