Point Transformations and Their Canonical Lifts

Intuition Beginner

A point transformation is the simplest kind of coordinate change: you pick new position coordinates $Q$ that are functions of the old position coordinates $q$, without mixing in the momenta at all. Switch from Cartesian $(x,y)$ to polar $(r,\theta )$. Rotate the axes. Switch from lab frame to centre-of-mass frame. These are all point transformations.

But in Hamiltonian mechanics, positions and momenta are paired. You cannot change the position coordinates without also changing the momenta -- the new momenta must be "compatible" with the new positions. There is exactly one correct way to do this: the canonical lift. The new momentum $P_i$ is the old momentum $p_j$ times the rate of change of $q^j$ with respect to $Q^i$. This is the chain rule at work, and it guarantees that the lifted transformation preserves the Hamiltonian structure.

The term "lift" comes from differential geometry: the configuration-space map $Q=f(q)$ "lifts" to a map on the full phase space that sits above it. The lift is automatic and unique -- you do not get to choose how the momenta transform. This makes point transformations the easiest canonical transformations to work with: specify the configuration-space change, and the phase-space change follows for free.

Every point transformation is canonical (after lifting), but not every canonical transformation is a point transformation. Canonical transformations can mix positions and momenta (like the exchange $Q=p$, $P=-q$), which no configuration-space coordinate change can do. Point transformations are a proper subclass -- the ones that respect the position/momentum splitting.

Why does this matter? Because most of the coordinate changes you actually use in physics are point transformations. Polar coordinates, spherical coordinates, rotating frames, normal modes -- all are point transformations. The canonical lift tells you exactly how momenta transform in each case, so you never have to guess or re-derive the momentum transformation from scratch.

Visual Beginner

Picture a flat configuration space (the plane, for a particle moving in two dimensions). A point transformation is a smooth reparameterisation of this plane: a new grid superimposed on the old one. The Cartesian grid $(x,y)$ is replaced by the polar grid $(r,\theta )$, or by a rotated Cartesian grid.

Now lift this to phase space. Phase space has four dimensions $(x,y,p_x,p_y)$. The lift replaces $(x,y)$ by the new position coordinates and simultaneously adjusts $(p_x,p_y)$ so that the "area element" $d{p}_x\wedge dx+d{p}_y\wedge dy$ (the symplectic form) is preserved. The momentum transformation is determined entirely by the position transformation.

For the polar case: $P_r=p_x\cos \theta +p_y\sin \theta$ (the radial component of momentum) and $P_{\theta }=r(-p_x\sin \theta +p_y\cos \theta )$ (the angular momentum). The canonical lift produces exactly these expressions from the Jacobian of the polar transformation.

Worked example Beginner

Transform the 2D harmonic oscillator from Cartesian to polar coordinates. The Hamiltonian is $H=\frac{1}{2}(p_x^2+p_y^2)+\frac{1}{2}{\omega }^2(x^2+y^2)$.

The point transformation is $r=\sqrt{x^2+y^2}$, $\theta =\arctan (y/x)$. The inverse is $x=r\cos \theta$, $y=r\sin \theta$.

The canonical lift gives the new momenta. For $p_r$: take $p_x$ times $dx/dr$ plus $p_y$ times $dy/dr$. Since $dx/dr=\cos \theta$ and $dy/dr=\sin \theta$:

$$p_r=p_x\cos \theta +p_y\sin \theta .$$

For $p_{\theta }$: take $p_x$ times $dx/d\theta$ plus $p_y$ times $dy/d\theta$. Since $dx/d\theta =-r\sin \theta$ and $dy/d\theta =r\cos \theta$:

$$p_{\theta }=-p_{x}r\sin \theta +p_{y}r\cos \theta .$$

Invert these relations: $p_x=p_r\cos \theta -(p_{\theta }/r)\sin \theta$ and $p_y=p_r\sin \theta +(p_{\theta }/r)\cos \theta$.

