02.12.05 · analysis / ode

Rectification (straightening) of a vector field

shipped3 tiersLean: nonepending prereqs

Anchor (Master): Arnold *Ordinary Differential Equations* Ch.2 §7 + Ch.4 §31 (the flow-box theorem as the geometric form of existence-uniqueness-smooth-dependence); Hartman *Ordinary Differential Equations* Ch. IV §1 (smoothing-the-flow-by-the-transverse-section construction with full Banach-space generality); Lang *Differential Manifolds* Ch. IV §2 (flow-box theorem on a smooth manifold); Hörmander *The Analysis of Linear Partial Differential Operators* Vol. I §8.1 (rectification as the gateway to the method of characteristics for first-order PDEs)

Intuition [Beginner]

A flowing river has currents that swirl, slow down, and speed up. But if you zoom in on a single point in the middle of the river where the water is moving — not stuck in an eddy and not stalled at the bank — the local flow looks like uniform motion along straight parallel lines. Pick the right coordinate system aligned to the current direction at that point, and locally the flow becomes "everything moves rightward at unit speed". The rectification theorem is the precise statement that every smooth vector field looks like this near any point where the field is non-zero.

The picture is geometric. Draw a small surface through your chosen point, transverse to the flow direction. Each starting point on this surface, when carried by the flow for various small times, traces out a curve. The two coordinates "starting point on the surface" and "time spent flowing" together label every nearby point. In these new coordinates, the flow is the simple translation in the time-direction. The vector field has been straightened.

Why does it matter? Local ODE theory split into existence, uniqueness, and smooth dependence on initial data is the calculational content. Rectification is the same content in geometric clothing: at every point where the field does not vanish, the flow is a uniform straight-line motion in disguise. Everything interesting about an ODE — singularities, periodic orbits, equilibria — happens at the points the theorem excludes.

Visual [Beginner]

A pair of side-by-side phase portraits. On the left, a smooth vector field on the plane near a non-equilibrium point: the arrows curve, vary in length, and may form loops or spirals further away. On the right, the same neighbourhood after the rectifying coordinate change: the arrows are all horizontal, all of the same length, pointing rightward. A small vertical line segment in the right picture marks the transverse section that the rectification used to build the new coordinates.

The picture conveys the structural content: every regular point of a smooth vector field has a neighbourhood in which the field is a uniform translation. The transverse section is the geometric input that builds the coordinate change. Outside this neighbourhood the field can do anything, but inside it the flow is straight-line motion.

Worked example [Beginner]

Rectify the planar vector field given by the system $\overset{x}{˙} = - y$ , $\overset{y}{˙} = x$ near the point $p = (1, 0)$ . This is the rotation flow whose trajectories are circles around the origin; the field vanishes at the origin and is non-zero everywhere else.

Step 1. Check that the field is non-zero at $p$ . At $p = (1, 0)$ the velocity vector is $(\overset{x}{˙}, \overset{y}{˙}) = (- 0, 1) = (0, 1)$ . The field points straight up at $p$ with unit length. So $p$ is a regular point and rectification applies.

Step 2. Choose a transverse section. The line segment $Σ = {(x, 0) : 1/2 < x < 3/2}$ passes through $p$ and is transverse to the flow at $p$ because the velocity $(0, 1)$ at $p$ has no horizontal component while $Σ$ runs horizontally.

Step 3. Compute the flow. The trajectories of the rotation system are circles: starting at $(r, 0)$ on $Σ$ and flowing for time $t$ , the trajectory is $(r cos t, r sin t)$ . So the flow at time $t$ of the point $(r, 0)$ is $g^{t} (r, 0) = (r cos t, r sin t)$ .

Step 4. Build the rectifying coordinates. Define new coordinates $(y^{1}, y^{2})$ on a neighbourhood of $p$ by $y^{1} = t$ (time flowed) and $y^{2} = r$ (starting radius on $Σ$ ). The map back to the original $(x_{1}, x_{2})$ coordinates is $x_{1} = y^{2} cos y^{1}$ , $x_{2} = y^{2} sin y^{1}$ . This is polar coordinates with angle $y^{1}$ and radius $y^{2}$ , defined on the neighbourhood $∣ y^{1} ∣ < 1$ , $1/2 < y^{2} < 3/2$ .

Step 5. Read off the flow in new coordinates. In $(y^{1}, y^{2})$ , the flow is $g^{t} (y^{1}, y^{2}) = (y^{1} + t, y^{2})$ , which is uniform translation in the $y^{1}$ direction. In these coordinates the rotation has been straightened into uniform horizontal motion: the trajectories are now horizontal lines along which $y^{1}$ varies and $y^{2}$ stays constant.

What this tells us: rectification turns the curved circular flow into uniform translation. The single trick — pick a transverse section, parametrise by "starting point and time flowed" — works near every non-zero point. The origin is excluded because the field vanishes there and there is nothing to straighten.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $M$ be a smooth $n$ -dimensional manifold and let $X$ be a $C^{1}$ vector field on $M$ . A point $p \in M$ is regular for $X$ when $X (p) \neq = 0$ ; otherwise it is an equilibrium (or singular point). The local flow of $X$ near $p$ is the unique $C^{1}$ map $g^{t} : U \to M$ defined on an open neighbourhood $U ∋ p$ for $∣ t ∣ < ε$ that satisfies $g^{0} = id$ and $\frac{d}{d t} g^{t} (q) = X (g^{t} (q))$ for every $q \in U$ .

A rectifying chart for $X$ at $p$ is a $C^{1}$ diffeomorphism $Φ : V \to W$ from an open neighbourhood $V \subseteq R^{n}$ of $0$ onto an open neighbourhood $W \subseteq M$ of $p$ with $Φ (0) = p$ and $$ \Phi_*(\partial / \partial y^1) = X|W, $$ where $\Phi* $d e n o t es t h e p u s h f or w a r d o f v ec t or f i e l d s . E q u i v a l e n tl y, in t h e l oc a l coor d ina t es$ (y^1, \ldots, y^n) $d e f in e d b y$ \Phi^{-1} $, t h e v ec t or f i e l d$ X $i s t h eco n s t an t coor d ina t e f i e l d$ \partial / \partial y^1$.

