← Part I · The problem
CLEBSCH · TUTORIAL 02 OF 03Fields, charts, and topology

Part II · put the representation to work

Build the field.
Find the seam.

Begin by carrying a known Clebsch triple through a steady flow. Then freeze time and inspect how the three scalar fields assemble the velocity one-form.

one-form, vorticity two-form, and displayed scalar

The torus experiment finally exposes what local vorticity cannot determine: two circulations around the two noncontractible cycles.

First experiment · Clebsch variables as evolving state

Carry the labels.
Let φ change.

PERIODIC TAYLOR–GREEN FLOW · ZERO HARMONIC PERIODSdeforming material-label grid
α contours β contours tracked patch
SHOW
NEXT QUESTION · WHAT DID THIS ZERO-PERIOD TEST LEAVE OUT? Add the two harmonic circulations on the flat torus → Keep the local vortex pattern while changing global particle winding around the x and y cycles.

Second experiment · Chapter 1 picture · dual pairing

Make the covector visible.

Before assembling a velocity one-form, see what a one-form actually does. Fix eta equals s d x, rotate the orange tangent vector, and watch the pairing count signed crossings. The cyan arrow is eta sharp; it appears only after choosing the Euclidean metric.

covector level linesmetric-dual η♯test vector v
THE NUMBER RETURNED BY THE COVECTOR
eta of v
η(v) = 0.688

The vector crosses the level lines in the positive direction.

WITHOUT A METRICThe pairing already exists.

eta at p acting on v only needs a vector and a covector at the same point.

WITH A METRICWe may draw η as an arrow.

The metric defines eta sharp, so in this Euclidean picture eta of v is the metric pairing.

ALONG A FLUID PATHAdd the local readings.

Set v equal to gamma dot and integrate: integral of u flat along gamma is circulation.

How the potentials are chosen

Do not choose three mysterious fields at once.

Start from the physical velocity you want. Split its specification into local vorticity and, when the surface has handles, global loop circulations. The Clebsch fields are a nonunique factorization of that data—not three additional observables.

  1. 01
    Name what is local and what is global.
    vorticity plus independent periods

    The two-form is local circulation density. The periods are additional state only for noncontractible cycles; a genus-g surface has 2g of them.

  2. 02
    Choose two label families whose crossing matches it.
    factor the target vorticity

    This factorization is local and highly nonunique. In a material formulation, these labels can instead be initialized once and transported with the fluid.

  3. 03
    Complete the velocity, including its periods.
    exact, label, and harmonic pieces

    Locally the closed remainder is dφ. Globally, h supplies any periods the chosen labels do not carry. An incompressible projection changes only the exact part.

Third experiment · local field construction

Change one ingredient at a time.

  1. 01

    Cross α and β. Where does vorticity appear?

  2. 02

    Add dφ. What changes while curl stays fixed?

  3. 03

    Project. Which exact component is removed?

LIVE CONTROLLED FIELDperiodic plane · two crossing label foliations
metric-dual arrows selected probe
drag to orbit · click the surface to move the probe
color: α · arrows: (dα)♯
STEP 01 · LABELS
alpha and beta are zero forms

SMOOTH VIEW: evaluate at a point · MESH VIEW: sample at vertices

IN THE PERIODIC-PLANE EXAMPLE
controlled planar labels

The angle θ controls whether the two label families change in the same direction or independently.

Fourth experiment · the global test

Hold vorticity at zero.
Change the circulation.

The construction lab controls local curl. This final experiment deliberately sets that local signal to zero, then asks whether the velocity can still carry something around a noncontractible loop.

THE TOPOLOGICAL OBSTRUCTION, MADE VISIBLE

Follow one circulation all the way around a torus.

The smooth one-form eta h equals c d theta has zero vorticity everywhere yet a nonzero circulation around the torus hole. Locally it looks like the differential of an angle. Globally that angle must jump somewhere. The four views below separate charts, Clebsch labels, the exact term d phi, and harmonic circulation.

LOCAL PATCH TESTNo vorticity.
d eta h is zero

Every small contractible loop has zero circulation because it bounds a patch with zero vorticity.

GLOBAL LOOP TESTOne visible period.
period around the theta cycle

The displayed loop is not a patch boundary, so Stokes does not force this number to vanish.

FULL TORUS STATEThere are two harmonic numbers.
two harmonic basis forms

This embedded-torus demo isolates the θ coefficient. The flat-torus lab exposes both coefficients and both particle windings.

drag to orbit · arrows show the metric dual of d theta
orange ring: the unavoidable branch cut of one angle field
ONE REAL-VALUED POTENTIAL

The angle tears at the orange seam.

one local angle potential

Approaching the seam from opposite sides gives values near 2π and 0. The physical one-form is smooth, but a naïve difference of the scalar samples sees a false jump of 2π.

WHY ONE GLOBAL φ IS IMPOSSIBLE

Exact forms have zero period.

closed-loop contradiction

A single-valued scalar returns to its starting value after a closed loop. The toroidal circulation does not. Therefore c d theta is closed but not globally exact.

HOW THE EXTRA CHART FIXES IT

Move the cuts and glue derivatives.

transition on each overlap component

Chart A fails only at one meridian; chart B puts its cut at another. On overlaps their angle values differ by a constant, so their differentials—and hence the velocity—agree exactly.

A CLEBSCH WORKAROUND

Two periodic labels can hide the cut.

a global Clebsch triple for this example

The pair (α,β) traces an ellipse in label space; it is not a coordinate chart on the torus. Its weighted differential αdβ carries the desired period plus an oscillation, and the exact term dφ cancels that oscillation. The sum is smooth and uniform.

WHAT A SOLVER USUALLY STORES

Keep topology as two state variables.

Hodge decomposition on the torus

Rather than manage multivalued potentials, a mesh solver can store two harmonic basis one-forms and two circulation coefficients. Vorticity determines the coexact part, not c₁ and c₂; pressure projection subtracts an exact form and therefore cannot remove them.

The distinction between closed and exact forms is the de Rham-cohomology obstruction. The torus has two independent degree-one classes, corresponding to its two noncontractible cycles; see the MIT notes on de Rham cohomology and the Hodge Laplacian. The form and metric conventions follow Exterior Calculus in Graphics. The real harmonic coefficients and their role in fluid reconstruction follow Fluid Cohomology. The linked flat-torus lab treats both real coefficients first and keeps quantized subtraction clearly separate as an optional experiment.

Optional Part III

Move the implementation details off the experiment page.

The vector/covector types, Stokes derivation, DEC storage map, gauge freedom, and solver checklist now live on a compact reference page.

OPEN THE REFERENCEForms, DEC, and the solver checklist →