05.09.12 · symplectic / kam

Casimir functions of degenerate Poisson structures

shipped3 tiersLean: none

Anchor (Master): Weinstein 1983 *The local structure of Poisson manifolds* (J. Differential Geom. 18, splitting theorem and symplectic foliation); Olver 1993 §6.2; Marsden-Ratiu 1994 §10-§13; Vaisman *Lectures on the Geometry of Poisson Manifolds* Ch. 1-2

Intuition Beginner

A Poisson bracket is a rule that takes two quantities on a phase space and returns a third, telling you how one quantity drags the other along as time runs. Pick an energy function, and the bracket converts it into a flow — the motion of the system. On an ordinary phase space every direction is reachable: with enough cleverly chosen energies you can push the state anywhere. But many real systems are not like that. Some quantities simply never move, no matter which energy you choose to drive the dynamics.

A quantity that the bracket can never change is called a Casimir. It is conserved not because of a clever symmetry you arranged, but because the geometry forbids any motion in its direction. Think of a marble rolling on the surface of a fixed sphere. The marble can wander anywhere on the sphere, but its distance from the center is locked: no push along the surface can change the radius. That radius is a Casimir.

Casimirs matter because they slice phase space into sheets. Each sheet is a level set of all the Casimirs, and the dynamics stays trapped inside one sheet forever. Understanding a degenerate system means first finding these locked quantities, then studying the motion sheet by sheet. The degeneracy that looks like a defect is really a bookkeeping gift: it hands you conserved quantities for free.

Visual Beginner

Picture ordinary three-dimensional space filled with nested spheres, one inside the next, like the layers of an onion. Each sphere is one sheet. A point can travel freely across the surface of its own sphere, but it can never hop to a larger or smaller sphere.

The radius of the sphere a point sits on is the Casimir. The bracket can generate any motion you like along a sphere — that is where all the interesting dynamics happen — but it generates exactly zero motion in the radial direction. The whole space is foliated, meaning cleanly stacked, into these two-dimensional spheres plus the single point at the center where the sphere shrinks to nothing. The picture shows the two faces of a degenerate Poisson structure at once: rich motion inside each sheet, total stillness across the stack.

Worked example Beginner

Take the rotation bracket on three coordinates $x_{1}, x_{2}, x_{3}$ , the bracket that drives the spinning of a rigid body. Its rules are ${x_{1}, x_{2}} = x_{3}$ , then ${x_{2}, x_{3}} = x_{1}$ , then ${x_{3}, x_{1}} = x_{2}$ . Each pair of coordinates brackets to the third, cycling around. We will check that the squared radius $C = x_{1}^{2} + x_{2}^{2} + x_{3}^{2}$ is a Casimir: the bracket cannot move it.

Step 1. The bracket spreads across products by the same rule as a derivative: ${C, x_{1}} = {x_{1}^{2}, x_{1}} + {x_{2}^{2}, x_{1}} + {x_{3}^{2}, x_{1}}$ .

Step 2. The first term is ${x_{1}^{2}, x_{1}} = 2 x_{1} {x_{1}, x_{1}} = 0$ , since anything brackets to zero with itself.

Step 3. The second term is ${x_{2}^{2}, x_{1}} = 2 x_{2} {x_{2}, x_{1}}$ . The rule gives ${x_{2}, x_{1}} = - x_{3}$ , so this term is $- 2 x_{2} x_{3}$ .

Step 4. The third term is ${x_{3}^{2}, x_{1}} = 2 x_{3} {x_{3}, x_{1}} = 2 x_{3} \cdot x_{2} = 2 x_{2} x_{3}$ .

Step 5. Add them: $0 - 2 x_{2} x_{3} + 2 x_{2} x_{3} = 0$ . So ${C, x_{1}} = 0$ . The same cancellation works for $x_{2}$ and $x_{3}$ .

What this tells us: the squared radius brackets to zero against every coordinate, so it brackets to zero against every quantity built from them. No flow can change it. The level sets $C = r^{2}$ are the spheres of radius $r$ , and the spinning motion of a rigid body is locked onto one such sphere — exactly the onion picture, now confirmed by arithmetic.

Check your understanding Beginner

Exercise (easy, multiple choice).

What is the defining property of a Casimir function $C$ on a Poisson manifold?

