Complex Analysis
Cauchy-Riemann Equations
For f = u + iv, holomorphicity ⇔ ∂u/∂x = ∂v/∂y and ∂u/∂y = −∂v/∂x
A complex function f(x + iy) = u(x, y) + iv(x, y) is holomorphic at a point iff its real and imaginary parts u, v satisfy the Cauchy-Riemann equations: ∂u/∂x = ∂v/∂y and ∂u/∂y = −∂v/∂x. Equivalently, the Jacobian of (u, v) is a scaled rotation — this geometrically means complex differentiation preserves angles (conformal). The CR equations imply both u and v are harmonic: u_xx + u_yy = 0 (similarly for v). First derived by d'Alembert (1752) for fluid dynamics, formalized by Cauchy (1814) and Riemann (1851 PhD thesis). Used in fluid dynamics (potential flow), electrostatics (potential u harmonic), and conformal mapping.
- Equations∂u/∂x = ∂v/∂y, ∂u/∂y = −∂v/∂x
- Equivalentf holomorphic
- Consequenceu, v harmonic
- Firstd'Alembert 1752
- FormalizedCauchy 1814, Riemann 1851
- Geometricconformal (angle-preserving)
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
Why the Cauchy-Riemann equations matter
- Fluid dynamics — complex potentials. 2D incompressible irrotational flow has a velocity potential φ and a stream function ψ that satisfy the CR equations; F(z) = φ + iψ is holomorphic, and Joukowski maps construct flow around airfoils by conformal transformation. Every introductory aerodynamics course solves the airfoil problem by CR + conformal map.
- Electrostatics — 2D potential problems. The electric potential u(x, y) in a charge-free region is harmonic. Its harmonic conjugate v gives the electric field lines (level sets of v are perpendicular to equipotentials of u). Capacitor edge fields, transmission line characteristic impedances, and microstrip designs use CR-based 2D solvers daily.
- Conformal mapping. The Riemann mapping theorem says every simply connected proper open subset of ℂ is conformally equivalent to the unit disc. CR is the tool that builds and certifies these maps; Schwarz-Christoffel formulas construct explicit polygons-to-half-plane maps.
- Harmonic function theory. CR makes the bridge between holomorphic and harmonic. Locally, every harmonic u has a harmonic conjugate v, so harmonic-function questions can be answered by complex methods. Hardy spaces, Bergman spaces, and boundary-value problems all run on this duality.
- Heat conduction (steady state). The temperature distribution in a thin plate at equilibrium satisfies Laplace's equation; isotherms and heat-flux lines form an orthogonal grid (a CR conjugate pair). Engineers solve heat-flow problems in 2D geometries via conformal maps.
- Computer graphics — texture mapping and parametrization. Conformal mesh parametrizations satisfy a discrete version of CR; they minimize angular distortion when unfolding 3D surfaces to 2D textures. Algorithms like Least-Squares Conformal Maps and Discrete Natural Conformal Parametrization are direct numerical CR solvers.
- Quantum mechanics — analytic continuation. Wave functions and Green's functions are routinely continued to complex energies; the CR equations are the structural constraint that makes this continuation well-defined. S-matrix theory and resonance scattering live in CR-constrained spaces.
Common misconceptions
- "Real-differentiable implies CR." No. Real differentiability as a map ℝ² → ℝ² gives an arbitrary linear approximation; CR forces that approximation to be a scaled rotation. f(z) = z̄ is C^∞ as a real map but fails CR (u_x = 1, v_y = −1).
- "Any harmonic function comes from a holomorphic one." Locally yes; globally only on simply connected domains. log|z| is harmonic on ℂ \ {0} but its harmonic conjugate is arg(z), which is multi-valued — exactly the obstruction to defining log z on a punctured plane.
- "CR is just notation." CR encodes a rigid structural constraint with deep consequences: holomorphic ⇒ analytic, conformality, the maximum modulus principle, Liouville's theorem. Real partial differential equations don't have these features; CR is what makes complex analysis qualitatively different.
- "CR + smoothness ⇒ holomorphic, anywhere." CR plus continuity at a point only gives complex differentiability there. Looman-Menchoff theorem: CR + continuity on an open set ⇒ holomorphic on that set. Without continuity, CR can hold pointwise without giving differentiability — pathological examples exist.
- "f(z) = |z|² is holomorphic — its real part is x² + y², which is differentiable." No: u = x² + y², v = 0, so u_x = 2x, v_y = 0 — CR fails everywhere except z = 0. f is real-analytic but holomorphic only at 0 (where the derivative trivially exists). Polynomials in z̄ are almost never holomorphic.
