A different question about a solution
On the previous rung you learned to take a system x' = A x and grind out a formula for it — eigenvalues, eigenvectors, a clean x(t) you could plug a number into. That is precise, and it is powerful. But it answers only one kind of question: *what is the solution equal to at time t?* This rung asks a different, often more useful question: *what does the solution do?* Does it settle down, blow up, circle forever, or run off to one side? You can answer all of that without ever writing the formula — and for the vast majority of real equations, where no formula exists, this is the only way left.
The shift is from arithmetic to geography. Instead of a table of values we will draw a *map* of the plane, and on that map every solution becomes a curve you can see at a glance. We restrict ourselves, for now, to two unknowns and to systems whose right-hand side does not depend on time on its own — an autonomous planar system, written x' = f(x), y' = g(x, y). 'Autonomous' is the key word: the rules of the flow are nailed to the *place*, not to the clock, and that is exactly what lets one fixed picture describe motion that starts at any moment.
The vector field: a wind over the plane
Here is the central reframe. Read x' = f(x) not as an equation to be solved but as an *instruction posted at every point*. Stand at any location (x, y) in the plane; the system hands you a single arrow — the velocity (x', y') = (f, g) — telling you which way to step next and how fast. Do this at every point and you have painted the plane with arrows. That field of arrows is the velocity vector field, and it is the whole differential equation laid bare as a picture: a steady wind you can feel at any spot before you have committed to any particular journey.
You met a thin cousin of this before. For a single first-order equation y' = f(x, y) you drew a slope field — little tilted dashes giving the slope dy/dx at each point. The vector field is the richer, 2-D version: each arrow carries not just a direction but a *length*, encoding speed, and it lives in a plane of two genuine unknowns rather than the x-versus-y graph of a single function. Where slopes were tilts to follow, arrows are gusts to ride. Same spirit, one dimension up, and far more to see.
Trajectories: solutions you can see
Now release a particle into that wind. Drop it at a starting point and let it always move exactly along the arrow beneath it; the curve it traces is a trajectory — also called an orbit — and it *is* a solution, drawn rather than written. A solution (x(t), y(t)) is a moving point; its trajectory is the path that point sweeps out as t runs. Notice what we threw away: the trajectory shows the *route* but not the *timetable*. Two particles can follow the very same curve, one dawdling and one sprinting, and on the phase plane they leave the identical track.
One fact governs the whole picture and is worth holding onto: trajectories never cross. Because the wind assigns exactly one arrow to each point, a particle arriving there has exactly one way to leave — there is no fork in the road. (This is uniqueness from the existence-and-uniqueness rung, now wearing geometric clothes; it can genuinely fail where f is not smooth enough to be Lipschitz, but for the linear and smooth systems here it holds.) So trajectories tile the plane like the grain in a piece of wood — they may crowd together, sweep around, even spiral, but two distinct ones can never truly meet. The lone exception is a point where the arrow is zero, and that special place is the hero of the next section.
Equilibria and the flow
What happens where the wind dies? A point where the velocity is zero, f(x, y) = 0 and g(x, y) = 0 at once, is an equilibrium point — a place of perfect stillness. A particle placed exactly there never moves, because there is no arrow to push it; the constant function sitting at that point is itself a solution, and a single dot is its whole trajectory. Equilibria are the skeleton of the entire portrait. Find them first, always: everything else in the plane is organized around how the wind behaves in their neighbourhoods.
Collect *all* the trajectories together and you have the flow of the system — the grand bookkeeping rule that takes any starting point and tells you where it has drifted after time t. The flow is the system's destiny machine: hand it a where-you-are-now and it returns a where-you-end-up. Near an equilibrium the flow reveals its personality, and that personality comes in a short menu of types — does the wind drain *into* the still point, gust *out* of it, swirl *around* it, or do something split? Naming and reading those types is the work of the guides ahead.
Here is the roadmap so the names are not a surprise when they land. For a linear system the lone equilibrium is the origin, and it falls into one of a handful of shapes: a node, where every nearby trajectory marches straight in or straight out; a saddle, which pulls in along one direction while flinging out along another; a spiral, where trajectories wind in or out while rotating; and a center, the delicate case of closed loops that circle forever, neither growing nor shrinking. The very last guide of the rung — the trace-determinant plane — packs all of these into a single diagram you read off two numbers of the matrix A.
The phase portrait: every future at once
Now assemble everything into one drawing. Mark the equilibria, sketch a representative handful of trajectories with little arrowheads showing the direction of travel, and you have a phase portrait — the signature image of this entire subject. Its power is breathtaking once you feel it: a single static picture displays the future of *every* possible initial condition simultaneously. Pick any starting dot, follow the curve through it, and you have read off that solution's whole fate without solving a thing. One sketch answers infinitely many what-ifs.
Two honest cautions keep this from becoming magical thinking. First, the portrait is qualitative: it shows shape and direction faithfully but throws away the timetable, so it cannot tell you *when* a particle reaches a given spot, only that it heads there. Second — and this is the great bridge to the nonlinear world — for a *nonlinear* system you study each equilibrium by replacing the curvy wind near it with the best straight-line wind, its linearization, and reading the matching linear portrait. That swap is trustworthy only at a hyperbolic equilibrium, where the real part of every eigenvalue is nonzero; right on a center, the linear loops can be a mirage that the true nonlinear flow quietly spirals through. Honesty about that boundary is what separates a real understanding from a pretty cartoon.
system: x' = f(x, y) (autonomous)
y' = g(x, y)
vector field: at each point (x,y), draw arrow (f, g) -> the wind
trajectory: release a point, follow the arrows -> one solution
(trajectories never cross; route, not timetable)
equilibrium: f = 0 AND g = 0 -> arrow is zero
types: node | saddle | spiral | center
phase portrait = equilibria + sample trajectories + direction arrows
= the future of EVERY initial condition, in one pictureWhere this rung is taking you
You now hold the four words the rest of the rung is built from — field, trajectory, equilibrium, portrait. The plan from here is to make each precise and then sketchable by hand. Next we zoom in on a single equilibrium and ask exactly how the flow behaves in its immediate neighbourhood; after that we name the full cast — node, saddle, spiral, center — and learn the eigenvalue fingerprint of each. The trace-determinant plane then compresses that whole classification into one master diagram, and the final guide turns all of it into a hand-drawing recipe.
Carry one sentence out of this guide above all others: a solution is not just a formula, it is a *path through a wind*, and the phase portrait shows every path at once. That single mental image is what makes the qualitative theory work — and it is precisely the image that survives when the formulas run out, which, for almost every equation worth caring about in the wild, they eventually do.