8

NOTE: The answer found here is not what I'm looking for.

The question is: Prove $P(z)=z^4+2z^3+3z^2+z+2$ has exactly two zeros in the right half plane. [Hint: Write $P(iy)=(y^2-2)(y^2-1)+iy(1-2y^2)$, and show that $\lim_{R\to\infty}arg\{P(iy)\}\Biggr|_{-R}^{R}=0$.]

I need to prove this (preferably using Rouche's theorem. Any other method involving the hint may also be accepted). Now, this can be shown to be true by taking two functions $f(z)=(z-1)^2(z+2)^2$ and $h(z)=z^4+2z^3-3z^2-4z+4$ which are analytic on and inside a closed semi-circular contour (radius $R$) encapsulating the right half-plane, and by showing that $|f(z)|>|h(z)|$ on the boundary of it. Then by Rouche's theorem, it is easy to verify that $P(z)$ has only 2 zeros (same as $f(z)$) in the right half-plane.

What I'm looking for is a solution involving the Hint! I know the Hint says that the imaginary axis of the domain gets mapped to the right half-plane (more precisely to the positive real numbers as $R\to\infty$). But I'm not sure how to use it in getting a solution.

Thanks in advance.

slhulk
  • 293
  • 1
    I don't think the second hint is correct. $\lim_{R\rightarrow\infty} P(iy) |{-R}^R$ is $\lim{R\rightarrow\infty} P(iR) - P(-iR)$ in which the real part vanishes, but imaginary part does not vanish. – Sungjin Kim Apr 27 '21 at 20:49
  • 1
    But, without the second hint, we can work with the first hint to get the solution. – Sungjin Kim Apr 27 '21 at 20:50
  • @SungjinKim I'm extremely sorry for the mistake in my question. It should be $\lim_{R\to\infty}arg{P(iy)}\Biggr|_{-R}^{R}=0$. Which I have corrected just now. – slhulk Apr 29 '21 at 07:29
  • 3
    The result follows easily from the hint but without using Rouche's thm. Are you sure you are supposed to use the hint and Rouche? – H. H. Rugh Apr 30 '21 at 21:51
  • @H.H.Rugh This question was categorized under questions that needed Rouche's theorem. But from the wording of the question, it may not use Rouche's theorem as well. Anyway, I'm looking for an answer involving the hint. I have edited the question. – slhulk May 01 '21 at 14:16

2 Answers2

6

For $R>0$, let ${\cal C}_R$ be the closed curve oriented anti-clockwise and consisting of the segment $S_R=[iR,-iR]$ and the half circle $C_R=\{Re^{it}: t\in [-\pi/2,\pi/2]\}$. It suffices to show that when $R$ goes to infinity, the index $\oint_{{\cal C}_R} \frac{P'(z)}{P(z)} \frac{dz}{2\pi i}$ is well-defined and equals two, or equivalently, writing $P(z)=|P(z)| e^{i\phi}$ that the angular increase $\oint d\phi = 4\pi$. Note that as long as $P(z)\neq 0$, $d\phi$ is well-defined as a differential form although $\phi$ is only defined mod $2\pi {\Bbb Z}$.

For $R$ large, the dominant term is $z^4$ and $(\pm iR)^4 = R^4 >0$ points in the direction of the positive real axis. Going along the half-circle we have $$\int_{C_R} \frac{P'(z)}{P(z)} \frac{dz}{2\pi i} = \int_{C_R} \frac{4}{z} \frac{dz}{2\pi i}+ O(\frac{1}{R}) = \int_{-\pi/2}^{\pi/2} {4} \frac{dt}{2\pi} +O(\frac{1}{R})= 2 + O(\frac{1}{R}) $$ Regarding the segment $S_R$, the hint implies that $P(iy)$ never points in the direction of the negative real axis. In order to see this, note that for $P(iy)$ to be real the imaginary part must be zero so that $y=0$ or $y^2=1/2$ but in any of these cases the real part is strictly positive (incidentally showing that $P(z)$ never vanishes). The angle thus takes values in $(-\pi,\pi)$ and from the above we infer that $\int_{S_R} d\phi = O(\frac{1}{R})$. Summing and letting $R\to +\infty$ we obtain the wanted result.

H. H. Rugh
  • 35,992
  • This answer is quite promising and makes sense. I'll wait for someone to come up with a more promising answer involving Rouche's theorem and if no one shows up this will be selected as the answer. – slhulk May 03 '21 at 07:55
1

Let $$Q(y) = P(iy) = (y^2-1)(y^2-2)+iy(1-2y^2).$$ Since fourth-order polynomial $\;P(z)\;$ has not roots with $\;\Re z=0,\;$ then it suffices to prove that $\;Q(y)\;$ has exactly two roots in the lower half-plane or in the upper one. Let us prove this for the upper half-plane.

Let $$U(y)=(y^2-1)(y^2-2),\quad V(y)=y(1-2y^2),$$ then $$\lim\limits_{R\to\pm\infty}\arg Q(y) = \lim\limits_{R\to\pm\infty}\arctan\dfrac{V(y)}{U(y)} = \lim\limits_{R\to\pm\infty}\arctan(\pm 0) = 0.$$ Since $\;Q(-y)=Q(y),\;$ then $\;\lim\limits_{R\to \infty}\arg Q(y)\bigg|_{-R++0i}^{R+0i} = 0.\;$

In accordance with the Rouche's theorem, the functions $\;Q(y) = U(y)+iV(y)\;$ and $\;Q_2(y) = U(y)+2iV(y)\;$ have the same quantity of roots in the upper half-plane.

At the same time, $$\arg Q_2(y) = \arctan\dfrac{2y-4y^3}{y^4-3y+2} = \arctan \dfrac{\dfrac2{y^3}-\dfrac3y -\dfrac1y}{1+\left(\dfrac2{y^3}-\dfrac3y\right)\dfrac1y}$$ $$= \arctan\dfrac{2-3y^2}{y^3}-\arctan\dfrac1y = \arg \left(y^3+i(2-3y^2)\right)+\arg(y-i)$$ $$ = \arg \left(i((iy)^3 +3(iy)^2+2)\right)+\arg(y-i),$$ wherein, in accordance with the Besou theorem, the polynomial $$t^3+3t^2+2=(t+r)(t^2-(r-3)t+r(r-3))$$ has one real root $\;t=-r,\; r>3\;$ and two complex roots with $\;\Re t>0.\;$

Since in the upper half-plane $\;\Re(iy)<0,\;$ then $\;Q_2(y)\;$ has in sum exactly two roots in the upper half-plane.

Proved!