A local transverse section through $p$ for $X$ is a $C^{1}$ embedding $σ : Σ_{0} \to M$ of a neighbourhood $Σ_{0} \subseteq R^{n - 1}$ of $0$ with $σ (0) = p$ and $X (p)$ not contained in the tangent space $d σ (0) (R^{n - 1}) \subseteq T_{p} M$ . The image $Σ = σ (Σ_{0})$ is a $C^{1}$ hypersurface meeting the flow line through $p$ transversally at $p$ .

The straightening map built from $X$ and $σ$ is $$ \Phi : (-\varepsilon, \varepsilon) \times \Sigma_0 \to M, \qquad \Phi(t, y) = g^t(\sigma(y)). $$ The rectification theorem states that $Φ$ is a local $C^{1}$ diffeomorphism near $(0, 0)$ and that the inverse coordinates on a neighbourhood of $p$ rectify $X$ to $\partial / \partial y^{1}$ .

Counterexamples to common slips

The hypothesis $X (p) \neq = 0$ is required. At an equilibrium the flow fixes $p$ , every transverse section through $p$ collapses to a point under the flow, and the straightening map cannot be a diffeomorphism — the linearisation $d Φ (0)$ degenerates.
The choice of transverse section is not unique. Different sections produce different rectifying charts that agree on the field $\partial / \partial y^{1}$ but differ in the transverse coordinates. The theorem produces a rectification, not a canonical one.
The theorem is local. The flow may rectify on a small neighbourhood but the rectifying chart need not extend globally; an integral curve may return arbitrarily close to itself, producing topological obstructions that the local theorem ignores.
Smoothness matches the field. A $C^{k}$ vector field admits a $C^{k}$ rectifying chart by the same construction, since the flow $g^{t}$ inherits $C^{k}$ joint regularity from the smooth-dependence-on-initial-conditions theorem.

Key theorem with proof [Intermediate+]

Theorem (rectification theorem; Arnold ODE Ch.2 §7). Let $M$ be a smooth $n$ -manifold and $X$ a $C^{k}$ vector field on $M$ for some $k \geq 1$ . For every regular point $p$ of $X$ (i.e., $X (p) \neq = 0$ ) there is a local $C^{k}$ diffeomorphism $Φ$ from a neighbourhood of $0$ in $R^{n}$ onto a neighbourhood of $p$ in $M$ such that $$ \Phi_*(\partial / \partial y^1) = X. $$ Equivalently, there exist local coordinates $(y^{1}, \dots, y^{n})$ near $p$ in which $X = \partial / \partial y^{1}$ .

Proof. The argument has three steps. First, reduce to Euclidean space by a chart. Second, build the candidate diffeomorphism from the flow and a transverse section. Third, apply the inverse function theorem to verify it is a local diffeomorphism.

Step 1: reduction to Euclidean space. Choose a smooth chart $ψ : U \to V \subseteq R^{n}$ around $p$ with $ψ (p) = 0$ . The vector field $X$ pushes forward under $ψ$ to a $C^{k}$ vector field $X = ψ_{*} X$ on $V \subseteq R^{n}$ with $X (0) = ψ_{*} (X (p)) \neq = 0$ . If the theorem holds on $R^{n}$ for $X$ at $0$ , the composition with $ψ$ produces the rectifying chart for $X$ at $p$ on $M$ . Reduction complete.

Step 2: transverse section and straightening map. Work on $R^{n}$ with $X (0) = v \neq = 0$ . By a linear change of coordinates we arrange $v = e_{1}$ , the first standard basis vector. The hyperplane $Σ_{0} = {0} \times R^{n - 1} \subseteq R^{n}$ passes through the origin and is transverse to $v$ . The local flow $g^{t}$ of $X$ exists on a neighbourhood of $0$ for $∣ t ∣ < ε$ , is $C^{k}$ in $(t, x)$ jointly by the smooth-dependence-on-initial-conditions theorem, and satisfies $g^{0} = id$ , $\frac{d}{d t} g^{t} (x) = X (g^{t} (x))$ .

Define the straightening map $$ \Phi : (-\varepsilon, \varepsilon) \times \Sigma_0 \to \mathbb{R}^n, \qquad \Phi(t, y) = g^t(0, y), $$ where $y = (y^{2}, \dots, y^{n}) \in Σ_{0}$ identifies points of $Σ_{0}$ with their last $n - 1$ coordinates and $(0, y) = (0, y^{2}, \dots, y^{n}) \in R^{n}$ . The map $Φ$ is $C^{k}$ in $(t, y)$ jointly.

Step 3: inverse function theorem. Compute the differential of $Φ$ at $(0, 0)$ . For the $t$ -derivative, $$ \frac{\partial \Phi}{\partial t}(0, 0) = \frac{d}{dt}\Big|{t=0} g^t(0) = \widetilde X(0) = e_1. $$ For the $y^{j}$ -derivative with $j \geq 2$ , $$ \frac{\partial \Phi}{\partial y^j}(0, 0) = \frac{d}{ds}\Big|{s=0} g^0(0, 0, \ldots, s, \ldots, 0) = e_j, $$ where the $s$ sits in slot $j$ and $g^{0} = id$ . So the Jacobian matrix of $Φ$ at $(0, 0)$ is $$ D\Phi(0, 0) = [e_1 \mid e_2 \mid \cdots \mid e_n] = I_n, $$ the $n \times n$ identity. By the inverse function theorem, $Φ$ is a $C^{k}$ local diffeomorphism on some neighbourhood of $(0, 0)$ in $R \times Σ_{0}$ onto its image, an open neighbourhood of $0$ in $R^{n}$ .

It remains to identify $Φ_{*} (\partial / \partial t) = X$ in the rectified coordinates. By construction, $$ \Phi_*\left(\frac{\partial}{\partial t}\right)\bigg|{\Phi(t, y)} = \frac{d}{ds}\bigg|{s=0} \Phi(t + s, y) = \frac{d}{ds}\bigg|{s=0} g^{t+s}(0, y) = \widetilde X(g^t(0, y)) = \widetilde X(\Phi(t, y)). $$ So $\Phi*(\partial / \partial t) = \widetilde X $t h r o ug h o u tt h e ima g e . R e namin g$ t = y^1 $an d t h esec t i o n coor d ina t es$ y = (y^2, \ldots, y^n) $, t h ec ha r t$ \Phi^{-1} $p r o v i d es l oc a l coor d ina t es in w hi c h$ \widetilde X = \partial / \partial y^1 $. C o m p os in g w i t h t h eor i g ina l c ha r t$ \psi $co m pl e t es t h e p r oo f o n$ M $.$ \square$

Bridge. The rectification theorem builds toward the geometric language of ODE theory: rather than thinking of $\overset{x}{˙} = X (x)$ as a calculational object, the theorem reframes local ODE theory as the assertion that every regular point of a vector field has a neighbourhood diffeomorphic to a constant horizontal flow on $R^{n}$ . This is exactly the geometric repackaging of the existence-uniqueness-smooth-dependence triad: existence becomes the existence of the flow $g^{t}$ , uniqueness becomes injectivity of $Φ$ on a small enough neighbourhood, and smooth dependence becomes joint regularity of $Φ$ . The foundational reason the theorem works is exactly the smooth-dependence theorem of Picard-Lindelöf applied to the flow $g^{t} (σ (y))$ : differentiating in $t$ recovers the field, differentiating in $y$ at $t = 0$ recovers the embedding of the transverse section.

