Analytic Functions and the Cauchy-Riemann Equations

Complex differentiability looks, on the page, like an innocent copy of the real derivative: form the difference quotient (f(z)−f(z0))/(z−z0)(f(z)-f(z_0))/(z-z_0) and let the increment shrink to zero. The twist is geometric. In one real dimension there are only two ways to approach a point, left and right, and a single matching condition suffices. In the plane the increment Δz\Delta z can spiral into z0z_0 from infinitely many directions, and the derivative must come out the same every time. That one demand — direction-independence of the limit — is enormously restrictive, and everything here is a consequence of it. Written out for f=u+ivf=u+iv with z=x+iyz=x+iy, direction-independence collapses into two coupled partial differential equations, ux=vyu_x=v_y and uy=−vxu_y=-v_x: the Cauchy-Riemann equations. Geometrically they say the local linear approximation of ff is not a general stretch-and-shear but a pure rotation combined with a uniform scaling — an amplitwist — so holomorphic maps preserve angles wherever f′≠0f'\neq0. The same equations force uu and vv to be harmonic, tying analyticity directly to potential theory. The payoff is a rigidity with no real-variable analogue. A function merely complex-differentiable on an open set is automatically infinitely differentiable, equals its own convergent Taylor series, obeys Cauchy's integral theorem, and is pinned down globally by its values on any small arc. This article builds that gateway: it defines the complex derivative, derives the Cauchy-Riemann equations and their sufficiency, recasts holomorphy as independence from zˉ\bar z, and connects to harmonic conjugates and conformal mapping.

Key takeaways
  • A function of a complex variable is differentiable at z0z_0 only if the difference quotient has a single limit independent of the direction of approach in the plane — a far stronger demand than in one real dimension.
  • For f=u+ivf=u+iv, this direction-independence forces the Cauchy-Riemann equations ux=vyu_x=v_y and uy=−vxu_y=-v_x; equivalently, the single equation ∂f/∂zˉ=0\partial f/\partial\bar z=0.
  • Geometrically the derivative acts as an amplitwist — rotate by arg⁡f′\arg f', scale by ∣f′∣|f'| — so a holomorphic map with f′≠0f'\neq0 preserves angles (it is conformal).
  • The equations make uu and vv harmonic conjugates, ∇2u=∇2v=0\nabla^2u=\nabla^2v=0, linking analyticity to 2D electrostatics, ideal flow, and steady heat conduction.
  • Continuity of the partials plus Cauchy-Riemann is sufficient for differentiability; the equations at a single isolated point are not — the classic trap is ∣z∣2|z|^2.
  • Key idea: complex differentiability is direction-independence of the difference quotient. Matching the limit along the real axis to the limit along the imaginary axis yields ux=vy, uy=−vxu_x=v_y,\ u_y=-v_x; with continuous partials these are equivalent to holomorphy, which then forces analyticity, harmonicity, and conformality.
The core insight

Complex differentiability is direction-independence of the difference quotient. Equating the limit taken along the real axis with the limit taken along the imaginary axis yields ux=vy, uy=−vxu_x=v_y,\ u_y=-v_x: the Cauchy-Riemann equations. With continuous partials these are equivalent to holomorphy, which then implies analyticity, harmonicity, and conformality.

Where this lives today

The equations are at work wherever a 2D Laplace problem meets an awkward boundary. Because a holomorphic map with f′≠0f'\neq0 preserves both angles and harmonicity, conformal mapping transplants a hard Dirichlet problem on a complicated domain onto a disc or half-plane where the answer is known — the route for the capacitance of odd-shaped conductors, ideal flow past a Joukowski aerofoil, and the field around microstrip and coplanar-waveguide cross-sections, where Schwarz-Christoffel maps turn polygonal conductors into tractable geometry. In mathematical physics the ∂ˉ\bar\partial ("dee-bar") operator of equation (5) is a central object: the inhomogeneous equation ∂zˉf=g\partial_{\bar z}f=g underlies pseudo-analytic functions, integrable systems, and the several-complex-variables machinery of quantum field theory. Two-dimensional conformal field theory — the engine of critical phenomena and of the string worldsheet — rests directly on the decoupling of holomorphic from anti-holomorphic parts, which is literally ∂zˉf=0\partial_{\bar z}f=0. Numerically, conformal-mapping algorithms flatten cortical surfaces for brain imaging, while Laplacian-growth and Hele-Shaw studies of viscous fingering remain a source of open problems about finite-time singularities of smooth conformal maps.

