• Aucun résultat trouvé

14-Zeroesofanalyticfunctions ComplexVariables

N/A
N/A
Protected

Academic year: 2022

Partager "14-Zeroesofanalyticfunctions ComplexVariables"

Copied!
5
0
0

Texte intégral

(1)

Complex Variables

14 - Zeroes of analytic functions

Jean-Baptiste Campesato November 4th, 2020

Reviews from Oct 16 – Zeroes Definition 1.

Let𝑈 ⊂ ℂbe open and𝑓 ∶ 𝑈 → ℂbe holomorphic/analytic. Let𝑧0 ∈ 𝑈 be such that𝑓 (𝑧0) = 0.

We define theorder of vanishing of𝑓 at𝑧0by𝑚𝑓(𝑧0) ≔min{𝑛 ∈ ℕ ∶ 𝑓(𝑛)(𝑧0) ≠ 0}. Note that𝑚𝑓(𝑧0) > 0since𝑓 (𝑧0) = 0.

Proposition 2. Let𝑈 ⊂ ℂbe open and𝑓 ∶ 𝑈 → ℂbe holomorphic/analytic. Let𝑧0 ∈ 𝑈 be such that𝑓 (𝑧0) = 0.

Denote the power series expansion of𝑓 at𝑧0by𝑓 (𝑧) =

+∞

𝑛=0

𝑎𝑛(𝑧 − 𝑧0)𝑛. Then𝑚𝑓(𝑧0) =min{𝑛 ∈ ℕ ∶ 𝑎𝑛 ≠ 0}.

Proposition 3. Let𝑈 ⊂ ℂbe open and𝑓 ∶ 𝑈 → ℂbe holomorphic/analytic. Then𝑧0is a zero of order𝑛of𝑓 if and only if there exists𝑔 ∶ 𝑈 → ℂholomorphic such that𝑓 (𝑧) = (𝑧 − 𝑧0)𝑛𝑔(𝑧)and𝑔(𝑧0) ≠ 0.

Reviews from Oct 16 – Analytic continuation

Theorem 4. Let𝑈 ⊂ ℂbe adomainand𝑓 ∶ 𝑈 → ℂbe a holomorphic/analytic function.

If there exists𝑧0∈ 𝑈 such that∀𝑛 ∈ ℕ≥0, 𝑓(𝑛)(𝑧0) = 0then𝑓 ≡ 0on𝑈.

Corollary 5. Let𝑈 ⊂ ℂbe adomainand𝑓 , 𝑔 ∶ 𝑈 → ℂbe holomorphic/analytic functions.

If𝑓 and𝑔coincide in the neighborhood of a point,

i.e. ∃𝑧0 ∈ 𝑈 , ∃𝑟 > 0, ∀𝑧 ∈ 𝐷𝑟(𝑧0) ∩ 𝑈 , 𝑓 (𝑧) = 𝑔(𝑧), then they coincide on𝑈,

i.e.∀𝑧 ∈ 𝑈 , 𝑓 (𝑧) = 𝑔(𝑧).

(2)

Isolated zeroes

It is actually possible to strengthen the previous results.

Theorem 6. Let𝑈 ⊂ ℂbe a domain and𝑓 ∶ 𝑈 → ℂbe a holomorphic/analytic function.

Then either𝑓 ≡ 0or the zeroes of𝑓 are isolated:

if𝑓 (𝑧0) = 0then there exists𝑟 > 0such that𝐷𝑟(𝑧0) ⊂ 𝑈 and∀𝑧 ∈ 𝐷𝑟(𝑧0) ⧵ {𝑧0}, 𝑓 (𝑧) ≠ 0.

Proof. Assume that𝑧0is a non-isolated zero of𝑓.

We know that𝑓 admits a power series expansion𝑓 (𝑧) =

+∞

𝑛=0

𝑎𝑛(𝑧 − 𝑧0)𝑛in a neighborhood of𝑧0.

Assume by contradiction that there exists a smallest𝑛 ∈ ℕ≥0such that𝑎𝑛≠ 0then𝑓 (𝑧) = (𝑧 − 𝑧0)𝑛𝑔(𝑧)where 𝑔is holomorphic and𝑔(𝑧0) ≠ 0.

For every𝑛 ∈ ℕ>0, ∃𝑤𝑛 ∈ (𝐷1𝑛(𝑧0) ∩ 𝑈 ) ⧵ {𝑧0}, 𝑓 (𝑤𝑛) = 0. But then𝑔(𝑤𝑛) = 0since𝑤𝑛 ≠ 𝑧0 Then, since𝑤𝑛−−−−−→

𝑛→+∞ 𝑧0, by continuity𝑔(𝑧0) = lim

𝑛→+∞𝑔(𝑤𝑛) = 0. Which leads to a contradiction.