This pattern appears again in the constant-rank theorem of multi-variable calculus, where a map of locally constant rank is straightened to a linear projection, and in the local theory of foliations, where a non-singular distribution is straightened to a sub-bundle of coordinate vector fields. The central insight is that local geometric structures with a non-degenerate first jet have a canonical local model — and the canonical local model is computed by the inverse function theorem. Putting these together, rectification identifies the local theory of a non-vanishing vector field with the local theory of a constant vector field on $R^{n}$ : every question about regular points reduces to a question about the constant translation flow $y^{1} \mapsto y^{1} + t$ . The bridge is the recognition that local ODE theory is the geometric content of the inverse function theorem applied to the flow.

Exercises [Intermediate+]

Exercise 3 (medium, symbolic).

Compute global rectifying coordinates for the linear shear field $X = \partial_{x} + x \partial_{y}$ on $R^{2}$ . Verify that the rectification extends to all of $R^{2}$ .

Hint

The integral curves satisfy $\overset{x}{˙} = 1$ , $\overset{y}{˙} = x$ . Integrate: $x (t) = x_{0} + t$ , $y (t) = y_{0} + x_{0} t + t^{2} /2$ . Choose the transverse section to be the $y$ -axis $Σ = {(0, y_{0})}$ .

Answer

The flow starting at $(0, y_{0}) \in Σ$ is $g^{t} (0, y_{0}) = (t, y_{0} + t^{2} /2)$ . So the straightening map is $Φ (t, y_{0}) = (t, y_{0} + t^{2} /2)$ , with inverse $(x, y) \mapsto (x, y - x^{2} /2)$ . The rectifying coordinates are $y^{1} = x$ , $y^{2} = y - x^{2} /2$ , defined globally on $R^{2}$ since the inverse map is a polynomial diffeomorphism. In these coordinates, $\partial / \partial y^{1} = \partial_{x} + x \partial_{y} = X$ and $\partial / \partial y^{2} = \partial_{y}$ , confirming the rectification. The field $X$ never vanishes on $R^{2}$ since $\partial_{x}$ alone is non-zero in every component, so every point is regular and the global rectification is consistent. Rubric: full credit for the coordinates $y^{1} = x, y^{2} = y - x^{2} /2$ plus verification that they rectify the field.

Exercise 4 (medium, short-answer).

Explain why the rectification of $X = - y \partial_{x} + x \partial_{y}$ near $(1, 0)$ cannot be extended to a chart containing the origin. Identify the obstruction explicitly.

Hint

The origin is an equilibrium. The flow lines near $(1, 0)$ are circles around the origin; what happens to a transverse section as you flow around?

Answer

The origin is an equilibrium ( $X (0, 0) = 0$ ), so rectification fails there. Moreover, the flow lines of $X$ are circles centred at the origin, periodic with period $2 π$ . A rectifying chart near $(1, 0)$ uses the polar coordinates $(θ, r)$ as $(y^{1}, y^{2})$ , but $y^{1} = θ$ is multi-valued on any disk containing the origin and only single-valued on a simply-connected neighbourhood that excludes the origin. The local rectification cannot extend past angle $2 π$ without identifying $y^{1} = 0$ and $y^{1} = 2 π$ , and it certainly cannot reach the origin where the field vanishes. The obstruction is the combination of an equilibrium at the origin and a topologically nontrivial periodic flow surrounding it. Rubric: full credit for naming the equilibrium plus the periodicity issue.

Exercise 5 (medium, symbolic).

Let $X = (1 + x^{2}) \partial_{x}$ on $R$ . Show that $X$ has no equilibria, write down the flow, and give a rectifying coordinate change.

Hint

Integrate $d t = d x / (1 + x^{2})$ . The resulting integral is the arctangent.

Answer

$X$ is never zero on $R$ since $1 + x^{2} \geq 1 > 0$ , so every point is regular. The ODE $\overset{x}{˙} = 1 + x^{2}$ separates as $d t = d x / (1 + x^{2})$ , integrating to $t = arctan x - arctan x_{0}$ , so $x (t) = tan (t + arctan x_{0})$ . The flow blows up in finite time: $x (t) \to \infty$ as $t \to π /2 - arctan x_{0}$ . The rectifying coordinate is $y^{1} = arctan x$ , with inverse $x = tan y^{1}$ for $y^{1} \in (- π /2, π /2)$ . In this coordinate, $\partial / \partial y^{1} = (\partial x / \partial y^{1}) \partial_{x} = sec^{2} y^{1} \cdot \partial_{x} = (1 + tan^{2} y^{1}) \partial_{x} = (1 + x^{2}) \partial_{x} = X$ . The rectification is valid on the open interval, but cannot extend to the whole real line because the flow itself has only finite-time existence in either direction starting from each point. Rubric: full credit for the flow plus the arctangent rectifying coordinate.

Exercise 6 (medium, short-answer).

Show that if $(y^{1}, \dots, y^{n})$ are rectifying coordinates for $X$ near a regular point $p$ , then $y^{2}, \dots, y^{n}$ are first integrals of $X$ on the rectified neighbourhood.

Hint

A function $f$ is a first integral of $X$ if $X (f) = 0$ everywhere on the domain. Compute $X (y^{j})$ for $j \geq 2$ in the rectified coordinates.