Substitute into $H$. The kinetic energy becomes $\frac{1}{2}(p_r^2+p_{\theta }^2/r^2)$ and the potential becomes $\frac{1}{2}{\omega }^2{r}^2$. So:

$$H=\frac{1}{2}\left(p_r^2+\frac{p_{\theta }^2}{r^2}\right)+\frac{1}{2}{\omega }^2{r}^2.$$

The angle $\theta$ is cyclic ($H$ does not depend on $\theta$), so $p_{\theta }$ is conserved. This is angular momentum conservation, recovered automatically from the canonical lift. The Hamiltonian in polar coordinates reveals the rotational symmetry that was hidden in the Cartesian form.

Check your understanding Beginner

Formal definition Intermediate+

Configuration-space diffeomorphisms

Let $Q$ be the configuration manifold and $\varphi :Q\to Q$ a diffeomorphism. In local coordinates, $\varphi$ is written $Q^i=f^i(q^1,\dots ,q^n)$. The Jacobian matrix of $\varphi$ at $q$ is

$$J_j^i(q)=\frac{\partial f^i}{\partial q^j}{|}_q\in \mathrm{GL}(n,\mathbb{R}).$$

The assumption that $\varphi$ is a diffeomorphism guarantees that $J(q)$ is invertible at every point. Singular coordinate changes (where the Jacobian degenerates) are excluded.

The tangent lift

The derivative (tangent map) $T\varphi :TQ\to TQ$ acts on tangent vectors: if $\dot{q}\in T_{q}Q$, then $T_q\varphi (\dot{q})=J\dot{q}$. In coordinates:

$${\dot{Q}}^i=J_j^i{\dot{q}}^j=\frac{\partial f^i}{\partial q^j}{\dot{q}}^j.$$

This is the familiar transformation law for velocities from Lagrangian mechanics 09.02.01. It expresses the fact that velocities are tangent vectors: they live in the tangent bundle $TQ$ and transform by the tangent map.

The cotangent lift

The cotangent lift $T^{\ast }\varphi :T^{\text{astQ}}\to T^{\text{astQ}}$ is the dual of the tangent lift on fibres. For a covector $p\in T_q^{\ast }Q$ (a momentum), the lifted momentum is defined by

$$\boxed{(T_{\varphi }^{\ast }p)(v)=p\left((T_q\varphi {)}^{-1}v\right)\text{for all }v\in T_{\varphi (q)}Q.}$$

In coordinates, this gives

$$P_i=p_j(J^{-1}{)}_i^j=p_j\frac{\partial q^j}{\partial Q^i}.$$

The cotangent lift is a diffeomorphism of $T^{\text{astQ}}$ that covers $\varphi$ (i.e., $\pi \circ T^{\ast }\varphi =\varphi \circ \pi$ where $\pi :T^{\text{astQ}}\to Q$ is the projection). It maps the fibre $T_q^{\ast }Q$ to the fibre $T_{\varphi (q)}^{\ast }Q$.

Time-dependent point transformations

Allow the transformation to depend explicitly on time: $Q^i=f^i(q,t)$. The canonical lift is the same: $P_i=p_j\partial q^j/\partial Q^i$ (with $t$ held fixed). The new Hamiltonian acquires an extra term:

$$K(Q,P,t)=H(q(Q,t),p(Q,P,t),t)+\frac{\partial F_2}{\partial t}$$

where $F_2(q,P,t)=P_i{f}^i(q,t)$ is the time-dependent generating function. The time derivative $\partial F_2/\partial t=P_i\partial f^i/\partial t$ accounts for the motion of the coordinate system.

Preservation of the canonical 1-form and symplectic form

The Liouville 1-form $\theta =p_{i}d{q}^i$ on $T^{\text{astQ}}$ satisfies $(T^{\ast }\varphi {)}^{\ast }\theta =\theta$ for any diffeomorphism $\varphi$. This is the coordinate-free statement that the cotangent lift preserves the canonical 1-form. Since $\omega =-d\theta$, it follows that $(T^{\ast }\varphi {)}^{\ast }\omega =\omega$: the cotangent lift is a symplectomorphism.

