Applications to Potential and Fluid Flow

Some of the prettiest results in classical field theory rest on a coincidence of dimension. In the plane, and only in the plane, the real and imaginary parts of an analytic function are automatically harmonic and mutually conjugate. That single fact turns three subjects — two-dimensional electrostatics, steady heat conduction, and ideal fluid flow — into one branch of complex analysis, since each is governed in a source-free region by the same $\nabla^2\phi=0$ . Instead of wrestling with an awkward boundary, one seeks a single analytic function, the complex potential $\Omega(z)=\phi+i\psi$ , whose geometry encodes the entire field. Two structural gifts make it work. The Cauchy–Riemann equations force $\phi$ (the potential) and $\psi$ (its conjugate, the stream function) to be harmonic and to have level curves that cross at right angles everywhere — equipotentials against field lines, isotherms against heat-flux lines, velocity potentials against streamlines. And conformal invariance means any analytic map with non-vanishing derivative preserves angles and carries harmonic functions to harmonic functions, so a complicated boundary can be straightened into a half-plane, disk, or strip, solved there, and mapped back. Singularities then carry the physics — a logarithm is a source, $i$ times a logarithm a vortex, an inverse power a dipole — while the Milne–Thomson circle theorem inserts a rigid boundary by reflection and the residue theorem reads off the flux and circulation that, through Kutta–Joukowski, becomes lift.

Key takeaways

In two dimensions any harmonic $\phi$ is the real part of an analytic complex potential $\Omega(z)=\phi+i\psi$ ; the conjugate $\psi$ is the stream function, and their level curves are everywhere orthogonal.
The derivative packages the field: $\Omega'(z)=u-iv$ , so the actual velocity (or electric field) is $\overline{\Omega'}$ , and flux and circulation are contour integrals by residues.
Conformal maps $w=f(z)$ with $f'\neq0$ leave Laplace's equation invariant, so a hard boundary is transformed to a standard one and pulled back — via logarithm, power, Möbius, and Joukowski maps.
Elementary singularities build every flow: source $\tfrac{m}{2\pi}\log z$ , vortex $-\tfrac{i\Gamma}{2\pi}\log z$ , dipole $\propto 1/z$ ; the circle theorem adds a cylinder by an image.
Flow past a cylinder plus the Joukowski map gives the classical aerofoil; the Kutta condition fixes the circulation and Kutta–Joukowski gives the lift $L=-\rho U\Gamma$ .
A strictly two-dimensional miracle: it needs Laplace (not Poisson), incompressible irrotational flow, and the plane — no analytic conjugate or conformal freedom exists in 3D.
Key idea: in the plane, harmonic = real part of analytic. Encode the field in a complex potential $\Omega=\phi+i\psi$ whose derivative $\Omega'=u-iv$ is the field; straighten boundaries with conformal maps and read off flux and circulation with residues.

The core insight

In the plane, harmonic $=$ real part of analytic. Encode the field in a complex potential $\Omega=\phi+i\psi$ whose derivative $\Omega'=u-iv$ is the field; use conformal maps to straighten boundaries and residues to read off flux and circulation.

Where this lives today

The method is no museum piece. In microwave design it still gives the fastest closed-form estimates of the characteristic impedance of coplanar waveguides and microstrip lines, where conformal maps of the cross-section beat a numerical solve for first-pass work. In aerodynamics, thin-aerofoil and panel methods remain the backbone taught before any Navier–Stokes solver, and the Kutta condition is still how codes select the correct circulation. Flettner rotor ships — the spinning cylinders of the worked example — are back in commercial service as fuel-saving wind-assist devices, their thrust set by $L=\rho U\Gamma$ . Hydrogeologists model well fields with superposed source/sink potentials and image singularities, and the wedge exponent $r^{\pi/\beta-1}$ governs high-voltage electrode design, lightning protection, and the corner-rounding that keeps IC interconnects from breaking down.