Answer

In rectifying coordinates, $X = \partial / \partial y^{1}$ . For each $j \geq 2$ , $$ X(y^j) = \frac{\partial y^j}{\partial y^1} = 0 $$ since $y^{j}$ depends only on $y^{j}$ , not on $y^{1}$ . So $y^{2}, \dots, y^{n}$ are first integrals of $X$ . Geometrically: the level sets ${y^{j} = c_{j} : j = 2, \dots, n}$ are the integral curves of $X$ , each labelled by a constant value of $(y^{2}, \dots, y^{n})$ . The first $n - 1$ rectifying coordinates parametrise the local family of integral curves. Rubric: full credit for the computation plus the geometric interpretation as labels on integral curves.

Exercise 7 (hard, short-answer).

State and prove a parametric version of the rectification theorem: a $C^{k}$ family $X_{λ}$ of vector fields depending on a parameter $λ$ in a neighbourhood of $λ_{0}$ , with $X_{λ_{0}} (p) \neq = 0$ , admits a $C^{k}$ rectifying chart depending $C^{k}$ -ly on $λ$ near $λ_{0}$ .

Hint

Repeat the proof of the rectification theorem with the joint $C^{k}$ flow $g_{λ}^{t} (q)$ in place of $g^{t} (q)$ . The smooth-dependence theorem includes parameters.

Answer

Form the joint flow $g_{λ}^{t} (q)$ of $X_{λ}$ at parameter $λ$ ; by the smooth-dependence theorem with parameters, $g_{λ}^{t} (q)$ is $C^{k}$ in $(t, q, λ)$ jointly on a neighbourhood of $(0, p, λ_{0})$ . Build the parametrised straightening map $$ \Phi(t, y, \lambda) = g^t_\lambda(\sigma(y)) $$ where $σ$ is a transverse section to $X_{λ_{0}}$ at $p$ . At $(0, 0, λ_{0})$ , $D_{(t, y)} Φ$ is invertible by the same Jacobian computation as in the basic theorem (the Jacobian depends continuously on $λ$ , and is the identity at $λ_{0}$ ). By the parametric inverse function theorem, $Φ$ is a $C^{k}$ local diffeomorphism in $(t, y)$ for $λ$ near $λ_{0}$ , with $C^{k}$ -dependence on $λ$ . The rectifying coordinates of $X_{λ}$ are the inverse $Φ^{- 1}$ , which by construction sends $X_{λ}$ to $\partial / \partial y^{1}$ for every $λ$ . The transverse section remains transverse to $X_{λ}$ for $λ$ near $λ_{0}$ by continuity of the transversality condition. Rubric: full credit for the parametric flow construction plus the parametric inverse function theorem step.

Exercise 8 (hard, short-answer).

Use rectification to derive the method of characteristics for the first-order linear PDE $a (x, y) u_{x} + b (x, y) u_{y} = 0$ on an open subset of $R^{2}$ where $(a, b) \neq = (0, 0)$ .

Hint

Recognise the PDE as $X (u) = 0$ where $X = a \partial_{x} + b \partial_{y}$ is the characteristic vector field. Apply rectification to $X$ .

Answer

The PDE is $X (u) = 0$ for $X = a \partial_{x} + b \partial_{y}$ . By hypothesis $X$ never vanishes on the domain, so every point is regular and the rectification theorem applies locally. Choose rectifying coordinates $(y^{1}, y^{2})$ on a neighbourhood with $X = \partial / \partial y^{1}$ . The PDE becomes $$ \frac{\partial u}{\partial y^1} = 0, $$ whose general solution is $u (y^{1}, y^{2}) = f (y^{2})$ for any $C^{1}$ function $f$ of one variable. Translating back to $(x, y)$ coordinates, $u (x, y) = f (ϕ (x, y))$ where $ϕ = y^{2}$ is the transverse rectifying coordinate, which is a first integral of $X$ — that is, constant along the integral curves of $X$ . The integral curves of $X$ are the characteristics of the PDE; the general solution is a function constant along each characteristic. Rubric: full credit for identifying the characteristic field, applying rectification, and concluding the general-solution form.

Lean formalization [Intermediate+]

Mathlib has the local-flow infrastructure for ODEs (Mathlib.Analysis.ODE.PicardLindelof, Mathlib.Analysis.ODE.Gronwall) and the inverse function theorem (Mathlib.Analysis.Calculus.InverseFunctionTheorem), but does not package the rectification theorem as a named result. The intended formalisation reads schematically:

import Mathlib.Analysis.ODE.PicardLindelof
import Mathlib.Analysis.Calculus.InverseFunctionTheorem.FDeriv
import Mathlib.Geometry.Manifold.MFDeriv.Basic

/-- A local transverse section to a C^1 vector field at a regular point. -/
structure TransverseSection {E : Type*} [NormedAddCommGroup E]
    [NormedSpace ℝ E] (X : E → E) (p : E) where
  embedding : (Fin (Nat.pred (Module.finrank ℝ E))) → E
  base_point : embedding 0 = p
  transverse : sorry  -- X(p) ∉ range (d embedding 0)

/-- Rectification theorem: a C^1 vector field is locally diffeomorphic to
    the constant vector field e₁ near every regular point. -/
theorem rectificationTheorem
    {E : Type*} [NormedAddCommGroup E] [NormedSpace ℝ E] [FiniteDimensional ℝ E]
    (X : E → E) (hX : ContDiff ℝ 1 X) (p : E) (hp : X p ≠ 0) :
    ∃ (V : Set (ℝ × E)) (W : Set E) (hV : IsOpen V) (hW : IsOpen W)
      (Φ : ℝ × E → E),
      (0, p) ∈ V ∧ p ∈ W ∧
      ContDiffOn ℝ 1 Φ V ∧
      Set.BijOn Φ V W ∧
      ∀ q ∈ W, fderiv ℝ Φ (Φ.symm q) (1, 0) = X q :=
  sorry  -- straightening map Φ(t, y) = g^t(σ(y)), apply IFT

The proof gap is the named transverse-section structure, the construction of the straightening map $Φ (t, y) = g^{t} (σ (y))$ from the local flow, and the application of the inverse function theorem with the Jacobian-equals-identity calculation at the base point. Each piece is formalisable from existing Mathlib infrastructure but has not been packaged. The manifold version of the theorem (with $E$ replaced by a smooth manifold and the field a section of the tangent bundle) is a further packaging target, predicated on Mathlib's existing SmoothManifoldWithCorners framework.