In coordinates, the proof is direct. The lifted transformation is $Q^i=f^i(q)$, $P_i=p_j\partial q^j/\partial Q^i$. The pulled-back 1-form is

$$(T^{\ast }\varphi {)}^{\ast }\theta =P_{i}d{Q}^i=p_j\frac{\partial q^j}{\partial Q^i}d{Q}^i=p_{j}d{q}^j=\theta .$$

The key step is the chain rule: $(\partial q^j/\partial Q^i)d{Q}^i=d{q}^j$. Since $\theta$ is preserved and $\omega =-d\theta$, the symplectic form is also preserved.

The generating function for a point transformation

The type-2 generating function for the cotangent lift of $Q^i=f^i(q)$ is

$$F_2(q,P)=P_i{f}^i(q).$$

Verification: $p_j=\partial F_2/\partial q^j=P_i\partial f^i/\partial q^j=P_i{J}_j^i$, which gives $P=p\cdot J^{-1}$ (i.e., $P_i=p_j(J^{-1}{)}_i^j$). And $Q^i=\partial F_2/\partial P_i=f^i(q)$. This is exactly the cotangent lift.

Point transformations are fibre-preserving

A point transformation maps the fibre $T_q^{\ast }Q$ to the fibre $T_{\varphi (q)}^{\ast }Q$: momenta at the old position go to momenta at the new position, with no mixing between different positions. In the language of bundles, the cotangent lift is a fibre-preserving map that covers the base diffeomorphism. Most canonical transformations are not fibre-preserving (they mix fibres), which is why point transformations are a proper subclass of all canonical transformations.

Verification of the symplectic condition

The symplectic condition for a canonical transformation 09.05.01 requires that the Jacobian $M$ of the full phase-space transformation satisfy $M^{T}JM=J$ where $J=\left(\begin{array}{cc}0& I\\ -I& 0\end{array}\right)$ is the symplectic matrix. For the cotangent lift $Q^i=f^i(q)$, $P_i=p_j\partial q^j/\partial Q^i$, the phase-space Jacobian is

$$M=\left(\begin{array}{cc}{J}_{\varphi }& 0\\ \ast & (J_{\varphi }^{-1}{)}^T\end{array}\right)$$

where $J_{\varphi }=\partial Q/\partial q$ and $\ast$ denotes second-derivative terms involving ${\partial }^2{q}^j/\partial Q^i\partial Q^k$. Computing $M^{T}JM$:

$$M^{T}JM=\left(\begin{array}{cc}{J}_{\varphi }^T& {\ast }^T\\ 0& J_{\varphi }^{-1}\end{array}\right)\left(\begin{array}{cc}0& I\\ -I& 0\end{array}\right)\left(\begin{array}{cc}{J}_{\varphi }& 0\\ \ast & (J_{\varphi }^{-1}{)}^T\end{array}\right).$$

The off-diagonal blocks of the result are $J_{\varphi }^T\cdot (J_{\varphi }^{-1}{)}^T=(J_{\varphi }^{-1}{J}_{\varphi }{)}^T=I$ and $J_{\varphi }^{-1}\cdot J_{\varphi }=I$. The diagonal blocks vanish after simplification. Hence $M^{T}JM=J$, confirming that the cotangent lift satisfies the symplectic condition.

Key derivation Intermediate+

Key derivation: The Hamiltonian transforms by substitution under a point transformation.

Under a point transformation $Q^i=f^i(q)$ with canonical lift $P_i=p_j\partial q^j/\partial Q^i$, the new Hamiltonian is obtained by substituting the old coordinates in terms of the new:

$$K(Q,P)=H(q(Q),p(Q,P)).$$

We verify that Hamilton's equations in the new variables follow from those in the old variables.

From $p_j=P_i\partial Q^i/\partial q^j$ and $Q^i=f^i(q)$, compute:

$${\dot{Q}}^i=\frac{\partial Q^i}{\partial q^j}{\dot{q}}^j=J_j^i\frac{\partial H}{\partial p_j}.$$

Now $\partial K/\partial P_i=(\partial H/\partial p_j)(\partial p_j/\partial P_i)=(\partial H/\partial p_j)J_j^i$, where we used $\partial p_j/\partial P_i=\partial Q^i/\partial q^j=J_j^i$. So ${\dot{Q}}^i=\partial K/\partial P_i$.

