Flows, orbits, and phase portraits

Let f be continuously differentiable on an open subset U of Euclidean space. The autonomous equation x dot equals f(x) has a unique local solution through every initial state. Write its value after elapsed time t as phi subscript t of x. Uniqueness says that evolving for s and then for t agrees with evolving for s plus t, whenever all the required solution segments exist. The identity at time zero and this composition law make solution curves into a local flow.

Autonomous means that the velocity depends only on state, not on the external clock. A periodically forced oscillator needs a forcing phase added to its state before it becomes autonomous; dropping that phase can make projected curves cross. Distinct geometric orbits of a unique autonomous flow cannot cross at a shared state. If two parametrized solutions reach that same state at different times, uniqueness identifies their future and past portions after shifting the clock, rather than creating two competing velocity choices.

The orbit through a point is the set of states visited over its maximal existence interval. Equilibria are singleton orbits where f vanishes. In one dimension, the sign of f supplies arrows between equilibria, and uniqueness prevents a solution from passing through an equilibrium in finite time and then leaving it. A positive scalar multiple of f has the same oriented geometric trajectories, while generally changing their speed. A time trace and a phase portrait therefore answer different questions about the same model.

A set is positively invariant if every solution starting there stays there for all future times for which the model is considered. Full invariance additionally includes backward evolution. A bounded interval whose boundary arrows point inward is a useful positive trapping region, but checking a picture at a handful of boundary points is not a proof for a higher-dimensional boundary. Long-time accumulation is described by an omega-limit set; compact containment supplies convergent subsequences, while an escaping orbit can have no finite accumulation point at all.

Even a polynomial vector field need not generate a flow for every real time. The scalar equation x dot equals x squared has solutions that blow up in finite forward time from positive initial states. Local existence and uniqueness are fully compatible with this failure of completeness. A standard continuation result says a maximal solution cannot stop at a finite time while remaining in a compact subset of U. It must instead escape every such compact subset, through unbounded growth or approach to the domain boundary.

This distinction matters in numerical work. A plot that stops at a finite axis limit may show a plotting choice, a solver failure, or actual mathematical blow-up. A trajectory that seems to settle near a stationary point may simply move very slowly. Check invariance, sign information, and existence estimates alongside the computation. Phase-space coordinates also matter: a second-order scalar equation requires both position and velocity as state, because position alone does not determine which direction the system will move next.