Hence∀𝑛 ∈ ℕ≥0, 𝑓(𝑛)(𝑧0) = 𝑛!𝑎𝑛= 0and𝑓 ≡ 0on𝑈 by Theorem4. ■ Corollary 7. Let𝑈 ⊂ ℂbe a domain and𝑓 , 𝑔 ∶ 𝑈 → ℂbe holomorphic/analytic functions.

If𝑓 − 𝑔admits a non-isolated zero

i.e. ∃𝑧0∈ 𝑈 , ∀𝑟 > 0, ∃𝑧 ∈ (𝑈 ∩ 𝐷𝑟(𝑧0)) ⧵ {𝑧0}, 𝑓 (𝑧) − 𝑔(𝑧) = 𝑓 (𝑧0) − 𝑔(𝑧0) = 0 then𝑓 and𝑔coincide on𝑈,

i.e.∀𝑧 ∈ 𝑈 , 𝑓 (𝑧) = 𝑔(𝑧).

Corollary 8. Let𝑈 ⊂ ℂbe a domain and𝑓 , 𝑔 ∶ 𝑈 → ℂbe holomorphic/analytic functions.

Let(𝑧𝑛)𝑛∈ℕbe a sequence of terms in𝑈 which is convergent to𝑧̃in𝑈 and such that∀𝑛 ∈ ℕ, 𝑓 (𝑧𝑛) = 0.

Then𝑓 ≡ 0on𝑈.

Remark 9. The fact that the limit𝑧 ∈ 𝑈̃ is very important.

Indeed, let𝑓 ∶ ℂ ⧵ {0} → ℂbe defined by𝑓 (𝑧) =sin(𝜋𝑧). Then𝑓 (1𝑛) = 0but𝑓 ≢ 0onℂ ⧵ {0}.

Hence, it is possible for the zeroes of𝑓 to accumulate at a point of the boundary of the domain (including

∞, see for instance𝑧𝑛= 𝜋𝑛for𝑓 =sin).

Homework 10. Let𝑈 ⊂ ℂbe a domain and𝑓 , 𝑔 ∶ 𝑈 → ℂbe holomorphic/analytic on𝑈. Prove that if𝑓 𝑔 ≡ 0on𝑈then either𝑓 ≡ 0or𝑔 ≡ 0.

Homework 11. Let𝑈 = 𝐷1(0). Find all the holomorphic functions𝑓 ∶ 𝑈 → ℂsatisfying respectively:

1. 𝑓 (1𝑛) = 𝑛12

2. 𝑓 (2𝑛1) = 𝑓 (2𝑛+11 ) = 1𝑛

Otherwise stated, if you attend MAT327, either𝑓 is constant equal to0or{𝑧 ∈ 𝑈 ∶ 𝑓 (𝑧) = 0}is discrete.

(3)

Reviews from Oct 23 – Poles

Theorem 12. Let𝑈 ⊂ ℂbe open and𝑧0∈ 𝑈. Assume that𝑓 ∶ 𝑈 ⧵ {𝑧0} → ℂis holomorphic/analytic. Then TFAE:

1. 𝑧0is a pole of𝑓, i.e. lim

𝑧→𝑧0|𝑓 (𝑧)| = +∞.

2. There exist𝑛 ∈ ℕ>0and𝑔 ∶ 𝑈 → ℂanalytic such that𝑔(𝑧0) ≠ 0and𝑓 (𝑧) = 𝑔(𝑧)

(𝑧 − 𝑧0)𝑛 on𝑈 ⧵ {𝑧0}.

3. 𝑧0is not a removable singularity of𝑓 and there exists𝑛 ∈ ℕ>0such that lim

𝑧→𝑧0(𝑧 − 𝑧0)𝑛+1𝑓 (𝑧) = 0.

Definition 13. The integer𝑛 > 0in (2) is uniquely defined and we say that𝑓 admits apole of order𝑛at𝑧0. Proposition 14. The order of the pole𝑧0is also:

• The order of vanishing of1/𝑓 at𝑧0.

• The smallest𝑛such that lim

𝑧→𝑧0(𝑧 − 𝑧0)𝑛+1𝑓 (𝑧) = 0.

The argument principle

Lemma 15(Logarithmic residue).

• If𝑧0is an isolated zero of𝑓 thenRes(𝑓

𝑓 , 𝑧0)is the order of𝑧0.

• If𝑧0is an isolated pole of𝑓 then−Res(𝑓

𝑓 , 𝑧0)is the order of𝑧0. Proof.

• Assume that 𝑓 (𝑧) = (𝑧 − 𝑧0)𝑚𝑔(𝑧) in a neighborhood of 𝑧0 where 𝑔 is analytic and 𝑔(𝑧0) ≠ 0. Then

𝑓(𝑧)

𝑓 (𝑧) = 𝑚(𝑧 − 𝑧0)−1+𝑔(𝑧)

𝑔(𝑧).