For the momentum equation, the generating function $F_2=P_i{f}^i(q)$ has $K=H+\partial F_2/\partial t=H$ (time-independent). Since the generating function guarantees symplecticity (proven in 09.05.03), Hamilton's equations are preserved: ${\dot{Q}}^i=\partial K/\partial P_i$ and ${\dot{P}}_i=-\partial K/\partial Q^i$. The new Hamiltonian is $K(Q,P)=H(q(Q),p(Q,P))$, obtained by substitution. This is the simplest rule for transforming a Hamiltonian: substitute the old variables in terms of the new ones.

Bridge. This derivation builds toward the normal-mode analysis of small oscillations 09.02.04, where a point transformation to normal coordinates decouples the coupled oscillator Hamiltonian into independent harmonic oscillators. The substitution rule appears again in action-angle theory 09.06.01, where the transformation to action-angle coordinates is not a point transformation but still preserves the Hamiltonian structure, requiring the full generating-function machinery of 09.05.03.

Worked examples Intermediate+

Cartesian to spherical coordinates

A particle moves in 3D with Hamiltonian $H=|\mathbf{p}{|}^2/(2m)+V(|\mathbf{q}|)$. The point transformation to spherical coordinates is $x=r\sin \theta \cos \phi$, $y=r\sin \theta \sin \phi$, $z=r\cos \theta$.

The canonical lift gives:

$$p_r=p_x\frac{\partial x}{\partial r}+p_y\frac{\partial y}{\partial r}+p_z\frac{\partial z}{\partial r}=p_x\sin \theta \cos \phi +p_y\sin \theta \sin \phi +p_z\cos \theta .$$ $$p_{\theta }=p_x\frac{\partial x}{\partial \theta }+p_y\frac{\partial y}{\partial \theta }+p_z\frac{\partial z}{\partial \theta }=p_{x}r\cos \theta \cos \phi +p_{y}r\cos \theta \sin \phi -p_{z}r\sin \theta .$$ $$p_{\phi }=p_x\frac{\partial x}{\partial \phi }+p_y\frac{\partial y}{\partial \phi }+p_z\frac{\partial z}{\partial \phi }=-p_{x}r\sin \theta \sin \phi +p_{y}r\sin \theta \cos \phi .$$

Note that $p_r=\mathbf{p}\cdot \hat{\mathbf{r}}$ (radial component of momentum), $p_{\theta }=r\mathbf{p}\cdot \hat{\mathbf{\theta }}$ (angular component scaled by $r$), and $p_{\phi }=r\sin \theta \mathbf{p}\cdot \hat{\mathbf{\phi }}$ (azimuthal component scaled by $r\sin \theta$). These are the canonical momenta, not the physical components -- the metric factors $r$ and $r\sin \theta$ appear because canonical momenta conjugate to angular coordinates carry dimensions of angular momentum.

The kinetic energy in the new coordinates: the spherical metric $d{s}^2=d{r}^2+r^{2}d{\theta }^2+r^2{\sin }^2\theta d{\phi }^2$ has inverse metric $g^{rr}=1$, $g^{\theta \theta }=1/r^2$, $g^{\phi \phi }=1/(r^2{\sin }^2\theta )$. So:

$$\frac{|\mathbf{p}{|}^2}{2m}=\frac{1}{2m}\left(p_r^2+\frac{p_{\theta }^2}{r^2}+\frac{p_{\phi }^2}{r^2{\sin }^2\theta }\right).$$

The Hamiltonian becomes $H=p_r^2/(2m)+L^2/(2m{r}^2)+V(r)$ where $L^2=p_{\theta }^2+p_{\phi }^2/{\sin }^2\theta$ is the squared angular momentum. Both $\theta$ and $\phi$ are cyclic for a central potential, so $p_{\phi }$ and $L^2$ are conserved.