- "CR works the same in higher dimensions." In ℂⁿ for n ≥ 2, the analogous CR system is overdetermined; not every smooth function in ℝ^{2n} can be made holomorphic. The Hartogs phenomenon and Levi pseudoconvexity govern when ℂⁿ-valued problems have solutions; this is the realm of several complex variables.
Statement and equivalent forms
Let f(z) = f(x + iy) = u(x, y) + i v(x, y) be a function on an open set U ⊂ ℂ. The following are equivalent (for f continuous on U):
- f is holomorphic on U: the limit f′(z) = lim_{h→0} (f(z+h) − f(z))/h exists at every z ∈ U.
- u and v have continuous first partials on U and satisfy ∂u/∂x = ∂v/∂y and ∂u/∂y = −∂v/∂x.
- The Wirtinger derivative ∂f/∂z̄ = (1/2)(∂f/∂x + i ∂f/∂y) vanishes identically on U.
- The Jacobian of f as a map ℝ² → ℝ² is at every point a scaled rotation matrix.
The equivalence between (1) and (2) is the historical statement; (3) is the convenient computational form (∂̄ f = 0); (4) is the geometric interpretation (conformality up to scale).
Why these specific PDEs?
Complex differentiability requires the limit (f(z+h) − f(z))/h to be the same regardless of how h → 0. Two natural directions: along the real axis (h = Δx) and along the imaginary axis (h = i Δy).
Real-axis approach:
f′(z) = lim_{Δx→0} [u(x+Δx, y) + i v(x+Δx, y) − u(x, y) − i v(x, y)] / Δx
= ∂u/∂x + i ∂v/∂x.
Imaginary-axis approach:
f′(z) = lim_{Δy→0} [u(x, y+Δy) + i v(x, y+Δy) − u(x, y) − i v(x, y)] / (i Δy)
= (1/i) (∂u/∂y + i ∂v/∂y)
= ∂v/∂y − i ∂u/∂y.
Equating real parts: ∂u/∂x = ∂v/∂y.
Equating imag parts: ∂v/∂x = −∂u/∂y.
That's the system. CR equations are not chosen for elegance — they are the precise compatibility conditions that force the directional derivative to be direction-independent. Approaches along arbitrary directions θ then automatically agree because the resulting Jacobian is a scaled rotation, and rotations applied to direction vectors commute with the rotation built into f.
Why u and v are harmonic
Differentiating both CR equations:
∂/∂x of (∂u/∂x = ∂v/∂y): u_xx = v_yx.
∂/∂y of (∂u/∂y = −∂v/∂x): u_yy = −v_xy.
Add: u_xx + u_yy = v_yx − v_xy = 0 (mixed partials equal, by Schwarz).
So Δu = 0; u is harmonic. Symmetric argument: Δv = 0.
Schwarz's theorem on equality of mixed partials needs C² regularity. Holomorphic functions are automatically C^∞ (proved separately via the Cauchy integral formula), so the regularity is free. Both u and v solve Laplace's equation; they form a harmonic conjugate pair.
A converse holds locally: given a harmonic u on a simply connected domain, the function v(x, y) = ∫ (−u_y dx + u_x dy) along any path from a fixed base point is well-defined (path-independence follows from u_xx + u_yy = 0), and (u, v) satisfies CR. So harmonic functions on simply connected domains lift to holomorphic functions, and the CR equations are the bridge.
Geometric meaning: conformality
The Jacobian of f = u + iv as a map ℝ² → ℝ² is
J = [[ u_x, u_y ],
[ v_x, v_y ]].
Under CR, u_y = −v_x and v_y = u_x. Letting a = u_x, b = v_x:
J = [[ a, −b ],
[ b, a ]].
This is the matrix representation of multiplication by the complex number a + bi: a scaled rotation by angle arg(a + bi) and factor |a + bi| = √(a² + b²). Geometrically, an infinitesimal displacement vector at z is rotated by a fixed angle (depending on z) and scaled by a fixed factor — independent of the direction of the vector.
Consequence: angles between curves at z are preserved. If two curves cross at z with directions v₁, v₂ making angle θ between them, the image curves at f(z) cross with directions Jv₁, Jv₂ also making angle θ. This is conformality. It fails wherever f′(z) = 0 (a ramification point, where the local map is z ↦ z^k for k ≥ 2 and angles are multiplied by k).
Worked examples
Example 1: f(z) = z² = (x + iy)² = x² − y² + 2ixy. So u = x² − y², v = 2xy. Compute partials:
u_x = 2x, u_y = −2y.
v_x = 2y, v_y = 2x.
Check: u_x = v_y ✓ (both 2x).
u_y = −v_x ✓ (−2y = −2y).
