Analytic Functions and the Cauchy–Riemann Equations
CSIR NET complex analysis: complex differentiability, Cauchy–Riemann equations in Cartesian and polar form, harmonic functions, conjugates, Milne-Thomson.
- 21 min
On this page · 13 sections
Prerequisites
- Complex Plane, Power Series and Elementary Functions: \( e^z \), \( \operatorname{Log} z \) and the branch cut.
- Metric Spaces and Several Variables: partial derivatives, the total derivative, and the classic functions whose partials exist without differentiability.
- Continuity, Differentiability and Bounded Variation: the one-variable derivative this page generalises.
The Hook: Two Directions, One Answer
Complex Differentiability
Let \( D \subseteq \mathbb{C} \) be open and \( f : D \to \mathbb{C} \). Then \( f \) is differentiable at \( z_0 \in D \) if
$$f'(z_0) = \lim_{h \to 0} \frac{f(z_0 + h) - f(z_0)}{h}$$exists, with \( h \to 0 \) through complex values in any manner.
- Sums, products, quotients (where the denominator is non-zero) and compositions of differentiable functions are differentiable, with the usual rules. Differentiable at \( z_0 \) implies continuous at \( z_0 \).
- Polynomials are differentiable everywhere; rational functions everywhere except at the zeros of the denominator.
Analytic (holomorphic). \( f \) is analytic at \( z_0 \) if it is differentiable at every point of some neighbourhood of \( z_0 \). It is analytic on an open set if it is differentiable at every point of the set. A function analytic on all of \( \mathbb{C} \) is entire.
The Cauchy–Riemann Equations
Write \( f(x + iy) = u(x, y) + i v(x, y) \).
Necessity. If \( f \) is differentiable at \( z_0 = x_0 + iy_0 \), let \( h \to 0 \) along the real axis and then along the imaginary axis:
$$f'(z_0) = u_x + i v_x = \frac{1}{i}(u_y + i v_y) = v_y - i u_y.$$Comparing real and imaginary parts:
$$\boxed{u_x = v_y, \qquad u_y = -v_x} \qquad \text{and} \qquad f'(z_0) = u_x + i v_x = v_y - i u_y.$$Sufficiency. CR alone is not enough (see the counter-examples below). The standard sufficient conditions:
- If \( u \) and \( v \) are differentiable at \( (x_0, y_0) \) as functions of two real variables (in the sense of the total derivative) and satisfy CR there, then \( f \) is differentiable at \( z_0 \).
- In particular, if the four partials \( u_x, u_y, v_x, v_y \) exist in a neighbourhood and are continuous at \( (x_0, y_0) \), and CR hold at \( (x_0, y_0) \), then \( f \) is differentiable at \( z_0 \).
- Hence: \( u, v \in C^1(D) \) satisfying CR throughout the open set \( D \) gives \( f \) analytic on \( D \).
(A deeper result, the Looman–Menchoff theorem, says that a continuous \( f \) whose partials exist and satisfy CR at every point of an open set is analytic there. Continuity on the whole open set is essential; it is not a pointwise test.)
Other forms of the equations
| Form | Equations | Derivative |
|---|---|---|
| Cartesian | \( u_x = v_y \), \( u_y = -v_x \) | \( f' = u_x + i v_x \) |
| Polar, \( z = re^{i\theta} \), \( r > 0 \) | \( u_r = \frac{1}{r} v_\theta \), \( v_r = -\frac{1}{r} u_\theta \) | \( f' = e^{-i\theta}(u_r + i v_r) \) |
| Complex (Wirtinger) | \( \dfrac{\partial f}{\partial \bar z} = \dfrac{1}{2}\left(f_x + i f_y\right) = 0 \) | \( f' = \dfrac{\partial f}{\partial z} = \dfrac{1}{2}(f_x - i f_y) \) |
| Single complex equation | \( f_y = i f_x \) |
The Wirtinger form gives the fastest test for functions written in \( z \) and \( \bar z \): a smooth \( f \) is analytic exactly where it “does not depend on \( \bar z \)”, i.e. where \( \partial f / \partial \bar z = 0 \). For \( f = z\bar z = \lvert z \rvert^2 \), \( \partial f/\partial \bar z = z \), zero only at 0.