A. $C$ is the energy that generates the dynamics B. The bracket of $C$ with every function is zero, so no flow can change $C$ C. $C$ has the largest possible bracket with the coordinates D. $C$ is any function that is constant in time for one chosen Hamiltonian

Hint

Think about the radius of the sphere: which functions can the bracket never move?

Answer

B. A Casimir brackets to zero against every function, so it is conserved by every Hamiltonian flow at once. Feedback-correct: this is exactly why the level sets of the Casimirs trap the dynamics. Feedback-wrong: option A names the Hamiltonian, not a Casimir; option D is too weak, since a single Hamiltonian conserves itself but is rarely a Casimir; option C inverts the property — Casimirs have zero bracket, not the largest.

Formal definition Intermediate+

Let $(M, π)$ be a Poisson manifold: a smooth manifold $M$ carrying a bivector field $π \in Γ (Λ^{2} T M)$ whose Schouten-Nijenhuis bracket vanishes, $[π, π]_{SN} = 0$ . The bracket of two functions is ${f, g} = π (df, d g)$ , and the vanishing condition is equivalent to the Jacobi identity for ${\cdot, \cdot}$ . The anchor (or sharp) map is the bundle map $$ \pi^\sharp : T^\ast M \to TM, \qquad \langle \beta, \pi^\sharp(\alpha) \rangle = \pi(\alpha, \beta), $$ the same map studied for the cotangent algebroid in 03.04.19. The Hamiltonian vector field of $f$ is $X_{f} = π^{♯} (df)$ , so ${f, g} = X_{f} (g) = ⟨ d g, π^{♯} (df)⟩$ .

A function $C \in C^{\infty} (M)$ is a Casimir function (Olver's distinguished function) of $(M, π)$ when $$ {C, f} = 0 \quad \text{for all } f \in C^\infty(M), \qquad \text{equivalently} \qquad \pi^\sharp(dC) = 0. $$ The equivalence is immediate: ${C, f} = ⟨ df, π^{♯} (d C)⟩$ for all $f$ vanishes precisely when $π^{♯} (d C) = 0$ , since the differentials $df$ separate covectors. Thus the Casimirs are exactly the functions whose differential lies pointwise in $ker π^{♯}$ .

The Poisson structure has constant rank $2 k$ on $M$ when the image $im π_{p}^{♯} \subseteq T_{p} M$ has dimension $2 k$ for every $p$ (the rank is always even because $π$ is a bivector). On an $m$ -manifold of constant rank $2 k$ , the corank is $m - 2 k$ . The structure is degenerate when $2 k < m$ , that is when $ker π^{♯} \neq = 0$ . The constant Casimirs always exist; the content of the degenerate case is the existence of $m - 2 k$ functionally independent non-constant Casimirs near each point.

Counterexamples to common slips

A function conserved by one chosen Hamiltonian is not a Casimir. Energy is conserved by its own flow yet typically moves under other Hamiltonians. A Casimir is conserved by every Hamiltonian simultaneously.
On a symplectic manifold ( $2 k = m$ , $π^{♯}$ invertible) the only Casimirs are the locally constant functions. Degeneracy is what makes non-constant Casimirs possible; a non-degenerate bracket has none.
Constant rank is a real hypothesis. The rotation bracket has rank $2$ away from the origin but rank $0$ at the origin, where the symplectic leaf is a single point. Casimir counting by the formula $m - 2 k$ is local to the constant-rank region.

Key theorem with proof Intermediate+

Theorem (Casimirs are leaf invariants and universal conserved quantities). Let $(M, π)$ be a Poisson manifold and $C \in C^{\infty} (M)$ a Casimir. Then $C$ is constant along every Hamiltonian flow, and $C$ is constant on every symplectic leaf — the maximal integral manifolds of the (possibly singular) distribution $D = im π^{♯}$ .

Proof. Let $f \in C^{\infty} (M)$ be arbitrary and let $φ_{t}$ denote the flow of $X_{f} = π^{♯} (df)$ . The rate of change of $C$ along this flow is $$ \frac{d}{dt}(C \circ \varphi_t) = X_f(C) = {f, C} = -{C, f} = 0, $$ where the last equality is the defining property of a Casimir. Hence $C \circ φ_{t} = C$ for all $t$ in the flow's domain, so $C$ is constant along every Hamiltonian trajectory.

