Chapter 1 gave us one variable evolving in time: $\dot{x} = f(x)$. Now we have two — a state $(x, y)$ — and time $t$ running alongside. The moment we stop treating $t$ as a background clock and start treating it as a third coordinate axis, something remarkable happens: the geometry of the ODE crystallises into a single differential 1-form. That form is the contact form. Everything in Books 3 and 4 lives inside it.
In Chapter 1 the state of the system was a single number $x \in \mathbb{R}$. The ODE $\dot{x} = f(x)$ 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: $(x, y) \in \mathbb{R}^2$. 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 $t$.
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 $\mathbb{R}^3$ 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 $(x_0, y_0)$ in the phase plane but at different times are now distinct curves in the extended space.
Here is where the geometry becomes beautiful. A first-order ODE $\dot{x} = f(x, y)$ — 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. Specifically, define:
A curve $(x(t), y(t))$ is a solution of the ODE if and only if it is tangent to the kernel of $\alpha$ — that is, if $\alpha(\dot{\gamma}) = 0$ along the curve $\gamma$.
The 1-form $\alpha$ is the contact form. It lives naturally on the 3-dimensional extended phase space $(x, y, t)$ — or, when the ODE depends on $t$, on the jet space $J^0(\mathbb{R}, \mathbb{R}) = \mathbb{R}^3_{(x, t, y)}$.
A contact structure on a 3-manifold $M$ is a maximally non-integrable smooth 2-plane field $\xi = \ker(\alpha)$, where $\alpha$ is a 1-form satisfying $\alpha \wedge d\alpha \neq 0$ everywhere.
The prototype: $M = \mathbb{R}^3_{(x, y, t)}$, $\alpha = dy - y'\, dx$, where $y' = dy/dx$ is the slope coordinate in the 1-jet space $J^1(\mathbb{R}, \mathbb{R})$.
The condition $\alpha \wedge d\alpha \neq 0$ is the non-integrability condition. It says the 2-plane field $\xi$ 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.
Let us be explicit about the dimension count. The state lives in $\mathbb{R}^2$ (two degrees of freedom: $x$ and $y$). Time is one additional dimension. So the extended phase space is $\mathbb{R}^2 \times \mathbb{R} = \mathbb{R}^3$.
This is precisely the 3-dimensional contact manifold where Book 3 was set. The helical attractor in Chapter 10 of Book 3 — with its three coordinates $(r, \theta, z)$ — is the same structure: a 2-dimensional state (radius $r$, angle $\theta$) 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 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. Going from 3D to 4D (Chapter 3) introduces the symplectic leaf. Going from 4D to 5D (Chapter 4) enters jet space $J^1$. Adding $t$ again at 5D reaches the 6D arena where GTCT acts. Each transition has a name and a theorem.
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 in the phase plane lifts to a helix: $(\cos t, -\sin t, t)$. The helix winds around the $t$-axis, advancing one full turn per period $2\pi$.
The contact form for this system is $\alpha = dy + x\, dx$ (since $f(x,y) = y$ gives $dy - y\, dx$, rewritten using the equation). The helical lift is tangent to $\ker(\alpha)$ at every point — you can verify: $\alpha(\dot{\gamma}) = \dot{y} + x\dot{x} = -x + x \cdot y$... which requires checking along the specific solution. The point is structural: the contact form is exactly the certificate that the curve is a solution.
Fig 2.1 — Left: phase plane (circles). Right: extended phase space (helices). Colour encodes time $t$. Each orbit in the phase plane lifts to a helix in 3D.
The operator chain $G = U \circ F \circ K \circ C$ from Book 3 does not disappear when we add a time axis — it reappears in the geometry of the contact form.
C (Compression) is the reduction to the 1-form $\alpha$: from the full ODE to the kernel distribution, discarding everything except the essential constraint.
K (Threshold) is the non-integrability condition $\alpha \wedge d\alpha \neq 0$: the moment the distribution refuses to integrate into a surface. This is the critical curvature $\kappa^*$ of contact geometry.
F (Fold) is the solution: the curve tangent to $\ker(\alpha)$. It builds the trajectory that respects the constraint.
U (Unfolding) is the projection back: from the helix in extended phase space down to the orbit in the phase plane — the visible output of the dynamics.
The same four operations that describe how a circadian clock keeps time (Book 3, Chapter 3) describe how an ODE generates its solutions in extended phase space. The contact form is the single object that makes this correspondence precise.
Near any point of a contact 3-manifold $(M, \ker\alpha)$, there exist local coordinates $(x, y, z)$ such that $\alpha = dz - y\, dx$.
Consequence: all contact 3-manifolds look the same locally. The only interesting geometry is global. This is why Chapter 3 (the helical attractor) is a genuinely 3-dimensional result — it is a global statement about the 3-manifold, not reducible to local coordinates.
We have the extended phase space $\mathbb{R}^3$ with a contact structure. Chapter 3 makes the manifold curved — instead of flat $\mathbb{R}^3$, the state space wraps into a genuine 3-manifold (the one from Book 3). The contact form gains curvature. Fixed points become Reeb orbits. The ODE becomes a flow on the manifold — the Reeb flow — and the attractor theorem of Book 3 is now a theorem about that flow.
We are not repeating Book 3. We are re-entering it from below, having built the floor it stands on.