ProofComplete

Proof of Darboux's Theorem

ProofConstruction via Moser's Method

We prove Darboux's theorem using Moser's deformation method, showing that any symplectic form can be transformed to standard form near a point.

Setup: Let (M,Ο‰)(M, \omega) be a 2n2n-dimensional symplectic manifold and p∈Mp \in M. We construct coordinates (qi,pi)(q_i, p_i) near pp with Ο‰=βˆ‘dqi∧dpi\omega = \sum dq_i \wedge dp_i.

Step 1: Linear algebra at pp

The tangent space (TpM,Ο‰p)(T_pM, \omega_p) is a symplectic vector space. By the symplectic basis theorem, there exist e1,…,en,f1,…,fn∈TpMe_1, \ldots, e_n, f_1, \ldots, f_n \in T_pM with:

Ο‰p(ei,ej)=0,Ο‰p(fi,fj)=0,Ο‰p(ei,fj)=Ξ΄ij\omega_p(e_i, e_j) = 0, \quad \omega_p(f_i, f_j) = 0, \quad \omega_p(e_i, f_j) = \delta_{ij}

Choose coordinates (x1,…,x2n)(x_1, \ldots, x_{2n}) centered at pp such that:

βˆ‚βˆ‚xi∣p=ei,βˆ‚βˆ‚xn+i∣p=fi\frac{\partial}{\partial x_i}\bigg|_p = e_i, \quad \frac{\partial}{\partial x_{n+i}}\bigg|_p = f_i

In these coordinates, Ο‰p=Ο‰0∣p\omega_p = \omega_0|_p where Ο‰0=βˆ‘i=1ndxi∧dxn+i\omega_0 = \sum_{i=1}^n dx_i \wedge dx_{n+i}.

Step 2: Taylor expansion near pp

Write Ο‰=Ο‰0+Ξ±\omega = \omega_0 + \alpha where Ξ±\alpha is a 2-form vanishing at pp (i.e., Ξ±p=0\alpha_p = 0).

Since ω\omega is closed, dω=0d\omega = 0, and ω0\omega_0 is closed, we have dα=0d\alpha = 0.

By the PoincarΓ© lemma, on a small neighborhood UU of pp, there exists a 1-form Ξ²\beta with dΞ²=Ξ±d\beta = \alpha.

Moreover, since Ξ±p=0\alpha_p = 0, we can choose Ξ²\beta with Ξ²p=0\beta_p = 0 (by subtracting an exact form if necessary).

Step 3: Apply Moser's method

Define a path of 2-forms:

Ο‰t=Ο‰0+tΞ±=Ο‰0+t dΞ²,0≀t≀1\omega_t = \omega_0 + t\alpha = \omega_0 + t \, d\beta, \quad 0 \leq t \leq 1

For tt sufficiently small, Ο‰t\omega_t is non-degenerate (since Ξ±p=0\alpha_p = 0 means Ξ±\alpha is small near pp).

We seek a vector field XtX_t satisfying:

LXtΟ‰t=βˆ’Ξ±\mathcal{L}_{X_t} \omega_t = -\alpha

By Cartan's formula:

LXtωt=d(ιXtωt)+ιXt(dωt)=d(ιXtωt)\mathcal{L}_{X_t} \omega_t = d(\iota_{X_t} \omega_t) + \iota_{X_t}(d\omega_t) = d(\iota_{X_t} \omega_t)

since dωt=0d\omega_t = 0. Thus we need:

d(ΞΉXtΟ‰t)=βˆ’dΞ²d(\iota_{X_t} \omega_t) = -d\beta

So we can take:

ΞΉXtΟ‰t=βˆ’Ξ²\iota_{X_t} \omega_t = -\beta

Since Ο‰t\omega_t is non-degenerate and Ξ²p=0\beta_p = 0, this uniquely defines XtX_t with Xt(p)=0X_t(p) = 0.

Step 4: Integrate the flow

Let Ο•t\phi_t be the flow of XtX_t: ddtΟ•t=Xtβˆ˜Ο•t\frac{d}{dt}\phi_t = X_t \circ \phi_t with Ο•0=id\phi_0 = \text{id}.

