When you cannot solve, you can still see
The four guides before this one were lucky. Growth and decay, Newton's cooling, mixing tanks, the logistic equation — each one separated cleanly and gave a tidy formula for the answer. But that luck is rare. The honest truth of this whole subject is that most differential equations have no closed-form solution at all; there is simply no formula in elementary functions to write down. So the real question becomes: if I can never solve y' = f(y), can I still say what eventually happens to a population, a temperature, an account balance? Astonishingly, often yes — and the tool is the phase line.
The trick works only for a special but enormously common shape: the autonomous equation y' = f(y), where the right-hand side depends on y alone and never explicitly on the time t. Almost every model in this rung is autonomous: dP/dt = k P, dP/dt = k P (1 - P/K), dT/dt = -k(T - T_room) all have a right side built only from the unknown itself. That one feature — no t on the right — is exactly what lets us forget about solving and instead read the future off a number line. (An equation with t on the right, like y' = y + sin t, does not get a clean phase line, because the rules of motion keep changing with time.)
Drawing the line: equilibria and the sign of f
Here is the whole construction. Draw a vertical number line for the variable y. First find the places where the system stands perfectly still: the values y* where f(y*) = 0. At such a point y' = 0, so y never moves — these are the equilibria, the constant equilibrium solutions y(t) = y*. For the logistic equation y' = r y (1 - y/K), setting the right side to zero gives y* = 0 (extinction) and y* = K (the carrying capacity). Mark those two dots on the line; they cut it into intervals.
Now comes the one genuine idea. On each interval between equilibria, f(y) has a single sign — it cannot change sign without passing through a zero, and the zeros are exactly the equilibria you already marked. So just test one point in each interval. Where f(y) > 0, the solution has y' > 0, so y is increasing: draw an arrow pointing up. Where f(y) < 0, y is decreasing: draw an arrow pointing down. This is the entire method of sign analysis of f(y), and the decorated number line you end up with is the phase line.
Logistic: y' = r y (1 - y/K) with r > 0, K > 0 y f(y) sign arrow meaning ----------------------------------------------------------- y > K f < 0 down falls back to K --- K --------------------------- equilibrium (sink) ----- 0 < y < K f > 0 up climbs toward K --- 0 --------------------------- equilibrium (source) --- y < 0 f < 0 down (runs away; non-physical)
Reading fate from arrows: stable, unstable, semistable
The arrows tell you everything about the long run. Look at an equilibrium and ask which way its neighbours flow. If both arrows point toward it — up from below, down from above — then any nearby start is pulled in, and the equilibrium is stable: a sink (or attractor). If both arrows point away, the smallest nudge sends the solution fleeing, and it is unstable: a source. For the logistic equation, y = K is a sink and y = 0 is a source — which is exactly why a small founding population grows toward the carrying capacity and stays there. This is the whole content of the stability of equilibria, read off without a single integral.
There is a third possibility that beginners often miss: the arrows can agree — both pointing up, or both down — across an equilibrium. Then solutions approach from one side and are pushed away on the other; the point is semistable. This happens precisely when f touches zero without crossing, like the double root in y' = (y - 2)^2. Stable to the left, unstable to the right. Naming all three honestly matters, because a real harvesting or threshold model often sits exactly at such a delicate semistable case, where the fate of the system hangs on which side of the line you start.
A faster test, and where it quietly fails
There is a shortcut to the up/down arrows that avoids testing intervals: the derivative test. At an equilibrium y*, look at the slope of f there. If f'(y*) < 0, then f is positive just below y* and negative just above, so arrows point inward — stable. If f'(y*) > 0, arrows point outward — unstable. This is the simplest possible case of linearization about an equilibrium: near y*, the equation behaves like y' = f'(y*) (y - y*), whose tiny deviations decay like e^(f'(y*) t) when f'(y*) is negative and blow up when it is positive.
But be honest about the gap in this shortcut. When f'(y*) = 0 the linear test says nothing at all — the linear part is flat, and the verdict hangs on the higher-order shape of f. The semistable case y' = (y - 2)^2 is exactly this: f'(2) = 0, yet the point is decidedly semistable. The lesson, which echoes a deep theorem you will meet later for higher dimensions, is that linearization is trustworthy only at a hyperbolic equilibrium — one where the linear slope is nonzero. At a degenerate point, throw away the shortcut and go back to the actual sign of f on each side.
When a knob changes the picture: a first taste of bifurcation
The real power shows when your model carries a parameter — a fishing quota, a drug dose, a control gain. Take a logistic fishery with constant harvesting: y' = r y (1 - y/K) - h, where h is the catch rate. For small h the right side still has two zeros, a high stable population and a low unstable threshold. As you crank h up, those two equilibria slide toward each other, collide, and vanish. Past that critical h the parabola never reaches zero, no equilibrium survives, and y crashes to zero no matter where you started — the fishery collapses suddenly, not gradually.
That qualitative jolt — equilibria appearing, merging, or changing stability as a parameter passes a threshold — is a bifurcation, and the phase line is the cleanest possible window onto it. Stack the phase lines for many values of h side by side and you get a bifurcation diagram, a map of fate versus parameter. This is your first encounter with bifurcation, the seam where smooth tuning produces an abrupt change in behaviour, and where 'stability without solving' pays off most: you would never have spotted the sudden collapse by staring at any single formula.
- Write the equation as y' = f(y) and confirm it is autonomous — no t on the right.
- Solve f(y) = 0 for every equilibrium y*; mark each as a dot on a vertical y-line.
- Test the sign of f(y) in each interval between dots; draw an up arrow where f > 0, a down arrow where f < 0.
- Classify each equilibrium from its surrounding arrows: both inward is a stable sink, both outward an unstable source, agreeing arrows a semistable point.
- If a parameter is present, repeat for several of its values and watch for equilibria colliding or flipping — that is a bifurcation.