Rotating frame

A frame rotating with angular velocity $\Omega$ about the $z$-axis has coordinates related to the lab frame by $x=X\cos \Omega t-Y\sin \Omega t$, $y=X\sin \Omega t+Y\cos \Omega t$, $z=Z$. This is a time-dependent point transformation.

The canonical lift gives the lab-frame momenta in terms of the rotating-frame momenta:

$$p_x=P_X\cos \Omega t-P_Y\sin \Omega t,p_y=P_X\sin \Omega t+P_Y\cos \Omega t,p_z=P_Z.$$

The generating function is $F_2=P_X(x\cos \Omega t+y\sin \Omega t)+P_Y(-x\sin \Omega t+y\cos \Omega t)+P_{Z}z$. Its time derivative is $\partial F_2/\partial t=\Omega (P_{X}Y-P_{Y}X)$, which equals $\Omega L_z$ where $L_z$ is the angular momentum in the rotating frame.

The new Hamiltonian is $K=H+\Omega L_z$. For a free particle ($H=|\mathbf{p}{|}^2/(2m)$), this becomes $K=|\mathbf{P}{|}^2/(2m)+\Omega (P_{X}Y-P_{Y}X)$, and Hamilton's equations yield $\ddot{X}=2\Omega \dot{Y}+{\Omega }^{2}X$ -- the equation of motion in the rotating frame with Coriolis and centrifugal forces.

Normal-mode diagonalisation

A 2D Hamiltonian is $H=\frac{1}{2}(p_x^2+p_y^2)+\frac{1}{2}{\omega }^2(x^2+y^2)+\lambda xy$. The quadratic potential has matrix $\left(\begin{array}{cc}{\omega }^2& \lambda \\ \lambda & {\omega }^2\end{array}\right)$ with eigenvalues ${\omega }^2\pm \lambda$ and eigenvectors $(1,1)/\sqrt{2}$ and $(1,-1)/\sqrt{2}$.

The diagonalising point transformation is $Q_1=(x+y)/\sqrt{2}$, $Q_2=(x-y)/\sqrt{2}$ (a 45-degree rotation). The lifted momenta are $P_1=(p_x+p_y)/\sqrt{2}$, $P_2=(p_x-p_y)/\sqrt{2}$. The kinetic energy $\frac{1}{2}(p_x^2+p_y^2)=\frac{1}{2}(P_1^2+P_2^2)$ is invariant under the rotation.

The transformed Hamiltonian is $K=\frac{1}{2}(P_1^2+P_2^2)+\frac{1}{2}{\omega }_1^2{Q}_1^2+\frac{1}{2}{\omega }_2^2{Q}_2^2$ with ${\omega }_1^2={\omega }^2+\lambda$ and ${\omega }_2^2={\omega }^2-\lambda$. The system decouples into two independent harmonic oscillators. This is the normal-mode decomposition via a point transformation.

Exercises Intermediate+

Lean formalization Intermediate+

The cotangent lift is a missing piece in Mathlib's differential-geometry layer. The specific ingredients needed are:

1. The cotangent lift $T^{\ast }\varphi :T^{\text{astQ}}\to T^{\text{astQ}}$ of a smooth map $\varphi :Q\to Q$, defined as the dual of the inverse tangent map on fibres.
2. The proof that $(T^{\ast }\varphi {)}^{\ast }\theta =\theta$ (the cotangent lift preserves the Liouville 1-form).
3. The consequence $(T^{\ast }\varphi {)}^{\ast }\omega =\omega$ (the cotangent lift is a symplectomorphism).
4. The generating function $F_2=P_i{f}^i(q)$ as a formal object, with the verification that its partial derivatives produce the cotangent lift.
5. The Jacobian interpretation: the momentum transformation $P_i=p_j\partial q^j/\partial Q^i$ as matrix multiplication by $(J_{\varphi }^{-1}{)}^T$.

Items (1)--(3) require the pullback of differential forms along smooth maps, which Mathlib's geometry layer is approaching but has not yet reached. Item (4) requires generating functions for canonical transformations, which are absent. Item (5) is a computation with Jacobian matrices that is within reach once (1) is established. The formalisation of the cotangent lift is a natural target for a Mathlib contribution.