Jacobian. For analytic \( f \), the Jacobian of \( (x, y) \mapsto (u, v) \) is
$$\det \begin{pmatrix} u_x & u_y \\ v_x & v_y \end{pmatrix} = u_x^2 + v_x^2 = \lvert f'(z) \rvert^2 \ge 0.$$So an analytic map with \( f' \ne 0 \) is a local orientation-preserving diffeomorphism that scales areas by \( \lvert f' \rvert^2 \). This is the root of conformality.
Where is it differentiable? The routine
- Write \( u \), \( v \); compute the four partials.
- Solve the CR equations: the solution set is the only place where \( f \) can be differentiable.
- Check a sufficient condition there (continuous partials is usually immediate).
- \( f \) is analytic only on the interior of that set: on an open set, not on a curve or at isolated points.
| \( f(z) \) | CR hold at | Differentiable at | Analytic on |
|---|---|---|---|
| \( \bar z \) | nowhere | nowhere | \( \varnothing \) |
| \( \lvert z \rvert^2 \) | \( z = 0 \) | \( z = 0 \) | \( \varnothing \) |
| \( x^2 + i y^2 \) | the line \( y = x \) | the line \( y = x \) | \( \varnothing \) |
| \( x^2 - y^2 + 2ixy = z^2 \) | everywhere | everywhere | \( \mathbb{C} \) |
| \( e^x(\cos y - i\sin y) = e^{\bar z} \) | nowhere | nowhere | \( \varnothing \) |
| \( \operatorname{Re} z \) | nowhere (\( u_x = 1 \ne 0 = v_y \)) | nowhere | \( \varnothing \) |
| \( \operatorname{Log} z \) | on \( \mathbb{C} \setminus (-\infty, 0] \) | there | \( \mathbb{C} \setminus (-\infty, 0] \) |
CR at a point without differentiability
| Function | Why CR hold at 0 | Why not differentiable at 0 |
|---|---|---|
| \( f = \sqrt{\lvert xy \rvert} \) (so \( v = 0 \)) | \( u = 0 \) on both axes, so \( u_x = u_y = 0 = v_x = v_y \) | along \( y = x \), \( \frac{f(h)}{h} = \frac{\lvert t \rvert}{t(1 + i)} \) has different limits as \( t \to 0^\pm \) |
| \( f(z) = \frac{z^5}{\lvert z \rvert^4} \), \( f(0) = 0 \) | \( f(x) = x \), \( f(iy) = iy \), so \( u_x = v_y = 1 \), \( u_y = v_x = 0 \) | \( \frac{f(h)}{h} = \frac{h^4}{\lvert h \rvert^4} = e^{4i\theta} \) depends on direction |
| \( f(z) = e^{-z^{-4}} \), \( f(0) = 0 \) | \( f \to 0 \) along both axes faster than any power | along \( z = re^{i\pi/4} \), \( z^{-4} = -r^{-4} \), so \( \lvert f \rvert = e^{r^{-4}} \to \infty \): not even continuous |
The moral for Part C: “partials exist and satisfy CR at \( z_0 \)” does not imply “differentiable at \( z_0 \)”.
Rigidity: When Analytic Functions Must Be Constant
Let \( f = u + iv \) be analytic on a domain \( D \) (open and connected). Each of the following forces \( f \) to be constant on \( D \):
| Hypothesis | One-line reason |
|---|---|
| \( f' \equiv 0 \) | \( u_x = u_y = v_x = v_y = 0 \), and \( D \) is connected |
| \( u \) constant (or \( v \) constant) | CR give \( v_x = v_y = 0 \) |
| \( f \) real-valued | \( v \equiv 0 \) is constant |
| \( \lvert f \rvert \) constant | \( u^2 + v^2 = c \); differentiate and use CR: \( (u^2 + v^2) u_x = 0 \), etc. |
| \( \bar f \) also analytic | \( f + \bar f = 2u \) is analytic and real-valued |
| \( f(D) \) lies on a line or a circle | compose with a map sending it to \( \mathbb{R} \); or use the open mapping theorem |
| \( v = \varphi(u) \) for a differentiable real \( \varphi \) | CR give \( u_x(1 + \varphi'(u)^2) = 0 \) |
| \( \operatorname{Arg} f \) constant | \( f(D) \) lies on a ray |
The last three are best seen through the open mapping theorem: a non-constant analytic function maps open sets to open sets, and a line, a circle or a ray contains no open set.
