Difficulty: Intermediate | Prerequisites: First-Order ODEs study notes, basic linear algebra (determinants)
Higher-order ODEs extend everything from first-order methods into equations involving y'', y''', and beyond. The core idea shifts from direct integration to finding characteristic roots, and the solution splits into two pieces: the homogeneous solution (from the left side) and the particular solution (from the right side). This is where the course gets substantially harder, and understanding the root cases is essential for everything that follows.
Higher-order ODEs (second order and above) with constant coefficients boil down to solving a characteristic polynomial for its roots, then writing the general solution based on whether those roots are real and distinct, repeated, or complex. For non-constant coefficients of the Euler-Cauchy type, a substitution y = x^m gives you the same three-case structure. When the right side is non-zero, you add a particular solution found either by educated guessing (undetermined coefficients) or by a systematic integral-based process (variation of parameters).
Characteristic equation (auxiliary equation)
The polynomial you get after substituting y = e^(λx) into a constant-coefficient ODE and dividing out e^(λx). For ay'' + by' + cy = 0, it is aλ² + bλ + c = 0. Think of it as converting a calculus problem into an algebra problem.
Constant coefficients
An ODE where the coefficients a, b, c in front of y'', y', y are plain numbers, not functions of x.
Euler-Cauchy equation
An ODE of the form x²y'' + bxy' + cy = 0, where the coefficients are powers of x matched to the order of derivative. The substitution y = x^m reduces it to a polynomial in m.
Wronskian
A determinant built from a set of functions and their derivatives, used to test whether solutions are linearly independent. If W ≠ 0, the solutions form a valid fundamental set.
Homogeneous solution (y_h)
The general solution to the equation with the right side set to zero. This captures the "natural behaviour" of the system.
Particular solution (y_p)
One specific solution to the full non-homogeneous equation. It accounts for the forcing term r(x) on the right side.
General solution (y_g)
The sum y_g = y_h + y_p. The homogeneous part carries the arbitrary constants; the particular part does not.
Method of undetermined coefficients (MoUC)
A technique where you guess the form of y_p based on the shape of r(x), plug it in, and solve for the unknown coefficients. Only works when r(x) is an exponential, polynomial, sine/cosine, or a combination.
Variation of parameters (VOP)
A more general technique that works for any r(x). You replace the constants in y_h with unknown functions u_i(x) and solve a system of equations to find them.
For equations of the form ay'' + by' + cy = 0, where a, b, c are constants.
Set y = e^(λx). Derive twice: y' = λe^(λx), y'' = λ²e^(λx).
Substitute. The equation becomes aλ²e^(λx) + bλe^(λx) + ce^(λx) = 0.
Divide by e^(λx). Since e^(λx) is never zero, you get the characteristic equation: aλ² + bλ + c = 0.
Solve for roots. Use the quadratic formula. The nature of the roots determines the solution form:
Case 1: Two distinct real roots (λ₁ ≠ λ₂)
y = c₁ e^(λ₁ x) + c₂ e^(λ₂ x)
Case 2: Complex roots (λ = α ± iβ)
y = c₁ e^(αx) cos(βx) + c₂ e^(αx) sin(βx)
Case 3: Repeated real root (λ₁ = λ₂ = λ)
y = c₁ e^(λx) + c₂ x e^(λx)
The extra factor of x in Case 3 is needed because the two solutions must be linearly independent, and e^(λx) alone would only give one.
For equations of the form x²y'' + bxy' + cy = 0 (where b and c are constants but the coefficients involve powers of x).
Set y = x^m. Derive: y' = mx^(m−1), y'' = m(m−1)x^(m−2).
Substitute. x² · m(m−1)x^(m−2) + bx · mx^(m−1) + cx^m = 0, which simplifies to m(m−1)x^m + bmx^m + cx^m = 0.
Divide by x^m. Require x > 0. The indicial equation is m(m−1) + bm + c = 0, or equivalently m² + (b−1)m + c = 0.
