01 / 20
\(\lambda\)

Parameter
Sensitivity

Discover how small changes in a parameter can cause qualitative shifts in the behavior of differential equations: from stable to unstable, from decay to growth.

Differential Equations

What Is a Parameter?

In a differential equation like \(\frac{dy}{dt} = \lambda y\), the variable \(y\) changes over time, but \(\lambda\) is a parameter: a constant that shapes the equation's behavior without being solved for.

Variable
The unknown function we solve for, e.g. \(y(t)\)
Parameter
A fixed constant that controls shape, e.g. \(\lambda\), \(k\), \(\mu\)

What Is Parameter Sensitivity?

Parameter sensitivity asks: if you nudge a parameter by a tiny amount, does the solution change gradually or does it change drastically? When a small shift in \(\lambda\) flips the entire qualitative picture, the system is sensitive at that value.

\(t\) \(y\) \(\lambda > 0\) \(\lambda = 0\) \(\lambda < 0\) \(y_0\)

Why Does It Matter?

Parameter sensitivity appears everywhere in science and engineering. A tiny change in a physical constant can mean the difference between a system that self-corrects and one that spirals out of control.

Ecology
A birth rate crossing a threshold shifts a population from decline to explosion
Engineering
A damping coefficient dropping below zero turns stable vibrations into resonance
Medicine
A drug dosage parameter crossing a level flips tumor growth from shrinking to expanding

Equilibria and Stability

An equilibrium is a constant solution \(y^*\) where \(\frac{dy}{dt} = 0\). The key question: if we perturb slightly away from \(y^*\), does the system return (stable) or diverge (unstable)?

1
Find equilibria: Set \(f(y, \lambda) = 0\) and solve for \(y^*\).
2
Test stability: Evaluate \(\frac{\partial f}{\partial y}\) at \(y^*\). If it is negative, the equilibrium is stable.
3
Track the parameter: As \(\lambda\) changes, watch whether \(\frac{\partial f}{\partial y}\big|_{y^*}\) crosses zero.

The Core Idea: Bifurcation

A bifurcation is the precise parameter value where the number or stability of equilibria changes. At a bifurcation point \(\lambda = \lambda_0\), the system's qualitative portrait shifts fundamentally.

\(\lambda\) \(y^*\) \(\lambda_0\) No equilibria Stable Unstable

Quantitative vs. Qualitative Changes

Not every change in a parameter is equally important. The crucial distinction: does the character of the solution change, or just its numbers?

Quantitative Change Qualitative Change
What shifts Numerical values of solutions Type, number, or stability of solutions
Example Decay rate gets faster Decay becomes growth
Phase portrait Arrows stretch or shrink Arrows reverse direction
Bifurcation? No Yes

How to Analyze Sensitivity

1
Write the ODE: Express the system as \(\frac{dy}{dt} = f(y, \lambda)\) with the parameter \(\lambda\) explicit.
2
Find equilibria: Solve \(f(y^*, \lambda) = 0\) for \(y^*\) in terms of \(\lambda\).
3
Linearize: Compute \(\frac{\partial f}{\partial y}\big|_{y^*}\) and express it as a function of \(\lambda\).
4
Identify critical values: Find \(\lambda\) where the derivative is zero or where equilibria appear/disappear.
5
Sketch the bifurcation diagram: Plot \(y^*\) vs. \(\lambda\), marking stable (solid) and unstable (dashed) branches.

Example 1: \(\frac{dy}{dt} = \lambda y\)

The simplest case. There is one equilibrium: \(y^* = 0\). The stability depends entirely on the sign of \(\lambda\).

\(\lambda < 0\)
\(y^* = 0\) is stable. Solutions decay to zero exponentially.
\(\lambda = 0\)
\(\frac{dy}{dt} = 0\) everywhere. Every point is an equilibrium: critical case.
\(\lambda > 0\)
\(y^* = 0\) is unstable. Solutions grow exponentially away from zero.

Example 2: Logistic Harvesting

Consider a population model with harvesting: \(\frac{dy}{dt} = y(1 - y) - h\), where \(h \geq 0\) is the harvest rate.

\(y\) \(\frac{dy}{dt}\) \(h = 0\) \(h = 0.15\) \(h = 0.25\) 0

Harvesting: Bifurcation Diagram

Plotting the equilibria \(y^*\) against the harvest rate \(h\) reveals the full picture. The equilibria satisfy \(y^*(1 - y^*) = h\), giving \(y^* = \frac{1 \pm \sqrt{1 - 4h}}{2}\).

\(h\) \(y^*\) \(h = \frac{1}{4}\) Stable branch Unstable branch No equilibria 0.5

Example 3: Transcritical Bifurcation

Consider \(\frac{dy}{dt} = \lambda y - y^2\). There are always two equilibria: \(y^* = 0\) and \(y^* = \lambda\). But their stability swaps as \(\lambda\) crosses zero.

\(\lambda\) \(y^*\) \(y^*=0\) stable \(y^*=0\) unstable \(y^*=\lambda\) stable \(y^*=\lambda\) unstable

Types of Bifurcations

Type What Happens Prototype ODE
Saddle-Node Two equilibria appear or vanish \(\frac{dy}{dt} = \lambda - y^2\)
Transcritical Two equilibria swap stability \(\frac{dy}{dt} = \lambda y - y^2\)
Pitchfork One equilibrium splits into three \(\frac{dy}{dt} = \lambda y - y^3\)

Reading the Phase Line

A phase line shows the direction of flow on the \(y\)-axis. Arrows point in the direction \(y\) moves. Equilibria are dots. When a parameter changes and arrows reverse: that is a qualitative change.

\(\lambda = -1\) Stable \(\lambda = 0\) Critical No flow No flow \(\lambda = +1\) Unstable

Common Mistakes

✕ Confusing Variables and Parameters
Students treat \(\lambda\) as a variable to solve for. Remember: \(\lambda\) is a fixed constant you analyze, not an unknown.
✕ Ignoring Stability
Finding equilibria is not enough. You must also determine whether each equilibrium is stable or unstable by checking the sign of \(\frac{\partial f}{\partial y}\).
✕ Missing the Critical Value
Jumping to conclusions about behavior without finding the exact \(\lambda\) where the qualitative change occurs. Always solve for the bifurcation point.

Question 1 of 5

For the ODE \(\frac{dy}{dt} = \lambda y\), at what value of \(\lambda\) does a qualitative change in the equilibrium at \(y^* = 0\) occur?

Question 2 of 5

In the harvesting model \(\frac{dy}{dt} = y(1 - y) - h\), what is the critical harvest rate \(h\) at which equilibria disappear?

Question 3 of 5

In a transcritical bifurcation, what happens to the equilibria as the parameter crosses the critical value?

Question 4 of 5

For \(\frac{dy}{dt} = f(y, \lambda)\), the equilibrium \(y^*\) is stable when:

Question 5 of 5

Consider \(\frac{dy}{dt} = \lambda y - y^3\). For \(\lambda > 0\), how many stable equilibria does this system have?