Harmonic Functions
A real \( C^2 \) function \( u \) on an open set is harmonic if
$$\Delta u = u_{xx} + u_{yy} = 0, \qquad \text{in polar form} \quad u_{rr} + \frac{1}{r}u_r + \frac{1}{r^2} u_{\theta\theta} = 0.$$Real and imaginary parts of analytic functions are harmonic. Analytic functions are \( C^\infty \) (a consequence of Cauchy’s integral formula), so \( u_{xx} = v_{yx} = v_{xy} = -u_{yy} \).
Harmonic conjugate. \( v \) is a harmonic conjugate of \( u \) on \( D \) if \( u + iv \) is analytic on \( D \).
| Fact | Detail |
|---|---|
| Uniqueness | on a domain, a conjugate is unique up to an additive real constant |
| Existence | every harmonic \( u \) on a simply connected domain has a conjugate |
| Can fail otherwise | \( u = \ln\lvert z \rvert \) on \( \mathbb{C} \setminus \{0\} \): a conjugate would be a continuous \( \arg z \), which does not exist on the punctured plane |
| Order matters | if \( v \) is a conjugate of \( u \), then \( -u \) (not \( u \)) is a conjugate of \( v \), since \( -i(u + iv) = v - iu \) |
| Products | \( uv \) is harmonic (\( uv = \frac{1}{2}\operatorname{Im}(f^2) \)) |
| Squares | \( u^2 \) is harmonic only if \( u \) is constant: \( \Delta(u^2) = 2\lvert \nabla u \rvert^2 \) |
| Orthogonality | where \( f' \ne 0 \), the level curves \( u = c_1 \) and \( v = c_2 \) cross at right angles |
| Mean value | \( u(a) = \frac{1}{2\pi} \int_0^{2\pi} u(a + re^{i\theta})\,d\theta \) on any closed disc in the domain |
| Maximum principle | a non-constant harmonic function on a domain has no local maximum or minimum |
Standard harmonic functions: \( x \), \( y \), \( xy \), \( x^2 - y^2 \), \( x^3 - 3xy^2 \), \( e^x \cos y \), \( e^x \sin y \), \( \ln(x^2 + y^2) \) on \( \mathbb{R}^2 \setminus \{0\} \), \( r^n \cos n\theta \), \( r^n \sin n\theta \), \( \frac{x}{x^2 + y^2} \). Not harmonic: \( x^2 + y^2 \), \( x^2 \), \( e^x \cos x \), \( \lvert z \rvert \).
Finding \( f \) from \( u \): three methods
1. Integrate the CR equations. From \( v_y = u_x \), integrate in \( y \); fix the “constant” \( \phi(x) \) using \( v_x = -u_y \).
2. Line integral. On a simply connected domain, \( v(x, y) = \int_{(x_0, y_0)}^{(x, y)} (-u_y\,dx + u_x\,dy) \), which is path-independent because \( u \) is harmonic.
3. Milne-Thomson method (the fastest). For \( f = u + iv \) analytic,
$$\boxed{f'(z) = u_x(z, 0) - i\,u_y(z, 0)} \qquad \text{or, if } v \text{ is given,} \qquad \boxed{f'(z) = v_y(z, 0) + i\,v_x(z, 0)}.$$Substitute \( x = z \), \( y = 0 \) in the partials, then integrate in \( z \). It works because both sides are analytic and agree on the real axis (the identity theorem; the method needs the domain to meet the real axis).
Memory Tricks
Don’t Confuse
| This | vs | That |
|---|---|---|
| differentiable at \( z_0 \) | analytic at \( z_0 \) (differentiable on a neighbourhood) | |
| CR hold at \( z_0 \) (necessary) | CR + continuous partials (sufficient) | |
| real-differentiable as a map of \( \mathbb{R}^2 \) (e.g. \( \bar z \)) | complex-differentiable | |
| \( u \) harmonic | \( u \) has a single-valued conjugate (needs simple connectivity, or luck) | |
| \( uv \) harmonic (always, if \( v \) is a conjugate) | \( u^2 \) harmonic (only if \( u \) is constant) | |
| conjugate of \( u \) is \( v \) | conjugate of \( v \) is \( -u \) | |
| \( f' = u_x + iv_x \) | \( f' = u_x + iu_y \) (wrong) |
Common Mistakes
Practice
All questions below are original practice questions written for this page. They are not from any CSIR NET paper. In Level 3, as in Part C of the Mathematical Sciences paper, one or more options may be correct.