For the leaf statement, recall that the characteristic distribution is $D_{p} = im π_{p}^{♯} = {X_{f} (p) : f \in C^{\infty} (M)}$ . The Jacobi identity makes $D$ involutive in the Stefan-Sussmann sense, and its integral manifolds are the symplectic leaves $L$ ; each $L$ carries the symplectic form $ω_{L}$ defined by $ω_{L} (X_{f}, X_{g}) = {f, g}$ , well posed because $π^{♯}$ restricts to an isomorphism $T_{p}^{*} M / ker π_{p}^{♯} \to T_{p} L$ . A tangent vector to $L$ at $p$ has the form $X_{f} (p)$ for some $f$ , and $$ dC_p\big(X_f(p)\big) = \langle dC_p, \pi^\sharp_p(df_p) \rangle = -\langle df_p, \pi^\sharp_p(dC_p) \rangle = 0 $$ because $π_{p}^{♯} (d C_{p}) = 0$ . Thus $d C$ annihilates every vector tangent to $L$ , so $d C ∣_{T L} = 0$ , and since $L$ is connected $C$ is constant on $L$ . The level sets of the Casimirs are therefore unions of symplectic leaves. $□$

Bridge. This theorem builds toward the structural decomposition of a Poisson manifold into symplectic leaves carrying the dynamics and a transverse direction recording the Casimirs; it appears again in the Lie-Poisson setting, where the leaves are coadjoint orbits and the Casimirs are the invariant polynomials, and again in symplectic reduction, where fixing the Casimirs is the first stage of cutting the phase space down to a symplectic quotient. The same kernel-of-the-anchor calculation organizes both the local splitting theorem and the global counting of conserved quantities, so a single bracket identity controls how degeneracy distributes a system's invariants.

Exercises Intermediate+

Exercise (hard, symbolic).

For the Lie-Poisson bracket on $so (3)^{*} ≅ R^{3}$ , ${x_{i}, x_{j}} = \sum_{k} ε_{ij k} x_{k}$ , take the rigid-body Hamiltonian $H = \frac{1}{2} (x_{1}^{2} / I_{1} + x_{2}^{2} / I_{2} + x_{3}^{2} / I_{3})$ . Derive the equations of motion $\overset{x}{˙}_{i} = {x_{i}, H}$ and confirm the Casimir $C = x_{1}^{2} + x_{2}^{2} + x_{3}^{2}$ is conserved along them.

Hint

Compute $\overset{x}{˙}_{1} = {x_{1}, H}$ term by term, then check $\dot{C} = 2 \sum_{i} x_{i} \overset{x}{˙}_{i}$ .

Answer

$\overset{x}{˙}_{1} = {x_{1}, H} = \sum_{i} \frac{x _{i}}{I _{i}} {x_{1}, x_{i}} = \frac{x _{2}}{I _{2}} {x_{1}, x_{2}} + \frac{x _{3}}{I _{3}} {x_{1}, x_{3}} = \frac{x _{2} x _{3}}{I _{2}} - \frac{x _{3} x _{2}}{I _{3}} = x_{2} x_{3} (\frac{1}{I _{2}} - \frac{1}{I _{3}})$ . Cyclically, $\overset{x}{˙}_{2} = x_{3} x_{1} (\frac{1}{I _{3}} - \frac{1}{I _{1}})$ and $\overset{x}{˙}_{3} = x_{1} x_{2} (\frac{1}{I _{1}} - \frac{1}{I _{2}})$ — the Euler equations. Then $\dot{C} = 2 (x_{1} \overset{x}{˙}_{1} + x_{2} \overset{x}{˙}_{2} + x_{3} \overset{x}{˙}_{3}) = 2 x_{1} x_{2} x_{3} [(\frac{1}{I _{2}} - \frac{1}{I _{3}}) + (\frac{1}{I _{3}} - \frac{1}{I _{1}}) + (\frac{1}{I _{1}} - \frac{1}{I _{2}})] = 0$ . The body's angular-momentum magnitude is conserved: the motion is the intersection of the Casimir sphere with the energy ellipsoid — Poinsot's construction.

Exercise (hard, short-answer).