The physical picture

The complex potential draws two grids at once: the equipotentials $\phi=\text{const}$ (equal voltage, temperature, or velocity potential) and the field lines $\psi=\text{const}$ (the streamlines). Because $\Omega$ is analytic, Cauchy–Riemann forces these families to cross at right angles everywhere away from singularities. The first figure shows this for the archetype: uniform flow past a cylinder, where the streamlines wrap the body and squeeze at the shoulders.

Figure

Streamlines of ideal flow past a circular cylinder with no circulation: the free stream of speed U wraps the body, stagnation points S sit fore and aft where the speed is zero, and the streamlines crowd at the shoulders where the surface speed peaks at 2U (θ = 90°)

Conformal mapping is the act of unbending a region. An analytic map with $f'\neq0$ preserves angles locally, so the orthogonal grid stays orthogonal; lengths and field magnitudes stretch by $|f'(z)|$ . Solve the trivial problem on the straightened boundary, then let the map carry the solution — singularities and all — back. The second figure shows the flagship: the Joukowski transform $w=z+a^2/z$ bends a circle into an aerofoil, and the circulation that spun the cylinder becomes the wing's lift.

Figure

The conformal punchline: the Joukowski map w = z + a²/z carries a circle in the z-plane, with uniform stream U and a bound vortex of circulation Γ, into a cambered aerofoil in the w-plane; Kutta–Joukowski turns the circulation into lift L = ρUΓ per unit span, perpendicular to the stream

Lift, in this picture, is a topological corollary: a wing works because the flow carries nonzero circulation — $\Omega'$ has a nonzero $1/z$ term — and the Kutta condition of smooth flow off the trailing edge fixes that term to its physical value.

The core equations

The method begins with one definition: bundle $\phi$ and its conjugate $\psi$ into a single analytic function of $z=x+iy$ . Analyticity is the whole content — it forces both parts harmonic and their gradients orthogonal.

\Omega(z)=\phi(x,y)+i\,\psi(x,y),\qquad z=x+iy \qquad (1)

where: $\Omega$ — complex potential, analytic in the source-free region; $\phi$ — physical potential (voltage, temperature, velocity potential); $\psi$ — harmonic conjugate, the stream function whose level curves are the field lines; $z$ — complex coordinate of $(x,y)$ . Analyticity is equivalent to the Cauchy–Riemann equations — the engine behind harmonicity, orthogonality, and the field formula.

\frac{\partial\phi}{\partial x}=\frac{\partial\psi}{\partial y},\qquad \frac{\partial\phi}{\partial y}=-\frac{\partial\psi}{\partial x} \qquad (2)

where: $\partial_x,\partial_y$ — partial derivatives. Cross-differentiating gives $\nabla^2\phi=\nabla^2\psi=0$ ; forming $\nabla\phi\cdot\nabla\psi=0$ gives orthogonal grids. The derivative turns the potential into a vector field. Because $\Omega$ is analytic the derivative is direction-independent, so take it along $x$ and let Cauchy–Riemann collapse the components.

\Omega'(z)=\frac{\partial\phi}{\partial x}+i\frac{\partial\psi}{\partial x}=u-iv \qquad (3)

where: $u,v$ — field components (velocity $(u,v)$ , or $(-E_x,-E_y)$ in electrostatics); $\Omega'$ — complex velocity. The actual field is the conjugate $\overline{\Omega'}=u+iv$ ; the $u-iv$ sign is the subject's most common error. Conformal invariance is the second pillar: an analytic map with non-vanishing derivative rescales the Laplacian by a positive factor, so harmonicity survives.

w=f(z)\ \text{analytic},\ f'(z)\neq0\ \Rightarrow\ \nabla_z^2\Phi=|f'(z)|^2\,\nabla_w^2\Phi \qquad (4)

where: $w=f(z)$ — conformal map from the physical $z$ -region to a model $w$ -region; $\Phi$ — any twice-differentiable function; $|f'|^2$ — positive area-stretch factor. So $\Phi$ is harmonic in $z$ iff harmonic in $w$ . Every flow is built from three singular potentials — logarithm (source/sink), $i$ times a logarithm (vortex), inverse power (dipole).

