01 / 21

Variation of
Parameters

A powerful method to find particular solutions of non-homogeneous linear ODEs: no guessing required, works when undetermined coefficients cannot.

Differential Equations

What Is Variation of Parameters?

Given a non-homogeneous linear ODE such as \( y'' + p(t)\,y' + q(t)\,y = g(t) \), we already know how to find the complementary solution \( y_c \). Variation of Parameters lets us build a particular solution \( y_p \) by allowing the constants in \( y_c \) to become functions of \( t \).

The Big Idea: Constants Become Functions

If the complementary solution is \( y_c = c_1 y_1 + c_2 y_2 \), we replace the constants \( c_1, c_2 \) with unknown functions \( u_1(t), u_2(t) \):

We then determine \( u_1'(t) \) and \( u_2'(t) \) by solving a system of equations derived from the original ODE. This is the heart of the method.

Why Does This Matter?

Universal
Works for any continuous forcing function \( g(t) \): no restriction to polynomials, exponentials, or sines.
Systematic
No guessing: the formulas for \( u_1' \) and \( u_2' \) follow directly from the Wronskian and \( g(t) \).
Extendable
Generalizes naturally to higher-order ODEs and even to systems of differential equations.

The Wronskian Determinant

The Wronskian of two solutions \( y_1, y_2 \) is the determinant that measures their linear independence:

If \( y_1 \) and \( y_2 \) form a fundamental set, then \( W \neq 0 \) on the interval of interest. This guarantees the method works.

Core Formulas

For the standard-form equation \( y'' + p(t)y' + q(t)y = g(t) \), the derivatives of the parameter functions are:

Flowchart: The Method

1. Standard Form: divide by \( a(t) \)
2. Solve homogeneous: \( y_1, y_2 \)
3. Compute Wronskian \( W \)
4. Find \( u_1' \) and \( u_2' \) via formulas
5. Integrate: \( u_1, u_2 \to y_p = u_1 y_1 + u_2 y_2 \)

Variation of Parameters vs. Undetermined Coefficients

Variation of Parameters Undetermined Coefficients
Applies to Any continuous \( g(t) \) Only specific forms of \( g(t) \)
Coefficients Can be variable: \( p(t), q(t) \) Must be constant
Guessing? No: systematic Yes: guess the form
Difficulty Integrals may be hard Algebra-heavy but routine
When to use \( g(t) = \sec t, \ln t, \frac{1}{t}, \dots \) \( g(t) = e^{at}, \sin(bt), t^n, \dots \)

How to Apply the Method

1
Write in standard form. Ensure the leading coefficient of \( y'' \) is \(1\). If not, divide the entire equation by it.
2
Find \( y_1 \) and \( y_2 \). Solve the associated homogeneous equation to get a fundamental set of solutions.
3
Compute the Wronskian. Calculate \( W = y_1 y_2' - y_2 y_1' \).
4
Find \( u_1' \) and \( u_2' \). Use \( u_1' = \frac{-y_2\,g}{W} \) and \( u_2' = \frac{y_1\,g}{W} \).
5
Integrate. Find \( u_1 = \int u_1'\,dt \) and \( u_2 = \int u_2'\,dt \). Then \( y_p = u_1 y_1 + u_2 y_2 \).
6
General solution. Combine: \( y = y_c + y_p = c_1 y_1 + c_2 y_2 + y_p \).

Example 1: Setup

Solve \( y'' + y = \sec t \).

Example 1: Solution

Example 2: Variable Forcing

Solve \( y'' - 2y' + y = \dfrac{e^t}{t} \), for \(t > 0\).

Example 2: Completing the Solution

Example 3: Third-Order Glimpse

For a third-order equation with solutions \( y_1, y_2, y_3 \), the particular solution is \( y_p = u_1 y_1 + u_2 y_2 + u_3 y_3 \), where the \( u_k' \) are found using a \(3 \times 3\) Wronskian system:

Each \( u_k' \) is obtained by Cramer's rule: replace the \( k \)-th column with \( \begin{pmatrix} 0 \\ 0 \\ g(t) \end{pmatrix} \) and divide by \( W \). The pattern extends to any order \( n \).

Common Mistakes

✕ Forgetting Standard Form
If the ODE is \( 2y'' + 4y' + 2y = g(t) \), you must divide by 2 first. The formulas assume the leading coefficient is \(1\).
✕ Sign Errors in \( u_1' \)
Remember: \( u_1' \) has a negative sign: \( u_1' = \frac{-y_2 g}{W} \). Dropping the minus flips the entire particular solution.
✕ Wrong Wronskian
Computing \( W = y_1 y_2 - y_1' y_2' \) instead of \( y_1 y_2' - y_2 y_1' \). Always use the correct cross-multiplication.

Summary Reference

Component Formula / Role
Standard Form \( y'' + p(t)y' + q(t)y = g(t) \)
Complementary Solution \( y_c = c_1 y_1 + c_2 y_2 \)
Wronskian \( W = y_1 y_2' - y_2 y_1' \)
\( u_1'(t) \) \( \dfrac{-y_2\,g}{W} \)
\( u_2'(t) \) \( \dfrac{y_1\,g}{W} \)
Particular Solution \( y_p = u_1 y_1 + u_2 y_2 \)
General Solution \( y = y_c + y_p \)

Question 1 of 5

In variation of parameters for \( y'' + p(t)y' + q(t)y = g(t) \), what must the leading coefficient of \( y'' \) equal before applying the formulas?

Question 2 of 5

If \( y_1 = e^{2t} \) and \( y_2 = e^{-t} \), what is the Wronskian \( W(y_1, y_2) \)?

Question 3 of 5

For \( y'' + y = \sec t \) with \( y_1 = \cos t,\; y_2 = \sin t \), what is \( u_2'(t) \)?

Question 4 of 5

Which forcing function requires variation of parameters (i.e., undetermined coefficients cannot handle it)?

Question 5 of 5

In the formula \( u_1' = \dfrac{-y_2\,g}{W} \), why is the Wronskian \( W \) guaranteed to be nonzero?