Advanced results Master

Fibre-preserving transformations and extended point transformations

A transformation $\Phi :T^{\text{astQ}}\to T^{\text{astQ}}$ is fibre-preserving if it maps each fibre $T_q^{\ast }Q$ into some fibre $T_{Q(q)}^{\ast }Q$ -- the new position depends only on the old position, not on the old momentum. Every point transformation is fibre-preserving, but the converse is not quite true: a fibre-preserving map $Q=f(q)$, $P_i=g_i(q,p)$ need not satisfy the canonical condition. The fibre-preserving canonical transformations are exactly the point transformations (with their canonical lifts), together with a class of momentum-dependent gauge transformations that preserve the fibre structure.

An extended point transformation allows the new positions to depend on a parameter: $Q^i=f^i(q,a)$ where $a$ is a real parameter. As $a$ varies, the transformation sweeps out a family of point transformations. The generator of this family is a Hamiltonian vector field on $T^{\text{astQ}}$ whose flow is a family of point transformations. For example, the family of rotations $Q=R(a)q$ (where $R(a)$ is a rotation matrix) is generated by the angular momentum $L=q\times p$.

Point transformations and the Lagrangian

Under a point transformation $Q=f(q)$, the Lagrangian transforms by substitution: $L^{\prime }(Q,\dot{Q})=L(q(Q),\dot{q}(Q,\dot{Q}))$. This is simpler than the Hamiltonian transformation because the Lagrangian depends on velocities (tangent vectors), which transform by the tangent map. The Euler-Lagrange equations are covariant under point transformations -- they have the same form in any coordinate system 09.02.01 -- which is one of the main advantages of the Lagrangian formulation.

The Hamiltonian transformation is more subtle because the canonical lift involves the cotangent map, which depends on the inverse Jacobian. However, the end result is the same: Hamilton's equations are covariant under canonical transformations, and point transformations (after lifting) are canonical. The extra work in the Hamiltonian formulation is the price of the symmetric $(q,p)$ structure that makes the Legendre transform and Poisson brackets available.

The cotangent lift in geometric mechanics

On a general smooth manifold $Q$, there is no canonical coordinate system, and the distinction between point transformations and general canonical transformations is coordinate-independent. A point transformation is a diffeomorphism $\varphi :Q\to Q$, and its cotangent lift $T^{\ast }\varphi :T^{\text{astQ}}\to T^{\text{astQ}}$ is the unique symplectomorphism that covers $\varphi$ and preserves the Liouville 1-form $\theta$. This characterisation is intrinsic: it does not depend on any choice of coordinates.

The group of diffeomorphisms $\mathrm{Diff}(Q)$ embeds into the group of symplectomorphisms $\mathrm{Symp}(T^{\text{astQ}},\omega )$ via the cotangent lift. The image of this embedding is the subgroup of exact symplectomorphisms that preserve the fibration $\pi :T^{\text{astQ}}\to Q$. Not every exact symplectomorphism is a cotangent lift (momentum translations are exact but not fibre-preserving), so the image is a proper subgroup. The quotient $\mathrm{Symp}(T^{\text{astQ}},\omega )/T^{\ast }\mathrm{Diff}(Q)$ measures the "non-point-transformation" part of the symplectomorphism group and is related to the topology of $Q$.

Connection to covariant mechanics and fibre bundles

The cotangent bundle $T^{\text{astQ}}$ is a fibre bundle over $Q$ with fibre ${\mathbb{R}}^n$. The Liouville 1-form $\theta =p_{i}d{q}^i$ is a canonical object on $T^{\text{astQ}}$: it exists independently of any choice of coordinates, and it encodes the pairing between covectors (momenta) and tangent vectors (velocities). The symplectic form $\omega =-d\theta$ is the exterior derivative of $\theta$.

The cotangent lift of $\varphi :Q\to Q$ is characterised by the property that it preserves $\theta$: $(T^{\ast }\varphi {)}^{\ast }\theta =\theta$. This is equivalent to saying that $T^{\ast }\varphi$ preserves the pairing between momenta and velocities: $(T^{\ast }\varphi p)(T\varphi v)=p(v)$ for all $p\in T_q^{\ast }Q$ and $v\in T_{q}Q$. This is the natural duality condition between the tangent and cotangent functors.

