The Cauchy–Riemann Equations
The complex
derivative exists only when the difference quotient gives the same limit from
every direction. Demanding agreement between just two directions —
along the real axis and along the imaginary axis — already pins down a remarkable pair of
equations. They are the Cauchy–Riemann equations, and they are the algebraic
heart of the whole subject.
Write f(z) = u(x, y) + i\,v(x, y) as usual, with
z = x + iy. We will compute f'(z) two
ways and force the answers to match. Both ways use
partial
derivatives of u and v.
Deriving the equations
Step 1 — approach along the real axis. Let
h = \Delta x \to 0. Only x changes, so
the quotient becomes a partial derivative in x:
f'(z) = \frac{\partial u}{\partial x} + i\,\frac{\partial v}{\partial x} = u_x + i\,v_x.
Step 2 — approach along the imaginary axis. Let
h = i\,\Delta y \to 0. Now only y
changes, and dividing by h = i\,\Delta y introduces a
1/i:
f'(z) = \frac{1}{i}\left(\frac{\partial u}{\partial y} + i\,\frac{\partial v}{\partial y}\right) = \frac{u_y + i\,v_y}{i}.
Step 3 — simplify using 1/i = -i. Multiply
through by -i:
f'(z) = -i\,(u_y + i\,v_y) = v_y - i\,u_y.
Step 4 — set the two expressions equal. For the derivative to exist, the
real-axis answer and the imaginary-axis answer must agree:
u_x + i\,v_x = v_y - i\,u_y.
Step 5 — match real and imaginary parts. Two complex numbers are equal
exactly when their real parts match and their imaginary parts match:
\underbrace{u_x = v_y}_{\text{real parts}}, \qquad \underbrace{v_x = -u_y}_{\text{imaginary parts}}.
These are the Cauchy–Riemann equations. Whenever a complex function is
differentiable, its real and imaginary parts are locked together by them.
Two worked checks
The squaring map passes. For
f(z) = z^2 we found
u = x^2 - y^2 and v = 2xy. Differentiate:
u_x = 2x, \quad v_y = 2x, \qquad u_y = -2y, \quad v_x = 2y.
Indeed u_x = v_y (both 2x) and
v_x = -u_y (since 2y = -(-2y)). Cauchy–Riemann
holds at every point — consistent with f'(z) = 2z.
Conjugation fails. For f(z) = \bar{z} = x - iy we
have u = x and v = -y:
u_x = 1, \qquad v_y = -1.
Since u_x = 1 \ne -1 = v_y, the first Cauchy–Riemann equation is
violated everywhere — confirming, with no limits at all, that
\bar{z} is nowhere differentiable.
For f(z) = u(x, y) + i\,v(x, y) to be complex-differentiable:
-
its parts must satisfy u_x = v_y and
u_y = -v_x (the Cauchy–Riemann equations);
-
the derivative is then f'(z) = u_x + i\,v_x;
-
z^2 satisfies them everywhere
(u_x = v_y = 2x); \bar{z} fails them
everywhere (u_x = 1, v_y = -1).
It is striking how little we assumed. We only required that two of the infinitely many
approach directions — the real and imaginary axes — give the same derivative. That alone
forces the Cauchy–Riemann equations. With the mild extra hypothesis that the partials are
continuous, those same equations turn out to be sufficient: any
(u, v) obeying them, with continuous partials, is genuinely
complex-differentiable. So Cauchy–Riemann is not just a necessary symptom of
differentiability — it very nearly is differentiability.
Check Cauchy–Riemann live
Pick a function with the function control — z^2 or
\bar{z} — and move the point with the x and y
sliders. The diagram computes u_x, v_y,
u_y, and -v_x at that point and flags
whether the two equations u_x = v_y and
u_y = -v_x hold (green) or not (red). The squaring map is always
green; conjugation is always red.