Differential Equations › Section B · First order differential equations › Topic B.8
Equilibrium solutions and stability
For an equation whose rate depends only on the current state, the constant solutions organise everything else. Find them, decide which attract and which repel, and you can predict the fate of every solution without solving the equation.
Autonomous equations
An equation of the form \(y' = f(y)\), with no explicit \(t\), is called autonomous: the rate of change depends only on the current state, not on the clock. Populations, cooling bodies and many chemical reactions are modelled this way. An equilibrium solution is a constant solution \(y = y^*\), and it exists exactly where \(f(y^*) = 0\), because a constant has derivative 0. The Calculus topics "Slope fields and Euler's method" and "The logistic equation and population models" met equilibria and phase lines. This topic turns them into a complete classification, including the borderline cases and what happens when a parameter changes.
The phase line
Draw a vertical \(y\)-axis and mark the equilibria. Between neighbouring equilibria, \(f\) is continuous and never zero, so it keeps one sign. Test one value in each gap: where \(f > 0\), draw an arrow up, since solutions increase; where \(f < 0\), draw an arrow down. This one-line diagram is the phase line. It works because, when \(f\) and \(f'\) are continuous, non-constant solutions of an autonomous equation are monotonic: a solution cannot cross an equilibrium, by uniqueness, so it moves steadily in one direction, approaching an equilibrium or heading to \(\pm\infty\).
Three kinds of equilibrium
An equilibrium is classified by what happens to solutions that start near it. It is asymptotically stable, often just called stable, if solutions starting close to it on both sides approach it as \(t \to \infty\): the arrows on both sides point toward it. It is unstable if solutions starting close to it on both sides move away: both arrows point away. It is semi-stable if solutions approach from one side and move away on the other: both arrows point the same way. A ball at the bottom of a bowl, on the top of a hill, and on a ledge partway down a slope are useful pictures of the three cases.
Worked example: a cubic
Find and classify the equilibria of \(y' = y^3 - 9y\). 1. Factor: \(f(y) = y(y - 3)(y + 3)\), so the equilibria are \(-3\), 0 and 3. 2. Test the gaps: \(f(-4) = -28 < 0\), \(f(-1) = 8 > 0\), \(f(1) = -8 < 0\) and \(f(4) = 28 > 0\). 3. Arrows: down below \(-3\), up between \(-3\) and 0, down between 0 and 3, up above 3. 4. Classify: arrows point toward 0 from both sides, so 0 is stable; they point away from \(-3\) and from 3, so both are unstable. 5. Consequence: every solution starting strictly between \(-3\) and 3 tends to 0, and every other non-equilibrium solution runs off to \(\pm\infty\).
The derivative test
Near an equilibrium, the linear approximation gives \(f(y) \approx f'(y^*)(y - y^*)\). So the gap \(u = y - y^*\) approximately obeys \(u' = f'(y^*)u\), whose solutions are \(u = u_0e^{f'(y^*)t}\). If \(f'(y^*) < 0\), small gaps shrink and \(y^*\) is stable; if \(f'(y^*) > 0\), they grow and \(y^*\) is unstable. For the cubic, \(f'(y) = 3y^2 - 9\) gives \(f'(0) = -9\) and \(f'(\pm 3) = 18\), agreeing with the phase line. If \(f'(y^*) = 0\), the test is inconclusive: the equilibrium could be of any type, and you must go back to the signs of \(f\).
Worked example: a semi-stable equilibrium
Classify the equilibria of \(y' = (2 - y)(y - 5)^2\). 1. Equilibria: \(y = 2\) and \(y = 5\). 2. Signs: \(f(0) = 50 > 0\), \(f(4) = -2 < 0\) and \(f(6) = -4 < 0\). The squared factor never changes sign, so only the factor \(2 - y\) flips the sign, at 2. 3. At 2 the arrows point up from below and down from above: stable. The derivative test agrees, since \(f'(2) = -9\). 4. At 5 both arrows point down: solutions above 5 fall toward it, while solutions just below fall away from it. So 5 is semi-stable. 5. Note \(f'(5) = 0\), so the derivative test could not have decided this case.
Reading long-run behaviour
The phase line answers the question most models ask: where does the solution with a given \(y(0)\) end up? Locate \(y(0)\) on the line. If it is an equilibrium, the solution stays there. Otherwise follow the arrow: the solution moves monotonically toward the next equilibrium in that direction, or to \(\pm\infty\) if there is none. It never reaches that equilibrium in finite time, because of uniqueness. For \(y' = (2 - y)(y - 5)^2\), a start at \(y(0) = 7\) decreases toward 5, while a start at \(y(0) = 4.9\) decreases toward 2. Two nearby starting values can therefore have very different fates.
Bifurcations
Many models contain a parameter, and equilibria can appear, merge or vanish as it changes. This is called a bifurcation. Take a fish stock with \(P' = 0.8P\left(1 - \frac{P}{500}\right) - H\), harvested at \(H\) fish per year. A growth term \(rP(1 - \frac{P}{K})\) peaks at \(\frac{rK}{4}\) when \(P = \frac{K}{2}\), here 100 fish per year at \(P = 250\). With \(H = 60\), the equilibria are about 92, unstable, and 408, stable: the stock settles near 408 if it starts above 92 and collapses if it starts below. As \(H\) rises to 100 the two merge into one semi-stable equilibrium at 250, and for \(H > 100\) there are none, so every population declines to zero.
Common mistakes to avoid
First, swapping the classifications: an equilibrium is stable when nearby arrows point toward it, not merely because \(f\) is positive somewhere. Second, trusting the derivative test when \(f'(y^*) = 0\); check the signs instead, as the semi-stable example required. Third, testing one side only and missing a semi-stable equilibrium. Fourth, applying a phase line to an equation that contains \(t\) explicitly, such as \(y' = y - t\). Fifth, claiming a solution reaches a stable equilibrium at some finite time; it only approaches it. Sixth, forgetting equilibria where \(f\) has a repeated factor.
Key takeaways
For an autonomous equation \(y' = f(y)\), equilibria are the zeros of \(f\). A phase line, with arrows from the sign of \(f\) between equilibria, classifies each one: stable if both arrows point in, unstable if both point out, semi-stable if both point the same way. The derivative test says \(f'(y^*) < 0\) means stable and \(f'(y^*) > 0\) unstable, and is silent when \(f'(y^*) = 0\). Non-constant solutions are monotonic and approach the next equilibrium in the arrow's direction or \(\pm\infty\). Changing a parameter can create or destroy equilibria in a bifurcation.
Flashcards
Flashcard
y′ = f(y): the rate depends only on y
Test your understanding
Question 1 of 5
How are the equilibria of y′ = y² − 4y classified?
Question 2 of 5
What kind of equilibrium is y = 1 for y′ = (y − 1)²?
Question 3 of 5
For y′ = cos y, what does the derivative test say about the equilibrium y = π/2?
Question 4 of 5
For y′ = y³ − 9y, what happens to the solution with y(0) = 2 as t → ∞?
Question 5 of 5
A stock obeys P′ = 0.5P(1 − P/800) − H. Above what harvest rate H does the stock collapse from every starting size?