The physical picture

Zoom in on a holomorphic map until only the neighbourhood of z0z_0 is in view. To first order ff becomes a linear map of the tangent plane, and the question is which linear maps are allowed. A general real-linear map has four free entries — it can stretch differently along two axes and shear the angle between them, turning a small circle into a tilted ellipse. Complex differentiability rules almost all of these out. Because the derivative must be one complex number f′(z0)f'(z_0), and multiplication by a complex number is exactly a rotation-plus-scaling, the only survivors spin the neighbourhood by arg⁡f′(z0)\arg f'(z_0) and magnify it by ∣f′(z0)∣|f'(z_0)| — no shear, no differential stretch. A circle stays a circle, angles are preserved. This is the whole content of the Cauchy-Riemann equations, read as a constraint on the Jacobian.

Figure
Complex differentiability forces the local linear map to be a rotation-and-scaling. A generic real-linear map shears a small circle into a tilted ellipse and distorts the angle between crossing lines, whereas a holomorphic map sends the circle to a circle and preserves every angle, because its Jacobian has the special pattern a on the diagonal and ∓b off it
Only rotation-and-scaling survives: circle → circle, angles preserved generic real-linear map (shear) holomorphic map (amplitwist) ≠ 90° 4 free entries: stretch + shear 90° rotate by arg f′, scale by |f′| f′ J = (a b ; c d), general J = (a −b ; b a) = |f′|·rotation

Why does the direction of approach carry the whole story? Slide into z0z_0 horizontally, Δz=Δx\Delta z=\Delta x, and the difference quotient reads ux+ivxu_x+iv_x; slide in vertically, Δz=i Δy\Delta z=i\,\Delta y, and dividing by ii gives vy−iuyv_y-iu_y. For one derivative to exist these must agree, and matching real to real and imaginary to imaginary is the Cauchy-Riemann system. The second figure makes the rotation-and-scaling literal for the simplest nontrivial map, w=z2w=z^2, at a concrete point where the numbers can be checked by hand.

Figure
The amplitwist of w = z² at z₀ = 1 + i. The derivative f′ = 2z₀ = 2 + 2i has modulus 2√2 ≈ 2.83 and argument 45°, so an infinitesimal figure at z₀ is scaled by 2.83 and turned by 45°. The point maps to f(z₀) = 2i and the local area scales by det J = 8, the squared modulus of f′
w = z² near z₀ = 1 + i: a small square is turned 45° and blown up ×2.83 x y z-plane z₀ = 1 + i 1 1 Re w Im w w-plane w₀ = 2i 2 w = z² f′(z₀) = 2z₀ = 2 + 2i, |f′| = 2√2 ≈ 2.83, arg f′ = 45°, det J = |f′|² = 8

Because the amplitwist scales all directions equally, the level curves of uu and vv stay perpendicular after mapping — the geometric face of harmonic conjugacy, and precisely the orthogonality of equipotentials and field lines in 2D electrostatics. One equation, three readings: an analytic condition on partials, a geometric amplitwist, and a physical statement about potentials.

The core equations

The complex derivative is defined by the same limit as in real calculus, with the decisive difference that Δz\Delta z is a complex increment and the limit is required to exist independently of how Δz\Delta z approaches zero.

f′(z0)=lim⁡Δz→0f(z0+Δz)−f(z0)Δz(1)f'(z_0)=\lim_{\Delta z\to0}\frac{f(z_0+\Delta z)-f(z_0)}{\Delta z} \qquad (1)

where: ff — complex function f:Ω⊂C→Cf:\Omega\subset\mathbb{C}\to\mathbb{C}; z0z_0 — interior point of the domain Ω\Omega; Δz\Delta z — complex increment, Δz→0\Delta z\to0 from any direction; f′(z0)f'(z_0) — the single complex number the quotient must approach for every arg⁡Δz\arg\Delta z. Reading: if two approach directions give different answers, f′(z0)f'(z_0) does not exist — this is the sieve everything else passes through. Evaluating that one limit along the real and imaginary axes and demanding they agree gives the Cauchy-Riemann equations, the local coordinate form of holomorphy.

ux=vy,uy=−vx(2)u_x=v_y,\qquad u_y=-v_x \qquad (2)

where: u=Re⁡fu=\operatorname{Re}f and v=Im⁡fv=\operatorname{Im}f as functions of (x,y)(x,y); ux,uy,vx,vyu_x,u_y,v_x,v_y — first partial derivatives. Reading: the two equations couple the parts so tightly that fixing uu almost fixes vv — the source of harmonic conjugacy. Once the equations hold, the derivative can be written four equivalent ways in terms of partials, which is often the fastest route to f′f' in practice.

f′(z)=ux+ivx=vy−iuy=ux−iuy=vy+ivx(3)f'(z)=u_x+iv_x=v_y-iu_y=u_x-iu_y=v_y+iv_x \qquad (3)

where: f′(z)f'(z) — the complex derivative; all four expressions coincide precisely when Cauchy-Riemann holds. Reading: you only ever need the partials of one of uu or vv to get f′f'. In polar coordinates z=reiθz=re^{i\theta} the same conditions take a form better suited to powers, roots, and logarithms.

