Hamiltonian mechanics

Hamiltonian mechanics describes phase-space evolution through a Hamiltonian and the symplectic structure.

A quantity conserved to leading order when a parameter varies slowly compared with the natural motion period. The action variable of a regular periodic Hamiltonian orbit is a standard example, as explained by adiabatic invariance of the action.

For slow variation of the field compared with the gyration period, the magnetic moment $m{v}_{\perp }^2/(2B)$ is approximately conserved. Thus transverse random energy scales with $B$, and remains unchanged only when the relevant field strength is unchanged.

The magnetic-moment invariant of rapid cyclotron motion in slowly varying fields. Here the per-unit-mass convention is used: the physical magnetic moment is $m\mu$. In drift kinetics, using $(\mu ,v_{\parallel })$ as coordinates removes the leading perpendicular acceleration and leaves a velocity-space measure proportional to the exact magnetic field strength.

In a slowly changing spherical gravitational potential, angular momentum is exactly conserved and the radial action is an adiabatic invariant. An initially circular orbit remains circular in the adiabatic limit, so $J_r=0$ and $L^2=GrM(r)$ give the displayed shell relation. If dark-matter shells do not cross, their enclosed dark mass is also conserved. A density profile alone does not specify this orbital model: eccentric-orbit populations require their full action distribution.

An inner NFW profile has $M_i(r_i)=A{r}_i^2$, where $A=2\pi {\rho }_s{r}_s$. If the final total circular speed is $V$, circular-shell adiabatic contraction gives $r_i=(V^2/(GA){)}^{1/3}{r}^{2/3}$ and $M_D^{\prime }=A^{1/3}(V^2/G{)}^{2/3}{r}^{4/3}$. Thus ${\rho }_D^{\prime }=A^{1/3}(V^2/G{)}^{2/3}{r}^{-5/3}/(3\pi )$. The total density producing the flat curve is $V^2/(4\pi G{r}^2)$, so the dark fraction tends to zero as $r^{1/3}$ despite its increase relative to the original cusp.

An elastic collision with a wall moving toward the particle at speed $V$ changes the particle speed by approximately $2V$ when recoil is negligible. The collision rate is approximately $v/\ell$ for $v\gg V$, and $\dot{\ell }=-2V$, yielding $\dot{v}/v=-\dot{\ell }/\ell$. For a freely moving finite wall of mass $M$, the neglected recoil changes the increment by order $(m/M)v$, so relative accuracy also requires $(m/M)(v/V)\ll 1$.

// Target: classical-mechanics.bigb

A dynamical system on a symplectic manifold whose vector field is generated by a Hamiltonian. In canonical variables it obeys Hamilton's equations.

The Hénon-Heiles system is a Hamiltonian system with two degrees of freedom and Hamiltonian functionIts Hamilton's equations give $\ddot{x}=-x-2xy$ and $\ddot{y}=-y-x^2+y^2$. Its potential energy combines an isotropic quadratic minimum with a cubic interaction.

The potential energy $V(x,y)=\frac{1}{2}(x^2+y^2)+x^{2}y-y^3/3$ in the Hénon-Heiles system has the following expression in polar coordinates:so it is invariant under rotation through $2\pi /3$. Solving $\nabla V=0$ gives a strict local minimum at the origin and three saddle points, at $(0,1)$ and $(\pm \sqrt{3}/2,-1/2)$. All three saddle points have value $1/6$. Their common level set factors exactly:Thus this level set is the union of the three lines $y=-1/2$ and $y=1\pm \sqrt{3}x$. They bound an equilateral triangle containing the local minimum. The Hessian matrix is the identity at the origin; at each of the other three stationary points it has eigenvalues $3$ and $-1$. For energies $0\le E<1/6$, the central accessible region is bounded; above that value the three saddle passages allow paths from the centre to arbitrarily large radius. This follows directly from the radial potential: in each direction with $\sin (3\theta )<0$, its radial maximum occurs at $r=-1/\sin (3\theta )$, with value $1/[6{\sin }^2(3\theta )]\ge 1/6$.

