Differential Equations › Section B · First order differential equations › Topic B.1
Linear equations and integrating factors
Every first-order linear equation can be solved by the same short procedure: multiply by one well-chosen function and the equation collapses into a single derivative. This topic makes that procedure routine and shows what the solution formula reveals about long-term behaviour.
Where linear equations fit
A first-order linear equation is one that can be arranged in the standard form
The Calculus topic "First-order linear equations" introduced the integrating factor and solved mixing-tank and circuit problems with it. Here the method is treated as the first general tool of this course. It works for every linear equation, with any continuous coefficients, and its result has a structure that explains how solutions behave in the long run. Linear equations are also the main class whose intervals of validity can be read off before solving, a fact the topic "Existence, uniqueness, and intervals of validity" makes precise.
The integrating factor
Divide by the coefficient of \(y'\) first, so that the equation is in standard form. Then multiply both sides by
Because \(\mu' = P\mu\), the left side \(\mu y' + P\mu y\) is exactly \((\mu y)'\) by the product rule. The equation becomes \((\mu y)' = \mu Q\), and one integration finishes it:
No constant is needed inside \(\mu\). Using \(Ke^{\int P\,dt}\) instead multiplies both sides by \(K\), which cancels. Only the constant from the final integration matters.
Worked example: a polynomial forcing term
Solve \(y' + 2y = 4t\) with \(y(0) = 1\). 1. Standard form already, with \(P = 2\), so \(\mu = e^{2t}\). 2. Multiply: \((e^{2t}y)' = 4te^{2t}\). 3. Integrate by parts, as in the Calculus topic "Integration by parts": \(\int 4te^{2t}\,dt = 2te^{2t} - e^{2t} + C\). 4. Divide by \(e^{2t}\): \(y = 2t - 1 + Ce^{-2t}\). 5. The condition gives \(-1 + C = 1\), so \(C = 2\) and \(y = 2t - 1 + 2e^{-2t}\). Check: \(y' = 2 - 4e^{-2t}\), and
Worked example: a variable coefficient
Solve \(ty' + y = \cos t\) with \(y(\pi) = \frac{2}{\pi}\), for \(t > 0\). 1. Standard form: \(y' + \frac{1}{t}y = \frac{\cos t}{t}\), so \(P = \frac{1}{t}\). 2. Then \(\mu = e^{\ln t} = t\), using \(t > 0\). 3. Multiplying the standard form by \(t\) gives back the original left side, which is already \((ty)'\). So \((ty)' = \cos t\) and \(ty = \sin t + C\). 4. Hence \(y = \frac{\sin t + C}{t}\). 5. The condition: \(\frac{0 + C}{\pi} = \frac{2}{\pi}\), so \(C = 2\) and \(y = \frac{\sin t + 2}{t}\). Check: \(ty = \sin t + 2\) has derivative \(\cos t\). The solution fails at \(t = 0\), where \(P\) does.
Transient and steady state
Look again at \(y = 2t - 1 + Ce^{-2t}\). The term \(Ce^{-2t}\) is the only place the initial condition enters, and it dies away; the term \(2t - 1\) remains. Engineers call these the transient and the steady state, even when, as here, the steady state is a moving line rather than a constant. Every solution, whatever its starting value, ends up hugging the line \(y = 2t - 1\), exactly the behaviour the direction field of such an equation shows. In general the solution formula splits into a particular solution plus \(Ce^{-\int P\,dt}\), and whether that second term decays, grows or stays bounded decides how much the starting value matters later.
Worked example: behaviour that depends on y(0)
Solve \(y' - y = -2e^{-t}\) with \(y(0) = y_0\), and describe the solutions as \(t \to \infty\). 1. \(P = -1\), so \(\mu = e^{-t}\) and \((e^{-t}y)' = -2e^{-2t}\). 2. Integrate: \(e^{-t}y = e^{-2t} + C\), so \(y = e^{-t} + Ce^{t}\). 3. The condition gives \(1 + C = y_0\), so \(C = y_0 - 1\). 4. Read the result: if \(y_0 > 1\) the term \(Ce^t\) drives \(y \to \infty\); if \(y_0 < 1\), \(y \to -\infty\); if \(y_0 = 1\), \(y = e^{-t} \to 0\). Check: \(y = e^{-t}\) gives \(y' - y = -e^{-t} - e^{-t} = -2e^{-t}\). Here the homogeneous part grows, so one starting value separates two fates.
When the integral cannot be done
The method always produces \(\int \mu Q\,dt\), but that integral may have no elementary antiderivative. Then write the answer with a definite integral. Choosing \(\mu(t) = e^{\int_{t_0}^{t} P(s)\,ds}\), so that \(\mu(t_0) = 1\), gives
For \(y' + 2ty = 1\) with \(y(0) = 0\), \(\mu = e^{t^2}\) and \(y = e^{-t^2}\int_0^t e^{s^2}\,ds\). That is an exact solution: you can differentiate it, as the fundamental theorem of calculus allows, and evaluate it numerically to any accuracy.
Reading the method as a checklist
Every linear problem follows the same five steps. 1. Divide by the coefficient of \(y'\) and identify \(P\) and \(Q\), with their signs. 2. Compute \(\mu = e^{\int P\,dt}\) and simplify it, using \(e^{\ln u} = u\) where it applies. 3. Multiply both sides and write the left side as \((\mu y)'\). 4. Integrate both sides, adding the constant at once, then divide everything by \(\mu\). 5. Apply the initial condition last, and note where \(P\) and \(Q\) fail to be continuous, because the solution cannot be trusted across those points.
Common mistakes to avoid
First, finding \(\mu\) before dividing by the coefficient of \(y'\): in \(ty' + y = \cos t\), \(P\) is \(\frac{1}{t}\), not 1. Second, dropping the sign of \(P\): \(y' - y\) has \(P = -1\) and \(\mu = e^{-t}\). Third, multiplying only the left side by \(\mu\). Fourth, adding the constant after dividing by \(\mu\), which gives \(2t - 1 + C\) instead of \(2t - 1 + Ce^{-2t}\). Fifth, writing \(e^{\ln t} = t\) without noting that it needs \(t > 0\).
Key takeaways
Put the equation in standard form \(y' + P(t)y = Q(t)\), multiply by \(\mu = e^{\int P\,dt}\), recognise the left side as \((\mu y)'\), integrate, and divide by \(\mu\) with the constant included. No constant is needed in \(\mu\). When \(\int \mu Q\,dt\) is not elementary, a definite-integral formula is still an exact answer. The solution is a particular solution plus \(Ce^{-\int P\,dt}\); when that term decays it is a transient and every solution approaches the steady state, and when it grows the long-run outcome depends on \(y(0)\).
Flashcards
Flashcard
y′ + P(t)y = Q(t)
Test your understanding
Question 1 of 5
What is an integrating factor for y′ + (3/t)y = t, with t > 0?
Question 2 of 5
Which function solves y′ + y = 3 with y(0) = 5?
Question 3 of 5
The solution of y′ − 3y = −6 with y(0) = y₀ is y = 2 + (y₀ − 2)e^(3t). Which starting values give a solution that stays bounded as t → ∞?
Question 4 of 5
Solving a linear equation leads to e^(2t)y = t + C. What is y?
Question 5 of 5
Every solution of y′ + 2y = 4t has the form y = 2t − 1 + Ce^(−2t). For large t, every solution is close to which function?