\Omega_{\text{source}}=\frac{m}{2\pi}\log(z-z_0),\quad \Omega_{\text{vortex}}=\frac{-i\Gamma}{2\pi}\log(z-z_0),\quad \Omega_{\text{dipole}}=\frac{\mu\,e^{i\alpha}}{2\pi(z-z_0)} \qquad (5)

where: $m$ — source strength (flux per unit length; $m<0$ a sink); $\Gamma$ — circulation; $\mu$ — dipole strength; $\alpha$ — dipole axis; $z_0$ — singularity location. A source has real log coefficient, a vortex imaginary. The residue theorem then does physical work: one contour integral of $\Omega'$ separates into circulation (real) and flux (imaginary).

\oint_C \Omega'(z)\,\mathrm dz = \Gamma + i\,Q \qquad (6)

where: $\Gamma$ — circulation $=\oint_C(u\,\mathrm dx+v\,\mathrm dy)$ ; $Q$ — volume flux $=\oint_C(u\,\mathrm dy-v\,\mathrm dx)$ ; $C$ — closed contour around the singularities. Expanding $(u-iv)(\mathrm dx+i\,\mathrm dy)$ separates tangential integral (real) from normal flux (imaginary). Two named results close the aerodynamic loop: the Joukowski map bends a cylinder into an aerofoil.

w=z+\frac{a^2}{z},\qquad \Omega(z)=U\!\left(z+\frac{a^2}{z}\right)-\frac{i\Gamma}{2\pi}\log z \qquad (7)

where: $a$ — generating-circle radius (sets chord and thickness); $U$ — free-stream speed; $\Gamma$ — circulation. The doublet $Ua^2/z$ makes the circle a streamline; the log threads circulation through without disturbing it. Finally the payoff: the lift per unit span is fixed entirely by circulation, independent of body shape, with zero drag.

L=-\rho\,U\,\Gamma \qquad (8)

where: $L$ — lift per unit span, perpendicular to the stream; $\rho$ — fluid density; $U$ — free-stream speed; $\Gamma$ — circulation. Magnitude $|L|=\rho U|\Gamma|$ ; the zero drag is d'Alembert's paradox.

Derivation AUniform flow past a circular cylinder with circulation

The goal is to build, by pure superposition, the complex potential for steady inviscid flow of speed $U$ past a cylinder of radius $a$ carrying circulation $\Gamma$ , then verify the circle is a streamline and locate the stagnation points. We demand three things at once: a uniform stream far away, the circle held as a streamline so no fluid crosses it, and a tunable circulation that leaves that condition intact — each met by adding one elementary singular potential, with the boundary checked by watching $\psi$ go constant on the circle.

\Omega(z)=U\!\left(z+\frac{a^2}{z}\right)-\frac{i\Gamma}{2\pi}\log z

Step 1. Start with the free stream $\Omega_\infty(z)=Uz$ , whose derivative $\Omega_\infty'=U$ is a uniform velocity $(U,0)$ . This gets the far field right but knows nothing of the cylinder. Why: superposition starts from the simplest object matching the boundary at infinity, then corrects the interior.

Step 2. Insert the cylinder with the Milne–Thomson circle theorem, adding the reflected term $\overline{\Omega_\infty(a^2/\bar z)}=Ua^2/z$ , a doublet at the origin: $\Omega=Uz+Ua^2/z$ . Why: the image exactly cancels the normal velocity on $|z|=a$ — the 2D analogue of an image charge.

Step 3. Verify the boundary: $z=re^{i\theta}$ gives $\psi=\operatorname{Im}\Omega=U(r-a^2/r)\sin\theta$ , which vanishes at $r=a$ for every $\theta$ . Why: $\psi=0$ all along $r=a$ makes the circle the streamline $\psi=0$ , so no fluid crosses it.

