Differential Equations › Section D · Laplace transforms › Topic D.5
Solving initial value problems
Here the Laplace transform earns its keep: transform the equation, solve an algebra problem for the unknown transform, and invert. The initial conditions enter at the first step, so the answer arrives with every constant already fixed.
Back to differential equations
The earlier topics of this section built the tools: transforms from a table, inverses by partial fractions and completing the square, and the derivative rule from Computing Laplace transforms. Now they combine. This topic sticks to linear equations with constant coefficients and continuous forcing, which the methods of the previous section could also solve. That is deliberate: every answer here can be checked by substituting it back in, or by solving it the older way. Discontinuous and impulsive forcing, where the transform has no real rival, come in later topics of this section.
Derivatives of any order
Repeating the integration by parts behind \(\mathcal{L}\{f'\} = sF - f(0)\) gives the general rule. If \(f, f', \dots, f^{(n-1)}\) are continuous and of exponential order, and \(f^{(n)}\) is piecewise continuous, then
The powers of \(s\) fall by one at each step while the derivative order rises. For \(n = 2\) this is \(s^2Y - sy(0) - y'(0)\). Every initial value appears with a minus sign, which is how the transform records where the solution starts.
The method in three steps
1. Transform both sides of the equation, using the derivative rule with the initial conditions substituted at once. 2. Solve the resulting algebraic equation for \(Y(s)\). Collect every term containing \(Y\) on one side, factor it out, and divide. 3. Invert \(Y(s)\) to get \(y(t)\), usually by partial fractions. There is no general solution with arbitrary constants and no separate step to fit the initial conditions; the inverse transform is the solution to the initial value problem. Before inverting, simplify \(Y(s)\) as far as possible, because a common denominator often means fewer partial fractions.
Worked example: a first order problem
Solve \(y' + 3y = 2e^{-t}\) with \(y(0) = 4\). 1. Transform: \(sY - 4 + 3Y = \frac{2}{s + 1}\). 2. Solve: \((s + 3)Y = 4 + \frac{2}{s + 1}\), so
3. Partial fractions split the second term into \(\frac{1}{s + 1} - \frac{1}{s + 3}\), so \(Y = \frac{3}{s + 3} + \frac{1}{s + 1}\). 4. Invert: \(y = 3e^{-3t} + e^{-t}\). Check: \(y(0) = 4\), and
Worked example: a damped oscillator
Solve \(y'' + 2y' + 5y = 0\) with \(y(0) = 1\) and \(y'(0) = 3\). 1. Transform:
2. Collect: \((s^2 + 2s + 5)Y = s + 5\), so \(Y = \frac{s + 5}{s^2 + 2s + 5}\). 3. The denominator has complex roots, so complete the square, \((s + 1)^2 + 4\), and write \(s + 5 = (s + 1) + 4\):
4. Invert: \(y = e^{-t}(\cos 2t + 2\sin 2t)\). Check: \(y(0) = 1\), and differentiating gives \(y'(0) = -1 + 4 = 3\). Note that the characteristic polynomial from Complex roots appears as the denominator.
Worked example: forcing from rest
Solve \(y'' - y' - 2y = 6e^t\) with \(y(0) = 0\) and \(y'(0) = 0\). 1. Transform: \((s^2 - s - 2)Y = \frac{6}{s - 1}\), since both initial values are zero. 2. Factor \(s^2 - s - 2 = (s - 2)(s + 1)\), so
3. Cover-up partial fractions: at \(s = 1\), \(\frac{6}{(-1)(2)} = -3\); at \(s = 2\), \(\frac{6}{(1)(3)} = 2\); at \(s = -1\), \(\frac{6}{(-2)(-3)} = 1\). 4. Invert: \(y = -3e^t + 2e^{2t} + e^{-t}\). Check: \(y(0) = -3 + 2 + 1 = 0\) and \(y'(0) = -3 + 4 - 1 = 0\). The term \(-3e^t\) is the particular solution undetermined coefficients would find.
How it compares with earlier methods
Undetermined coefficients and variation of parameters find a general solution first and then fit constants to the initial data, which for a second order equation means solving two equations at the end. The Laplace method folds the initial values in at the start, so it delivers the specific solution in one pass. The price is the partial fraction work in step 3, which grows with the order of the equation and the complexity of the forcing. For simple homogeneous equations the characteristic equation is quicker. The transform wins when the forcing is complicated, piecewise or impulsive.
Initial conditions away from zero
The derivative rule uses values at \(t = 0\). If the data are given at \(t = t_0\), shift time first. Let \(\eta = t - t_0\) and \(w(\eta) = y(\eta + t_0)\); then \(w\) has the same derivatives, and its initial values are at \(\eta = 0\). For example, solve \(y' + 2y = 4\) with \(y(1) = 5\). The new problem is \(w' + 2w = 4\), \(w(0) = 5\). Transforming gives \(W = \frac{2}{s} + \frac{3}{s + 2}\), so \(w = 2 + 3e^{-2\eta}\) and \(y = 2 + 3e^{-2(t - 1)}\). If the forcing contains \(t\), rewrite it with \(t = \eta + t_0\) before transforming.
Transfer functions and poles
For \(ay'' + by' + cy = g(t)\) starting from rest, with \(y(0) = y'(0) = 0\), the transformed equation gives
The transfer function \(H(s)\) belongs to the system and does not depend on the input. Its poles, the roots of \(as^2 + bs + c\), are the roots of the characteristic equation. If every pole has a negative real part, the free response decays to zero, and the system is called stable. Engineers analyse circuits and control loops this way, with capacitors and inductors replaced by expressions in \(s\).
Common mistakes to avoid
Dropping or mis-signing initial values: \(\mathcal{L}\{y''\}\) is \(s^2Y - sy(0) - y'(0)\), and with \(y'(0) = -1\) the last term becomes \(+1\). Forgetting to multiply by the coefficient: \(2y'\) transforms to \(2(sY - y(0))\), so the \(y(0)\) term is doubled too. Transforming the forcing wrongly, as \(\frac{1}{s + 1}\) for \(e^t\). Inverting \(\frac{6}{s^2 + 9}\) as \(6\sin 3t\) instead of \(2\sin 3t\). Applying the derivative rule to data given at \(t_0 \ne 0\) without shifting first.
Key takeaways
Transform the equation using \(\mathcal{L}\{y'\} = sY - y(0)\) and \(\mathcal{L}\{y''\} = s^2Y - sy(0) - y'(0)\), solve for \(Y(s)\), and invert. The initial conditions are built in, so no constants remain to be found. Expect the characteristic polynomial in the denominator, and use partial fractions and completing the square to invert. Shift time when the data are not at \(t = 0\). From rest, \(Y = H(s)G(s)\), and the poles of the transfer function decide whether the free response decays.
Flashcards
Flashcard
Transform, solve for Y(s), invert
Test your understanding
Question 1 of 5
Transforming y′ + 4y = 0 with y(0) = 5 and solving gives which Y(s)?
Question 2 of 5
Solve y″ + 9y = 0 with y(0) = 0 and y′(0) = 6 by Laplace transforms.
Question 3 of 5
For y″ + 3y′ + 2y = g(t) starting from rest, where are the poles of the transfer function?
Question 4 of 5
An equation has data y(2) = 1 and y′(2) = 0. How should you prepare it for Laplace transforms?
Question 5 of 5
Solve y″ - y = 0 with y(0) = 1 and y′(0) = 1 by Laplace transforms.