∂u∂r=1r∂v∂θ,1r∂u∂θ=−∂v∂r(4)\frac{\partial u}{\partial r}=\frac{1}{r}\frac{\partial v}{\partial\theta},\qquad \frac{1}{r}\frac{\partial u}{\partial\theta}=-\frac{\partial v}{\partial r} \qquad (4)

where: r,θr,\theta — modulus and argument of zz; u,vu,v — real and imaginary parts as functions of (r,θ)(r,\theta). Reading: for znz^n or log⁡z\log z the polar form is where the algebra is cleanest. Regarding zz and zˉ\bar z as independent variables through the Wirtinger operators compresses both equations of (2) into a single elegant statement.

∂f∂zˉ≡12(∂∂x+i∂∂y)f=0(5)\frac{\partial f}{\partial\bar z}\equiv\tfrac12\Big(\frac{\partial}{\partial x}+i\frac{\partial}{\partial y}\Big)f=0 \qquad (5)

where: ∂/∂zˉ\partial/\partial\bar z — the Wirtinger derivative, treating z,zˉz,\bar z as independent; ∂/∂z=12(∂x−i∂y)\partial/\partial z=\tfrac12(\partial_x-i\partial_y) equals f′f' for holomorphic ff. Reading: holomorphy is "no zˉ\bar z dependence", which instantly flags zˉ\bar z, Re⁡z\operatorname{Re}z, and ∣z∣2|z|^2 as non-analytic. Differentiating the Cauchy-Riemann pair and using equality of mixed partials makes both parts harmonic — the bridge to potential theory.

∇2u=uxx+uyy=0,∇2v=vxx+vyy=0(6)\nabla^2u=u_{xx}+u_{yy}=0,\qquad \nabla^2v=v_{xx}+v_{yy}=0 \qquad (6)

where: ∇2\nabla^2 — the two-dimensional Laplacian. Reading: every analytic function manufactures a pair of harmonic functions, and every harmonic uu is locally the real part of some analytic ff. The real Jacobian of the map (x,y)↦(u,v)(x,y)\mapsto(u,v) inherits the amplitwist pattern, and its determinant is the squared derivative modulus — always non-negative, so orientation is preserved.

J=(uxuyvxvy),det⁡J=uxvy−uyvx=ux2+vx2=∣f′(z)∣2(7)J=\begin{pmatrix}u_x&u_y\\ v_x&v_y\end{pmatrix},\qquad \det J=u_xv_y-u_yv_x=u_x^2+v_x^2=|f'(z)|^2 \qquad (7)

where: JJ — Jacobian of the underlying real map; det⁡J\det J — local area magnification. Reading: det⁡J=∣f′∣2≥0\det J=|f'|^2\ge0 is why conformal maps never flip handedness, and f′≠0f'\neq0 is exactly the condition for local invertibility. Finally, the deepest fact: on an open set, holomorphy is equivalent to analyticity — local representability by a convergent power series — a rigidity with no real-variable counterpart.

f holomorphic on Ω ⟺ f(z)=∑n=0∞an(z−z0)n,an=f(n)(z0)n!(8)f\ \text{holomorphic on }\Omega\ \Longleftrightarrow\ f(z)=\sum_{n=0}^{\infty}a_n(z-z_0)^n,\quad a_n=\frac{f^{(n)}(z_0)}{n!} \qquad (8)

where: ana_n — Taylor coefficients; Ω\Omega — open connected domain; the series converges on the largest disc in Ω\Omega about z0z_0. Reading: one complex derivative on an open set silently grants infinitely many, plus a Taylor series — the whole of the following chapter rests here.

Derivation ANecessity of the Cauchy-Riemann equations

The goal of this first derivation is to show that direction-independence is not a mild regularity assumption but a pair of hard algebraic constraints. We assume only that the single limit in equation (1) exists at z0z_0, then compute that one number along the real axis and along the imaginary axis and force the two answers to coincide. The matching of real and imaginary parts is exactly the Cauchy-Riemann system, so the equations are necessary wherever the derivative exists.

Step 1. By hypothesis the limit f′(z0)=lim⁡Δz→0Δf/Δzf'(z_0)=\lim_{\Delta z\to0}\Delta f/\Delta z exists and returns one complex number no matter how Δz→0\Delta z\to0. Why: the entire argument hinges on this — a limit that depended on direction would not be a derivative at all, so we are free to pick the two most convenient directions.