Solve for roots. The same three cases apply, but with x^m in place of e^(λx):
Case 1: Distinct real roots (m₁ ≠ m₂)
y = c₁ x^(m₁) + c₂ x^(m₂)
Case 2: Complex roots (m = α ± iβ)
y = c₁ x^α cos(β ln x) + c₂ x^α sin(β ln x)
Case 3: Repeated root (m₁ = m₂ = m)
y = c₁ x^m + c₂ ln(x) x^m
Notice the parallel: where constant coefficients use e^(λx) and xe^(λx), Euler-Cauchy uses x^m and ln(x)x^m. The ln(x) plays the role that x plays in the repeated-root case for constant coefficients.
The Wronskian tests whether a set of solutions is linearly independent (and therefore forms a valid fundamental set).
For n functions f₁, f₂, ..., f_n, build the matrix where row k contains the (k−1)-th derivatives of each function. The Wronskian is the determinant of this matrix.
Row 0: f₁(x), f₂(x), ..., f_n(x)
Row 1: f₁'(x), f₂'(x), ..., f_n'(x)
...
Row n−1: f₁^(n−1)(x), f₂^(n−1)(x), ..., f_n^(n−1)(x)
W ≠ 0: The functions are linearly independent. You have a valid fundamental set.
W = 0: The functions may be linearly dependent (on the interval in question). Your proposed solutions do not form a complete set.
The same approach extends beyond second order. For an n-th order equation with constant coefficients, the characteristic polynomial has degree n.
Case 1: All roots distinct. y = c₁e^(λ₁x) + c₂e^(λ₂x) + ... + c_n e^(λ_n x)
Case 2: A root λ repeated k times. That root contributes k terms: c₁e^(λx) + c₂xe^(λx) + c₃x²e^(λx) + ... + c_k x^(k−1)e^(λx). Each additional repetition multiplies in another power of x.
Case 3: Complex roots. Two distinct complex conjugate pairs α₁ ± iβ₁ and α₂ ± iβ₂ give four terms: c₁e^(α₁x)cos(β₁x) + c₂e^(α₁x)sin(β₁x) + c₃e^(α₂x)cos(β₂x) + c₄e^(α₂x)sin(β₂x). If a complex pair is repeated, multiply in powers of x, just as with repeated real roots: xc₃e^(αx)cos(βx) + xc₄e^(αx)sin(βx).
Used when the equation has the form ay'' + by' + cy = r(x) and r(x) is an exponential, polynomial, sine/cosine, or a combination of these.
Step 1: Solve the homogeneous part. Find y_h by solving ay'' + by' + cy = 0 using the characteristic equation.
Step 2: Guess the particular solution. The guess is based on the form of r(x):
Exponential r(x) = e^(kx): Guess y_p = Ae^(kx).
Polynomial r(x) = ax² + bx + c: Guess y_p = Ax² + Bx + C (include all powers down to the constant).
Trigonometric r(x) = sin(kx) or cos(kx): Guess y_p = A sin(kx) + B cos(kx). Always include both sin and cos, even if only one appears in r(x).
Combination (addition): Add the individual guesses together.
Combination (multiplication): Include all cross-products (e.g. if r(x) = xe^(2x), guess y_p = (Ax + B)e^(2x)).
Bump rule: If any term of your guess already appears in y_h, multiply the entire guess by x (or x² if it appears twice, etc.) until there is no overlap.
Step 3: Derive and plug in. Compute y_p' and y_p'', substitute into the original equation.
Step 4: Solve for coefficients. Match coefficients on both sides. Assemble y_g = y_h + y_p.
A general method that works for any r(x), not just exponentials, polynomials and trig functions. It is more involved than MoUC.
Find y_h. Solve the homogeneous equation to get y_h = c₁f(x) + c₂g(x).
Set up y_p. Replace the constants with unknown functions: y_p = u₁(x)f(x) + u₂(x)g(x).
First derivative with constraint. Derive y_p. To simplify, impose the condition u₁'f(x) + u₂'g(x) = 0. This makes y_p' = u₁f'(x) + u₂g'(x).
Second derivative. Derive again: y_p'' = u₁'f'(x) + u₁f''(x) + u₂'g'(x) + u₂g''(x).
Substitute and simplify. Plug y_p, y_p', y_p'' into ay'' + by' + cy = r(x). Because f and g are homogeneous solutions, many terms cancel, leaving a system:
u₁'f(x) + u₂'g(x) = 0
u₁'f'(x) + u₂'g'(x) = r(x)/a
Solve the system. Use Cramer's rule or substitution to find u₁' and u₂'.