State, with justification, why on a connected open set of constant rank $2 k$ in an $m$ -manifold one can choose local coordinates in which exactly $m - 2 k$ of them are Casimirs, and name the theorem responsible.

Hint

This is the local normal form of a Poisson structure. Which 1983 theorem provides it?

Answer

Weinstein's splitting theorem (1983) gives, near a regular point, coordinates $(q_{1}, \dots, q_{k}, p_{1}, \dots, p_{k}, c_{1}, \dots, c_{m - 2 k})$ in which $π = \sum_{i} \partial_{q_{i}} \land \partial_{p_{i}}$ , with no $c$ -terms. The $(q, p)$ block is the canonical symplectic factor of the leaf; the $c_{j}$ have $π^{♯} (d c_{j}) = 0$ , so each $c_{j}$ is a Casimir, and there are exactly $m - 2 k$ of them. Functional independence is the statement that their differentials span $ker π^{♯}$ pointwise. Thus the corank counts the local Casimirs, and the level sets of the $c_{j}$ are the symplectic leaves.

Lean formalization Intermediate+

Mathlib has no Poisson-manifold structure, so neither the Casimir condition $π^{♯} (d C) = 0$ nor the symplectic foliation can be stated. the unit metadata field Mathlib gap analysis records the specific missing infrastructure: the bivector field, the Schouten-Nijenhuis bracket, the integrability condition $[π, π]_{SN} = 0$ , the anchor map, and the singular-foliation API behind Weinstein's splitting theorem. This unit ships with lean_status: none; the correctness gate is the named human reviewer, who verifies the bracket computations ${C, x_{i}} = 0$ and the leaf-invariance argument by hand.

Advanced results Master

Beyond the leaf-invariance theorem, three structural facts make Casimirs the organizing invariants of a degenerate Poisson manifold.

The first is Weinstein's splitting theorem (1983). Near any point $p$ where the rank is locally constant, equal to $2 k$ , there is a coordinate chart in which $$ (M, \pi) ;\cong; \big(S, \omega_{\mathrm{can}}\big) \times \big(N, \pi_N\big), $$ where $S$ is a $2 k$ -dimensional symplectic factor carrying the canonical bracket and $N$ is a transverse Poisson manifold whose bracket vanishes at the image of $p$ . At a regular point $π_{N} \equiv 0$ , so $N$ contributes $m - 2 k$ Casimir coordinates and $S$ is the symplectic leaf through $p$ . The transverse factor $N$ , with its possibly nonzero $π_{N}$ , is the transverse Poisson structure of the leaf; its isomorphism class is a leaf invariant, and its linearization at $p$ recovers a Lie algebra — the isotropy data that distinguishes regular from singular leaves.

The second is the Lie-Poisson identification. For a finite-dimensional Lie algebra $g$ , the dual $g^{*}$ carries the canonical Lie-Poisson bracket $$ {f, g}(\mu) = \big\langle \mu, [df_\mu, dg_\mu] \big\rangle, \qquad \mu \in \mathfrak{g}^\ast, $$ where $d f_{μ}, d g_{μ} \in g^{**} ≅ g$ . The symplectic leaves are the coadjoint orbits of any Lie group $G$ integrating $g$ , with the Kirillov-Kostant-Souriau form, and a function $C$ on $g^{*}$ is a Casimir precisely when it is $Ad^{*}$ -invariant: $C (Ad_{g}^{*} μ) = C (μ)$ for all $g \in G$ . The polynomial Casimirs are then the $G$ -invariant polynomials on $g^{*}$ , the image under symmetrization of the center $Z (U (g))$ of the universal enveloping algebra — the classical shadow of the Casimir operators of representation theory. For $so (3)$ this center is generated by the single quadratic invariant $x_{1}^{2} + x_{2}^{2} + x_{3}^{2}$ , recovering the rigid-body example; for $su (n)$ one obtains $n - 1$ independent invariants of degrees $2, 3, \dots, n$ .

The third is the role of Casimirs in reduction. Fixing the values of all Casimirs selects a symplectic leaf, the maximal submanifold on which the bracket restricts to a non-degenerate symplectic structure. Dynamics generated by any Hamiltonian descends to this leaf, where the standard machinery of symplectic geometry — Liouville's theorem, action-angle coordinates, the Arnold-Liouville integrability criterion — applies. Reduction by Casimirs is therefore the zeroth step of Marsden-Weinstein reduction: before quotienting by a symmetry group, one already restricts to a leaf cut out by the center of the Poisson algebra.