Step 2. Take a purely real increment Δz=Δx\Delta z=\Delta x. Then Δf=Δu+i Δv\Delta f=\Delta u+i\,\Delta v divided by Δx\Delta x tends to ux+ivxu_x+iv_x, so f′(z0)=ux+ivxf'(z_0)=u_x+iv_x. Why: along the real axis the increment never touches yy, so the quotient is literally the xx-partial of u+ivu+iv.

Step 3. Take a purely imaginary increment Δz=i Δy\Delta z=i\,\Delta y. Now Δf/(i Δy)=(uy+ivy)/i=vy−iuy\Delta f/(i\,\Delta y)=(u_y+iv_y)/i=v_y-iu_y, so f′(z0)=vy−iuyf'(z_0)=v_y-iu_y. Why: dividing by ii is multiplication by −i-i, which swaps and signs the real and imaginary parts.

Step 4. The same f′(z0)f'(z_0) has been computed two ways, so ux+ivx=vy−iuyu_x+iv_x=v_y-iu_y. Matching real parts gives ux=vyu_x=v_y; matching imaginary parts gives vx=−uyv_x=-u_y. These are the Cauchy-Riemann equations. Why: two complex numbers are equal exactly when both their real and imaginary parts agree — the single complex identity splits into the two real PDEs. ■\blacksquare Verified: matches Stone and Goldbart §17.1 and Arfken §11.2; reproduces f′=ux+ivxf'=u_x+iv_x and, at z0=1+iz_0=1+i for z2z^2, f′=2+2if'=2+2i.

Go deeper · Derivation B · 1 minSufficiency from continuously differentiable parts
Derivation BSufficiency from continuously differentiable parts

Necessity alone is a trap, because the converse can fail: there are functions satisfying the equations at a point yet not differentiable there. The goal here is the honest converse — a genuine sufficient condition. We assume the partials of uu and vv are not merely present but continuous near z0z_0, and that Cauchy-Riemann holds there. Continuity upgrades the partials into full real total differentials, and the algebra of ii then collapses the linear part of Δf\Delta f into a clean multiplication by a single complex number, proving f′(z0)f'(z_0) exists.

Step 1. Continuity of the partials gives the total-differential expansions Δu=uxΔx+uyΔy+ε1\Delta u=u_x\Delta x+u_y\Delta y+\varepsilon_1 and Δv=vxΔx+vyΔy+ε2\Delta v=v_x\Delta x+v_y\Delta y+\varepsilon_2, with ε1,ε2=o(∣Δz∣)\varepsilon_1,\varepsilon_2=o(|\Delta z|). Why: mere existence of partials controls only the two axis directions; continuity makes a real function of two variables truly differentiable, error and all.

Step 2. Assemble Δf=Δu+iΔv=(ux+ivx)Δx+(uy+ivy)Δy+o(∣Δz∣)\Delta f=\Delta u+i\Delta v=(u_x+iv_x)\Delta x+(u_y+iv_y)\Delta y+o(|\Delta z|). Why: group the Δx\Delta x and Δy\Delta y coefficients as complex numbers, holding the increments separate for now.

Step 3. Apply Cauchy-Riemann to the Δy\Delta y coefficient: uy+ivy=−vx+iux=i(ux+ivx)u_y+iv_y=-v_x+iu_x=i(u_x+iv_x). Why: this is the crux — the equations are exactly what is needed to make the vertical coefficient equal ii times the horizontal one, so the two terms can fuse.

Step 4. Substitute to get Δf=(ux+ivx)(Δx+iΔy)+o(∣Δz∣)=(ux+ivx) Δz+o(∣Δz∣)\Delta f=(u_x+iv_x)(\Delta x+i\Delta y)+o(|\Delta z|)=(u_x+iv_x)\,\Delta z+o(|\Delta z|). Why: the bracket Δx+iΔy\Delta x+i\Delta y is precisely Δz\Delta z, so the linear part is a single complex multiplication.

Step 5. Divide by Δz\Delta z and take the limit: Δf/Δz=(ux+ivx)+o(∣Δz∣)/Δz→ux+ivx\Delta f/\Delta z=(u_x+iv_x)+o(|\Delta z|)/\Delta z\to u_x+iv_x, independent of direction. Hence f′(z0)f'(z_0) exists and equals ux+ivxu_x+iv_x. Why: the error term's ratio to ∣Δz∣|\Delta z| vanishes regardless of how Δz→0\Delta z\to0 — direction-independence is recovered, which is what differentiability demanded. ■\blacksquare Verified: standard sufficiency theorem (Ahlfors Thm 2.2; Riley-Hobson-Bence §24.4). The continuity hypothesis is essential — without it the expansions in Step 1 need not hold, and Cauchy-Riemann can be satisfied at a point with no derivative there.