The autonomous planar system $\dot{x}=y-2y^3$, $\dot{y}=-x$ preserves $K$. Its origin is a center equilibrium and $(0,\pm 1/\sqrt{2})$ are saddle equilibria. The level $K=1/4$ is $x=\pm (y^2-1/2)$: its two inner arcs are heteroclinic orbits enclosing periodic motion, while its outer branches connect saddles to escaping trajectories. Since $|\dot{y}|\sim y^2$ on escaping levels, infinity is reached in finite time. Adding $\alpha x/2$ to $\dot{x}$ gives $\dot{K}=\alpha x^2$, so positive $\alpha$ produces outward crossing of central levels and breaks the equal-energy heteroclinic connections.

A one-dimensional particle with $H=p^2/2+g/x^2$ admits conserved time-translation, dilation and special-conformal charges. Their Poisson brackets form the sl2R Lie algebra. The homogeneity of the inverse-square potential is what allows the symmetry.

For the extended canonical variables with $\{t,E\}=-1$ and $\{x,p\}=1$, the charge $K=t^{2}E-txp+x^2/2$ generates a special conformal transformation. Its time derivative is $2te(E-H)$, so it is conserved on the constraint surface. With a time-dependent parameter $\beta$, the multiplier varies as $\delta e=-2te\beta$ and the action variation is $\dot{\beta }K-d(\beta x^2/2)/d\lambda$.

A principal function $S(q,t)$ obeys the Hamilton-Jacobi equation ${\partial }_{t}S+H(q,{\partial }_{q}S,t)=0$. A complete solution generates canonical momentum components $p_j={\partial }_{j}S$ and local canonical coordinates constant along motion. For a time-independent Hamiltonian, $S=W(q)-Et$ reduces the equation to $H(q,{\partial }_{q}W)=E$. Additive separation of $W$ yields conserved separation constants, subject to local regularity and independence.

A complete integral depends on as many independent parameters as configuration coordinates and supplies a locally invertible canonical transformation to constants of motion. For a type-one generating function $G(q,Q,t)$, require a nondegenerate mixed-derivative matrix in the relevant local chart. Solving $G_t+H(q,G_q,t)=0$ makes the transformed Hamiltonian zero, so both new canonical coordinates and new canonical momenta are constant. An arbitrary parameter-deficient solution does not ensure such a transformation.

A phase-space action treats canonical variables as independent fields. Varying them gives Hamilton's equations; Lagrange multipliers can impose constraints.

The worldline einbein imposes the relativistic mass-shell condition. With the Minkowski metric of signature $(-,+,\dots ,+)$, its first-class constraint is $\phi =(p^2+M^2)/2$.

A pseudoclassical spinning particle supplements ordinary position and momentum by odd Grassmann variables. An even multiplier imposes $p^2=0$ and an odd multiplier imposes $p\cdot \zeta =0$. The odd symplectic term gives $\{{\zeta }^m,{\zeta }^n{\}}_D=-i{\eta }^{mn}$. Quantization represents these variables by a Clifford algebra, producing a spinor wavefunction.

Quantizing the odd Dirac brackets yields $\{{\hat{\zeta }}^m,{\hat{\zeta }}^n{\}}_{+}={\eta }^{mn}$. The constraint $p\cdot \zeta =0$ is then the massless Dirac equation on a spinor. Its square enforces $p^2=0$, matching the even constraint. The spin bilinear is represented by $-i[{\gamma }^m,{\gamma }^n]/4$ in this action convention.

The odd particle variables form an even antisymmetric tensor through their bilinear. Adding it to orbital angular momentum gives a conserved Lorentz Noether charge. The particle and odd-variable equations make the two time derivatives cancel. After spinning-particle Dirac quantization, the bilinear becomes the spinor Lorentz generator rather than an ordinary commuting classical spin vector.