Step 4. Add circulation via a vortex $-\tfrac{i\Gamma}{2\pi}\log z$ , whose $\psi=-\tfrac{\Gamma}{2\pi}\ln r$ is constant on $r=a$ . Why: contributing a constant to $\psi$ on the boundary, the vortex adds circulation without breaking the streamline condition — $\Gamma$ is genuinely free.

Step 5. Locate stagnation points from $\Omega'(z)=U(1-a^2/z^2)-\tfrac{i\Gamma}{2\pi z}=0$ : on $|z|=a$ they sit at $\sin\theta_s=-\Gamma/(4\pi Ua)$ , so two lie on the cylinder only while $|\Gamma|\le 4\pi Ua$ ; at $\Gamma=0$ they return to $z=\pm a$ . Why: the bound marks where the two points merge and leave the surface, beyond which a standing vortex detaches. *Quick checks: $\psi=0$ on $r=a$ confirms the circle is a streamline; setting $\Gamma=0$ recovers symmetric flow with stagnation at $z=\pm a$ ; and the bound $|\Gamma|\le4\pi Ua$ marks where the two stagnation points merge and leave the surface.*

Go deeper · Derivation B · 1 minThe Kutta–Joukowski lift theorem via Blasius

Derivation BThe Kutta–Joukowski lift theorem via Blasius

The goal is to prove that the net force on any closed two-dimensional body in steady irrotational flow is a pure lift $L=-\rho U\Gamma$ , perpendicular to the stream and set entirely by the circulation, with identically zero drag. The tool is Blasius's first theorem, which folds the Bernoulli pressure into a compact contour integral of $(\Omega')^2$ . We expand the complex velocity far from the body, square it, and let the residue theorem extract the one surviving term — the $1/z$ term carrying the circulation.

F_x-iF_y=\frac{i\rho}{2}\oint_C\big(\Omega'(z)\big)^2\,\mathrm dz

Step 1. Expand the far-field velocity as a Laurent series $\Omega'(z)=U+\dfrac{-i\Gamma/2\pi}{z}+\dfrac{c_2}{z^2}+\cdots$ , with no $1/z$ source term since the closed body emits no net flux. Why: for a closed body the only singular structure at large $z$ is the circulation, so the $1/z$ coefficient is $-i\Gamma/2\pi$ .

Step 2. Square it through $1/z$ : $(\Omega')^2=U^2+\dfrac{-iU\Gamma/\pi}{z}+O(z^{-2})$ . Why: only the $1/z$ coefficient has a residue; $U^2$ and the $O(z^{-2})$ tail contribute nothing.

Step 3. Apply the residue theorem on a large enclosing circle: $\oint_C(\Omega')^2\,\mathrm dz=2\pi i\left(-\dfrac{iU\Gamma}{\pi}\right)=2U\Gamma$ . Why: $\Omega'$ is analytic in the fluid, so $C$ can be deformed to infinity and only the far-field residue survives.

Step 4. Assemble the force: $F_x-iF_y=\dfrac{i\rho}{2}(2U\Gamma)=i\rho U\Gamma$ , so $F_x=0$ and $F_y=-\rho U\Gamma$ . Why: the vanishing real part is d'Alembert's paradox (no drag); the imaginary part is the entire lift.

Step 5. Read off $L=F_y=-\rho U\Gamma$ : perpendicular to the stream, proportional to circulation, independent of body shape. Why: shape-independence is why one formula lifts a spinning cylinder, a cambered aerofoil, or a flat plate alike — only $\Gamma$ matters. *Quick checks: $F_x=0$ is d'Alembert's paradox (no drag), $|L|=\rho U|\Gamma|$ is the lift magnitude, and the sign of $F_y=-\rho U\Gamma$ follows directly from the $\Omega'=u-iv$ convention.*

Go deeper · Derivation C · 1 minField enhancement at a conducting wedge by a power map

Derivation CField enhancement at a conducting wedge by a power map

The goal is to find how the electrostatic field behaves near the apex of a grounded conducting wedge of interior angle $\beta$ , exposing why sharp edges concentrate field and blunt corners suppress it. The strategy is pure conformal mapping: a power map straightens the wedge into a half-plane where the solution is trivial, and pulling it back reintroduces a fractional power of $r$ whose exponent is set by the angle. The entire edge-singularity phenomenology — lightning rods, corona discharge, re-entrant corners — falls out of the single exponent $\pi/\beta-1$ .