Synthesis. The Casimirs of a degenerate Poisson structure are simultaneously the kernel of the anchor map, the transverse coordinates of the splitting normal form, the center of the Poisson algebra, the $Ad^{*}$ -invariants in the Lie-Poisson case, and the data whose level sets are the symplectic leaves. These five descriptions are one object viewed from five directions: the corank $m - 2 k$ counts them, the splitting theorem produces them as coordinates, the leaf-invariance theorem confirms the dynamics cannot move them, the enveloping-algebra center supplies the invariant polynomials for Lie-Poisson models, and symplectic reduction consumes them to recover an honest symplectic phase space on each leaf. Degeneracy, far from being an obstruction, is the mechanism that endows a Poisson system with its conserved geometry.

Full proof set Master

Proposition (Casimirs form the center of the Poisson algebra and are closed under the bracket of the algebra). Let $(M, π)$ be a Poisson manifold and let $Cas (M) = {C \in C^{\infty} (M) : {C, f} = 0 \forall f}$ . Then $Cas (M)$ is a subalgebra of $C^{\infty} (M)$ under pointwise multiplication, it coincides with the center of the Lie algebra $(C^{\infty} (M), {\cdot, \cdot})$ , and any smooth function of Casimirs is again a Casimir.

Proof. That $Cas (M)$ is the center of the Lie algebra is the definition rewritten: $C$ lies in the center exactly when ${C, f} = 0$ for all $f$ , which is the Casimir condition.

For closure under multiplication, let $C_{1}, C_{2} \in Cas (M)$ and $f \in C^{\infty} (M)$ . The Leibniz rule for the Poisson bracket gives $$ {C_1 C_2, f} = C_1 {C_2, f} + {C_1, f} C_2 = C_1 \cdot 0 + 0 \cdot C_2 = 0, $$ so $C_{1} C_{2} \in Cas (M)$ . The same Leibniz rule with $f$ in the first slot shows sums and scalar multiples remain Casimirs, so $Cas (M)$ is a subalgebra.

For closure under smooth composition, let $C_{1}, \dots, C_{r} \in Cas (M)$ and $Φ \in C^{\infty} (R^{r})$ , and set $C = Φ (C_{1}, \dots, C_{r})$ . Using $X_{f} = π^{♯} (df)$ and the chain rule, $$ {C, f} = X_f\big(\Phi(C_1,\dots,C_r)\big) = \sum_{j=1}^{r} \frac{\partial \Phi}{\partial t_j}(C_1,\dots,C_r) , X_f(C_j) = \sum_{j=1}^{r} \frac{\partial \Phi}{\partial t_j} , {C_j, f} = 0, $$ since each ${C_{j}, f} = - {f, C_{j}} = 0$ . Hence $C \in Cas (M)$ .

It remains to note that $d C = \sum_{j} (\partial_{j} Φ) d C_{j}$ lies pointwise in $ker π^{♯}$ , because each $d C_{j}$ does and $ker π_{p}^{♯}$ is a linear subspace; this re-proves $π^{♯} (d C) = 0$ directly and shows the Casimirs are exactly the functions whose differentials are sections of $ker π^{♯}$ . On a constant-rank region this kernel is a subbundle of rank $m - 2 k$ , so the maximal number of functionally independent Casimirs near a point is $m - 2 k$ , attained by the transverse coordinates of the splitting normal form. $□$

Connections Master

The anchor map $π^{♯}$ whose kernel defines the Casimirs is the anchor of the cotangent Lie algebroid 03.04.19; Casimirs are precisely the functions whose differentials are flat sections of this algebroid, and the symplectic foliation is its orbit foliation.

The Poisson bracket and the integrability condition $[π, π]_{SN} = 0$ on which every computation here rests are established in 05.02.02, and the Hamiltonian vector field $X_{f} = π^{♯} (df)$ used throughout the proofs is the object of 05.02.01.

The Lie-Poisson identification of Casimirs with $Ad^{*}$ -invariants makes this unit the bridge from degenerate brackets to the Euler-Arnold equations 05.09.05, where the rigid-body Hamiltonian flows on a coadjoint orbit, and to the master-symmetry construction 05.09.11, whose commuting hierarchies live on the symplectic leaves cut out by Casimirs in the bi-Hamiltonian picture.

