# How many roots of a polynomial have positive real part?

I am given an exercise with three polynomials, and we have to find the number of roots of the first one that lie in the unit disk, the number of roots that lie in some region, e.g. those that lie in unit disk, or have $1<|root|<2$, etc.

I succeeded with the first two, by Rouché's theorem (in short: if |f|>|g| on boundary of D, then f and f+g have same number of roots in D.). But now, the function in question is $z^4+8z^3+3z^2+8z+3$, and I shall find # of those roots that have positive real part.

But I cannot find a way to divide the half-plane into appropriate regions, and then the function into a sum of appropriate functions to make this work.

Does anybody have an idea?

-
Have you been taught either the rule of signs or the Routh-Hurwitz criterion? – J. M. Nov 10 '11 at 13:02
no, none of those. I will look those up. – Marie. P. Nov 10 '11 at 13:10

What follows is a standard solution using the argument principle. There are many clever approaches that will work, by choosing the right places to use Rouché's theorem and using some inequalities, but this general approach almost never fails.

Argument principle: The argument principle states that over a closed curve, the number of zeros minus the number of poles is the change in argument over the curve divided by $2\pi$. This requires of course that there are no zeros or poles on the boundary. We can write it as $$2\pi(n-m)=\Delta_C \arg(f).$$

To use this, first note that there are no real roots with positive real part since all the coefficients are positive. Since all the coefficients are real, every root in the first quadrant will correspond to a root in the fourth quadrant, so we need only consider the roots in the first quadrant.

Now, notice that there are no roots on the imaginary axis since

$$(ix)^4+8(ix)^3+3(ix)^2+8(ix)+3=x^4-8ix^3-3x^2 +8ix+3=0$$ implies that both $$x^4-3x^2+3=0,\text{ and } 8x-8x^3=0$$ which is impossible.

Consider the contour which is a pizza slice of the right quarter circle or Radius $R$ in the first quadrant. (That is, start at zero, go along the positive real axis until $R$, and then follow the circle or radius $R$ until the imaginary axis, and then return back to the origin) We need only find the change in argument of the polynomial along this curve to know how many zeros are inside. (Since it has no poles.)

Computing the change in argument: We break this down into the three segments.

Real line: On the real line, our polynomial it is always positive, so the change in argument is zero.

Circular arc: On the circular arc, when we take $R$ larger and larger, the first term $z^4$ will dominate so that the change in argument there is the same as it is for $z^4$ in the limit. This gives a change in argument of $2\pi$.

Imaginary axis The imaginary axis is hardest to deal with. However, since the real part is $$x^4-3x^2+3=\left(z^2-\frac{3}{2}\right)^2+\frac{3}{4}$$ which is always positive, we see that it cannot leave the right half plane, and cannot wind around the origin. Hence only the final point and initial point matter, so we see the change in argument is $0$.

Hence as $R\rightarrow \infty$, we see that the change in argument is $2\pi$ so that there is exactly 1 zero. This means that there are two zeros in the right half plane.

Hope that helps,

-
Thank you very much. I could not have come up with that by myself! – Marie. P. Nov 10 '11 at 19:05
What do you mean by "final point and initial point"? Do you mean $0$ and $Ri$ or $0$ and $0$ (the latter case meaning the initial and final point of your contour)? – The Substitute Apr 28 '15 at 1:49