In the language of category theory, the cotangent lift makes $T^{\ast }$ a contravariant functor from the category of smooth manifolds and diffeomorphisms to the category of symplectic manifolds and symplectomorphisms. The contravariance (order reversal of composition) reflects the duality between tangent and cotangent spaces: the tangent functor $T$ is covariant, the cotangent functor $T^{\ast }$ is contravariant.

Natural symplectic structure and the tautological 1-form

The existence of a canonical symplectic structure on $T^{\text{astQ}}$ is a deep fact. The Liouville 1-form $\theta$ (also called the tautological 1-form or the canonical 1-form) is defined intrinsically by the condition: for $\alpha \in T_q^{\ast }Q$ and $\xi \in T_{\alpha }(T^{\text{astQ}})$,

$${\theta }_{\alpha }(\xi )=\alpha ({\pi }_{\ast }\xi )$$

where $\pi :T^{\text{astQ}}\to Q$ is the bundle projection and ${\pi }_{\ast }$ is its derivative. In words: the Liouville 1-form at a point $\alpha$ in the cotangent bundle evaluates a tangent vector $\xi$ by first projecting $\xi$ down to a tangent vector on $Q$, then pairing with the covector $\alpha$. This is a "tautological" construction: the covector $\alpha$ is used to evaluate vectors at its own base point.

The symplectic form $\omega =-d\theta$ is exact, meaning it is the exterior derivative of a globally defined 1-form. This is a special property of cotangent bundles: a general symplectic manifold $(M,\omega )$ need not have $\omega =d\alpha$ for any global 1-form $\alpha$. The exactness of $\omega$ on $T^{\text{astQ}}$ is what makes the cotangent lift an exact symplectomorphism.

Synthesis. Point transformations and their canonical lifts provide the simplest bridge between configuration-space geometry and phase-space symplectic geometry. The central insight is that every diffeomorphism of the base lifts uniquely to an exact symplectomorphism of the cotangent bundle, and the lift is determined by the natural duality between tangent and cotangent vectors. The foundational reason this works is the existence of the canonical 1-form $\theta$ on $T^{\text{astQ}}$, whose preservation under the cotangent lift guarantees symplecticity. Putting these together, point transformations are the building blocks from which more general canonical transformations are constructed -- every canonical transformation can be decomposed (locally) into a point transformation followed by a fibre-wise linear transformation, and the generating-function machinery of 09.05.03 provides the global framework for this decomposition.

Connections Master

• Canonical transformations 09.05.01 are the general class of which point transformations are a special (fibre-preserving) case. The generating-function formalism in 09.05.03 provides the general machinery that specialises to $F_2=P_i{f}^i(q)$ for point transformations.
• The Euler-Lagrange equations 09.02.01 are covariant under point transformations, which is one of the foundational motivations for the Lagrangian formulation and the reason point transformations are natural in mechanics.
• Small oscillations and normal modes 09.02.04 are found by a point transformation to normal coordinates that diagonalises the quadratic Hamiltonian. The normal-mode decomposition is the most important physical application of point transformations in the intermediate curriculum.
• Action-angle variables 09.06.01 are a canonical transformation that is not a point transformation. The contrast between point transformations (which preserve the fibre structure) and action-angle transformations (which mix positions and momenta) illustrates the full power of the canonical transformation framework.
• Hamilton's equations 09.04.02 are covariant under the canonical lift of any point transformation: the form of the equations is preserved, and only the functional form of the Hamiltonian changes.

Historical & philosophical context Master

The transformation of momenta under coordinate changes was understood implicitly by Lagrange, who derived the equations of motion in generalised coordinates and noted that the conjugate momenta transform as covariant vectors. Lagrange's Mecanique analytique (1788) treated generalised coordinates as the natural language for mechanics, and the transformation law for momenta followed from the definition $p_i=\partial L/\partial {\dot{q}}^i$ via the chain rule. The conceptual distinction between tangent vectors (velocities, which transform by $J$) and cotangent vectors (momenta, which transform by $(J^{-1}{)}^T$) was not yet explicit in Lagrange's work.