CR holds everywhere; f is entire (holomorphic on all of ℂ).
f′(z) = u_x + i v_x = 2x + i 2y = 2(x + iy) = 2z. ✓
Example 2: f(z) = z̄ = x − iy. u = x, v = −y. Partials u_x = 1, v_y = −1. CR fails: u_x = 1 ≠ −1 = v_y. So z̄ is nowhere holomorphic, despite being smooth as a real map.
Example 3: f(z) = e^z = e^x cos(y) + i e^x sin(y). u = e^x cos(y), v = e^x sin(y). Partials:
u_x = e^x cos(y) = v_y ✓.
u_y = −e^x sin(y) = −v_x ✓.
CR holds everywhere; e^z is entire and f′ = u_x + iv_x = e^x cos(y) + i e^x sin(y) = e^z.
Example 4: f(z) = log z = log|z| + i arg(z). u = (1/2) log(x² + y²), v = arctan(y/x) (on a branch). Compute u_x = x/(x² + y²), v_y = (1/x) · 1/(1 + y²/x²) = x/(x² + y²). CR's first equation holds. u_y = y/(x² + y²), v_x = (−y/x²) · 1/(1 + y²/x²) = −y/(x² + y²) = −v_x. CR's second equation holds. log z is holomorphic on its branch, with f′(z) = u_x + iv_x = x/(x²+y²) + i (−y/(x²+y²)) = (x − iy)/|z|² = z̄/|z|² = 1/z. ✓
Polar form and Wirtinger calculus
In polar coordinates z = r e^{iθ}, write f = u(r, θ) + i v(r, θ). The CR equations become
∂u/∂r = (1/r) ∂v/∂θ, ∂v/∂r = −(1/r) ∂u/∂θ.
Useful for functions naturally defined in polar form (logarithms, powers, Bessel functions). The Laplacian in polar coordinates becomes Δu = u_rr + (1/r) u_r + (1/r²) u_θθ; harmonicity of u still follows from CR.
Wirtinger derivatives provide a more compact reformulation. Define
∂/∂z = (1/2) (∂/∂x − i ∂/∂y),
∂/∂z̄ = (1/2) (∂/∂x + i ∂/∂y).
Then f is holomorphic iff ∂f/∂z̄ = 0. This single equation packages both CR equations in one complex equation. ∂f/∂z, when ∂f/∂z̄ = 0, equals f′(z). Wirtinger calculus generalizes cleanly to several variables (∂̄ operator and Dolbeault cohomology) and underlies modern complex geometry and several-complex-variable theory.
Where the CR equations show up
- Aerodynamics — Joukowski airfoils. The Joukowski map z ↦ z + 1/z (with z = a + r e^{iθ}, a ≠ 0) sends the unit circle (slightly offset) to a closed curve with a pointed trailing edge — an airfoil. Flow past a cylinder (analytically computed) maps via CR-respecting transformation to flow past the airfoil, giving lift and circulation in closed form. Every introductory fluid dynamics course derives lift this way.
- Geophysics — gravity and magnetic anomalies. 2D anomalies in geological surveys are modeled by line sources whose potentials are harmonic. CR-conjugate pairs (gravity vs. its derivative; magnetic intensity vs. tilt angle) are computed by Hilbert transforms, which are CR-induced operators on the boundary of the half-plane.
- Electrical engineering — Smith charts. The Smith chart is a conformal map of the right half of the impedance plane to the unit disc, used to design impedance-matching networks. Its conformality (preservation of angles) is exactly the CR-driven feature that lets engineers read off transmission line parameters geometrically.
- Computer graphics — Möbius and conformal parametrizations. Discrete CR-equation solvers (Least-Squares Conformal Maps, ABF, LSCM++) parametrize 3D meshes onto the plane with minimal angle distortion. Texturing, retopology, and surface reconstruction pipelines depend on these CR-derived energies.
- Cardiology — heart conformal maps. Cortical and cardiac surfaces are flattened to discs or spheres for atlas registration via discrete CR. Brain imaging (FreeSurfer, BrainSuite) uses CR-respecting parametrizations to compare anatomy across subjects.
- Mathematical physics — Riemann-Hilbert problems. Inverse scattering for integrable PDEs (KdV, NLS) is recast as a Riemann-Hilbert problem: find a piecewise-holomorphic function with prescribed jumps across a contour. CR equations on each piece are the structural constraint; the jump condition encodes the scattering data.