Advanced results [Master]

Theorem (rectification on a smooth manifold). Let $M$ be a $C^{k}$ manifold and $X$ a $C^{k}$ vector field on $M$ . For every regular point $p$ of $X$ there is a $C^{k}$ chart $(Φ, V)$ at $p$ with $Φ : V \to U \subseteq R^{n}$ , $Φ (p) = 0$ , and $\Phi_ X = \partial / \partial y^1 $o n$ U $. T h ec ha r t i sco n s t r u c t e df r o m t h e l oc a l f l o w$ g^t $o f$ X $an d an y t r an s v er se$ C^k $h y p er s u r f a ce$ \Sigma $t h r o ug h$ p $v ia$ \Phi^{-1}(t, y) = g^t(\sigma(y))$.*

The manifold version reduces to the Euclidean version by composition with a smooth chart, exactly as in Step 1 of the proof of the basic theorem. The local flow $g^{t}$ exists on $M$ as a $C^{k}$ map on a neighbourhood of ${0} \times {p}$ in $R \times M$ by the manifold version of Picard-Lindelöf. The transverse hypersurface $Σ$ is an embedded $C^{k}$ submanifold of codimension one through $p$ meeting $X (p)$ transversally. The straightening map is the composition of the section embedding with the flow, and the inverse function theorem applied at $p$ produces the chart.

Theorem (first integrals near a regular point). Let $X$ be a $C^{k}$ vector field on $M$ ( $k \geq 1$ ) with $X (p) \neq = 0$ . There is a neighbourhood $V$ of $p$ on which there exist $n - 1$ independent $C^{k}$ first integrals $ϕ^{2}, \dots, ϕ^{n} : V \to R$ of $X$ , meaning $X (ϕ^{j}) = 0$ for every $j \in {2, \dots, n}$ and the differentials $d ϕ^{2}, \dots, d ϕ^{n}$ are linearly independent at every point of $V$ . The level sets ${ϕ^{2} = c_{2}, \dots, ϕ^{n} = c_{n}}$ are exactly the integral curves of $X$ on $V$ .

The proof is immediate from rectification: in rectifying coordinates the functions $ϕ^{j} = y^{j}$ for $j \geq 2$ are first integrals because $\partial y^{j} / \partial y^{1} = 0$ , and their differentials $d y^{j}$ are linearly independent. This is the maximal possible number of locally independent first integrals on $R^{n}$ : any further first integral would have to be a function of $ϕ^{2}, \dots, ϕ^{n}$ by the implicit function theorem and the linear independence of the differentials.

Theorem (relative rectification of a non-vanishing distribution). A $C^{k}$ involutive distribution $D$ of rank $r$ on $M$ is, near every point, of the form $D = span (\partial / \partial y^{1}, \dots, \partial / \partial y^{r})$ in suitable local coordinates. (This is the Frobenius theorem.) The rectification theorem is the rank-one special case: a single non-vanishing vector field is an involutive line distribution, and Frobenius reduces to the existence of a single straightening coordinate.

Frobenius reduces to rectification when $r = 1$ because every line distribution is involutive (Lie brackets vanish for rank-one distributions of sections proportional to a single vector field), and the rectification chart provides the straightening. Higher-rank Frobenius requires an additional argument: a sequence of rectifications, with each step rectifying one coordinate at a time and using the involutivity hypothesis to ensure the successive rectifications are compatible. The base case is the rectification theorem; the induction step is the involutivity-driven extension.

Theorem (rectification implies local uniqueness of integral curves). Let $X$ be a $C^{1}$ vector field on $M$ and $p$ a regular point. Any two integral curves of $X$ through $p$ agree on the intersection of their domains in a neighbourhood of $p$ .

In rectifying coordinates near $p$ the field is $\partial / \partial y^{1}$ , whose integral curves are the horizontal lines $y^{j} = c_{j}$ for $j \geq 2$ . Two integral curves through $p$ must lie on the same horizontal line and have the same parametrisation up to a time translation, hence agree wherever both are defined. The local rectification picture makes uniqueness of integral curves geometrically evident.

Theorem (method of characteristics for first-order PDEs; Arnold §31, Hörmander §8.1). Let $a^{1}, \dots, a^{n} : U \subseteq R^{n} \to R$ be $C^{1}$ functions with $(a^{1} (x), \dots, a^{n} (x)) \neq = 0$ on $U$ , and let $X = \sum_{i = 1}^{n} a^{i} \partial / \partial x^{i}$ . The general $C^{1}$ solution of the homogeneous first-order linear PDE $X (u) = 0$ on $U$ is locally of the form $u = f (ϕ^{2}, \dots, ϕ^{n})$ where $ϕ^{2}, \dots, ϕ^{n}$ are the $n - 1$ rectifying coordinates transverse to $X$ and $f$ is an arbitrary $C^{1}$ function of $n - 1$ variables. The integral curves of $X$ are the characteristics; the general solution is constant along each characteristic.

The proof is the application of rectification to $X$ followed by integration of $\partial u / \partial y^{1} = 0$ . The first-order PDE $X (u) = 0$ becomes the elementary equation $\partial u / \partial y^{1} = 0$ in rectifying coordinates, whose solutions are precisely the functions independent of $y^{1}$ . The method extends to inhomogeneous and quasilinear PDEs by introducing the dependent variable $u$ as an extra coordinate and rectifying in $n + 1$ dimensions.

Theorem (flow-equivariance of rectifying charts). Let $Φ, Φ^{'} : V \to W$ be two rectifying charts for $X$ at $p$ , both with $Φ (0) = Φ^{'} (0) = p$ and both sending $\partial / \partial y^{1}$ to $X$ . Then the transition map $Φ^{- 1} \circ Φ^{'} : V \to V$ has the form $(y^{1}, y^{2}, \dots, y^{n}) \mapsto (y^{1} + h (y^{2}, \dots, y^{n}), ψ (y^{2}, \dots, y^{n}))$ for some $C^{k}$ functions $h, ψ$ , where $ψ$ is a diffeomorphism of the transverse part.

Both charts identify $X$ with $\partial / \partial y^{1}$ , so the transition map commutes with the time-translation flow $(y^{1}, y) \mapsto (y^{1} + t, y)$ . A diffeomorphism commuting with horizontal translation must add an arbitrary function of the transverse coordinates to $y^{1}$ and act on the transverse coordinates by an arbitrary diffeomorphism. The freedom is parametrised by $(h, ψ)$ ; the transverse-section choice fixes a representative in the equivalence class.

