Kelvin's Circulation Theorem

What the theorem says

Take any closed loop drawn in a moving fluid and let it be a material loop — a loop made of the very same fluid particles at every instant, so it stretches, twists, and drifts along with the flow. Around that loop measure the circulation, the line integral of velocity:

Γ = ∮_C v · dl

Kelvin's theorem says that for an ideal fluid this quantity does not change as the loop is carried by the flow:

DΓ/Dt = 0

where D/Dt is the material (substantial) derivative — the rate of change following the fluid. The circulation you measure now is the circulation you will measure a minute later, no matter how badly the loop has been mangled in between. Three conditions must hold:

• Inviscid. The fluid has no viscosity (μ = 0), so there are no tangential stresses to add or remove circulation.
• Barotropic. Density is a function of pressure alone, ρ = ρ(p), so the pressure force can be written as the gradient of a single-valued function and contributes nothing around a closed loop.
• Conservative body forces. Any external force per unit mass derives from a potential, f = −∇Φ (gravity being the archetype), so it too integrates to zero around a closed loop.

Why it matters

Kelvin's theorem is the backbone of ideal-fluid dynamics. It tells you that vorticity is not created or destroyed inside an ideal fluid — it can only be moved around, stretched, and tilted. That single fact organizes an enormous amount of physics:

• Vortices are permanent. A smoke ring, a tornado, the trailing vortices behind a jetliner — each carries a locked-in circulation that the flow cannot erase. This is why wake turbulence forces air-traffic controllers to space aircraft by several kilometres.
• Lift exists at all. The Kutta–Joukowski theorem ties lift directly to bound circulation, and Kelvin's theorem dictates that this circulation is paid for by a starting vortex shed into the wake.
• Irrotational flow stays irrotational. If a flow starts with zero vorticity everywhere (Γ = 0 around every loop), it stays that way for all time — the justification for potential-flow theory and the whole apparatus of complex-variable aerodynamics.
• It sets a baseline for the real world. Every place the theorem fails — boundary layers, breaking waves, atmospheric fronts — is a place where viscosity or baroclinicity is doing something interesting. The theorem tells you exactly where to look.

How it works — the derivation step by step

Start from the Euler equation for an inviscid fluid with a conservative body force:

Dv/Dt = -(1/ρ)∇p - ∇Φ

Now differentiate the circulation following the material loop. There are two contributions: the change in velocity of each element, and the change in the line element dl itself as the loop deforms.

DΓ/Dt = ∮ (Dv/Dt)·dl + ∮ v·(D(dl)/Dt)

Step 1 — the second integral vanishes. A material line element evolves as D(dl)/Dt = (dl·∇)v, so v·D(dl)/Dt = v·(dl·∇)v = dl·∇(½|v|²). Integrating a perfect differential (the gradient of ½|v|²) around a closed loop gives zero.

Step 2 — substitute the Euler equation. The first integral becomes

∮ (Dv/Dt)·dl = -∮ (1/ρ)∇p·dl - ∮ ∇Φ·dl

Step 3 — the gravity term vanishes. ∮ ∇Φ·dl = 0 because Φ is single-valued; a conservative force does no net work around a closed loop.

Step 4 — the pressure term vanishes if barotropic. When ρ = ρ(p) we can define a pressure function P(p) = ∫ dp/ρ so that (1/ρ)∇p = ∇P. Then ∮ ∇P·dl = 0 as well. Every term on the right is zero, leaving

DΓ/Dt = 0.  ∎

Symbols and units. Γ = circulation (m²/s); v = fluid velocity (m/s); dl = directed line element along the loop (m); ρ = mass density (kg/m³); p = pressure (Pa = N/m²); Φ = body-force potential per unit mass (J/kg = m²/s²); P = ∫dp/ρ = specific pressure function (m²/s²); D/Dt = material derivative (1/s times the quantity); ω = ∇×v = vorticity (1/s); ν = μ/ρ = kinematic viscosity (m²/s).

Circulation, vorticity, and Stokes' theorem

Circulation and vorticity are two views of the same thing. Stokes' theorem converts the loop integral into an area integral over any surface S spanning the loop:

Γ = ∮_C v·dl = ∬_S (∇×v)·dA = ∬_S ω·dA

So circulation is the flux of vorticity through the loop. Kelvin conserving Γ around a material loop is exactly the statement that the vorticity flux through a material surface is constant — the vortex lines are dragged along with the fluid, never crossing fluid particles. This is why the ideal-fluid vorticity equation takes the elegant "frozen-in" form

Dω/Dt = (ω·∇)v

with no source term: the right-hand side only stretches and tilts existing vorticity, it never creates it. Vortex stretching (ω·∇)v is why a spinning column that is pulled thinner spins faster — the same reason a bathtub vortex accelerates as water necks down into the drain, conserving its circulation while shrinking its radius.

Worked example — the starting vortex and airfoil lift

Consider a wing initially at rest in still air. Every loop you can draw has Γ = 0, so a giant material loop enclosing the wing and all the air it will ever touch has Γ_total = 0. Kelvin's theorem says this total must remain zero forever.

