• 沒有找到結果。

By the Mean-Value Theorem (for real functions of a real variable) u(z + h

N/A
N/A
Protected

Academic year: 2022

Share "By the Mean-Value Theorem (for real functions of a real variable) u(z + h"

Copied!
7
0
0

加載中.... (立即查看全文)

全文

(1)

Analyticity and the Cauchy Riemann Equations

Proposition Suppose fx and fy exist in a neighborhood of z. If fx and fy are continuous at z and fy = ifx there, then f is differentiable at z.

Proof Let f = u + iv, h = ξ + iη. By the Mean-Value Theorem (for real functions of a real variable)

u(z + h) − u(z)

h = u(x + ξ, y + η) − u(x, y) ξ + iη

= u(x + ξ, y + η) − u(x + ξ, y)

ξ + iη +u(x + ξ, y) − u(x, y) ξ + iη

= η

ξ + iηuy(x + ξ, y + θ1η) + ξ

ξ + iηux(x + θ2ξ, y)

= η

ξ + iηuy(z1) + ξ

ξ + iηux(z2) and

v(z + h) − v(z)

h = η

ξ + iηvy(x + ξ, y + θ3η) + ξ

ξ + iηvx(x + θ4ξ, y)

= η

ξ + iηvy(z3) + ξ

ξ + iηvx(z4) for some 0 < θk < 1, k = 1, 2, 3, 4. Thus

f (z + h) − f (z)

h = η

ξ + iηuy(z1) + ivy(z3) + ξ

ξ + iηux(z2) + ivx(z4) Since fy = ifx at z,

fx(z) = ξ + iη

ξ + iηfx(z) = ξ

ξ + iηfx(z) + η

ξ + iηfy(z), and thus

f (z + h) − f (z)

h − fx(z) = f (z + h) − f (z)

h − ξ

ξ + iηfx(z) − η

ξ + iηfy(z)

= η

ξ + iηuy(z1) − uy(z) + i vy(z3) − vy(z)

+ ξ

ξ + iηux(z2) − ux(z) + i vx(z4) − vx(z).

Since lim

h→0|zk − z| = 0 for each 1 ≤ k ≤ 4,

η ξ + iη

,

η ξ + iη

≤ 1 and fx = ux + ivx and fy = uy+ ivy are continuous at z, we have

h→0lim

f (z + h) − f (z)

h − fx(z) = 0.

Examples

(2)

(a) Let f (z) = |z|2 = x2+ y2, z ∈ C.

Since fx = 2x, fy = 2y are continuous for all z ∈ C, f is differentiable if and only if fy = ifx

by the above proposition. Hence f is differentiable only at z = 0.

(b) The converse of the above proposition is not true. For example, consider

f (x, y) =

xy(x + iy)

x2+ y2 if z 6= 0

0 if z = 0.

Note that f = 0 on both axes so thatfx = fy = 0 at the origin but since limz→0

f (z) − f (0)

z = lim

(x,y)→(0,0)

xy x2+ y2

y=αx= α

1 + α2 depends on α, the limit does not exist and f is not differentiable at z = 0. In fact,fx and fy

fx(x, y) =

2xy3+ iy2(y2− x2)

(x2+ y2)2 if z 6= 0

0 if z = 0,

fy(x, y) =

x2(x2− y2) + 2ix3y

(x2 + y2)2 if z 6= 0

0 if z = 0

are not continuous at z = 0.

(c) Consider the real-valued function f : R2 → R defined by

f (x, y) =

x2+ y2 sin

 1

x2+ y2



if (x, y) 6= (0, 0)

0 if (x, y) = (0, 0).

Note that f is real-differentiable at every point (x, y) ∈ R2,

fx(x, y) =

 2x sin

 1

x2 + y2



− 2x

x2+ y2 cos

 1

x2 + y2



if (x, y) 6= (0, 0)

0 if (x, y) = (0, 0),

and

fy(x, y) =

 2y sin

 1

x2+ y2



− 2y

x2+ y2 cos

 1

x2+ y2



if (x, y) 6= (0, 0)

0 if (x, y) = (0, 0)

with fx(0, 0) = 0 = fy(0, 0), but fx, fy are not bounded (hence not continuous) in a neighborhood of (0, 0).

Remark (An algebraic interpretation of the Cauchy-Riemann equations) Consider the complex polynomial

p(z) =

n

X

k=0

ckzk.

