Chern–Simons forms and transgression
Derive the exact difference between two characteristic forms without mistaking a local primitive for a global trivialization.
On this page
The question for today
What you’ll understand
- Derive a global relative transgression for two connections.
- Verify the local cubic Chern–Simons formula with wedge signs.
- State the qualifications needed for gauge-invariant secondary integrals.
Chern–Weil theory says that changing a connection does not change its characteristic class. Transgression makes the exact difference explicit. It also produces secondary quantities whose integrals can carry geometry invisible to the primary class. Two related constructions require careful separation: the relative form for two connections on one bundle is global on the base, while the familiar expression built from a single matrix potential is local unless a trivialization is supplied. Neither construction says that every characteristic form is globally exact on its base. We keep smooth connections and the curvature convention .
Two connections give a global difference form
Let denote local expressions for two connections on the same principal bundle, and set . Their difference is an adjoint-valued one-form, because the inhomogeneous gauge terms cancel. The affine path consists of connections with curvatures . For a symmetric adjoint-invariant polynomial of degree , differentiating its curvature evaluation gives an exterior derivative. The Bianchi identity removes the covariant derivatives of the remaining curvature factors, and invariance replaces the final covariant derivative by an ordinary derivative.
Integrating this identity in the path parameter yields the global relative transgression below. Both connections and their difference transform together, so its local expressions glue on the base. The formula compares two characteristic representatives of the same class, rather than supplying a primitive of either representative separately. Taking a nonexistent global zero connection on a nontrivial bundle would invalidate that inference. A zero local potential can be chosen in a trivialization, but its values in other frames include an inhomogeneous term.
A single potential gives a local primitive
For the matrix invariant , a potential on a trivializing chart gives the cubic expression below. Its derivative is the characteristic four-form. Products mean matrix multiplication combined with wedge products, and trace cyclicity has graded signs when forms are moved past one another. In an abelian line group the cubic term vanishes, but the quadratic term need not vanish: a one-form can wedge nontrivially with its exterior derivative.
The corresponding form can be defined globally on the principal total space using the principal connection form, but generally it is not basic and therefore does not descend as a single form on the base. Local primitives on different charts can disagree by gauge terms. This is entirely compatible with a globally nonexact characteristic form. In a globally trivial bundle with a chosen trivialization, one does obtain a global base expression; the choice is then part of the construction, and its behavior under changing that choice must be analyzed.
Secondary integrals and global qualifications
On a closed oriented three-manifold, an integrally normalized Chern–Simons construction can define an invariant modulo integers under suitable compact-group and characteristic-class hypotheses. Large gauges may shift a chosen real-valued representative by a period. For example, the standard compact simply connected theory uses an integral invariant bilinear form, or equivalently the appropriate degree-four integral characteristic data. An arbitrary real scaling of a trace polynomial does not automatically have integer periods, and general groups can require additional integral refinements.
Boundaries introduce further data and boundary terms, so a closed-manifold assertion cannot simply be transferred to a manifold with boundary. Likewise a vanishing primary form in dimension three does not force a Chern–Simons three-form to vanish: its derivative is zero there for dimensional reasons. Relative transgression, local primitives, and differential refinements each retain different information. Keeping their domains and normalizations explicit makes the secondary theory useful without turning a local algebraic identity into an incorrect global exactness claim.
Work it through
An abelian primitive in four dimensions
On oriented with volume , take on a trivial line. Compute , the unnormalized Chern–Simons form, and its derivative.
Since the line algebra is abelian, and . Squaring gives .
The cubic term vanishes, and . In the second term moving dw past dx and dy produces two sign changes.
Differentiating the first term gives minus one volume form. Differentiating the second gives another minus volume form, because moving dz across dx and dy again gives positive sign. Their sum is times the oriented volume, equal to .
The identity dCS₃=F² holds with every sign explicit. The global primitive exists here because the bundle and the chosen matrix potential are globally trivialized on ℝ⁴.
The living notebook
Explore the model
Interactive lab · function
A parameter integral of a derivative
Pause/Reset, set coefficient a to 1 and position x to 1: predict value 1, slope 2, and signed area 1/3. Reset and set both controls to coefficient 2 and position 1: all three readouts double. Parameter changes retain time. This scalar quadratic illustrates integration along a parameter family; it does not evaluate differential forms, wedge products, or a Chern–Simons invariant.
- y at chosen x
- 0.608
- Tangent slope · 2ax
- 1.35
- Signed area · ax³/3
- 0.182
This lab uses only y = ax². Playback traces the curve within the fixed integration interval from 0 to x; it does not estimate numerical error or change the chosen point. Parameter changes pause playback and recompute the model at the current elapsed time. Reset restores both parameters and returns time to zero. Playback starts paused. Adjust either parameter, then choose Play or Step.
Look forIntegrating a variation can recover a difference, while the objects and global hypotheses still matter.
Pause & reason
Check your understanding
Your turn
Try it yourself
On the oriented unit cube in , let with real constant . Compute its unnormalized CS₃ integral. What does its zero exterior derivative imply, and what does it fail to imply?
Take a moment to make a prediction or work it out on paper. Open a hint when you need one.
Give me a hint
For a line connection CS₃=A∧dA. Put the surviving three-form into dx∧dy∧dz order before integrating.
Compare with the solution
We have . Thus , so its cube integral is . Its derivative vanishes because every four-form on a three-dimensional base is zero. This does not force the three-form or its integral to vanish. The cube has boundary, the displayed expression is unnormalized, and no closed-manifold gauge-invariance or integer-period conclusion follows from this calculation.
Keep the idea
Take it with you
- Two connections on one bundle have a global relative transgression.
- A local single-potential primitive need not descend to a global base form.
- Secondary integrals require explicit normalization, gauge, and boundary hypotheses.
Read further
Original sources and deeper explanations to follow at your own pace.
Optional progress, saved in this browser when available. Revisit any lesson at any time.
Lesson not marked complete.