We conclude using that 𝑔𝑔 is holomorphic in a neighborhood of𝑧0.

• 𝑧0is a pole of order𝑚of𝑓 if and only if it is a zero of order𝑚of 1

𝑓. We conclude using that Res(

(1/𝑓 ) (1/𝑓 ), 𝑧0

)= −Res (

𝑓 𝑓 , 𝑧0

)

■ The previous lemma holds at∞:

Lemma 16.

• Ifis an isolated zero of𝑓 thenRes(𝑓

𝑓 , ∞)is the order of∞.

• Ifis an isolated pole of𝑓 then−Res(𝑓

𝑓 , ∞)is the order of.

Proof. ∞is an isolated zero (resp. pole) of order𝑚of𝑓 if and only if0is an isolated zero (resp. pole) of order𝑚of𝑔(𝑧) = 𝑓 (1/𝑧).

Then𝑚 =Res(𝑔

𝑔 , 0) =Res(−1𝑧2 𝑓(1/𝑧)

𝑓 (1/𝑧), 0) =Res(𝑓

𝑓 , ∞). ■

(4)

Theorem 17(The argument principle).

Let𝑈 ⊂ ℂbe open. Let𝑆 ⊂ 𝑈be finite. Let𝑓 ∶ 𝑈 ⧵ 𝑆 → ℂbe holomorphic/analytic.

Let𝛾 ∶ [𝑎, 𝑏] → ℂbe piecewise smooth positively oriented simple closed curve on𝑈which doesn’t pass through a zero or a pole of𝑓 and such that its inside is entirely included in𝑈.

Then 1

2𝑖𝜋 ∫𝛾 𝑓(𝑧)

𝑓 (𝑧)d𝑧 = 𝑍𝑓 ,𝛾− 𝑃𝑓 ,𝛾 where

𝑍𝑓 ,𝛾 is the number of zeroes of𝑓 enclosed in𝛾 counted with their multiplicites/orders,

𝑃𝑓 ,𝛾is the number of poles of𝑓 enclosed in𝛾 counted with their multiplicites/orders.

Proof. We apply Cauchy’s residue theorem to 𝑓𝑓 and then we use the above lemma:

1 2𝑖𝜋 ∫𝛾

𝑓(𝑧) 𝑓 (𝑧)d𝑧 =

𝑧∈Inside(𝛾)∑ Res(

𝑓 𝑓 , 𝑧

)=

𝑧zero of𝑓

Res( 𝑓

𝑓 , 𝑧0 )+

𝑧pole of𝑓

Res( 𝑓

𝑓 , 𝑧0

)= 𝑍𝑓 ,𝛾− 𝑃𝑓 ,𝛾

Remark 18. The value 1 2𝑖𝜋 ∫𝛾

𝑓(𝑧)

𝑓 (𝑧)d𝑧involved in the previous slide is equal to the number of counterclock- wise turns made by𝑓 (𝑧)as𝑧goes through𝛾.

Indeed, if we set ̃𝛾(𝑡) = 𝑓 ∘ 𝛾then

𝛾 𝑓(𝑧)

𝑓 (𝑧)d𝑧 =

̃𝛾 1 𝑤d𝑤.

Assume for instance that ̃𝛾 ∶ [0, 1] → ℂis defined by ̃𝛾(𝑡) = 𝑧0+ 𝑟𝑒2𝑖𝜋𝑛𝑡where𝑛 ∈ ℤ.

Then 1 2𝑖𝜋 ∫̃𝛾

1

𝑤d𝑤 = 𝑛which is the number of counterclockwise turns made by ̃𝛾around𝑧0. Then the conclusion of the previous statement can be rewritten as

changes of arg(𝑓 (𝑧))as𝑧goes through𝛾

2𝜋 = 𝑍𝑓 ,𝛾− 𝑃𝑓 ,𝛾

That’s why it is calledthe argument principle.

Rouché’s theorem

Theorem 19(Rouché’s theorem – version 1).

Let𝑈 ⊂ ℂbe open,𝑓 , 𝑔 ∶ 𝑈 → ℂbe two holomorphic/analytic functions on𝑈, and𝛾 ∶ [𝑎, 𝑏] → ℂbe a piecewise smooth simple closed curve on𝑈whose inside is also included in𝑈.

Assume that

∀𝑡 ∈ [𝑎, 𝑏], |𝑔(𝛾(𝑡))| < |𝑓 (𝛾(𝑡))|

Then𝑓 and𝑓 + 𝑔have the same number of zeroes inside𝛾, counted with multiplicities.

Proof. For𝑡 ∈ [0, 1], set𝜑𝑡(𝑧) = 𝑓 (𝑧) + (1 − 𝑡)𝑔(𝑧)andℎ(𝑡) = 1 2𝑖𝜋 ∫𝛾

𝜑𝑡(𝑧) 𝜑𝑡(𝑧)d𝑧.

