Cotangent bundles and the canonical form
Construct the tautological one-form intrinsically and derive the momentum transformation forced by a change of configuration coordinates.
On this page
The question for today
What you’ll understand
- Define the tautological one-form without relying on a coordinate choice.
- Derive the canonical symplectic form as minus its exterior derivative.
- Compute a cotangent lift and distinguish closed from exact one-form translations.
A configuration manifold records positions, while its cotangent bundle attaches possible momenta to each position. The surprising fact is that this bundle already carries a canonical symplectic structure without choosing a metric, a connection or a kinetic energy. The construction uses only evaluation of a covector on a projected tangent vector. It therefore survives coordinate changes automatically. We call the tautological one-form θ and set ω=−dθ, making its local expression agree with our fixed convention dq∧dp. The minus sign is deliberate rather than a cosmetic notation change.
Evaluate the momentum on a projected velocity
Let π:T*Q→Q be the bundle projection. A point α_q in T*Q is itself a covector at q. For a tangent vector V based at α_q, define θ(V) by applying α_q to dπ(V). This separates two levels of tangent information: V moves through phase space, while dπ(V) records only its configuration-space component. In local configuration coordinates q_j with momentum coefficients p_j, the definition gives θ=Σp_jdq_j. Although the coefficients change between charts, the intrinsic evaluation rule defines one global one-form.
Taking minus its exterior derivative yields Σdq_j∧dp_j, a nondegenerate closed form on every cotangent chart and therefore globally. It is exact even when Q has complicated topology, because the global θ provides a primitive for −ω. This does not conflict with the impossibility of exact symplectic forms on compact boundaryless manifolds: a positive-dimensional cotangent bundle has noncompact vector-space fibers. A metric may identify momenta with velocities for a mechanical Hamiltonian, but that extra identification is not part of the canonical form itself.
A configuration change determines its momentum change
If a configuration diffeomorphism sends q to Q=F(q), the same covector must have equal evaluations in both descriptions. Thus PᵀdQ=pᵀdq, giving P=(DF(q))^{-T}p. The corresponding cotangent lift preserves θ exactly, and hence preserves ω. Transforming position while leaving momentum numerically unchanged generally fails this rule. In particular, nonlinear coordinate changes introduce momentum couplings through the Jacobian. This is why canonical momentum is naturally a covector rather than simply a second copy of the position coordinates.
A different family of transformations translates momenta by a one-form β on Q. Pulling back θ adds π*β, so pulling back ω subtracts π*dβ. Consequently the translation is symplectic exactly when β is closed. If β is exact, its graph comes from a global generating potential; if it is closed but not exact, a symplectic translation can still exist without a single-valued global potential. This distinction keeps local calculus and global topology separate, just as closed contraction forms did for Hamiltonian vector fields.
Canonical geometry does not prescribe a mechanical model
Choosing H(q,p)=one half the squared momentum norm requires an inverse metric on Q. Adding a potential changes trajectories without changing the canonical form. Likewise, external forcing can alter the vector field while leaving the underlying phase space available as a mathematical description. The canonical structure organizes the possible Hamiltonian equations; it does not assert that every physical process on the bundle is conservative. Constraints may require restriction and reduction rather than simply deleting momentum coordinates from the form.
The zero section and individual cotangent fibers have half the phase-space dimension and zero restricted symplectic form. They are basic Lagrangian submanifolds, developed next. Graphs of covector fields offer another bridge: their pullback of θ is the covector field itself, so their restricted ω is minus its exterior derivative. This gives a direct calculation of when such graphs are Lagrangian. The same intrinsic evaluation formula will later construct momentum maps for lifted symmetries, linking configuration-space actions to conserved phase-space quantities.
Work it through
A nonlinear coordinate change mixes momenta
On Q=R², set Q₁=q₁ and Q₂=q₂+q₁². Find the cotangent lift and verify its tautological-form identity.
The Jacobian is , so its inverse transpose sends p to P₁=p₁−2q₁p₂ and P₂=p₂.
Compute P₁dQ₁+P₂dQ₂=(p₁−2q₁p₂)dq₁+p₂(dq₂+2q₁dq₁)=p₁dq₁+p₂dq₂.
Taking minus the exterior derivative proves preservation of the full symplectic form. The map is global because its configuration inverse is q₁=Q₁,q₂=Q₂−Q₁² and the momentum transformation is invertible.
The momentum correction is essential. Keeping P=p would add an unwanted term to the tautological form and would generally fail to preserve ω.
The living notebook
Explore the model
Interactive lab · matrix
Visualize a configuration shear before lifting it
Pause and Reset, then set a=1,b=1. Predict the configuration map (q₁,q₂)↦(q₁+q₂,q₂). Its cotangent lift would map momenta to (p₁,p₂−p₁), the inverse transpose. For a fresh comparison Reset and set a=2,b=0; predict q₁ doubles while its paired momentum halves. The instrument shows only the two-dimensional configuration matrix, not the four-dimensional cotangent lift; the momentum predictions must be checked analytically.
- Area scale · det A
- 1.4
- Singular values · σmax, σmin
- 1.597, 0.876
- Condition number · κ₂(A)
- 1.823
The selected matrix is A = [[a,b],[0,1]]. Playback interpolates geometry between the identity and A; readouts always describe A. Its determinant is a and its singular values come from AᵀA. Positive a makes every supported matrix invertible. This two-dimensional family is not a general eigensolver or a learned network. 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 forCanonical momentum transforms by the inverse transpose, compensating for configuration stretching and shear.
Pause & reason
Check your understanding
Your turn
Try it yourself
On T*S¹ with angular coordinate q modulo 2π, translate momentum by a nonzero constant k. Is the map symplectic? Can the one-form k dq be a global differential dS of a single-valued real function?
Take a moment to make a prediction or work it out on paper. Open a hint when you need one.
Give me a hint
Compute d(k dq), then integrate k dq once around the circle.
Compare with the solution
The form k dq is globally defined and closed, so the momentum translation is symplectic. Its integral around the circle is 2πk, nonzero. The integral of dS around any closed loop is zero for a single-valued real S, so k dq is not exact. Local potentials kq exist on angular charts, but their values do not return consistently after one full turn. Thus a global canonical transformation can exist without this type of global generating potential.
Keep the idea
Take it with you
- The tautological form is intrinsic evaluation of momentum on projected motion.
- Our canonical form is −dθ, giving dq∧dp locally.
- Closed fiber translations are symplectic; exactness is an additional global condition.
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.