With $d{s}^2=-2d{x}^{+}d{x}^{-}+d{x}_i^2$, $x^{+}=t$, and $p_{-}\ne 0$, the mass-shell condition gives $H=-p_{+}=-(p_i^2+{\mu }^2)/(2p_{-})$. In the future-directed sector $p^{+}=-p_{-}>0$, this is $(p_i^2+{\mu }^2)/(2p^{+})$. Its Schrödinger equation is equivalent to the Klein-Gordon equation after multiplication by $2p_{-}$.

An einbein specifies a one-dimensional metric tensor along a worldline. In the relativistic particle phase-space action it is the Lagrange multiplier for the mass-shell condition and transforms as $\delta e=\dot{ϵ}$ under the corresponding canonical gauge transformation. After eliminating the momentum, its geometric interpretation depends on the normalization of the particle action.

For the relativistic particle action $S=(m/2)\int [g_{\mu \nu }{\dot{x}}^{\mu }{\dot{x}}^{\nu }/e-e]d\lambda$, varying the worldline einbein gives $g_{\mu \nu }{\dot{x}}^{\mu }{\dot{x}}^{\nu }=-e^2$. On the positive-einbein branch the proper time satisfies $ds=ed\lambda$. A reparametrization with ${\lambda }^{\prime }=s$ sets the transformed einbein to one, making the geodesic equation affinely parametrized. This gauge choice is made after retaining the einbein mass shell constraint; fixing the action first would discard that equation. If the parameter interval is held fixed at both ends, its integrated proper length is a modulus and cannot be arbitrarily changed by endpoint-preserving transformations.

The constant mode of the worldline einbein is not removed by gauge transformations whose parameters vanish at the ends of a fixed interval. This Schwinger proper-time parameter remains integrated over after gauge fixing. For $\phi =(p^2+M^2)/2$, the geometric on-shell proper length is $M\int edt$.

On a unit interval with gauge parameters vanishing at its ends, $e=s+\dot{ϵ}$ separates the worldline einbein into its invariant average $s$ and a canonical gauge transformation. The Jacobian for nonconstant modes is $\det {\partial }_t$ between the appropriate boundary-condition spaces. On a circle its prime excludes the constant gauge-parameter zero mode. This Faddeev-Popov determinant can be represented by anticommuting $b,c$ fields with an action proportional to $\int b\dot{c}dt$.

Phase space coordinates are canonical when their Poisson brackets have the form $\{q^I,p_J\}={\delta }_J^I$, with the position-position and momentum-momentum brackets zero.

A canonical pair consists of a coordinate and its canonical momentum, with Poisson bracket $\{q,p\}=1$.

The leapfrog integrator alternates half-steps for momentum with a full step for position and a final momentum half-step. It is reversible, symplectic, and second-order accurate.

Hamiltonian flow is the phase-space evolution generated by Hamilton's equations for a fixed Hamiltonian.

For $H_{\omega }=(|v{|}^2+{\omega }^2|x{|}^2)/2$, the displayed linear Hamiltonian flow solves $\dot{X}=V$, $\dot{V}=-{\omega }^{2}X$. Its inverse is $S_{-t}$, its determinant is one and it preserves $H_{\omega }$. The zero-frequency limit is the shear $(x,v)\mapsto (x+tv,v)$.

For homogeneous Hamiltonian Liouville equation transport with autonomous $H$, compact initial support remains in the fixed energy sublevel determined by its maximum initial energy. If these sublevels are compact, this proves uniform compact support without solving the flow. The harmonic energy $H_{\omega }=(|v{|}^2+{\omega }^2|x{|}^2)/2$ has this property exactly when $\omega \ne 0$; at zero frequency spatial free streaming can escape every compact set.

If $H$ is $C^2$ and its spatial Hessian matrix is uniformly bounded on every finite time interval, its Hamiltonian vector field is globally Lipschitz continuous in phase space with at most linear growth. The Picard-Lindelöf theorem and Gronwall inequality give a unique Hamiltonian flow for all finite forward and backward times. Bounds on ${\nabla }_{z}H(t,0)$ follow from time continuity on compact intervals.