Synthesis. Rectification is the foundational reason every non-equilibrium point of a smooth vector field has a canonical local model, and that model is the uniform translation flow on $R^{n}$ . The central insight is that the local flow $g^{t}$ combined with a transverse section $σ$ assembles into a diffeomorphism $Φ (t, y) = g^{t} (σ (y))$ whose Jacobian is the identity at the base point — and by the inverse function theorem, this is enough to make $Φ$ a local diffeomorphism that rectifies $X$ to $\partial / \partial y^{1}$ . Putting these together, rectification gives the geometric form of local ODE theory: existence-uniqueness-smooth-dependence is exactly the assertion that the flow plus a transverse section produces a rectifying chart, and every other local property of the field is read off from the constant-field picture in rectified coordinates. The bridge is the recognition that local geometric structures with a non-degenerate first jet — non-vanishing vector fields, full-rank linear maps, immersions of constant rank, involutive distributions — all have a canonical local model computed by the inverse function theorem applied at the base point.

The theorem identifies several pictures that look distinct at first inspection. The smooth-dependence-on-initial-conditions theorem of Picard-Lindelöf is identified with the joint $C^{k}$ regularity of $Φ (t, y) = g^{t} (σ (y))$ . The local uniqueness of integral curves is identified with the injectivity of $Φ$ on a small enough neighbourhood. The existence of $n - 1$ local first integrals is identified with the rectifying coordinates $y^{2}, \dots, y^{n}$ — these are the constants of motion that label the integral curves. The method of characteristics for first-order PDEs is identified with the rectification of the characteristic vector field followed by integration of the rectified equation. Rectification generalises to higher rank as the Frobenius theorem on involutive distributions, where a single straightening coordinate is replaced by $r$ coordinates spanning the distribution. The pattern recurs in the constant-rank theorem of multi-variable calculus, where a map of locally constant rank is straightened to a linear projection, and in symplectic geometry as the Darboux theorem, where a non-degenerate closed two-form is straightened to the standard symplectic form. Each of these is the inverse function theorem applied to a structure with a non-degenerate first jet at a point.

Full proof set [Master]

Theorem (rectification theorem), proof. Given in the Intermediate-tier section: reduce to $R^{n}$ by a chart, build the straightening map $Φ (t, y) = g^{t} (σ (y))$ from the local flow and a transverse section, compute that the Jacobian at the origin is the identity, apply the inverse function theorem to conclude $Φ$ is a local diffeomorphism, and identify $Φ_{*} (\partial / \partial t) = X$ by direct computation. $□$

Theorem (rectification on a smooth manifold), proof. Let $ψ : V_{0} \to V \subseteq R^{n}$ be a $C^{k}$ chart at $p$ with $ψ (p) = 0$ . The pushforward $X = ψ_{*} X$ is a $C^{k}$ vector field on $V$ with $X (0) = ψ_{*} (X (p)) \neq = 0$ since $ψ$ is a diffeomorphism. Apply the rectification theorem on $R^{n}$ to $X$ at $0$ , producing a rectifying chart $Φ : U_{0} \to U \subseteq R^{n}$ with $Φ_{*} (\partial / \partial y^{1}) = X$ . The composite $ψ^{- 1} \circ Φ$ is a $C^{k}$ chart on $M$ at $p$ that rectifies $X$ to $\partial / \partial y^{1}$ . The construction is functorial in the chart $ψ$ : a different choice of $ψ$ produces a rectifying chart related by a transition map of the form recorded in the flow-equivariance theorem. $□$

Theorem (first integrals near a regular point), proof. Let $(y^{1}, \dots, y^{n})$ be rectifying coordinates near $p$ . Define $ϕ^{j} = y^{j}$ for $j \in {2, \dots, n}$ . Since $X = \partial / \partial y^{1}$ in these coordinates, $$ X(\phi^j) = \frac{\partial y^j}{\partial y^1} = 0 $$ for each $j$ . The differentials $d ϕ^{j} = d y^{j}$ are linearly independent on the chart domain because the $d y^{j}$ are coordinate differentials. The level sets ${ϕ^{2} = c_{2}, \dots, ϕ^{n} = c_{n}}$ are the horizontal lines in rectified coordinates, which are exactly the integral curves of $\partial / \partial y^{1} = X$ . Any further first integral $ϕ$ on the chart would satisfy $\partial ϕ / \partial y^{1} = 0$ , hence $ϕ$ depends only on $(y^{2}, \dots, y^{n}) = (ϕ^{2}, \dots, ϕ^{n})$ ; the implicit function theorem expresses $ϕ$ as a $C^{k}$ function of $(ϕ^{2}, \dots, ϕ^{n})$ . $□$

Theorem (rectification implies local uniqueness), proof. Let $γ_{1}, γ_{2}$ be two integral curves of $X$ through $p$ with $γ_{1} (0) = γ_{2} (0) = p$ , defined on a common interval $(- δ, δ)$ . In rectifying coordinates $(y^{1}, \dots, y^{n})$ near $p$ , the field $X = \partial / \partial y^{1}$ has integral curves $t \mapsto (t + c^{1}, c^{2}, \dots, c^{n})$ . The condition $γ_{i} (0) = p$ corresponds to $γ_{i} (0) = (0, \dots, 0)$ in rectified coordinates, fixing $c^{j} = 0$ for all $j$ . So both curves are $t \mapsto (t, 0, \dots, 0)$ on a sufficiently small interval, and they coincide. The argument applies on any rectified neighbourhood of $p$ , hence the curves agree on the intersection of their domains restricted to a neighbourhood of $p$ . The local uniqueness extends to a global uniqueness statement on the maximal interval of existence by the connectedness argument standard in Picard-Lindelöf. $□$

Theorem (method of characteristics), proof. Let $X = \sum_{i} a^{i} \partial / \partial x^{i}$ be the non-vanishing characteristic field. By rectification, near any point of $U$ there are local coordinates $(y^{1}, \dots, y^{n})$ with $X = \partial / \partial y^{1}$ . The PDE $X (u) = 0$ becomes $\partial u / \partial y^{1} = 0$ , whose solutions are precisely the $C^{1}$ functions of $(y^{2}, \dots, y^{n})$ . Writing $ϕ^{j} = y^{j}$ for $j \geq 2$ (first integrals of $X$ ), the general local solution is $u (x) = f (ϕ^{2} (x), \dots, ϕ^{n} (x))$ for an arbitrary $C^{1}$ function $f$ of $n - 1$ variables. The integral curves of $X$ are the level sets ${ϕ^{2} = c_{2}, \dots, ϕ^{n} = c_{n}}$ in $U$ — the characteristics. The solution is constant along each characteristic because $f \circ (ϕ^{2}, \dots, ϕ^{n})$ is constant where $(ϕ^{2}, \dots, ϕ^{n})$ is constant. $□$