Level 1: Foundation
- Is \( f(z) = \bar z \) differentiable at any point?
- Write the Cauchy–Riemann equations in polar form.
- Find a harmonic conjugate of \( u = x^2 - y^2 \).
Answers
- No. \( u_x = 1 \), \( v_y = -1 \): the first CR equation fails everywhere.
- \( u_r = \frac{1}{r}v_\theta \), \( v_r = -\frac{1}{r}u_\theta \) (for \( r > 0 \)).
- \( v = 2xy \) (plus any real constant), since \( z^2 = x^2 - y^2 + 2ixy \).
Level 2: Part B Style
Practice Q1. The function \( u(x, y) = ax^3 + bxy^2 \) is harmonic on \( \mathbb{R}^2 \) if and only if:
(a) \( b = 3a \) (b) \( b = -3a \) (c) \( a = -3b \) (d) \( a = b = 0 \)
Answer: (b)
\( u_{xx} = 6ax \), \( u_{yy} = 2bx \). \( \Delta u = (6a + 2b)x = 0 \) for all \( x \) iff \( b = -3a \). For \( a = 1 \) this is \( x^3 - 3xy^2 = \operatorname{Re}(z^3) \). Option (d) is sufficient but not necessary.
Practice Q2. Let \( f = u + iv \) be entire with \( u(x, y) = e^x(x\cos y - y\sin y) \) and \( f(0) = 0 \). Then \( f(i\pi) \) equals:
(a) \( i\pi \) (b) \( -i\pi \) (c) \( -\pi \) (d) 0
Answer: (b)
By Milne-Thomson (worked example above), \( f(z) = ze^z \). So \( f(i\pi) = i\pi e^{i\pi} = -i\pi \).
Practice Q3. Let \( f(x + iy) = x^2 + iy^2 \). Then \( f \) is:
(a) analytic on \( \mathbb{C} \) (b) differentiable only on the line \( y = x \), and analytic nowhere (c) differentiable only at \( z = 0 \) (d) differentiable nowhere
Answer: (b)
\( u_x = 2x \), \( v_y = 2y \), \( u_y = 0 = -v_x \). CR hold iff \( x = y \). The partials are continuous everywhere, so \( f \) is differentiable at each point of the line. The line contains no open set, so \( f \) is analytic nowhere.
Practice Q4. If \( v = e^x \sin y \) is the imaginary part of an entire function \( f \) with \( f(0) = 1 \), then \( f(z) \) is:
(a) \( e^z \) (b) \( e^{\bar z} \) (c) \( e^z + i \) (d) \( ie^z \)
Answer: (a)
Milne-Thomson with \( v \): \( f'(z) = v_y(z, 0) + i v_x(z, 0) = e^z \cos 0 + i e^z \sin 0 = e^z \). So \( f = e^z + c \) with \( c \) real (adding \( ic \) would change \( v \)); \( f(0) = 1 \) gives \( c = 0 \).
Level 3: Part C Style
Practice Q5. Let \( D \) be a domain and \( f = u + iv \) analytic on \( D \). Which statements are true?
(a) If \( u \) is constant on \( D \), then \( f \) is constant. (b) If \( \lvert f \rvert \) is constant on \( D \), then \( f \) is constant. (c) If \( v = u^2 \) on \( D \), then \( f \) is constant. (d) \( u^2 \) is harmonic on \( D \).
Answer: (a), (b) and (c)
- (a) True. CR give \( v_x = -u_y = 0 \), \( v_y = u_x = 0 \); \( D \) is connected.
- (b) True. If \( u^2 + v^2 = c \ne 0 \), differentiating and using CR gives \( c\,u_x = c\,u_y = 0 \); if \( c = 0 \) then \( f \equiv 0 \).
- (c) True. \( v_y = 2u u_y \) and \( v_x = 2u u_x \). CR: \( u_x = 2u u_y \) and \( u_y = -2u u_x \). Substituting, \( u_x = -4u^2 u_x \), so \( u_x(1 + 4u^2) = 0 \), hence \( u_x = 0 \) and then \( u_y = 0 \).