With the displayed signed Lie-Poisson bracket and evolution $\dot{F}=\{F,H\}$, the equations are ${\dot{x}}_1=2(C-B)x_2{x}_3$, ${\dot{x}}_2=2(A-C)x_3{x}_1$, and ${\dot{x}}_3=2(B-A)x_1{x}_2$. The energy and the Casimir function of a Poisson manifold $x_1^2+x_2^2+x_3^2$ are conserved. For $A=1/(2I_1)$, $B=1/(2I_2)$, $C=1/(2I_3)$ with positive inertias, these are the Euler equations for a torque-free rigid body. Reversing the bracket while fixing $H$ reverses time orientation.

An integral of motion is a function on phase space that remains constant along every trajectory. If it has no explicit time dependence, $I(q,p)$ is an integral exactly when its Poisson bracket with the Hamiltonian vanishes: $\{I,H\}=0$.

For $\Phi (R,\varphi -\Omega t,z)$ with constant pattern angular velocity $\Omega$, ordinary specific energy satisfies $\dot{E}={\partial }_t\Phi =-\Omega {\partial }_{\varphi }\Phi$ and specific axial angular momentum satisfies ${\dot{L}}_z=-{\partial }_{\varphi }\Phi$. Their combination $E_J$ is conserved. This is the rotating-frame integral of motion; in the circular restricted three-body problem, the usual Jacobi constant is an affine rescaling of it. Thus time dependence in an inertial frame does not alone imply the absence of all integrals.

An isolating integral is a regular single-valued integral of motion whose fixed value restricts accessible phase space to a lower-dimensional level set. Functional independence of several such integrals means their differentials are linearly independent on the region considered. A central potential has specific orbital energy and the three components of specific angular momentum as four generically independent integrals. A Kepler orbit has a fifth independent integral supplied by its fixed apsidal direction, encoded by the Laplace-Runge-Lenz vector; its vector components obey two constraints with energy and angular momentum. Circular and radial orbits are degenerate cases of this count.

In three-dimensional autonomous mechanics, a regular common level set of $N$ independent isolating integrals has dimension $6-N$ in phase space. If an orbit is dense in a $d$-dimensional invariant torus and the integrals characterize its closure, $N=6-d$. A generic three-frequency regular orbit has three integrals; a spherical nonclosed rosette has energy and three components of specific angular momentum, with two-dimensional closure and four integrals. A periodic nonstationary orbit has one-dimensional closure and can be isolated by five independent orbit labels on the invariant family. The maximum is five wherever the Hamiltonian vector field is nonzero, since all integral differentials annihilate that field. A single resonant orbit need not yield an additional smooth single-valued integral on an open neighborhood; global superintegrability is stronger than orbital closure.

A conserved quantity is a quantity that remains constant along the time evolution of a system. In Hamiltonian mechanics, a possibly time-dependent phase-space function $F$ is conserved exactly when ${\partial }_{t}F+\{F,H\}=0$.

The canonical momentum conjugate to a generalized coordinate $q_i$ is $p_i=\partial L/\partial {\dot{q}}_i$.

For an axisymmetric magnetic vector potential, the conserved angular canonical momentum is $p_{\theta }=\gamma m{\rho }^2\dot{\theta }+q\rho A_{\theta }$. A particle arriving parallel to a solenoid axis from a field-free region has $p_{\theta }=0$. In a uniform field, $A_{\theta }=B\rho /2$ and $\dot{\theta }=-qB/(2\gamma m)$ wherever $\rho \ne 0$. The mechanical angular momentum changes through the fringe while the canonical quantity remains constant. Static magnetic fields also conserve $\gamma m{c}^2$; the azimuthal kick therefore reduces axial kinetic motion rather than increasing total speed.