\Omega(z)=A\,z^{\pi/\beta},\qquad |\mathbf E|=|\Omega'|\propto r^{\,\pi/\beta-1}

Step 1. Map the wedge $0\le\arg z\le\beta$ to the half-plane $0\le\arg w\le\pi$ with $w=z^{\pi/\beta}$ . Why: in the half-plane both conductor faces lie on one straight line, so a solution with both at the same potential is immediate.

Step 2. Take the half-plane solution $\Omega=Aw$ and pull it back: $\Omega(z)=A\,z^{\pi/\beta}$ . Why: composing analytic map with analytic solution stays analytic, so $\phi=\operatorname{Re}\Omega$ is harmonic with the wedge faces as equipotentials.

Step 3. Differentiate: $\Omega'(z)=A\,\tfrac{\pi}{\beta}\,z^{\pi/\beta-1}$ , so $|\mathbf E|=|\Omega'|\propto r^{\pi/\beta-1}$ with $r=|z|$ . Why: $f'$ vanishes or blows up at the apex whenever $\beta\neq\pi$ — the conformal-critical behaviour that concentrates the field.

Step 4. Knife-edge limit $\beta\to2\pi$ : exponent $\tfrac12-1=-\tfrac12$ , so $|\mathbf E|\propto r^{-1/2}$ diverges. Why: this inverse-square-root blow-up is the field enhancement that makes points spark and lightning rods work.

Step 5. Flat-wall limit $\beta=\pi$ : exponent $0$ , so $|\mathbf E|$ is constant and the map is the identity. Why: no corner, no conformal singularity, uniform field — the fractional power is entirely a corner artefact. *Quick checks on the exponent $\pi/\beta-1$ : a right-angled interior corner ( $\beta=\pi/2$ ) gives $r^{+1}$ (field vanishes at the apex), a re-entrant corner ( $\beta=3\pi/2$ ) gives $r^{-1/3}$ (field diverges), and the knife-edge ( $\beta\to2\pi$ ) gives $r^{-1/2}$ .*

Limiting cases

The method is exact only where the governing equation is truly Laplace's and the geometry truly two-dimensional. The table collects the boundaries of validity.

Case	Result	Meaning
Three dimensions	$\nabla^2\phi=0,\ \mathbf x\in\mathbb R^3$	no single-valued harmonic conjugate; conformal maps reduce to similarities and inversions (Liouville) — the method fails
Compressible or rotational flow	$\nabla\cdot\mathbf u\neq0$ or $\nabla\times\mathbf u\neq0$	$\phi,\psi$ no longer both harmonic; viscosity, shocks, and bulk vorticity break analyticity
Charge or sources in the bulk	$\nabla^2\phi=-\rho/\varepsilon_0\neq0$	Laplace becomes Poisson; add a particular solution, and only source-free regions admit holomorphic $\Omega$
Conformal critical point	$f'(z_0)=0$	angles multiply rather than preserve, smooth boundaries can corner, and $\|\Omega'\|$ may vanish or diverge

Common trapComplex potentials work in three dimensions too — just add a z-component

They do not, and this is the deepest trap in the subject. The method is a strictly two-dimensional miracle for two reasons. First, only in the plane does every harmonic function possess a single-valued harmonic conjugate, so only in the plane is a harmonic $\phi$ automatically the real part of an analytic function — there is no "complex potential" in three dimensions. Second, Liouville's theorem says the only angle-preserving maps in three or more dimensions are compositions of translations, rotations, dilations, and inversions, leaving essentially no conformal freedom to straighten a boundary. Everything special to the plane — the conjugate function, residue bookkeeping of flux and circulation, deforming any simply connected region to a disk — is gone. A milder trap is sign convention: whether $\Omega'=u-iv$ or $u+iv$ , and whether increasing $\psi$ means flow left or right, is a choice — fix one and check it against a known source.