(3)

Look at complex-valued polynomials of real variables x, y:

p(x, y) =

n

X

m=0

X

k+`=m,k, `≥0

ck, `xky` with ck, `= ∂k+`p

∂xk∂y`(0, 0) (∗)

and take a ”new basis” as z = x + iy and ¯z = x − iy. Write the polynomial in z, ¯z:

p(z, ¯z) =

n

X

m=0

X

k+`=m,k, `≥0

dk, `zk` (†)

We claim that there’s a 1 − 1 correspondence between polynomials given by (∗) and (†). In fact, since z = x + iy and ¯z = x − iy implies that x = z + ¯z

2 , y = z − ¯z

2i and by the Chain Rule

∂z = ∂x

∂z

∂x +∂y

∂z

∂z = 1 2

∂x − i ∂

∂y, ∂

∂ ¯z = ∂x

∂ ¯z

∂x +∂y

∂ ¯z

∂y = 1 2

∂x + i ∂

∂y.

Claim We have the equalities

∂ ¯z(zk`) = `zk`−1

∂z(zk`) = kzk−1` Proof of Claim We simply check

∂ ¯z(zk`) = 1 2

∂x + i ∂

∂y (x + iy)k(x − iy)`

= 1

2k(x + iy)k−1(x − iy)`+ `(x + iy)k(x − iy)`−1 + i

2ki(x + iy)k−1(x − iy)`− i`(x + iy)k(x − iy)`−1

= 1

2kzk−1`+ `zk`−1 + i

2kizk−1`− i`zk`−1

= `zk`−1

Claim (equivalent form of CR equations) f (z) satisfies CR equations ⇐⇒ ∂

∂ ¯zf = 0.

Proof of Claim Since

∂f

∂ ¯z = 1 2

∂f

∂x + i∂f

∂y

 =⇒ fy = ifx ⇐⇒ ∂f

∂ ¯z = 0.

Remark The conclusion is that

p(z, ¯z) =

n

X

m=0

X

k+`=m, k, `≥0

dk, `zk`

is holomorphic if and only if dk, `= 0 for ` > 0 (e.g. no ¯z’s). Alternatively, p(x, y) =

n

X

m=0

X

k+`=m, k, `≥0

ck, `xky`

(4)

is holomorphic if it can be written as p(z) =

n

X

k=0

ckzk.

Let’s give some basic hints that holomorphic functions behave a bit like power series:

Proposition

(a) If f = u + iv is analytic in a region D and u is constant, then f is constant.

Proof Since u is constant,

ux = uy = 0 in D, and, by the Cauchy-Riemann equations,

vx = vy = 0 in D.

By the Mean Value Theorem, u and v are constants along any line segment in D. So f = u + iv is constant on D.

(b) If f is analytic in a region D and if |f | is constant there,then f is constant.

Proof If |f | = 0, the proof is immediate. Otherwise u2+ v2 ≡ C 6= 0 on D.

Taking the partial derivatives with respect to x and y, we see that u ux+ v vx ≡ 0

u uy+ v vy ≡ 0 Making use of the Cauchy-Riemann equations, we obtain

u ux− v uy ≡ 0 v ux+ u uy ≡ 0 so that

(u2+ v2)ux ≡ 0 on D

and ux = vy ≡ 0. Similarly, uy and vx are identically zero, hence f is constant.

The Functions ez, sin z, cos z, log z, zα

In general, either one of the following methods will be used to define a complex analogue of a real analytic function.

(i) we determine the same fundamental defining properties;

(ii) we use its power series to obtain a holomorphic extension.

Definitions

(1) To define an extension f (z) = ez of the real exponential function f (x) = ex, we observe that

(5)

(i) e0 = 1, d

dxex = ex, . . . , which implies for each fixed x1 ∈ R, d

dx ex+x1

ex = exex+x1 − ex+x1ex

(ex)2 = 0 =⇒ ex+x1

ex = constant = ex1 =⇒ ex+x1 = exex1 implies that f (x) = ex is infinitely differentiableand satisfies that

f (0) = 1, df

dx = f and f (x + x1) = f (x)f (x1) for all x, x1 ∈ R.

(ii) ex=

X

k=0

xk

k! is convergent for all x ∈ R,implies that f (x) = ex is a real analytic function defined by a convergent power series for all x ∈ R.

So we should define f (z) such that

(i) f (x) = ex for all z = x + i 0 ∈ R and f (z1+ z2) = f (z1)f (z2) for all z1, z2 ∈ C.

This implies that

f (x + iy) = f (x)f (iy) = exf (iy) ∀x, y ∈ R and, by setting x = 0 and using f (0) = 1, we have

f (iy) = A(y) + iB(y),

where A, B are real-valued functions satisfying A(0) = 1, B(0) = 0, and f (x + iy) = exA(y) + iexB(y).