Theorem (flow-equivariance of rectifying charts), proof. Let $Φ, Φ^{'} : V \to W$ be two rectifying charts as in the statement. The composition $Ψ = Φ^{- 1} \circ Φ^{'} : V \to V$ is a $C^{k}$ diffeomorphism fixing the origin. Both charts identify $X$ with $\partial / \partial y^{1}$ , so $Ψ$ intertwines the horizontal translation flow: $Ψ \circ T_{t} = T_{t} \circ Ψ$ where $T_{t} (y^{1}, y) = (y^{1} + t, y)$ . Writing $Ψ (y^{1}, y) = (Ψ^{1} (y^{1}, y), Ψ^{⊥} (y^{1}, y))$ , the intertwining condition reads $Ψ^{1} (y^{1} + t, y) = Ψ^{1} (y^{1}, y) + t$ and $Ψ^{⊥} (y^{1} + t, y) = Ψ^{⊥} (y^{1}, y)$ . The first condition forces $Ψ^{1} (y^{1}, y) = y^{1} + h (y)$ for some $h$ . The second forces $Ψ^{⊥} (y^{1}, y) = ψ (y)$ independent of $y^{1}$ . The diffeomorphism property of $Ψ$ on $V$ forces $ψ$ to be a local diffeomorphism on the transverse part. $□$

Connections [Master]

Vector field 02.12.01. The rectification theorem is a structural statement about smooth vector fields on smooth manifolds: every regular point of a smooth vector field has a neighbourhood diffeomorphic to a constant horizontal flow on $R^{n}$ . The unit 02.12.01 introduces the vector field and its integral curves in a coordinate-free framework; rectification refines this picture by exhibiting the canonical local model at every non-equilibrium point.
Phase flow 02.12.02. The straightening map $Φ (t, y) = g^{t} (σ (y))$ is built from the local flow $g^{t}$ of the field combined with a transverse section. Without the flow there is no straightening map to construct; without rectification, the flow is just an existence-uniqueness fact rather than a geometric local model. The two units are dual viewpoints on local ODE theory: the flow is the dynamical object, rectification is the geometric repackaging of its smooth dependence.
Smooth manifold 03.02.01. The rectification theorem is most cleanly stated on a smooth manifold, where the rectifying chart $Φ^{- 1}$ is a coordinate chart in the smooth atlas. The reduction to Euclidean space by a chart is the standard manifold-theoretic move, and the rectifying coordinates produced are local smooth coordinates around the regular point. The unit 03.02.01 provides the framework in which the theorem is naturally stated and where the inverse function theorem and the smooth-dependence theorem are available.
Implicit and inverse function theorems 02.05.04. The proof of rectification reduces to the inverse function theorem applied to the straightening map at the base point, where the Jacobian is computed to be the identity. This is exactly the pattern of every local-normal-form theorem in differential topology: a structure with a non-degenerate first jet at a point admits a canonical local model whose existence is verified by an inverse-function-theorem argument on the appropriate candidate diffeomorphism.
Frobenius theorem on involutive distributions. Rectification is the rank-one case of Frobenius: a non-vanishing vector field is an involutive line distribution, and the rectifying chart is the Frobenius chart. The general Frobenius theorem on an involutive rank- $r$ distribution produces $r$ straightening coordinates, with the rank-one case reducing to ordinary rectification. The connection makes precise the sense in which Frobenius is "many rectifications at once".
First integrals / conserved quantities 02.12.12. Rectification is the local-existence theorem for first integrals in disguise: the $n - 1$ rectifying coordinates $y^{2}, \dots, y^{n}$ are functionally independent first integrals of $X$ near the regular point, with the level-set partition ${y^{2} = c_{2}, \dots, y^{n} = c_{n}}$ recovering the family of integral curves. The unit 02.12.12 takes this corollary as its main theorem, packages the maximality count $n - 1$ and the implicit-function-theorem-based functional-independence criterion, and extends to the Hamiltonian setting where Poisson-commutation ${F, H} = 0$ replaces $X (ϕ) = 0$ . The present unit is the geometric foundation; 02.12.12 is the dedicated conserved-quantity treatment with the Noether and Liouville-Arnold upgrades.

Historical & philosophical context [Master]

The geometric picture of a flow near a regular point goes back to Cauchy's 1820s lectures at the École Polytechnique on the existence theorem for ODEs (collected in the Œuvres complètes Série II Tome XI), where the polygon method exhibits local solutions as nearly-straight broken-line approximations to the flow ^{[Cauchy 1820s]}. Cauchy's framing was calculational rather than geometric, but the underlying observation — that a smooth flow near a regular point looks like uniform motion — is implicit in the polygon construction. Liouville's 1838 note on the variation of arbitrary constants (J. Math. Pures Appl. 3, 342-349) ^{[Liouville 1838]} introduced an early form of the one-parameter group concept under the language of "arbitrary constants", treating the flow as a family of diffeomorphisms parametrised by time. The geometric viewpoint emerged in Poincaré's 1881-1886 mémoire Sur les courbes définies par une équation différentielle (J. de Math. Pures et Appl.) ^{[Poincaré 1881-1886]}, which framed local and global ODE theory in terms of phase portraits, distinguishing regular points (where flows look straight) from singular points (where the local model is the linearisation), and introducing transverse sections as the natural geometric input to the local analysis.

The modern statement of rectification as a stand-alone theorem — separating the geometric content (existence of straightening coordinates) from the calculational content (the polygon-method or Picard-iteration proof of existence and uniqueness) — was crystallised by Arnold in Ordinary Differential Equations (Russian first edition 1971, English translation 1973) ^{[Arnold 1973]}. Arnold's Ch.2 §7 calls the result the flow-box theorem and presents it as the geometric form of the existence-uniqueness-smooth-dependence package, arguing that this is the right way to state local ODE theory once one is doing geometry rather than calculation. The same theorem appears in Hartman's Ordinary Differential Equations (1964) Ch. IV §1 ^{[Hartman 1964]} in the Banach-space generality, in Spivak's Calculus on Manifolds (1965) Ch.5 ^{[Spivak 1965]} as the smooth-manifold version, and in Lang's Differential Manifolds (1972) Ch. IV §2 ^{[Lang 1972]} with the rectifying chart $Φ (t, y) = g^{t} (σ (y))$ named explicitly. The theorem extends to first-order PDEs via the method of characteristics, treated in Hörmander's The Analysis of Linear Partial Differential Operators Vol. I (1983) §8.1 ^{[Hörmander 1983]} as the geometric foundation of the characteristic-curve method. Through Arnold's framing, rectification became one of the recognisable Russian-school local-normal-form theorems, alongside the Morse lemma, Darboux's theorem on symplectic manifolds, and the Frobenius theorem on involutive distributions — all instances of the same pattern: a structure with a non-degenerate first jet has a canonical local model verifiable by an inverse-function-theorem argument.