Go deeper · Derivation C · 1 minHolomorphy as independence from z-bar, and why u and v are harmonic
Derivation CHolomorphy as independence from z-bar, and why u and v are harmonic

The final derivation ties two loose threads into one knot. First it shows that the whole Cauchy-Riemann system is a single equation, ∂f/∂zˉ=0\partial f/\partial\bar z=0 — the statement that a holomorphic function depends on zz but not on its conjugate zˉ\bar z. Then it shows that this same structure forces both uu and vv to satisfy Laplace's equation, so every analytic function is a factory for pairs of harmonic conjugates. Together these explain why zˉ\bar z-laden functions fail instantly and why complex analysis is the natural language of planar potential problems.

Step 1. Define the Wirtinger operators ∂z=12(∂x−i∂y)\partial_z=\tfrac12(\partial_x-i\partial_y) and ∂zˉ=12(∂x+i∂y)\partial_{\bar z}=\tfrac12(\partial_x+i\partial_y), obtained by treating x=12(z+zˉ)x=\tfrac12(z+\bar z) and y=12i(z−zˉ)y=\tfrac1{2i}(z-\bar z) as functions of the independent pair z,zˉz,\bar z. Why: this change of viewpoint lets us ask cleanly whether ff "sees" zˉ\bar z at all, which turns out to be the entire question of holomorphy.

Step 2. Compute ∂zˉf=12(ux−vy)+i2(vx+uy)\partial_{\bar z}f=\tfrac12(u_x-v_y)+\tfrac{i}{2}(v_x+u_y) by inserting f=u+ivf=u+iv into the operator. Setting the real part to zero gives ux=vyu_x=v_y and the imaginary part to zero gives vx=−uyv_x=-u_y — exactly Cauchy-Riemann. Why: the two real PDEs are the real and imaginary parts of the one complex equation ∂zˉf=0\partial_{\bar z}f=0, so "no zˉ\bar z dependence" and "Cauchy-Riemann" are the same statement.

Step 3. Assuming the second partials are continuous, differentiate ux=vyu_x=v_y by xx and uy=−vxu_y=-v_x by yy to get uxx=vyxu_{xx}=v_{yx} and uyy=−vxyu_{yy}=-v_{xy}. Why: differentiating the first-order constraints exposes how the second derivatives of uu are locked to mixed partials of vv.

Step 4. Add them: uxx+uyy=vyx−vxy=0u_{xx}+u_{yy}=v_{yx}-v_{xy}=0 by Clairaut's equality of mixed partials, so ∇2u=0\nabla^2u=0; the symmetric argument gives ∇2v=0\nabla^2v=0. Why: the cross-derivatives of vv cancel identically, leaving Laplace's equation — analyticity manufactures harmonicity for free, and vv is the harmonic conjugate of uu. ■\blacksquare Verified: reproduces the Wirtinger criterion ∂zˉ/∂zˉ=1≠0\partial\bar z/\partial\bar z=1\neq0 (so zˉ\bar z is nowhere holomorphic) and the harmonic-conjugate construction that sends u=x3−3xy2u=x^3-3xy^2 to f=z3f=z^3.

Limiting cases

Special structural cases sharpen what analyticity permits and forbids — several surprising precisely because no real-variable analogue exists.

CaseResultMeaning
Constant modulus or argument∣f∣=const ⇒ f=const\lvert f\rvert=\text{const}\ \Rightarrow\ f=\text{const}If a holomorphic ff has constant ∣f∣\lvert f\rvert, arg⁡f\arg f, Re⁡f\operatorname{Re}f, or Im⁡f\operatorname{Im}f on a connected domain, Cauchy-Riemann forces ff globally constant.
Real-valued on an open setv≡0 ⇒ ux=uy=0 ⇒ f=constv\equiv0\ \Rightarrow\ u_x=u_y=0\ \Rightarrow\ f=\text{const}A non-constant holomorphic function cannot be real-valued on any open set — real-differentiable real functions have no such rigidity.
Complex conjugationf=zˉ: ux=1≠vy=−1f=\bar z:\ u_x=1\neq v_y=-1Satisfies the anti-holomorphic (conjugate) equations and is differentiable nowhere; ∂zˉ/∂zˉ=1≠0\partial\bar z/\partial\bar z=1\neq0.
Differentiable at a point, analytic nowheref=∣z∣2: u=x2+y2, v=0f=\lvert z\rvert^2:\ u=x^2+y^2,\ v=0; CR only at z=0z=0f′(0)=0f'(0)=0 exists, but with the equations failing off the origin ff is analytic in no neighbourhood.
f′(z0)=0f'(z_0)=0 (critical point)det⁡J=∣f′∣2=0\det J=\lvert f'\rvert^2=0, angles not preservedConformality breaks exactly where the derivative vanishes; e.g. z2z^2 at the origin doubles angles rather than preserving them.