Using the Cauchy-Riemann equations fy = ifx, A(0) = 1 and B(0) = 0, we get (A0(y) = −B(y)

B0(y) = A(y) =⇒

(A00(y) + A(y) = 0

B00(y) + B(y) = 0 =⇒ A(y) = cos y and B(y) = sin y.

This implies that f (z) = ez should be defined by

ez = ex(cos y + i sin y) ⇐⇒ eiy = cos y + i sin y for all z = x + iy ∈ C.

(ii) f (z) =

X

k=0

zk

k! for all z ∈ C.

Since

lim

k→∞

1 k!

1/k

= 0,

the radius of convergence is ∞, f (z) =

X

k=0

zk

k! is convergent and differentiable for all z ∈ C. Since

f (0) = 1, f0(z) = f (z), · · · ∀ z ∈ C, by the term-by-term differentiation theorem, and

d dz

f (z + w)

f (z) = f (z)f (z + w) − f (z + w)f (z)

[f (z)]2 = 0 for a fixed w ∈ C,

(6)

we obtain that

f (z)f (w) = f (z + w) for all z, w ∈ C which implies that f (z) =

X

k=0

zk

k! = ez for all z ∈ C.

(2) Let f (z) = sin z, cos z for z ∈ C.

Since

eiz = 1 + (iz) + (iz)2

2! + · · · = 1 − z2 2! +z4

4! + · · · + i z − z3 3! +z5

5! + · · ·, we have

eix =

X

k=0

(−1)k

(2k)! x2k + i

X

k=0

(−1)k

(2k + 1)!x2k+1=: cos x + i sin x for x ∈ R and

cos x = eix+ e−ix

2 and sin x = eix − e−ix

2i for x ∈ R.

So sin z, cos z should be defined by cos z = eiz+ e−iz

2 and sin z = eiz− e−iz

2i for z ∈ C.

We can verify the usual properties that

∂z cos z = − sin z, ∂

∂z sin z = cos z,

cos 0 = 1, sin 0 = 0, cos(−z) = cos z, sin(−z) = − sin z, etc.

Check the addition formulas for cos(z1 + z2) and sin(z1+ z2).

(3) Let w = log z ⇐⇒ z = ew for z ∈ C.

For w = a + bi ∈ C, since

z = ew = ea(cos b + i sin b) = re,

log z should be defined by

log z = log r + iθ + i(2πn), z ∈ C so the complex log is multi-valued.

By writing z = re = r(cos θ + i sin θ), we define the principal branch of log z as follows.

For each z ∈ C \ {z = x + 0 · i ∈ C | x ≤ 0}, the principal branch of log z is defined by log z = log |z| + iArg(z), where Arg(z) ∈ (−π, π).

So the principal branch of log z has imaginary part in (−π, π).

Claim log z is holomorphic given a branch cut, and

∂z log z = 1

z for all z ∈ C \ {z = x + 0 · i ∈ C | x ≤ 0}.

(7)

Proof of Claim Cosider

h→0lim

log(z + h) − log z

h .

Setting w1 = log(z + h), w = log z, since log is continuous given a branch cut, we have lim

h→0w1 = w and thus

lim

h→0

log(z + h) − log z

h = lim

h→0

w1− w

ew1 − ew = 1

wlim1→w

ew1 − ew w1− w

= 1 ew = 1

z.

Intuitively, log z looks like a helix above C \ {0}; we call this the Riemann surface for log z.

(4) For α ∈ R, let f (z) = zα for z ∈ C.

Define

zα = eα log z = eα(log z+2πni), for z ∈ C

which is potentially multi-valued. eα log z is the principal branch; now examine e2πnαi. If α /∈ Q, e2πnαi takes infinitely many values.

If α = k

l ∈ Q, e2πnαi takes l values (e.g. eπni= ±1 for α = 1 2).

For example, when considering√

z, we still need a branch cut

re =√

reiθ/2, θ ∈ (−π, π).

參考文獻

相關文件

Mean Value Theorem... Tayor Theorem

(Cauchy condensation theorem) Let (a n ) be a nonnegative non increasing sequence of real numbers.. The proof follows from the

Verify that the function satisfies the hypotheses of the Mean Value Theorem on the

It is straightforward to show that this probability function satisfies the Kolmogorov Axioms.. If X is a discrete random variable, then X

In Section 3, the shift and scale argument from [2] is applied to show how each quantitative Landis theorem follows from the corresponding order-of-vanishing estimate.. A number

This contradiction shows that the given equation can’t have two distinct real roots, so it has exactly one

This contradiction shows that the given equation can’t have two distinct real roots, so it has exactly one

The Seed project, REEL to REAL (R2R): Learning English and Developing 21st Century Skills through Film-making in Key Stage 2, aims to explore ways to use film-making as a means