Worked examples

1 · Magnus lift on a Flettner rotor (fully worked)

A Flettner rotor ship carries a vertical spinning cylinder of radius $a=1.25\ \mathrm m$ and height $h=12\ \mathrm m$ , spun at $N=200\ \mathrm{rpm}$ in a beam wind $U=15\ \mathrm{m/s}$ in air ( $\rho=1.225\ \mathrm{kg/m^3}$ ). Treating it as ideal 2D flow past a rotating cylinder, estimate the circulation, lift per unit span, total side force, and check the stagnation points still lie on the cylinder.

Step 1. Convert the spin: $\omega=2\pi N/60=\mathbf{20.94\ rad/s}$ , so the surface speed is $v_s=a\omega=1.25\times20.94=\mathbf{26.2\ m/s}$ .

Step 2. A surface moving at $v_s$ supports a bound circulation of surface speed times circumference: $\Gamma=2\pi a\,v_s=2\pi a^2\omega=2\pi(1.25)^2(20.94)=\mathbf{205.6\ m^2/s}$ , the $\Gamma$ entering equation (7).

Step 3. Apply Kutta–Joukowski: $|L|=\rho U\Gamma=1.225\times15\times205.6=\mathbf{3.78\times10^3\ N/m}$ . Over the height the side force is $F=|L|\,h=3778\times12=\mathbf{45.3\ kN}$ — several tonnes of thrust from a spinning drum, the basis of rotor-ship propulsion.

Step 4. Check the stagnation bound $|\Gamma|\le4\pi Ua$ : the bound is $4\pi(15)(1.25)=\mathbf{235.6\ m^2/s}$ , and $205.6<235.6$ , so two stagnation points remain on the cylinder at $\sin\theta_s=-0.873$ , i.e. $\theta_s\approx-60.8^\circ$ and $-119.2^\circ$ . Γ = 205.6 m²/s, L = 3.78 kN/m, side force ≈ 45.3 kN, two surface stagnation points.

2 · Field enhancement at a corner

A grounded conductor shows three profiles: a flat wall, a right-angle interior corner ( $\beta=\pi/2$ ), and a re-entrant right-angle corner ( $\beta=3\pi/2$ , the outside of an L). Compare the field magnitude as $r\to0$ .

Step 1. Use $|\mathbf E|\propto r^{\pi/\beta-1}$ . Flat wall $\beta=\pi$ : exponent $0$ , so $|\mathbf E|=\text{const}$ — no enhancement.

Step 2. Interior right angle $\beta=\pi/2$ : exponent $2-1=+1$ , so $|\mathbf E|\propto r\to0$ — suppressed at the bottom of a notch, because charge avoids concave corners.

Step 3. Re-entrant corner $\beta=3\pi/2$ : exponent $\tfrac23-1=-\tfrac13$ , so $|\mathbf E|\propto r^{-1/3}\to\infty$ — the field diverges at a sharp $90^\circ$ edge, strengthening to $r^{-1/2}$ at a knife-edge. Flat: constant; interior corner: → 0; exterior corner: → ∞ as r^(−1/3).

3 · Capacitance of a coaxial cable

Two long coaxial conductors have inner radius $a$ , outer radius $b$ , vacuum between. Use the line-charge complex potential to find the capacitance per unit length, and evaluate it for $b/a=\mathrm e$ .

Step 1. A line charge $\lambda$ has $\Omega=\dfrac{\lambda}{2\pi\varepsilon_0}\log z$ , so $\phi=\dfrac{\lambda}{2\pi\varepsilon_0}\ln r$ and the conductor-to-conductor difference is $\Delta\phi=\dfrac{\lambda}{2\pi\varepsilon_0}\ln(b/a)$ . Hence $C=\lambda/\Delta\phi=\dfrac{2\pi\varepsilon_0}{\ln(b/a)}$ — the standard coaxial result from the logarithmic potential alone, with no differential equation.