For the affine action $S=-\frac{m}{2}\int g_{\mu \nu }{\dot{x}}^{\mu }{\dot{x}}^{\nu }d\tau$ with signature $(-+++)$, differentiating with respect to velocity gives the displayed canonical momentum. Future-oriented proper time in a conformally flat metric yields $P_i^c=-a\gamma m{v}_i$, whereas the conventional future-directed mechanical momentum has $q_i=a\gamma m{v}_i$. Changing the action's overall sign interchanges these canonical labels but leaves its classical geodesics unchanged. Both labels are conserved by spatial translation invariance, and either gives the same physical momentum magnitude proportional to $a^{-1}$. The sign should be stated when comparing a worldline canonical covector with a positive-frequency field-mode label.

If the spatially integrated Lagrangian changes by $dB[\Phi ]/dt$, where $B$ depends on fields and their spatial derivatives but not velocities, the canonical momentum components change by ${\pi }_a\mapsto {\pi }_a+\delta B/\delta {\Phi }_a$. This follows by varying $dB/dt$ and using spatial integration by parts. The canonical one-form changes by the exact variation $\delta B$, so the symplectic form and canonical commutation relations are preserved. For the two Feynman gauge Maxwell densities, $B=\int A_0{\partial }_i{A}_i{d}^{3}x$ gives shifts $\Delta {\pi }^0={\partial }_i{A}_i$ and $\Delta {\pi }^i=-{\partial }_i{A}_0$. Boundary flux must vanish for this spatial functional calculation.

Phase space has one position coordinate $q_i$ and one conjugate momentum $p_i$ for every degree of freedom; a point $(q,p)$ specifies an instantaneous mechanical state.

Velocity space is the space of velocity vectors at a fixed spatial position. Integrating a galactic distribution function over velocity space produces mass density; its normalized moments determine mean velocity and velocity dispersion.

For a gyrophase-independent phase-space distribution function, $d^{3}v=2\pi v_{\perp }d{v}_{\perp }d{v}_{\parallel }$. With first adiabatic invariant $\mu =v_{\perp }^2/(2B)$, this gives the measure in the title. When the magnetic field changes, both the measure and any explicit factor $\mu B$ in a moment must be perturbed.

The symplectic area enclosed by a closed orbit in a two-dimensional phase plane. With the orientation of Hamiltonian motion it is $\oint pdq=2\pi I$, where $I$ is the action variable.

An integral curve of a vector field in phase space, representing the positions and momenta of a moving system. A periodic one-dimensional Hamiltonian orbit encloses the area used to define the action variable.

The Hamiltonian is a function of canonical position and momentum that generates time evolution through Hamilton's equations.

For canonical coordinates $(q_i,p_i)$ and Hamiltonian $H$,

In canonical coordinates,

Functions $F_i$ for which $\{F_i,F_j\}=0$ for all pairs. Their Hamiltonian vector fields commute, and each is constant along the flows of the others. When they are integrals of motion of a Hamiltonian, they are first integrals in involution.

Second-class constraints ${\chi }_a$ with invertible bracket matrix can be eliminated with a modified bracket. For an ordinary even constraint set it is $\{F,G{\}}_D=\{F,G\}-\{F,{\chi }_a\}(C^{-1}{)}^{ab}\{{\chi }_b,G\}$, $C_{ab}=\{{\chi }_a,{\chi }_b\}$. Graded constraints require the matching graded sign convention.

A Dirac bracket on a graded phase space eliminates second-class constraints while respecting the graded Poisson bracket. The fermionic kinetic term $i{\psi }^m{\dot{\psi }}_m/2$ gives $\{{\psi }^m,{\psi }^n{\}}_D=-i{\eta }^{mn}$. Quantization then gives $\{{\hat{\psi }}^m,{\hat{\psi }}^n{\}}_{+}={\eta }^{mn}$.

