From One Variable to Two
In Chapter 1 the state of the system was a single number x ∈ ℝ. The ODE ẋ = f(x) told us how that number changed. The phase portrait was a line — a 1-dimensional picture with arrows pointing left or right depending on the sign of f.
Now suppose the system requires two numbers to describe its state: (x, y) ∈ ℝ². Think of a pendulum: x is the angle, y is the angular velocity. Or an electrical circuit: x is charge, y is current. The ODE becomes a pair:
The phase portrait is now a 2-dimensional picture — the phase plane. Each point (x, y) has an arrow attached to it, pointing in the direction (f, g). A solution is a curve in this plane that is everywhere tangent to those arrows.
This is already richer than 1D. Limit cycles can exist. Spirals can wind in or out. Two fixed points can be connected by a special curve — a heteroclinic orbit — that takes infinite time to traverse. But we are not stopping at the phase plane. We are adding t.
Promoting Time to a Coordinate
In System (2.1), time t is invisible. It is the parameter along the solution curves, but it does not appear as an axis. The phase plane only shows where the system goes, not when.
The extended phase space makes time visible. We add a third axis and consider the space ℝ³ with coordinates (x, y, t). A solution of System (2.1) is now a curve in this 3-dimensional space — it traces out a path (x(t), y(t), t) that moves upward in t as time advances.
The phase plane tells you the shape of the dynamics. The extended phase space tells you the shape and the timing simultaneously. Nothing is lost; everything is made explicit.
The projection of an extended-phase-space curve back onto the (x, y) plane recovers the usual phase portrait. But the lift to (x, y, t) space carries strictly more information: two solutions that pass through the same point (x₀, y₀) in the phase plane but at different times are now distinct curves in the extended space.
The Contact Form
Here is where the geometry becomes beautiful. A first-order ODE ẋ = f(x, y) — or equivalently dy/dx = f(x, y) — defines, at every point of the extended space (x, y, t), a preferred direction: the tangent to any solution through that point. A collection of preferred directions is called a distribution. We can encode this distribution as the kernel of a differential 1-form:
The contact form (prototype) associated to the ODE is α = dy − f(x, y) dx. A curve γ is a solution of the ODE if and only if it is tangent to the kernel of α — that is, α(γ̇) = 0 along γ.
The 1-form α lives naturally on the 3-dimensional extended phase space (x, y, t) — or, when the ODE depends on t, on the jet space J⁰(ℝ, ℝ) = ℝ³. The structure it defines has a name:
A contact structure on a 3-manifold M is a maximally non-integrable smooth 2-plane field ξ = ker(α), where α is a 1-form satisfying α ∧ dα ≠ 0 everywhere. The prototype: M = ℝ³, α = dy − y′ dx, where y′ = dy/dx is the slope coordinate in the 1-jet space J¹(ℝ, ℝ).
The condition α ∧ dα ≠ 0 is the non-integrability condition. It says the 2-plane field ξ cannot be tangent to any surface — not even locally. This is what makes contact geometry different from foliation theory, and what makes it the right language for ODEs: a solution touching two points of the plane field cannot "rest" in a surface between them; it must twist. That twist is the geometry of the ODE.
Why 2D+t Is Really 3D
Let us be explicit about the dimension count. The state lives in ℝ² (two degrees of freedom: x and y). Time is one additional dimension. So the extended phase space is ℝ² × ℝ = ℝ³.
This is precisely the 3-dimensional contact manifold where Book 3 was set — and the arena of Chapter 10 of this volume. The helical attractor of Chapter 10 — with its three coordinates (r, θ, z) — is the same structure: a 2-dimensional state (radius r, angle θ) evolving in a time-like direction z. Book 3 arrived at the contact 3-manifold from biology. Book 4 arrives at it from the ODE. Same room, two doors.
The staircase. The dimensional ladder of this book is not an escalator — each step changes the geometry, not just the number of axes. Going from 2D to 2D+t (= 3D) introduces contact structure (this chapter, made global in Chapter 3). Going from 3D to 4D introduces the symplectic leaf. Going from 4D to 5D enters jet space J¹. Adding t again at 5D reaches the 6D arena where GTCT acts. Each transition has a name and a theorem.
Worked Example: The Linear Oscillator
Take the simplest 2D system: the harmonic oscillator.
In the phase plane, the solutions are circles centred at the origin: (x(t), y(t)) = (cos t, −sin t), or any rotation thereof. Every solution is periodic; there are no fixed points except (0, 0), and it is a centre — neither attracting nor repelling.
In the extended phase space (x, y, t), each circular orbit lifts to a helix: γ(t) = (cos t, −sin t, t). The helix winds around the t-axis, advancing one full turn per period 2π.
The contact certificate for this system is cleanest in the rescaled form. Along solutions, dy/dx = ẏ/ẋ = −x/y, so the prototype form is α = dy + (x/y) dx; multiplying by y gives the equivalent form α̃ = x dx + y dy, with the same kernel away from y = 0. The verification is one line:
α̃ = ½ d(x² + y²) = ½ d(r²) // kernel = circles of constant radius
The geometry is transparent: α̃ is half the differential of r², so its kernel consists of directions along which the radius does not change — exactly the circular orbits. The contact form is the certificate that the curve is a solution; the helical lift carries that certificate into three dimensions, where the timing of the motion becomes part of the picture.