Integrate. Integrate u₁' and u₂' to get u₁(x) and u₂(x). No constants of integration are needed here (those are already in y_h).
Assemble. y_g = y_h + y_p. Apply initial conditions if given.
\text{Characteristic equation: } a\lambda^2 + b\lambda + c = 0\text{Euler-Cauchy indicial equation: } m(m-1) + bm + c = 0\text{Wronskian: } W(f_1, \ldots, f_n) = \det\begin{pmatrix} f_1 & \cdots & f_n \\ f_1' & \cdots & f_n' \\ \vdots & & \vdots \\ f_1^{(n-1)} & \cdots & f_n^{(n-1)} \end{pmatrix}\text{General solution (non-homogeneous): } y_g = y_h + y_pStudents often forget the bump rule in MoUC. If your guess for y_p contains a term that already appears in y_h, the guess will produce zero when substituted. Multiplying by x fixes this.
In Euler-Cauchy equations, students sometimes use y = e^(λx) instead of y = x^m. The substitution must match the equation type: constant coefficients use exponentials, Euler-Cauchy uses powers of x.
When applying variation of parameters, students sometimes add constants of integration when integrating u₁' and u₂'. Those extra constants are redundant because they are already captured in y_h.
A common error is confusing the structure of the complex-root solution for Euler-Cauchy with the constant-coefficient version. Euler-Cauchy uses x^α cos(β ln x), not e^(αx) cos(βx).
⚠️ The three root cases for constant coefficients are tested on every exam. You must be able to write the general solution from the roots without hesitation.
⚠️ MoUC guessing rules are a favourite exam topic. Expect a question where r(x) partially overlaps with y_h, requiring the bump rule.
⚠️ Variation of parameters often appears as a longer problem worth more marks. The system of equations and the integration steps are where marks are lost.
⚠️ Euler-Cauchy problems are less common but reliably appear at least once. The indicial equation is quick to set up if you know the form.
Fill in the blank: the characteristic equation for 3y'' + 2y' – y = 0 is ___. (3λ² + 2λ – 1 = 0)
True or false: for the Euler-Cauchy equation x²y'' + 5xy' + 4y = 0, you set y = e^(λx). (False, set y = x^m.)
Fill in the blank: if the characteristic roots are λ = 2 ± 3i, the general solution involves e^(2x) multiplied by ___ and ___. (cos(3x) and sin(3x))
True or false: MoUC works when r(x) = ln(x). (False, MoUC only handles exponentials, polynomials, and sine/cosine. Use VOP instead.)
Fill in the blank: y_g = ___ + ___. (y_h + y_p)
Q: Solve y'' – 5y' + 6y = 0.
A: Characteristic equation: λ² – 5λ + 6 = 0, giving λ = 2 and λ = 3 (distinct real roots). General solution: y = c₁e^(2x) + c₂e^(3x).
Q: For y'' + 4y = 3e^(2x), find the form of y_p using MoUC.
A: The homogeneous solution is y_h = c₁cos(2x) + c₂sin(2x) (roots λ = ±2i). Since e^(2x) does not appear in y_h, guess y_p = Ae^(2x). No bump needed.
Q: The indicial equation for an Euler-Cauchy problem gives m = 3 ± 2i. Write the general solution.
A: y = c₁ x³ cos(2 ln x) + c₂ x³ sin(2 ln x).
Q: Why would you choose VOP over MoUC for y'' + y = sec(x)?
A: MoUC only handles exponentials, polynomials, and sine/cosine on the right side. sec(x) does not fit any of those forms, so variation of parameters is required.
The homogeneous solutions found here become the building blocks for systems of ODEs (Doc 3), where eigenvalues of a matrix play the same role as characteristic roots. Laplace transforms (Doc 3) provide an alternative route to the same solutions, particularly useful when initial conditions are given. Numerical methods such as Euler's method (Doc 4) approximate these solutions when an analytic answer is impractical.
Related Terms / Search Tags
second order ODE, characteristic equation, auxiliary equation, constant coefficients, Euler-Cauchy, Cauchy-Euler, indicial equation, Wronskian, linear independence, fundamental set, undetermined coefficients, MoUC, variation of parameters, VOP, particular solution, homogeneous solution, bump rule, repeated roots, complex roots, MAT 2388, higher order differential equations