An even Poisson bracket on a space with commuting and anticommuting variables obeys $\{F,G\}=-(-1{)}^{|F||G|}\{G,F\}$ and the graded Leibniz rule. In particular the bracket of two odd functions is symmetric. Quantization turns the odd-odd bracket into an anticommutator.

The Poisson bracket satisfies

If $F$ and $G$ have no explicit time dependence and satisfy $\{F,H\}=\{G,H\}=0$, the Jacobi identity gives $\{\{F,G\},H\}=0$. Their Poisson bracket is therefore also conserved.

The derivative $D_t={\partial }_t+\{\cdot ,H\}$ is a derivation of the Poisson bracket: $D_t\{f,g\}=\{D_{t}f,g\}+\{f,D_{t}g\}$. The time derivative part obeys the product rule, and the Hamiltonian part obeys it by the Jacobi identity. Therefore explicitly time-dependent conserved quantities also have a conserved Poisson bracket.

Along a trajectory generated by the Hamiltonian $H$,Thus a time-independent observable is conserved exactly when its Poisson bracket with $H$ vanishes.

For a time-independent Hamiltonian $H$, the explicit time partial derivative of a smooth time-dependent conserved quantity is again a conserved quantity. Here conservation means the identity $D_{t}f=f_t+\{f,H\}=0$ on phase space, with the standard Poisson bracket. Differentiating this identity at fixed canonical coordinates givesBecause $H_t=0$, this is exactly $D_t(f_t)=0$, proving the assertion. Iteration gives the same result for all higher explicit time partial derivatives for which the differentiations are defined. By Hamiltonian energy balance, requiring $H$ itself to be a conserved quantity on all trajectories is equivalent to $H_t=0$.

Along Hamilton's equations, spatial derivatives of the Hamiltonian function cancel in its total time derivative. Thus $dH/dt={\partial }_{t}H$. An autonomous Hamiltonian function is conserved; explicit time dependence generally prevents conservation even though phase space is still preserved.

For an autonomous Lagrangian $L(x,\dot{x})$, define its action $A=\int Ldt$ and Hamiltonian $H=\dot{x}{L}_{\dot{x}}-L$. The chain rule givesThus the identity holds along forced trajectories as well as solutions of the homogeneous Euler-Lagrange equation. Explicit time dependence adds $-{\partial }_{t}L$; stochastic work requires a consistent regularization or the Stratonovich chain rule.

The function $D=q\cdot p$ generates the canonical scaling $(q,p)\mapsto (e^{t}q,e^{-t}p)$. An observable $F$ is invariant under this scaling exactly when

For $z_j=q_j+i{p}_j$ and ${\bar{z}}_j=q_j-i{p}_j$,This follows from ${\partial }_{q_j}={\partial }_{z_j}+{\partial }_{{\bar{z}}_j}$ and ${\partial }_{p_j}=i{\partial }_{z_j}-i{\partial }_{{\bar{z}}_j}$.

For an anti-Hermitian matrix $A$, the real quadratic HamiltoniansatisfiesThus $A\mapsto -I_A$ realizes the matrix commutator as a Poisson bracket; it is the quadratic moment-map construction for the unitary action.

For a functional $H[u]$, its variational derivative is defined byafter integrating derivatives of $v$ by parts and discarding the boundary terms.

For a complex field $\psi$, the canonical Hamiltonian evolution conventionis accompanied by its complex conjugate equation.

An evolution equation has Hamiltonian form when a Hamiltonian operator applied to the variational derivative of a functional gives its time derivative. The induced functional bracket must satisfy the Poisson properties, including Jacobi.

A canonical transformation preserves the symplectic form and hence all Poisson brackets. In one degree of freedom, an invertible map $(q,p)\mapsto (Q,P)$ is canonical exactly when $\{Q,P{\}}_{q,p}=1$.