Step 2. For $b/a=\mathrm e$ the logarithm is $1$ , so $C=2\pi\varepsilon_0=\mathbf{55.6\ pF/m}$ . A dielectric $\varepsilon_r$ multiplies this; a polyethylene coax ( $\varepsilon_r\approx2.3$ , $b/a\approx3.6$ ) gives roughly $100\ \mathrm{pF/m}$ , matching datasheets. C = 2πε₀/ln(b/a); for b/a = e, C = 55.6 pF/m.

Physics of the portrait

The complex potential is where nineteenth-century complex analysis meets twentieth-century aerodynamics — universal after Riemann's mapping theorem, and the foundation of flight after Joukowski's lift theorem.

Origins · 1 mincomplex potential and how this was discovered

complex potential

Ω = φ + iψ, the analytic function whose real part is the potential and whose imaginary part maps the field lines · two-dimensional potential theory, conformal mapping, and the aerodynamics of lift

The idea of encoding a plane field in one analytic function grew out of the mathematics of the 1850s. Bernhard Riemann, in his 1851 thesis, proved the mapping theorem that underwrites the whole method — any simply connected proper subdomain of the plane maps conformally to the unit disk. The physical exploitation came from the German school: Helmholtz and Kirchhoff built the theory of vortex motion and free-streamline flows in the 1860s. But potential flow past a smooth body embarrassingly predicted zero drag and zero lift — d'Alembert's 1752 paradox — and no one could say why wings worked. The resolution was circulation. Independently, Martin Kutta (1902) and Nikolai Zhukovsky (Joukowski, 1906) found that a wing's lift equals $\rho U\Gamma$ , with the otherwise-free circulation fixed by smooth flow off the sharp trailing edge — the Kutta condition. Joukowski's transform $w=z+a^2/z$ bent the tractable circle into a cambered aerofoil, and Blasius in 1910 recast the force as a contour integral of $(\Omega')^2$ , making lift a residue calculation. Milne-Thomson later packaged the boundary step as the circle theorem — turning pure complex analysis into the working theory of the wing.

History

1752

— d'Alembert notes that ideal flow predicts zero drag on a closed body (the paradox)

1851

— Riemann proves the conformal mapping theorem for simply connected domains

1902

— Kutta fixes a wing's circulation by the smooth-trailing-edge condition

1906

— Joukowski proves the lift theorem L = ρUΓ and introduces his aerofoil transform

1940

— Milne-Thomson publishes the circle theorem for inserting a rigid cylinder by reflection

Going further

Texts. Stone & Goldbart, Mathematics for Physics, Ch. 17–18 (analytic functions and conformal mapping) · Arfken, Weber & Harris, Mathematical Methods for Physicists, 7th ed., Ch. 11 · Hassani, Mathematical Physics, 2nd ed., Ch. 11 · Byron & Fuller, Mathematics of Classical and Quantum Physics, Ch. 6 · Nakahara, Geometry, Topology and Physics, §1.5 · Batchelor, An Introduction to Fluid Dynamics, Ch. 6 · Acheson, Elementary Fluid Dynamics, Ch. 4–5 · Jackson, Classical Electrodynamics, 3rd ed., §2. Key self-consistency checks for this material: cross-differentiating the Cauchy–Riemann equations yields $\nabla^2\phi=0$ ; $\psi=0$ on $r=a$ confirms the circle is a streamline; evaluating $\oint_C(\Omega')^2\,\mathrm dz$ by residues recovers $L=-\rho U\Gamma$ ; the stagnation bound $|\Gamma|\le4\pi Ua$ locates where surface stagnation points merge; and the Flettner rotor gives $\Gamma=205.6\ \mathrm{m^2/s}$ , $L=3.78\ \mathrm{kN/m}$ , with corner field exponents $0,+1,-\tfrac{1}{3},-\tfrac{1}{2}$ .

Prerequisites. Analytic Functions and the Cauchy–Riemann Equations · Harmonic Functions and Conjugates · Elementary Complex Functions · Complex Integration and Contour Integrals · Zeros, Poles, and Isolated Singularities. Leads to. Cauchy's Integral Formula and Derivatives · Laurent Series · Consequences: Liouville, Maximum Modulus, Morera.

