Finding particular solutions by intelligent guessing
By the end of this lesson, you will be able to:
For second-order linear ODEs with constant coefficients and special forcing functions, we can find the general solution as a sum:
where:
The key idea: we guess the form of \(y_p\) based on what \(g(t)\) looks like, then substitute into the full equation to find the unknown coefficients.
This method only works for constant-coefficient equations with special forcing functions (polynomials, exponentials, sines, cosines, or products/sums of these). For general functions, use Variation of Parameters instead.
The method of undetermined coefficients emerged in the 17th century as mathematicians realized that certain forcing functions naturally suggest particular solution forms. Rather than deriving these forms from first principles each time, we simply guess and verify—a principle of elegant mathematical economy.
"O you who have believed, seek help through patience and prayer. Indeed, Allah is with the patient."
The method of undetermined coefficients works by matching the form of \(g(t)\) with an appropriate trial form for \(y_p\). We study four main cases:
Case 1: Polynomial Forcing
If \(g(t) = P_n(t)\) (polynomial of degree \(n\)), try:
Exception: If \(1\) is a root of the characteristic equation (i.e., \(c=0\)), multiply by \(t\). If \(1\) is a repeated root, multiply by \(t^2\).
Case 2: Exponential Forcing
If \(g(t) = K e^{\alpha t}\), try:
Exception: If \(e^{\alpha t}\) appears in \(y_c\), multiply by \(t\). If \(te^{\alpha t}\) also appears in \(y_c\), multiply by \(t^2\).
Case 3: Trigonometric Forcing
If \(g(t) = K_1 \cos(\omega t) + K_2 \sin(\omega t)\), try:
Exception: If \(\cos(\omega t)\) and \(\sin(\omega t)\) appear in \(y_c\), multiply by \(t\).
Case 4: Superposition (Sum/Product)
If \(g(t) = g_1(t) + g_2(t)\), find \(y_{p1}\) for \(g_1\) and \(y_{p2}\) for \(g_2\), then:
Always check: does your trial form overlap with \(y_c\)? If yes, multiply by \(t\) (or \(t^2\) for repeated roots). Ignoring this leads to linearly dependent solutions and failure of the method.
| \(g(t)\) | First Guess | If Resonance? |
|---|---|---|
| Polynomial degree \(n\) | \(A_n t^n + \cdots + A_0\) | Multiply by \(t\) (or \(t^2\)) |
| \(e^{\alpha t}\) | \(Ae^{\alpha t}\) | Use \(Ate^{\alpha t}\) (or \(At^2e^{\alpha t}\)) |
| \(\cos(\omega t)\) or \(\sin(\omega t)\) | \(A\cos(\omega t) + B\sin(\omega t)\) | Use \(t(A\cos + B\sin)\) |
| \(g_1(t) + g_2(t)\) | \(y_{p1} + y_{p2}\) | Combine with care |
The principle is simple: if a function of a certain form (e.g., \(e^{\alpha t}\)) is a solution to \(ay'' + by' + cy = 0\), then \(te^{\alpha t}\) might be needed for \(y_p\). By building in these multiplicities from the start, we avoid contradictions and find a unique solution.
A linear constant-coefficient system \(ay'' + by' + cy = g(t)\) acts like a filter or amplifier on the input signal \(g(t)\). The particular solution \(y_p\) represents how the system processes the input. The system either:
| Input \(g(t)\) | Particular Solution \(y_p\) | Gain Interpretation |
|---|---|---|
| \(e^{\alpha t}\) | \(Ae^{\alpha t}\) | Scalar gain \(A\); system multiplies the exponential |
| \(\sin(\omega t)\) | \(A\sin(\omega t) + B\cos(\omega t)\) | Amplitude change plus phase shift |
| \(t^n\) (polynomial) | Polynomial of degree \(n\) | Polynomial structure preserved; coefficients scaled |
| \(e^{rt}\) (where \(r\) is a root) | \(Ate^{rt}\) or \(At^2e^{rt}\) | Resonance! Response grows linearly or quadratically |
When \(g(t)\) is itself a solution of the homogeneous equation (that is, when the forcing frequency matches a natural frequency of the system), we have resonance. The system can't simply amplify the input because that would already be a homogeneous solution! Instead, the response grows: we multiply by \(t\) or \(t^2\) to create a genuinely new solution.
This is exactly the concept of resonance in mechanical vibrations (undamped springs driven at their natural frequency) and electrical circuits (RLC circuits driven at their resonant frequency). The system is being driven at its own natural frequency, and the energy builds up unboundedly—hence the \(t\) factor.
The general solution decomposes into two essential parts:
The general solution \(y(t) = y_c(t) + y_p(t)\) combines both: the forced response (driven by \(g(t)\)) plus the system's own natural dynamics.
This perspective helps you understand not just how to find \(y_p\), but what it means physically: every particular solution is the system's way of responding to a given input.
Click each example to see the full solution. Watch for the critical overlap checking!
Step 1: Find \(y_c\). Characteristic equation: \(r^2 - 5r + 6 = 0\). Factor: \((r-2)(r-3) = 0\), so \(r = 2, 3\).