A canonical transformation preserves the symplectic form, equivalently the canonical Poisson brackets, and therefore the Hamiltonian form of every Hamiltonian system. Preservation of just one selected flow is weaker. For $H=(q_1^2+p_1^2)/2$ on two coordinate pairs, $Q_1=q_1$, $P_1=p_1$, $Q_2=2q_2$, $P_2=p_2$ preserves both that physical Hamiltonian and its equations, but $\{Q_2,P_2\}=2$. Thus it is not canonical. Universally preserving Hamiltonian form instead forces $DFJ(DF{)}^T=J$ by applying the chain rule to arbitrary Hamiltonian gradients.

The accompanying momentum map is $p=P+2\lambda XP/(m{\omega }^2)$. Its Poisson bracket is one up to second-order corrections. Substitution into $p^2/(2m)+m{\omega }^2{x}^2/2+\lambda x^3$ cancels all terms linear in $\lambda$, leaving a harmonic Hamiltonian to that order. The map is local and perturbative, not an exact global removal of an unbounded cubic potential.

With $\vartheta =\omega t+\delta$, a weak cubic potential produces a nonzero mean position and a second harmonic at first order. There is no first-order frequency shift. Higher-order frequency changes can accumulate phase over sufficiently long times, so the first-order expression is not a uniform all-time approximation.

For $p>0$, $Q=\sqrt{2p}\sinh q$ and $P=\sqrt{2p}\cosh q$ satisfy $\{Q,P\}=1$. They transform $p\cosh 2q$ into $(Q^2+P^2)/2$. Their inverse is $p=(P^2-Q^2)/2$ and $q=\mathrm{artanh}(Q/P)$, so the chart covers only $P>|Q|$, not the whole oscillator plane. Oscillator evolution beyond that wedge is not a global real continuation of the original variables.

For $H=\omega \cdot y+\frac{1}{2}|y{|}^2+ϵV(x)$ with a Diophantine frequency vector, solve ${\mathcal{D}}_{\omega }u=⟨V⟩-V$ and apply translated conjugacy of a Diophantine vector field to $w=ϵ\nabla u$. Set $\eta =ϵ\nabla u+a$ and use the symplectic cotangent lift with a closed momentum shift. The transformed Hamiltonian is exactly $E+\omega \cdot y^{\prime }+\frac{1}{2}{y}^{\prime T}D{\chi }^{-1}D{\chi }^{-T}{y}^{\prime }+\frac{1}{2}|\eta \circ \chi {|}^2$, where $E=ϵ⟨V⟩+\omega \cdot a$. Since $a=O({ϵ}^2)$, the last term is an angular potential of order ${ϵ}^2$ and the quadratic coefficient is identity plus $O(ϵ)$. The construction is on a smaller angular strip and momentum ball.

Independent functions $Q_1,\dots ,Q_n$ on a $2n$-dimensional symplectic manifold can locally form the position half of canonical coordinates exactly when $\{Q_i,Q_j\}=0$ throughout a neighbourhood. Their commuting Hamiltonian vector fields span the fibres of $Q$. Flow from a transverse section gives coordinates $(Q,t)$ with $\omega =\sum d{t}_i\wedge d{Q}_i+\beta (Q)$. Write $\beta =d\alpha$ locally and take $P_i=-t_i-{\alpha }_i(Q)$. Then $\omega =\sum d{Q}_i\wedge d{P}_i$.

A function $S(q,P)$ generates a canonical transformation through

The type-two generating function for a canonical transformation $F_2(q,P)=\cos q\sin P$ defines $p=-\sin q\sin P$ and $Q=\cos q\cos P$. On any branch with $0<q<\pi$ and $\cos P\ne 0$, its inverse in mixed coordinates isThe square root equals $\sin q$ on this branch. Differentiating $F_2$ proves the generating equations, and $F_{2,qP}=-\sin q\cos P\ne 0$ proves local invertibility by the inverse function theorem. Also $d(F_2-PQ)=pdq-PdQ$, so the symplectic form obeys $dq\wedge dp=dQ\wedge dP$. This proves the transformation is canonical locally. It is not a single global chart across the inverse-trigonometric branch boundaries.