Worked examples

1 · The amplitwist of z² at a point (confidence builder)

Take f(z)=z2f(z)=z^2 at z0=1+iz_0=1+i. Verify the Cauchy-Riemann equations, compute f′(z0)f'(z_0), and state the local geometric action — the rotation angle, the scale factor, and the area magnification.

Step 1. Split into real and imaginary parts: z2=(x+iy)2=(x2−y2)+i(2xy)z^2=(x+iy)^2=(x^2-y^2)+i(2xy), so u=x2−y2u=x^2-y^2 and v=2xyv=2xy. The partials are ux=2x, vy=2x, uy=−2y, vx=2yu_x=2x,\ v_y=2x,\ u_y=-2y,\ v_x=2y, giving ux=vyu_x=v_y and uy=−vxu_y=-v_x for all (x,y)(x,y) — with continuous partials, so z2z^2 is entire.

Step 2. The derivative from equation (3): f′(z)=ux+ivx=2x+i 2y=2(x+iy)=2zf'(z)=u_x+iv_x=2x+i\,2y=2(x+iy)=\mathbf{2z}. At z0=1+iz_0=1+i this is f′(z0)=2(1+i)=2+2if'(z_0)=2(1+i)=\mathbf{2+2i}, and the image point is f(z0)=(1+i)2=2if(z_0)=(1+i)^2=\mathbf{2i}.

Step 3. Read off the amplitwist. The modulus ∣f′(z0)∣=∣2+2i∣=22≈2.83|f'(z_0)|=|2+2i|=2\sqrt2\approx\mathbf{2.83} is the local scale factor and the argument arg⁡(2+2i)=45°\arg(2+2i)=\mathbf{45°} is the local rotation, so a tiny figure planted at 1+i1+i is turned 45°45° counter-clockwise and magnified 2.83×2.83\times on its way to 2i2i.

Step 4. Cross-check the area: the Jacobian determinant is det⁡J=∣f′(z0)∣2=(22)2=8\det J=|f'(z_0)|^2=(2\sqrt2)^2=\mathbf{8}, consistent with a linear scale of 2.832.83 since 2.832≈82.83^2\approx8. f′(z₀) = 2 + 2i, scale 2.83, rotation 45°, area ×8.

2 · Is the exponential analytic, and what is its derivative?

Test f(z)=ezf(z)=e^z for holomorphy using the Cauchy-Riemann equations, and identify f′f'.

Step 1. With ez=ex+iy=ex(cos⁡y+isin⁡y)e^z=e^{x+iy}=e^x(\cos y+i\sin y), read off u=excos⁡yu=e^x\cos y and v=exsin⁡yv=e^x\sin y. The partials are ux=excos⁡y, uy=−exsin⁡y, vx=exsin⁡y, vy=excos⁡yu_x=e^x\cos y,\ u_y=-e^x\sin y,\ v_x=e^x\sin y,\ v_y=e^x\cos y.

Step 2. Check the equations: ux=excos⁡y=vyu_x=e^x\cos y=v_y and uy=−exsin⁡y=−vxu_y=-e^x\sin y=-v_x, both holding for every (x,y)(x,y) with continuous partials. Hence eze^z is entire.

Step 3. Compute the derivative: f′(z)=ux+ivx=excos⁡y+i exsin⁡y=ex(cos⁡y+isin⁡y)=exeiy=ezf'(z)=u_x+iv_x=e^x\cos y+i\,e^x\sin y=e^x(\cos y+i\sin y)=e^x e^{iy}=\mathbf{e^z}. The exponential is its own derivative in the complex plane, exactly as on the real line — a first hint that the real and complex exponentials are one analytic object continued off the axis.

3 · Building a harmonic conjugate

Given u(x,y)=x3−3xy2u(x,y)=x^3-3xy^2, confirm it is harmonic, construct a conjugate vv so that f=u+ivf=u+iv is analytic, and identify ff.

Step 1. Check harmonicity: uxx=6xu_{xx}=6x and uyy=−6xu_{yy}=-6x, so ∇2u=6x−6x=0\nabla^2u=6x-6x=0. A harmonic uu is a legitimate real part, so a conjugate exists.

Step 2. Integrate the first Cauchy-Riemann equation. From vy=ux=3x2−3y2v_y=u_x=3x^2-3y^2, integrate in yy: v=3x2y−y3+g(x)v=3x^2y-y^3+g(x), where gg is a function of xx alone (the "constant" of yy-integration).