Step 2: Examine \(g(t)\). We have \(g(t) = 2t\), a polynomial of degree 1.
Step 3: Write trial form. For polynomial degree 1, try \(y_p = At + B\). Check overlap: are \(1\) and \(t\) in \(y_c\)? No (only exponentials). So this form is fine.
Step 4: Derivatives. \(y_p' = A\), \(y_p'' = 0\).
Step 5: Substitute.
Step 6: Match coefficients.
Step 7: General solution.
The first two terms represent the natural (homogeneous) response, decaying to zero. The last two terms form the particular solution—the steady-state response to the polynomial forcing.
Step 1: Find \(y_c\). Characteristic equation: \(r^2 - 7r + 12 = 0\). Factor: \((r-3)(r-4) = 0\), so \(r = 3, 4\).
Step 2: Examine \(g(t)\). We have \(g(t) = 4e^{2t}\), an exponential with \(\alpha = 2\).
Step 3: Write trial form. For exponential, try \(y_p = Ae^{2t}\). Check: is \(e^{2t}\) in \(y_c\)? No (only \(e^{3t}\) and \(e^{4t}\)). So this form is fine.
Step 4: Derivatives. \(y_p' = 2Ae^{2t}\), \(y_p'' = 4Ae^{2t}\).
Step 5: Substitute.
Step 6: Solve for \(A\). \(A = 2\).
Step 7: General solution.
Step 1: Find \(y_c\). Same as Example 2: \(y_c = C_1 e^{3t} + C_2 e^{4t}\).
Step 2: Examine \(g(t)\). We have \(g(t) = 5e^{4t}\), an exponential with \(\alpha = 4\).
Step 3: Write trial form — WITH RESONANCE CHECK. Naively, try \(y_p = Ae^{4t}\). But wait! Is \(e^{4t}\) in \(y_c\)? Yes! It appears as \(C_2 e^{4t}\). This is resonance. We must multiply by \(t\):
Step 4: Derivatives.
Step 5: Substitute.
Step 6: Solve for \(A\). The constant term gives \(A = 5\). (The coefficient of \(t\) is automatically zero—this confirms our choice was correct!)
Step 7: General solution.
Without the \(t\) factor, we would have gotten \(0 = 5\), a contradiction! The resonance check saved us. This is why Step 3 must always include overlap verification.
Step 1: Find \(y_c\). Characteristic equation: \(r^2 - 8r + 16 = 0\). Factor: \((r-4)^2 = 0\), so \(r = 4\) (double root).
Step 2: Examine \(g(t)\). We have \(g(t) = 2e^{4t}\).
Step 3: Resonance check. Both \(e^{4t}\) and \(te^{4t}\) appear in \(y_c\)! This is double resonance. We must multiply by \(t^2\):
Step 4: Derivatives.
Step 5: Substitute.
Step 6: Solve for \(A\). Constant term: \(2A = 2 \Rightarrow A = 1\). (Coefficient of \(t\) is automatically zero.)
Step 7: General solution.
Step 1: Find \(y_c\). Characteristic equation: \(r^2 - 2r + 1 = 0\). Factor: \((r-1)^2 = 0\), so \(r = 1\) (double root).
Step 2: Examine \(g(t)\). We have trig: \(g(t) = 5\cos(2t) + 10\sin(2t)\) with \(\omega = 2\).
Step 3: Resonance check. Do \(\cos(2t)\) or \(\sin(2t)\) appear in \(y_c = (C_1 + C_2 t)e^{t}\)? No. So try:
Step 4: Derivatives.
Step 5: Substitute.
Step 6: Match coefficients.
Solving: From the second equation, \(4A - 3B = 10\). Multiply by 3: \(12A - 9B = 30\). Multiply the first by 4: \(-12A - 16B = 20\). Add: \(-25B = 50\), so \(B = -2\). Then \(4A + 6 = 10\), so \(A = 1\).
Step 7: General solution.
Step 1: Find \(y_c\). Characteristic equation: \(r^2 + 4 = 0\). Roots: \(r = \pm 2i\), with \(\omega_0 = 2\).
Step 2: Examine \(g(t)\). We have \(g(t) = 8\cos(2t)\) with \(\omega = 2\).
Step 3: Resonance check. Both \(\cos(2t)\) and \(\sin(2t)\) appear in \(y_c\)! This is trig resonance (natural frequency matches forcing frequency). Multiply by \(t\):
Step 4: Derivatives (product rule!).
And \(y_p'' = 4B\cos(2t) - 4A\sin(2t) + \text{(terms with } t \text{)}\). [Detailed expansion: apply product rule twice.]
Step 5: Substitute into \(y'' + 4y = 8\cos(2t)\). After expansion and matching coefficients:
Step 6: Solve. \(4B = 8 \Rightarrow B = 2\), and \(-4A = 0 \Rightarrow A = 0\).
Step 7: General solution.
The term \(2t\sin(2t)\) grows without bound! This is resonance — when the forcing frequency matches the natural frequency, energy accumulates and amplitude grows linearly with time. This is why bridges must avoid matching the natural frequency of pedestrians' footsteps.