Historical & philosophical context Master

The degenerate bracket and its distinguished functions trace to Sophus Lie's theory of function groups in the second volume of the Theorie der Transformationsgruppen ^{[Lie 1890]}, where a closed system of functions under the bracket carries a distinguished subsystem that brackets to zero with all the rest — the functions later named Casimirs after the physicist Hendrik Casimir, whose 1931 work on the quantum rigid rotor used the corresponding invariant operator. Lie already understood that the corank of the bracket counts these distinguished functions and that they foliate the underlying space.

The modern geometric account is Weinstein's 1983 paper on the local structure of Poisson manifolds ^{[Weinstein 1983]}, which proved the splitting theorem and made the symplectic foliation a precise object, and Olver's 1993 treatment ^{[Olver 1993]}, which placed degenerate Poisson structures and their distinguished functions at the service of symmetry analysis for differential equations. The Lie-Poisson identification of Casimirs with coadjoint invariants, and its use in rigid-body and fluid reduction, is developed in Marsden-Ratiu ^{[Marsden-Ratiu 1994]}.

Bibliography Master

@article{Weinstein1983LocalStructure,
  author  = {Weinstein, Alan},
  title   = {The local structure of {Poisson} manifolds},
  journal = {Journal of Differential Geometry},
  volume  = {18},
  number  = {3},
  pages   = {523--557},
  year    = {1983}
}

@book{Olver1993,
  author    = {Olver, Peter J.},
  title     = {Applications of {Lie} Groups to Differential Equations},
  edition   = {2nd},
  series    = {Graduate Texts in Mathematics},
  volume    = {107},
  publisher = {Springer},
  year      = {1993}
}

@book{MarsdenRatiu1994,
  author    = {Marsden, Jerrold E. and Ratiu, Tudor S.},
  title     = {Introduction to Mechanics and Symmetry},
  series    = {Texts in Applied Mathematics},
  volume    = {17},
  publisher = {Springer},
  year      = {1994}
}

@book{Vaisman1994,
  author    = {Vaisman, Izu},
  title     = {Lectures on the Geometry of {Poisson} Manifolds},
  series    = {Progress in Mathematics},
  volume    = {118},
  publisher = {Birkh\"auser},
  year      = {1994}
}

@book{Lie1890,
  author    = {Lie, Sophus},
  title     = {Theorie der Transformationsgruppen, Zweiter Abschnitt},
  publisher = {Teubner},
  address   = {Leipzig},
  year      = {1890}
}

Prerequisites

05.02.02
03.04.19
05.02.01

Tier anchors

beginner: the quantities a Hamiltonian flow can never change because the bracket cannot move them — energy shells you cannot leave
intermediate: Olver *Applications of Lie Groups to Differential Equations* 2nd ed. §6.2 (degenerate Poisson structures, distinguished functions); Marsden-Ratiu *Introduction to Mechanics and Symmetry* §10.3-§10.4
master: Weinstein 1983 *The local structure of Poisson manifolds* (J. Differential Geom. 18, splitting theorem and symplectic foliation); Olver 1993 §6.2; Marsden-Ratiu 1994 §10-§13; Vaisman *Lectures on the Geometry of Poisson Manifolds* Ch. 1-2

References

Olver — Applications of Lie Groups to Differential Equations, 2nd ed. · Springer GTM 107, §6.2 degenerate Poisson structures, Casimir (distinguished) functions, symplectic foliation
Weinstein — The local structure of Poisson manifolds · J. Differential Geom. 18 (1983), splitting theorem and the symplectic foliation
Marsden-Ratiu — Introduction to Mechanics and Symmetry · Springer TAM 17, §10-§13, Lie-Poisson structures, coadjoint orbits, rigid-body reduction
Vaisman — Lectures on the Geometry of Poisson Manifolds · Birkhäuser PM 118, Ch. 1-2, characteristic foliation and Casimir functions
Lie — Theorie der Transformationsgruppen, vol. 2 · Teubner 1890, function groups and the origin of the degenerate bracket and its distinguished functions

Estimated time

beginner: 18m
intermediate: 45m
master: 80m