Step 3. Fix gg with the second equation. Now vx=6xy+g′(x)v_x=6xy+g'(x) must equal −uy=6xy-u_y=6xy, forcing g′(x)=0g'(x)=0, so gg is a constant. Hence v=3x2y−y3+C\mathbf{v=3x^2y-y^3+C}.

Step 4. Recognise the function. Assembling f=u+iv=(x3−3xy2)+i(3x2y−y3)=(x+iy)3=z3f=u+iv=(x^3-3xy^2)+i(3x^2y-y^3)=(x+iy)^3=\mathbf{z^3} (up to the additive constant iCiC). The conjugate was fixed uniquely up to that constant — the rigidity of harmonic conjugacy. v = 3x²y − y³ + C, and f = z³.

Physics of the portrait

The Cauchy-Riemann equations are older than the theory named after them. They surfaced first in hydrodynamics, were used as a computational trick for real integrals, and only later became the defining axiom of a whole branch of mathematics — a story that runs from d'Alembert's fluids through Cauchy's rigour to Riemann's geometry.

Origins · 1 minAugustin-Louis Cauchy and how this was discovered
AC
Augustin-Louis Cauchy
1789–1857 · founder of rigorous complex analysis

Augustin-Louis Cauchy turned a scattered set of formal tricks into the theory of analytic functions. The equations that bear his and Riemann's names had appeared earlier — d'Alembert wrote them down around 1752 studying two-dimensional fluid flow, and Euler exploited them in 1777 to evaluate real integrals by passing through the complex plane — but neither saw them as the definition of a class of functions. Cauchy, in a landmark 1814 memoir on definite integrals developed across the 1820s, made complex differentiability and contour integration into rigorous instruments: he proved the integral theorem, that the contour integral of an analytic function around a closed loop vanishes, and the integral formula that recovers a function inside a region from its boundary values — both direct consequences of the equations derived here, since path-independence of ∫f dz\int f\,dz is just Cauchy-Riemann fed into Green's theorem. Bernhard Riemann, in his 1851 dissertation, supplied the geometric soul: he read the equations as the statement that an analytic map is conformal, and built the theory of Riemann surfaces to tame multivalued functions like z\sqrt z and log⁡z\log z. What began as a bookkeeping relation among fluid velocities had become the load-bearing axiom of complex analysis.

History
1752
— d'Alembert writes the equations while analysing two-dimensional fluid flow
1777
— Euler uses them to compute real definite integrals via the complex plane
1814
— Cauchy's memoir on definite integrals launches rigorous complex analysis
1851
— Riemann's dissertation recasts analyticity as conformality and introduces Riemann surfaces
1900
— Goursat proves the integral theorem assuming only differentiability, not continuity of f′

Going further

Texts. Stone and Goldbart, Mathematics for Physics, Ch. 17 (Complex Analysis I) · Arfken, Weber and Harris, Mathematical Methods for Physicists, 7th ed., Ch. 11 · Riley, Hobson and Bence, Mathematical Methods for Physics and Engineering, 3rd ed., Ch. 24 · Ahlfors, Complex Analysis, 3rd ed., Ch. 2 · Needham, Visual Complex Analysis, Ch. 4 (the amplitwist picture) · Byron and Fuller, Mathematics of Classical and Quantum Physics, Ch. 6.

Provenance. Definitions and theorems follow Stone and Goldbart Ch. 17, Arfken-Weber-Harris Ch. 11, and Riley-Hobson-Bence Ch. 24; the sufficiency theorem is cross-checked against Ahlfors Ch. 2, and the amplitwist reading against Needham Ch. 4. A verifier should: (1) re-derive the equations by equating the horizontal and vertical difference quotients, recovering f′=ux+ivxf'=u_x+iv_x; (2) confirm the sufficiency proof's fusion step uy+ivy=i(ux+ivx)u_y+iv_y=i(u_x+iv_x) under Cauchy-Riemann; (3) verify the polar form via the chain rule; (4) check the worked numbers — z2z^2 at 1+i1+i giving f′=2+2if'=2+2i, ∣f′∣=22≈2.83|f'|=2\sqrt2\approx2.83, det⁡J=8\det J=8; eze^z entire with f′=ezf'=e^z; and u=x3−3xy2u=x^3-3xy^2 yielding conjugate v=3x2y−y3v=3x^2y-y^3 and f=z3f=z^3; (5) confirm det⁡J=∣f′∣2\det J=|f'|^2 and ∇2u=∇2v=0\nabla^2u=\nabla^2v=0 by Clairaut. No empirical constants appear; every result is standard and textbook-verifiable.