Ask the tutor: Want to watch the cylinder streamlines morph into a lifting aerofoil as circulation is dialed up, or work a conformal map of your own boundary step by step?

Why is the physical velocity the conjugate of Ω′ rather than Ω′ itself?

Because of the minus sign Cauchy–Riemann injects. Differentiating $\Omega=\phi+i\psi$ along $x$ gives $\Omega'=\phi_x+i\psi_x$ , and $\psi_x=-\phi_y=-v$ turns this into $\Omega'=u-iv$ — the conjugate of the velocity $u+iv$ , so the true velocity is $\overline{\Omega'}$ . It matters because the sign controls the direction of circulation and lift: get it backwards and your wing pushes down. Fix the convention once and test it on a known case — a source should give outward flow — before trusting any result.

What exactly is the stream function ψ, and why are its level curves streamlines?

The stream function is the harmonic conjugate of the velocity potential — the imaginary part of $\Omega$ . Its defining property is $u=\psi_y$ , $v=-\psi_x$ , so $\mathbf u\cdot\nabla\psi=\psi_y\psi_x-\psi_x\psi_y=0$ : the velocity is always tangent to curves of constant $\psi$ , which are therefore the streamlines. A bonus: the difference $\Delta\psi$ between two streamlines equals the volume flux between them, which is why a solid boundary must be one streamline — no fluid crosses it.

How does the Milne–Thomson circle theorem actually insert a cylinder?

It is a reflection trick, the fluid cousin of the method of images. Given any flow $\Omega_\infty(z)$ analytic outside a circle $|z|=a$ , the theorem states that $\Omega(z)=\Omega_\infty(z)+\overline{\Omega_\infty(a^2/\bar z)}$ has the circle as a streamline. The added term is the image reflected in the circle; on $|z|=a$ one has $a^2/\bar z=z$ , so the two pieces are conjugates there and their imaginary parts cancel, forcing $\psi=\text{const}$ . For a uniform stream $\Omega_\infty=Uz$ the image is the doublet $Ua^2/z$ — how a point-dipole appears at the cylinder centre.

Where does d'Alembert's paradox come from, and is it really wrong?

The paradox is the ideal-flow prediction that a body moving steadily through an inviscid fluid feels no drag — the real part of the Blasius integral vanishes. It is correct within its assumptions, so not a mistake but a warning that those assumptions omit something. Real fluids have viscosity, which however small creates a boundary layer that can separate, shed a wake, and produce pressure drag that does not vanish as viscosity goes to zero. Lift, by contrast, survives because it depends only on circulation — which is why Kutta–Joukowski is quantitatively useful while the zero-drag result is not.

Why does a sharp edge concentrate the electric field so strongly?

Because the conformal map that straightens the corner has a derivative that blows up there. Near a wedge apex of angle $\beta$ the potential behaves like $z^{\pi/\beta}$ , so $|\Omega'|$ scales as $r^{\pi/\beta-1}$ . When the conductor fills more than half the plane the exponent is positive and the field is suppressed (charge avoids concave notches); at an exterior or re-entrant edge it is negative and the field diverges as $r\to0$ , reaching $r^{-1/2}$ at a knife-edge. This is the mathematics behind lightning rods, corona discharge, and rounding off high-voltage electrode corners.

Can conformal mapping handle multiply connected regions, like flow past two cylinders?

Only with extra care. The Riemann mapping theorem guarantees a map to the disk only for simply connected regions. A region with holes — two bodies, or an annulus — is multiply connected and maps instead to a canonical domain like an annulus or slit plane, where a conformal modulus (a real invariant such as the radius ratio) obstructs further simplification. Physically this is extra freedom: each hole can carry its own circulation, undetermined until fixed by extra physics such as Kutta conditions or Kelvin's theorem. Practical multi-body problems use numerical conformal mapping or superposed image systems.