I want to solve the following :

(i) Show that $z^4+2z^2-z+1$ has exactly one root per plane quadrant.

My idea to prove (i) is by using Rouche theorem, by considering 4 cuts of the complex plane which are given as:

$$ \Gamma_{1}= \{z / Re(z)>0 \:\mbox{and}\: Im(z)>0 \}\\ \Gamma_{2}= \{z / Re(z)<0 \:\mbox{and}\: Im(z)>0 \}\\ \Gamma_{3}= \{z / Re(z)<0 \:\mbox{and}\: Im(z)<0 \}\\ \Gamma_{1}= \{z / Re(z)>0 \:\mbox{and}\: Im(z)<0 \}$$

So by having $f(z)=z^4+2z^2-z+1$ how can I choose g(z) in order to have proceed by Rouche theorem in each $\Gamma_{i}$ for $i=1,2,3,4$ Im justasking by an example in some quadrant in order to solve for the other quadrants . Also I was wondering if by solving for $\Gamma_{1}$ and $\Gamma_{2}$ we have the same for $\Gamma_{4}$ and $\Gamma_{3}$ respectively by some conjugate complex argument?Thanks:)

  • $\begingroup$ Rouche's theorem require a simple closed contour. $\endgroup$ – Mohammad W. Alomari May 17 '16 at 22:42
  • $\begingroup$ So Rouche´s theorem seems not quite a good way to solve the problem? :P $\endgroup$ – fgp May 17 '16 at 22:45
  • $\begingroup$ Your choice is wrong, $\Gamma_i$, $i=1,2,3,4$ $\endgroup$ – Mohammad W. Alomari May 17 '16 at 22:51
  • 1
    $\begingroup$ Can you be more helpful @mwomath? $\endgroup$ – fgp May 17 '16 at 22:55

In the view that $$ f(z)=(z^2+1)^2-z $$ is nearly a biquadratic expression, it makes sense to seek a nearby factorized expression of the form \begin{align} g(z)&=(z^2+c)^2-(az+b)^2=(z^2+c+az+b)(z^2+c-az-b) \\&=z^4+2cz^2+c^2-a^2z^2-2abz-b^2 \end{align} which then requires to satisfy $2c-a^2\approx 2$, $2ab\approx 1$, $c^2-b^2\approx 1$ as close as possible.

One may try $a=\frac23$ and thus $b=\frac34$, $c=1+\frac12a^2=\frac{11}9$, giving a constant term of $c^2-b^2=1+\frac{40}{81}-\frac{9}{16}=1+\frac{640-729}{16·81}=1-\frac{89}{1296}$ thus $$ g(z)=f(z)-\frac{89}{1296} $$ On the real axis, $$ f(x)=x^4+\frac32x^2+\frac12(x-1)^2+\frac12\ge\frac12>|f(x)-g(x)| $$ On the imaginary axis $$ |f(iy)|=\sqrt{(y^2-1)^4+y^2}\ge \max(|y|,|y^2-1|^2)>\frac1{4}>|f(iy)-g(iy)| $$ for $|y|>\frac{1}{4}$ or $|y|^2<\frac1{2}$, that is, for the whole imaginary axis. Thus both polynomial have the same number of roots in each of the quadrants. (If necessary, add segments of the circle of radius $2$ to the boundary, as $|f(z)|\ge(4-1)^2-2=7>|f(z)-g(z)|$ there.


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.