The systematic treatment of canonical transformations as symplectomorphisms is due to Sophus Lie (1890s), who introduced the concept of a contact transformation (the time-dependent version of a canonical transformation) and showed that they form a group. Lie recognised that point transformations are the subgroup of canonical transformations that preserve the fibration $T^{\text{astQ}}\to Q$, and he derived the cotangent lift as the natural dual of the tangent map. Lie's work on transformation groups placed the coordinate-change machinery of Lagrangian mechanics within the framework of group theory.

The cotangent lift was formalised in the language of fibre bundles by Charles Ehresmann (1950s) and developed into the modern theory of symplectic geometry by Arnold, Weinstein, and others in the 1970s. The intrinsic characterisation of the cotangent lift as the unique $\theta$-preserving lift of a base diffeomorphism is due to Weinstein (1971). This characterisation makes the cotangent lift a functor from the category of manifolds and diffeomorphisms to the category of symplectic manifolds and symplectomorphisms.

The distinction between point transformations and general canonical transformations has physical significance. Point transformations preserve the position/momentum splitting and can be interpreted as mere reparameterisations of space. General canonical transformations mix positions and momenta and have no direct interpretation in terms of spatial geometry. The action-angle transformation of an integrable system, which converts positions to angles and momenta to actions, is the paradigmatic example of a canonical transformation that is far from a point transformation. The philosophical point is that the Hamiltonian formulation allows coordinate changes that have no classical (spatial) interpretation but are deeply meaningful dynamically -- a flexibility that is essential for the transition to quantum mechanics, where the canonical commutation relations are the fundamental structure and the position/momentum distinction is a choice of representation.

The connection to quantum mechanics is direct. Under canonical quantisation, the position coordinates $q^i$ become operators ${\hat{q}}^i$ and the momentum coordinates $p_i$ become operators ${\hat{p}}_i$ satisfying $[{\hat{q}}^i,{\hat{p}}_j]=i\hslash {\delta }_j^i$. A point transformation $Q^i=f^i(q)$ lifts to a unitary transformation $\hat{U}$ on the Hilbert space such that $\hat{U}{\hat{q}}^i{\hat{U}}^{-1}=f^i(\hat{q})$ and $\hat{U}{\hat{p}}_i{\hat{U}}^{-1}$ is the quantum version of the canonical lift. The quantum lift is the natural representation-theoretic counterpart of the classical cotangent lift.

Bibliography Master

• Lagrange, J. L., Mecanique analytique (1788). Generalised coordinates and the transformation of conjugate momenta.
• Lie, S., Theorie der Transformationsgruppen (1888--1893). Contact transformations, the group of canonical transformations, and the cotangent lift.
• Ehresmann, C., "Les connexions infinitesimales dans un espace fibre differentiable," Colloque de topologie (1950). Fibre bundles and the functorial framework for tangent and cotangent lifts.
• Weinstein, A., "Symplectic manifolds and their Lagrangian submanifolds," Advances in Mathematics 6 (1971), 329--346. The intrinsic characterisation of the cotangent lift.
• Arnold, V. I., Mathematical Methods of Classical Mechanics, 2nd ed. (Springer GTM 60, 1989), section 45.
• Marsden, J. E. & Ratiu, T. S., Introduction to Mechanics and Symmetry, 2nd ed. (Springer TAM 17, 1999), Ch. 2.3.
• Goldstein, H., Poole, C. P. & Safko, J., Classical Mechanics, 3rd ed. (Pearson, 2002), Ch. 9.2.
• Landau, L. D. & Lifshitz, E. M., Mechanics, 3rd ed. (Course of Theoretical Physics Vol. 1, Pergamon, 1976), section 45.
• Taylor, J. R., Classical Mechanics (University Science Books, 2005), Ch. 13.5.
• Abraham, R. & Marsden, J. E., Foundations of Mechanics, 2nd ed. (Addison-Wesley, 1978), section 3.2.