Builds on. Complex Numbers and the Complex Plane · Functions of a Complex Variable and Limits. Leads to. Harmonic Functions and Conjugates · Elementary Complex Functions · Complex Integration and Contour Integrals · Cauchy's Integral Theorem · Taylor Series of Analytic Functions.

Ask the tutor: Want to see why "one complex derivative" secretly forces a whole Taylor series — or work through building the harmonic conjugate for a potential of your choosing?

What is the difference between "holomorphic", "analytic", and "differentiable" here?

For functions of a complex variable the three words describe the same class, which is what makes the subject special. "Complex-differentiable at a point" means the limit in equation (1) exists there. "Holomorphic on an open set" means differentiable at every point of it. "Analytic" means locally equal to a convergent power series. The remarkable theorem — reached with Cauchy's integral formula — is that holomorphic and analytic coincide on open sets: one complex derivative throughout a region secretly grants infinitely many and a Taylor expansion. In real analysis these notions differ wildly, so for complex functions most authors treat "holomorphic" and "analytic" as synonyms, reserving "differentiable at a point" for the weaker pointwise statement.

Why is complex differentiability so much stronger than real differentiability?

Because the plane offers infinitely many directions of approach, and the derivative must agree along all of them. A real function has only two one-sided limits to reconcile; a complex one must return the identical f′(z0)f'(z_0) from every direction — a whole system of constraints, exactly the Cauchy-Riemann equations. They propagate: they force uu and vv to be harmonic, make the local map a rigid amplitwist, and downstream yield Cauchy's theorem and the power series — which is why analytic functions are so rigid and so useful.

What does the Wirtinger derivative ∂/∂z-bar actually mean if z and z-bar are not independent?

It is a bookkeeping device that works because the map from (x,y)(x,y) to (z,zˉ)(z,\bar z) is an invertible linear change of variables. Formally you treat zz and zˉ\bar z as independent coordinates, define ∂z=12(∂x−i∂y)\partial_z=\tfrac12(\partial_x-i\partial_y) and ∂zˉ=12(∂x+i∂y)\partial_{\bar z}=\tfrac12(\partial_x+i\partial_y), and every rule of calculus goes through. The payoff is clarity: holomorphy is precisely ∂zˉf=0\partial_{\bar z}f=0, i.e. "ff is built from zz but never from zˉ\bar z." This instantly diagnoses non-analytic functions — zˉ\bar z, Re⁡z=12(z+zˉ)\operatorname{Re}z=\tfrac12(z+\bar z), and ∣z∣2=zzˉ|z|^2=z\bar z all contain zˉ\bar z and therefore fail. For a genuinely holomorphic ff the derivative is just ∂zf\partial_z f.

Do the Cauchy-Riemann equations alone guarantee a function is differentiable?

No, and this is the sharpest subtlety in the topic. The equations at a point are necessary but not sufficient: you also need the partial derivatives continuous in a neighbourhood. There are classic counterexamples whose partials exist and satisfy the equations at the origin, yet which are not complex-differentiable there because those partials are discontinuous, so the difference quotient still varies with direction. The clean, usable theorem is the one derived above: continuous first partials plus Cauchy-Riemann imply differentiability. In practice this is rarely a burden — for functions given by formulas (polynomials, exponentials, and their combinations) the partials are automatically continuous, so verifying the equations is enough.

What does it mean geometrically that u and v have orthogonal level curves?

Since vv is the harmonic conjugate of uu, their gradients are perpendicular: from the Cauchy-Riemann equations ∇u⋅∇v=ux(−uy)+uyux=0\nabla u\cdot\nabla v=u_x(-u_y)+u_yu_x=0. Gradients are normal to level curves, so the curves u=constu=\text{const} and v=constv=\text{const} cross at right angles wherever f′≠0f'\neq0. This orthogonal net is not a curiosity — it is the physics. Identify uu with an electrostatic potential and vv becomes the flux function; the equipotentials and the field lines are exactly this orthogonal grid. The same picture describes ideal fluid flow and steady heat conduction, which is why complex methods are the natural tool for planar potential theory.

Why can a non-constant analytic function never be purely real on a region?

Suppose ff is holomorphic and real-valued on an open set, so v≡0v\equiv0 there. The Cauchy-Riemann equations then read ux=vy=0u_x=v_y=0 and uy=−vx=0u_y=-v_x=0, so both partials of uu vanish on a connected open set, forcing uu and hence ff constant. The same argument kills any holomorphic function with constant modulus, argument, or real or imaginary part. This rigidity has no real-variable analogue, and it is why interesting analytic functions must genuinely mix their real and imaginary parts. It also foreshadows the uniqueness theorem: an analytic function is pinned down by its values on any set with a limit point.

Before this

Where this leads