History
Jean d'Alembert wrote down what we now call the Cauchy-Riemann equations in 1752 in the context of fluid dynamics — specifically, in studying 2D incompressible flow, where they appear as the condition that the velocity field admits both a potential and a stream function. Euler used closely related identities. The equations were not, at the time, viewed as a definition of complex differentiability; they were a useful PDE system in real-variable applications.
Augustin-Louis Cauchy, working in the 1810s and 1820s, developed complex integration and observed that the equations characterize complex differentiability. His 1814 manuscript on definite integrals contained an early version. Bernhard Riemann's 1851 doctoral dissertation, "Grundlagen für eine allgemeine Theorie der Functionen einer veränderlichen complexen Grösse," took the equations as the foundational definition of analytic functions and built complex analysis on top — making them the central object of the field. The naming "Cauchy-Riemann equations" stuck by the late 19th century.
Frequently asked questions
Why do these specific PDEs characterize complex differentiability?
The complex derivative f′(z) = lim_{h→0} (f(z+h) − f(z))/h must give the same value regardless of the direction along which h tends to zero. Approach along the real axis (h = Δx): f′ = ∂u/∂x + i ∂v/∂x. Approach along the imaginary axis (h = i Δy): f′ = ∂v/∂y − i ∂u/∂y. Equating real and imaginary parts gives ∂u/∂x = ∂v/∂y and ∂v/∂x = −∂u/∂y — exactly the Cauchy-Riemann equations. They are precisely what's needed for the limit to be direction-independent.
What does it mean geometrically (conformal/angle-preserving)?
Treat f = u + iv as a map ℝ² → ℝ². Its Jacobian matrix is [[u_x, u_y], [v_x, v_y]]. Under the CR equations, this becomes [[a, −b], [b, a]] with a = u_x, b = v_x. Such matrices are scaled rotations: a + bi acts on vectors in the plane by scaling by |a + bi| and rotating by arg(a + bi). So at every point where f′(z) ≠ 0, infinitesimal vectors are rotated by a fixed angle and scaled by a fixed factor. Angles between curves at z are preserved — that's exactly conformality.
Why are u and v harmonic?
Differentiate ∂u/∂x = ∂v/∂y with respect to x: u_xx = v_yx. Differentiate ∂u/∂y = −∂v/∂x with respect to y: u_yy = −v_xy. Mixed partials are equal (assuming sufficient smoothness, which holomorphic functions enjoy automatically), so u_xx + u_yy = v_yx − v_xy = 0. Thus Δu = 0; u is harmonic. The argument is symmetric in v: Δv = 0. Both real and imaginary parts of any holomorphic function satisfy Laplace's equation. This is why complex analysis underlies 2D potential theory.
What's the difference between real differentiable and complex differentiable?
Real differentiable as a map ℝ² → ℝ² (Fréchet differentiable) means f has a linear approximation L: ℝ² → ℝ² at each point. Any L will do — rotations, shears, reflections, arbitrary 2×2 matrices. Complex differentiable is much more restrictive: the linear approximation must be of the form 'multiply by a fixed complex number' — i.e., a scaled rotation. Generic real-differentiable maps fail Cauchy-Riemann; only the special ones whose derivative matrix is a similarity matrix at every point are holomorphic. f(z) = z̄ is real-differentiable (it's the reflection (x,y) ↦ (x,−y)) but not complex differentiable: u_x = 1, v_y = −1 — CR fails.
How are they used in fluid dynamics?
For 2D incompressible irrotational flow, the velocity field (vₓ, v_y) satisfies ∇·v = 0 (incompressible) and ∇×v = 0 (irrotational). Define a velocity potential φ with v = ∇φ; then Δφ = 0 (from incompressibility). Define a stream function ψ with vₓ = ∂ψ/∂y, v_y = −∂ψ/∂x; then Δψ = 0 (from irrotationality). The pair (φ, ψ) satisfies Cauchy-Riemann, so F(z) = φ + iψ is holomorphic — the complex potential. Streamlines are level sets of ψ, equipotentials are level sets of φ; they meet at right angles by conformality. Joukowski's airfoil maps and Karman-Trefftz transforms construct flow around airfoils by conformal maps.
What is a harmonic conjugate?
Given a harmonic function u on a simply connected domain, a harmonic conjugate v is a function such that f = u + iv is holomorphic; equivalently, v satisfies the CR equations with u. The conjugate is unique up to an additive constant. Construction: integrate ∂v/∂x = −∂u/∂y along x at fixed y, then determine the y-dependence from ∂v/∂y = ∂u/∂x. Existence requires simple connectedness — on a non-simply-connected domain, the conjugate may be multi-valued (this is exactly the obstruction in defining log z on ℂ \ {0}). Every harmonic function locally has a holomorphic completion, but globally only on simply connected domains.