- (d) False in general. \( \Delta(u^2) = 2(u_x^2 + u_y^2) \); for \( f = z \), \( u^2 = x^2 \) has Laplacian 2.
Practice Q6. Let \( u(x, y) = \ln(x^2 + y^2) \) on \( \Omega = \mathbb{C} \setminus \{0\} \). Which statements are true?
(a) \( u \) is harmonic on \( \Omega \). (b) \( u \) has a harmonic conjugate on \( \Omega \). (c) \( u \) has a harmonic conjugate on \( \mathbb{C} \setminus (-\infty, 0] \). (d) \( u \) is the real part of a function analytic on \( \Omega \).
Answer: (a) and (c)
- (a) True. \( u = 2\ln r \), and \( \frac{d^2}{dr^2}(2\ln r) + \frac{1}{r}\frac{d}{dr}(2\ln r) = -\frac{2}{r^2} + \frac{2}{r^2} = 0 \).
- (b) False. A conjugate \( v \) would make \( v - 2\operatorname{Arg} z \) constant on the cut plane, so \( v \) would jump by \( 4\pi \) across the negative axis, contradicting continuity on \( \Omega \).
- (c) True. \( v = 2\operatorname{Arg} z \): \( u + iv = 2\operatorname{Log} z \) is analytic on the cut plane.
- (d) False. It is the same statement as (b).
Practice Q7. Define \( f(z) = \frac{z^5}{\lvert z \rvert^4} \) for \( z \ne 0 \) and \( f(0) = 0 \). Which statements are true?
(a) \( f \) is continuous at 0. (b) The partial derivatives of \( u = \operatorname{Re} f \) and \( v = \operatorname{Im} f \) exist at \( (0, 0) \) and satisfy the Cauchy–Riemann equations there. (c) \( f \) is differentiable at 0. (d) \( f \) is analytic at 0.
Answer: (a) and (b)
- (a) True. \( \lvert f(z) \rvert = \lvert z \rvert \to 0 \).
- (b) True. \( f(x) = x \) and \( f(iy) = i^5 y = iy \) for real \( x, y \). So \( u(x, 0) = x \), \( u(0, y) = 0 \), \( v(x, 0) = 0 \), \( v(0, y) = y \): \( u_x = 1 = v_y \), \( u_y = 0 = -v_x \).
- (c) False. \( \frac{f(h)}{h} = \frac{h^4}{\lvert h \rvert^4} = e^{4i\theta} \) for \( h = \lvert h \rvert e^{i\theta} \): equal to 1 along \( \theta = 0 \) and to \( -1 \) along \( \theta = \frac{\pi}{4} \).
- (d) False. It is not even differentiable at 0.
Quick Revision Table
| Idea | Fact |
|---|---|
| Differentiable | quotient limit exists from all directions |
| Analytic at \( z_0 \) | differentiable on a neighbourhood |
| CR (Cartesian) | \( u_x = v_y \), \( u_y = -v_x \); \( f' = u_x + iv_x \) |
| CR (polar) | \( u_r = \frac{1}{r}v_\theta \), \( v_r = -\frac{1}{r}u_\theta \) |
| CR (complex) | \( \partial f/\partial \bar z = 0 \) |
| Sufficient | CR + continuous partials (or real-differentiable \( u, v \)) |
| Jacobian | \( \lvert f' \rvert^2 \) |
| Forced constant | \( u \), \( v \), \( \lvert f \rvert \) or \( \arg f \) constant; \( \bar f \) analytic; \( f \) real |
| Harmonic | \( \Delta u = 0 \); real and imaginary parts of analytic \( f \) |
| Conjugate | exists on simply connected domains; \( u \to v \to -u \) |
| Milne-Thomson | \( f'(z) = u_x(z, 0) - iu_y(z, 0) \) |
Key Takeaways
What’s Next
- Cauchy’s Theorem and Its Consequences: integration turns the CR equations into infinite differentiability, Liouville and the maximum modulus principle.
- Conformal and Möbius Maps: the geometric meaning of \( f' \ne 0 \).
- Complex Plane, Power Series and Elementary Functions for the branch of \( \log \) used above.
- Complex Analysis: Quick Revision.