By construction:

ddt(Ο•tβˆ—Ο‰t)=Ο•tβˆ—(LXtΟ‰t+Ο‰Λ™t)=Ο•tβˆ—(βˆ’Ξ±+Ξ±)=0\frac{d}{dt}(\phi_t^* \omega_t) = \phi_t^*(\mathcal{L}_{X_t} \omega_t + \dot{\omega}_t) = \phi_t^*(-\alpha + \alpha) = 0

Therefore Ο•tβˆ—Ο‰t=Ο•0βˆ—Ο‰0=Ο‰0\phi_t^* \omega_t = \phi_0^* \omega_0 = \omega_0.

At t=1t = 1:

Ο•1βˆ—Ο‰1=Ο•1βˆ—Ο‰=Ο‰0\phi_1^* \omega_1 = \phi_1^* \omega = \omega_0

So Ο•1\phi_1 is a local diffeomorphism near pp taking Ο‰\omega to Ο‰0\omega_0.

Step 5: Define Darboux coordinates

The coordinates (q1,…,qn,p1,…,pn)=(x1βˆ˜Ο•1,…,x2nβˆ˜Ο•1)(q_1, \ldots, q_n, p_1, \ldots, p_n) = (x_1 \circ \phi_1, \ldots, x_{2n} \circ \phi_1) are Darboux coordinates:

Ο‰=(Ο•1βˆ’1)βˆ—Ο‰0=βˆ‘i=1ndqi∧dpi\omega = (\phi_1^{-1})^* \omega_0 = \sum_{i=1}^n dq_i \wedge dp_i

This completes the proof.

β– 
Remark

The proof is constructive: given coordinates where Ο‰\omega agrees with Ο‰0\omega_0 at a point, we construct a diffeomorphism making them agree in a neighborhood. The key steps are:

  1. Linear algebra to match at a point
  2. PoincarΓ© lemma to write difference as exact
  3. Moser's method to flow to standard form
ExampleExplicit Construction for $n=1$

For a 2-dimensional symplectic manifold (M,Ο‰)(M, \omega), let p∈Mp \in M and choose coordinates (x,y)(x, y) with:

Ο‰=f(x,y) dx∧dy,f(0,0)=1\omega = f(x, y) \, dx \wedge dy, \quad f(0, 0) = 1

We want new coordinates (q,p)(q, p) with Ο‰=dq∧dp\omega = dq \wedge dp.

Set Ο‰t=(1+t(fβˆ’1))dx∧dy\omega_t = (1 + t(f-1)) dx \wedge dy. Solve:

ΞΉXtΟ‰t=βˆ’Ξ²\iota_{X_t} \omega_t = -\beta

where dΞ²=(fβˆ’1)dx∧dyd\beta = (f-1) dx \wedge dy. One can take Ξ²=g(x,y)dx\beta = g(x,y) dx where βˆ‚gβˆ‚y=fβˆ’1\frac{\partial g}{\partial y} = f - 1.

Integrating XtX_t gives Darboux coordinates explicitly.

DefinitionAlternative Proof via Induction

An alternative proof proceeds by induction on nn:

  1. Choose e1,f1e_1, f_1 with Ο‰(e1,f1)=1\omega(e_1, f_1) = 1
  2. Construct coordinates (q1,p1,…)(q_1, p_1, \ldots) with Ο‰(e1)=dq1\omega(e_1) = dq_1, Ο‰(f1)=dp1\omega(f_1) = dp_1
  3. Show Ο‰=dq1∧dp1+Ο‰β€²\omega = dq_1 \wedge dp_1 + \omega' where Ο‰β€²\omega' depends only on remaining coordinates
  4. Apply induction to Ο‰β€²\omega'

This proof is more elementary but less geometric than Moser's method.

Remark

The non-degeneracy of Ο‰t\omega_t for small tt is ensured by:

det⁑(Ο‰t)=(1+tβ‹…(correction))ndet⁑(Ο‰0)β‰ 0\det(\omega_t) = (1 + t \cdot (\text{correction}))^n \det(\omega_0) \neq 0

near pp, since the correction term vanishes at pp. This is why the proof works locally but requires care for global statements.