Since $F_2$ is independent of time, the transformed Hamiltonian is obtained by substitution, with zero generating-function time correction. An added function of time alone is an irrelevant Hamiltonian freedom, since it leaves Hamilton's equations unchanged.

A function $\Phi (q,Q)$ generates a canonical transformation through

Liouville integrability means having $n$ almost-everywhere independent first integrals in mutual Poisson involution.

For a Hamiltonian phase space of dimension $2n$, Liouville integrability means $n$ functionally independent conserved quantities with pairwise vanishing Poisson brackets. This is the Hamiltonian specialization of integrable systems.

A superintegrable Hamiltonian system with $n$ degrees of freedom has more than $n$ functionally independent integrals of motion, including an integrable set of $n$ in involution. At a regular nonzero Hamiltonian vector field, at most $2n-1$ independent integral differentials can annihilate the flow. A maximally superintegrable system attains that bound. The three-dimensional Kepler orbit has five independent integrals through energy, angular momentum and the Laplace-Runge-Lenz vector. A closed orbit on one isolated resonance is weaker than maximal superintegrability on an open phase-space region.

On the regular set of two uncoupled oscillators with irrational frequency ratio, every continuous globally defined integral of motion is constant on each fixed-action flat torus, because every orbit is dense there. A differentiable integral is consequently a function of the two actions, so its differential of a smooth map lies in their span. Local angle combinations can still give extra local integrals; they fail to be single-valued on the torus.

For a compact connected regular level of $n$ independent Poisson-commuting functions, the periods of their joint flow form a full-rank discrete subgroup $\Lambda \subset {\mathbb{R}}^n$. The level is ${\mathbb{R}}^n/\Lambda$. A local smooth basis of these periods determines action integrals and the normalized periodic fields used to construct action-angle variables.

A Stäckel potential allows additive separation of the Hamilton-Jacobi equation in an orthogonal coordinate system. In axisymmetric prolate spheroidal coordinates, the relative-potential form is $\psi =[G(\lambda )-G(\mu )]/(\lambda -\mu )$. It gives energy, axial angular momentum and an independent third separation integral. Here $G$ is a one-variable generating function, not the gravitational constant. Generic bounded orbits are regular invariant-torus motions; resonance and coordinate degeneracy require separate treatment.

For fixed $E,l,I$, separated motion obeys $p_{\tau }^2=[E\tau +G(\tau )-I+(\alpha -\beta )l^2/(2(\tau +\alpha ))]/[2(\tau +\alpha )(\tau +\beta )]$. Allowed intervals are where this is nonnegative, with turning points and coordinate-fold endpoints treated separately. A generic nonresonant bound orbit with $l\ne 0$ circulates azimuthally and densely samples a projected volume bounded by confocal spheroids and hyperboloids. It is not uniformly distributed in volume. Resonant or zero-angular-momentum orbits can project to lower-dimensional regions; integrability alone does not imply that every orbit fills a three-dimensional shell.

For $H=\frac{1}{2}\sum p_j^2/h_j^2-\psi$, $l=p_{\varphi }$ and a Stäckel potential, multiplying the separated Hamilton-Jacobi equation by $\lambda -\mu$ gives $2(\tau +\alpha )(\tau +\beta )p_{\tau }^2-(\alpha -\beta )l^2/[2(\tau +\alpha )]-G(\tau )-E\tau =-I$ for both $\tau =\lambda ,\mu$. Eliminating $E$ gives $I=\frac{1}{2}[\mu v_{\lambda }^2+\lambda v_{\mu }^2+(\lambda +\mu +\alpha )v_{\varphi }^2]-[\mu G(\lambda )-\lambda G(\mu )]/(\lambda -\mu )$. Physical components are $v_j=h_j{\dot{q}}_j$ and $p_j=h_j{v}_j$. Separation proves conservation locally; direct Poisson brackets extend it across regular coordinate patches.