The functionℎis continuous since𝜑𝑡doesn’t vanish on𝛾, indeed for𝑧 ∈ 𝛾

|𝜑𝑡(𝑧)| ≥ |𝑓 (𝑧)| + (1 − 𝑡)|𝑔(𝑧)| ≥ |𝑓 (𝑧)| − |𝑔(𝑧)| > 0

Henceℎis a continuous function taking values inℤ(by the principle argument), so it is constant.

Henceℎ(0) = ℎ(1), i.e. 𝑍𝑓 +𝑔,𝛾− 𝑃𝑓 +𝑔,𝛾 = 𝑍𝑓 ,𝛾− 𝑃𝑓 ,𝛾 by the principle argument.

But these functions have no poles in the inside of𝛾, hence𝑍𝑓 +𝑔,𝛾 = 𝑍𝑓 ,𝛾. ■

(5)

Theorem 20(Rouché’s theorem – version 2).

Let𝑈 ⊂ ℂbe open,𝑓 , 𝑔 ∶ 𝑈 → ℂbe two holomorphic/analytic functions on𝑈, and𝛾 ∶ [𝑎, 𝑏] → ℂbe a piecewise smooth simple closed curve on𝑈whose inside is also included in𝑈.

Assume that

∀𝑧 ∈ 𝛾, |𝑓 (𝑧) − 𝑔(𝑧)| < |𝑓 (𝑧)|

Then𝑓 and𝑔have the same number of zeroes inside𝛾, counted with multiplicities.

Proof. That’s an immediate consequence of the previous version since𝑧0 is a zero of order𝑛of 𝑔iff it is a

zero of order𝑛of−𝑔. ■

Theorem 21(Rouché’s theorem – version 3).

Let𝑈 ⊂ ℂbe open,𝑓 , 𝑔 ∶ 𝑈 → ℂbe two holomorphic/analytic functions on𝑈, and𝛾 ∶ [𝑎, 𝑏] → ℂbe a piecewise smooth simple closed curve on𝑈whose inside is also included in𝑈.

Assume that

∀𝑧 ∈ 𝛾, |𝑓 (𝑧) + 𝑔(𝑧)| < |𝑓 (𝑧)|

Then𝑓 and𝑔have the same number of zeroes inside𝛾, counted with multiplicities.

Proof. Since𝑧0is a zero of order𝑛of𝑔iff it is a zero of order𝑛of−𝑔. ■

We already proved the Fundamental Theorem of Algebra (or d’Alembert–Gauss theorem) using Liouville’s theorem (Oct 21): a non-constant complex polynomial admits at least one root.

Here is another proof using Rouché’s theorem.

Theorem 22. A complex polynomial of degree𝑛has exactly𝑛complex roots (counted with multiplicity).

Proof. Assume that𝑃 (𝑧) = 𝑎𝑛𝑧𝑛+ 𝑄(𝑧)where𝑄is a polynomial of degree< 𝑛and𝑎𝑛≠ 0.

If we take𝑅 > 0big enough then|𝑄(𝑧)| < |𝑎𝑛𝑧𝑛|on𝛾 ∶ [0, 1] → ℂdefined by𝛾(𝑡) = 𝑅𝑒2𝑖𝜋𝑡.

By Rouché’s theorem,𝑃 (𝑧) = 𝑎𝑛𝑧𝑛+𝑄(𝑧)and𝑎𝑛𝑧𝑛have the same number of zeroes counted with multiplicity.

Références

Documents relatifs

In section 2, we succeeded in describing the Dynamics on the invariant tori having a com- pletely irrational rotation vector. If we want to extend this to all the tori, we need to

The estimates of the derivatives with respect to the parameter t are proved by induction on the degree m using the following lemma which follows from (1.4) (1.5). This lemma is

The fundamental theorem of calculus reduces the problem ofintegration to anti- differentiation, i.e., finding a function P such that p'= f. We shall concentrate here on the proof of

For now, we’ll just say that the correct way to define improper integrals is as a limit of well–defined approximating integrals. The approximating integrals have restricted domains

A Liouville-type theorem in the same spirit as Theorem 1.1 has been established for periodic solutions of the hyperelastic rod equation in [2], when the physical parameter γ of

Since (sin(α 1 + 2q)) is positive and negative over certain intervals we can see that the function clearly has two rises and two falls, also, since cos(α 1 + 2q) has zeros and

(2012) ont montré qu'il n'y a pas de corrélation significative entre les indices climatiques, la température et l’abondance du zooplancton; ils suggèrent que les facteurs

Indeed, positive correlations were also found within these two pairs, these correlations being seen as the result of the common approach dimension of mastery-approach and