Now accelerate the wing to speed U. At the sharp trailing edge the flow cannot whip around an infinitely thin corner at infinite speed, so nature enforces the Kutta condition: the flow leaves the trailing edge smoothly. Achieving that smooth departure requires a nonzero bound circulation Γ_bound around the wing. But the big enclosing loop still has to read zero — so the fluid sheds an equal and opposite starting vortex of circulation −Γ_bound, which rolls off the trailing edge and is left behind in the wake.

Γ_bound + Γ_starting = 0   ⟹   Γ_starting = -Γ_bound

The bound circulation is what produces lift, through the Kutta–Joukowski theorem:

L' = ρ · U · Γ_bound

where L' is lift per unit span (N/m), ρ is air density (≈1.225 kg/m³ at sea level), U is flight speed (m/s), and Γ_bound is the bound circulation (m²/s). A quick estimate: a light aircraft wing carrying L' ≈ 2000 N/m at U = 55 m/s needs Γ_bound = L'/(ρU) ≈ 2000/(1.225·55) ≈ 30 m²/s. Every time such a wing starts to move it flings a ~30 m²/s starting vortex into the air behind it — a vortex you can sometimes see disturbing dust on a runway.

Quantity	Symbol / formula	Meaning
Bound circulation	Γ_bound	Circulation around the wing that produces lift
Starting vortex	Γ_starting = −Γ_bound	Shed into the wake to keep total Γ = 0
Lift per span	L' = ρ·U·Γ_bound	Kutta–Joukowski theorem
Total circulation	Γ_total = 0	Conserved by Kelvin's theorem
Trailing wingtip vortices	±Γ_bound	Close the vortex system in 3D (Helmholtz: tubes can't end)

When the theorem breaks down

Drop any one of the three assumptions and circulation is no longer conserved. Keeping the viscous and baroclinic terms, the full circulation budget reads:

DΓ/Dt = -∮ (1/ρ)∇p·dl + ν ∮ ∇²v·dl

and by Stokes' theorem the pressure line integral converts to the baroclinic area term:

DΓ/Dt = ∬_S (1/ρ²)(∇ρ × ∇p)·dA + ν ∮ ∇²v·dl

• Viscosity (the ν term). Real fluids diffuse vorticity. The kinematic viscosity of air is ν ≈ 1.5 × 10⁻⁵ m²/s and of water ≈ 1.0 × 10⁻⁶ m²/s. This is exactly how boundary layers, skin-friction drag, and the eventual decay of every real vortex arise. The relevant vorticity spreads over a distance √(νt), so at high Reynolds number the flow behaves ideally except in thin layers.
• Baroclinicity (the ∇ρ × ∇p term). When surfaces of constant density are tilted relative to surfaces of constant pressure — as at a coastline where cool sea air meets warm land air — the baroclinic torque manufactures circulation from nothing. Sea breezes, thunderstorm outflows, and large-scale ocean gyres all owe their spin to this term. It is the leading reason the atmosphere is not an ideal barotropic fluid.
• Non-conservative forces. In a conducting fluid the magnetic Lorentz force (1/ρ)(J×B) is generally not a gradient, so it drives circulation — the basis of magnetohydrodynamic dynamos. Body forces that cannot be written as −∇Φ break the theorem directly.

Assumption dropped	Circulation source	Physical example
Inviscid	ν∮∇²v·dl (viscous diffusion)	Boundary layers, drag, vortex decay
Barotropic	∬(1/ρ²)(∇ρ×∇p)·dA (baroclinic torque)	Sea breeze, storm outflow, ocean gyres
Conservative force	∮ f·dl ≠ 0	MHD dynamos (Lorentz force)

History

Hermann von Helmholtz laid the groundwork in 1858 with his vortex theorems, showing that in an ideal fluid vortex lines move with the material, vortex tubes keep constant strength, and such tubes cannot terminate inside the fluid. In 1869 William Thomson (Lord Kelvin) gave the sharp integral statement — the conservation of circulation around a material loop — that made Helmholtz's picture rigorous and quantitative. Kelvin was so taken with permanent, indestructible vortices that he proposed his "vortex atom" theory, imagining atoms as knotted vortex rings in the ether. The physics turned out to be wrong, but the mathematics seeded knot theory and the entire modern field of vortex dynamics. Today Kelvin's theorem is the first result every aerodynamicist, oceanographer, and plasma physicist learns about why spin is so hard to destroy.

Common mistakes

• Applying it to a fixed loop. The theorem is only about material loops that move with the fluid. Circulation around a loop fixed in space changes freely as vortices drift through it.
• Forgetting the barotropic requirement. "Inviscid" is not enough. A stratified atmosphere or ocean is inviscid to good approximation yet routinely generates circulation baroclinically. Barotropic (ρ = ρ(p)) is a separate, often-violated condition.
• Thinking it forbids all vortex decay. It only forbids decay in an ideal fluid. Real vortices decay through viscosity — the very term the theorem sets to zero — but slowly, which is why they look permanent.
• Confusing conserved Γ with conserved ω. The vorticity ω at a point is not constant — vortex stretching changes it. What is conserved is the circulation (vorticity flux) around a material loop, which lets ω grow as the loop shrinks.
• Believing lift needs no wake. By Kelvin's theorem, starting lift is impossible without shedding an equal and opposite starting vortex. "Lift with no circulation change anywhere" violates conservation of total Γ.