Step 1: Find \(y_c\). Characteristic equation: \(r^2 - 4r = r(r - 4) = 0\), so \(r = 0, 4\).
Step 2: Decompose \(g(t)\). We have \(g(t) = te^{t} + \cos(2t)\). Split:
For \(te^{t}\): This is a polynomial times exponential with \(\alpha = 1\). Try \(y_{p1} = (At + B)e^{t}\). Check: is \(e^{t}\) in \(y_c\)? No. Proceed.
Compute derivatives and substitute into \(y'' - 4y' = te^{t}\). After algebra: \(A = -\frac{1}{3}\), \(B = -\frac{2}{9}\).
For \(\cos(2t)\): Try \(y_{p2} = C\cos(2t) + D\sin(2t)\). Check: are trig in \(y_c\)? No. Proceed.
Compute and substitute into \(y'' - 4y' = \cos(2t)\):
Matching: \(-4C - 8D = 1\) and \(8C - 4D = 0\). Solve: \(D = 2C\) from the second, so \(-4C - 16C = 1 \Rightarrow C = -\frac{1}{20}\), \(D = -\frac{1}{10}\).
General solution (superposition):
When \(g(t) = g_1 + g_2\), we can solve for \(y_{p1}\) and \(y_{p2}\) separately, then combine. This reduces a complex problem to simpler pieces.
Step 1: Find \(y_c\). Characteristic equation: \(r^2 + 3r + 2 = (r+1)(r+2) = 0\), so \(r = -1, -2\).
Step 2: Examine \(g(x)\). Polynomial times trig. The highest degree is 1 (from \(20x\)), with trig \(\cos(x)\) and \(\sin(x)\) at \(\omega = 1\).
Step 3: Trial form. For \((ax + b)[\cos(\omega x) + \text{const} \sin(\omega x)]\), try:
Check: are any of \(\cos(x), \sin(x), x\cos(x), x\sin(x)\) in \(y_c\)? No (only exponentials). Proceed.
Step 4 & 5: Derivatives and substitution. [Lengthy algebra omitted.] After computing \(y_p'\) and \(y_p''\) and substituting into the full equation, matching coefficients of \(\cos(x)\) and \(\sin(x)\) (and their \(x\) multiples):
Step 7: General solution.
When \(g(t)\) is a product (polynomial \(\times\) trig, or polynomial \(\times\) exponential), combine the forms: include all coefficients for all powers of the polynomial, applied to each trig/exp term.
These problems are similar in style and difficulty to past exam questions. Click each problem to reveal the step-by-step solution.
Step 1: Find $y_c$ (homogeneous solution). Characteristic equation: $r^2 - 4r + 3 = 0$.
Step 2: Check for resonance. Is $e^{2t}$ in $y_c$? No. So we do not need to multiply by $t$.
Step 3: Guess the particular solution. Try $y_p = Ae^{2t}$.
Step 4: Compute derivatives. $y_p' = 2Ae^{2t}$, $y_p'' = 4Ae^{2t}$.
Step 5: Substitute into the equation.
Step 6: General solution.
Step 1: Find $y_c$. Characteristic equation: $r^2 + 1 = 0 \Rightarrow r = \pm i$.
Step 2: Check for resonance. The forcing is $\cos(t)$. Are $\cos(t)$ and $\sin(t)$ in $y_c$? Yes! This is resonance. Multiply by $t$.
Step 3: Guess the particular solution. Try $y_p = t(A\cos(t) + B\sin(t))$.
Step 4: Compute derivatives (using product rule).
Step 5: Substitute into $y'' + y = 3\cos(t)$.
Step 6: Match coefficients. $-2A = 0 \Rightarrow A = 0$, and $2B = 3 \Rightarrow B = \frac{3}{2}$.
Step 7: General solution.
Explore how changing the equation affects \(y_c\), \(y_p\), and the general solution. Watch the resonance case!
Blue curve: \(y_c(t)\) — complementary solution (homogeneous response). Notice how it decays or grows depending on eigenvalues.
Green curve: \(y_p(t)\) — particular solution (steady-state or forced response). Depends entirely on \(g(t)\).
Red curve: \(y(t) = y_c + y_p\) — the complete general solution with \(C_1 = C_2 = 0\) (you can adjust with initial conditions).
Test your understanding with 5 multiple-choice problems. Choose the correct general form for \(y_p\) or identify the solution.
For the equation \(y'' + 2y' + y = t^2 + 1\), what is the complementary solution \(y_c\)?
For \(y'' - 3y' + 2y = 5e^{2t}\), should we include a factor of \(t\) in \(y_p\) (i.e., resonance)?
If the characteristic equation is \((r-2)^2(r+1) = 0\) and \(g(t) = 3e^{2t}\), which is the correct trial form for \(y_p\)?
For \(y'' + y = 2\cos(t)\), the characteristic equation gives \(r = \pm i\). What is the correct trial form for \(y_p\)?
For \(y'' - 4y' + 3y = t + e^{3t}\), how should we split the problem using superposition?