Bibliography [Master]

@book{Arnold1973ODE,
  author    = {Arnold, Vladimir I.},
  title     = {Ordinary Differential Equations},
  publisher = {MIT Press},
  year      = {1973},
  note      = {Translated from the 1971 Russian edition by Richard A. Silverman}
}

@book{Hartman1964ODE,
  author    = {Hartman, Philip},
  title     = {Ordinary Differential Equations},
  publisher = {John Wiley \& Sons},
  year      = {1964}
}

@book{Spivak1965CalculusManifolds,
  author    = {Spivak, Michael},
  title     = {Calculus on Manifolds: A Modern Approach to Classical Theorems of Advanced Calculus},
  publisher = {W. A. Benjamin},
  year      = {1965}
}

@book{Lang1972DifferentialManifolds,
  author    = {Lang, Serge},
  title     = {Differential Manifolds},
  publisher = {Addison-Wesley},
  year      = {1972}
}

@book{Hormander1983LPDO,
  author    = {H{\"o}rmander, Lars},
  title     = {The Analysis of Linear Partial Differential Operators I: Distribution Theory and Fourier Analysis},
  publisher = {Springer-Verlag},
  series    = {Grundlehren der mathematischen Wissenschaften},
  volume    = {256},
  year      = {1983}
}

@article{Poincare1881,
  author  = {Poincar{\'e}, Henri},
  title   = {M{\'e}moire sur les courbes d{\'e}finies par une {\'e}quation diff{\'e}rentielle},
  journal = {J. de Math{\'e}matiques Pures et Appliqu{\'e}es},
  volume  = {7},
  year    = {1881},
  pages   = {375--422},
  note    = {Continued in volumes 8 (1882), 1 (1885), 2 (1886).}
}

@article{Liouville1838,
  author  = {Liouville, Joseph},
  title   = {Note sur la th{\'e}orie de la variation des constantes arbitraires},
  journal = {J. Math. Pures Appl.},
  volume  = {3},
  year    = {1838},
  pages   = {342--349}
}

@book{Cauchy1820sLectures,
  author    = {Cauchy, Augustin-Louis},
  title     = {{\OE}uvres compl{\`e}tes, S{\'e}rie II Tome XI},
  publisher = {Gauthier-Villars},
  year      = {1913},
  note      = {Collected lectures at the {\'E}cole Polytechnique on the existence theorem for ODEs, originally delivered in the 1820s.}
}

@book{Strogatz2015NonlinearDynamics,
  author    = {Strogatz, Steven H.},
  title     = {Nonlinear Dynamics and Chaos: With Applications to Physics, Biology, Chemistry, and Engineering},
  publisher = {Westview Press},
  edition   = {2},
  year      = {2015}
}

@book{Teschl2012ODE,
  author    = {Teschl, Gerald},
  title     = {Ordinary Differential Equations and Dynamical Systems},
  publisher = {American Mathematical Society},
  series    = {Graduate Studies in Mathematics},
  volume    = {140},
  year      = {2012}
}

Prerequisites

02.05.04 pending
03.02.01 pending
02.12.01 pending
02.12.02 pending

Tier anchors

beginner: Arnold *Ordinary Differential Equations* Ch.1 §1 (the radioactive-decay flow as the prototype straight-line motion); Strogatz *Nonlinear Dynamics and Chaos* §6.0 informal phase-portrait reading near a regular point
intermediate: Arnold *Ordinary Differential Equations* Ch.2 §7 (flow-box theorem statement and the transverse-section construction); Spivak *Calculus on Manifolds* Ch.5 (rectification via the local flow)
master: Arnold *Ordinary Differential Equations* Ch.2 §7 + Ch.4 §31 (the flow-box theorem as the geometric form of existence-uniqueness-smooth-dependence); Hartman *Ordinary Differential Equations* Ch. IV §1 (smoothing-the-flow-by-the-transverse-section construction with full Banach-space generality); Lang *Differential Manifolds* Ch. IV §2 (flow-box theorem on a smooth manifold); Hörmander *The Analysis of Linear Partial Differential Operators* Vol. I §8.1 (rectification as the gateway to the method of characteristics for first-order PDEs)

References

TODO_REF
Arnold — Ordinary Differential Equations · Ch.2 §7 statement and the transverse-section construction; Ch.4 §31 smooth-dependence input
TODO_REF
Spivak — Calculus on Manifolds · Ch.5 rectification of a vector field via the local flow on a smooth manifold
TODO_REF
Hartman — Ordinary Differential Equations · Ch. IV §1 flow-box theorem under $C^1$ hypothesis with full Banach-space generality
TODO_REF
Lang — Differential Manifolds · Ch. IV §2 flow-box theorem on a smooth manifold; the diffeomorphism $\Phi(t, y) = g^t(\sigma(y))$ as the rectifying chart
TODO_REF
Hörmander — The Analysis of Linear Partial Differential Operators I · §8.1 method of characteristics for first-order PDEs as global rectification along characteristic curves
TODO_REF
Cauchy — Lectures at the École Polytechnique on the existence theorem for ODEs · Œuvres complètes Série II Tome XI, 1820s lectures on the polygon method and the local picture of solutions
TODO_REF
Liouville — Note sur la théorie de la variation des constantes arbitraires · J. Math. Pures Appl. 3 (1838) 342-349; early flow / one-parameter-group concept in the language of arbitrary constants
TODO_REF
Poincaré — Mémoire sur les courbes définies par une équation différentielle · J. de Math. Pures et Appl., 1881-1886; geometric phase-portrait viewpoint, regular versus singular points, transverse sections

Reviewer

TBD

Estimated time

beginner: 15m
intermediate: 35m
master: 70m