Differential Equations › Section D · Laplace transforms › Topic D.6

Nonconstant coefficient problems

Laplace transforms are built for constant coefficients, but when a coefficient is a simple multiple of t the method still works, with a twist: instead of an algebraic equation for Y(s), you get a differential equation for it.

Beyond constant coefficients

Every equation in Solving initial value problems had constant coefficients, and that is what made step 2 pure algebra: each derivative of \(y\) became a power of \(s\) times \(Y\). Equations whose coefficients depend on \(t\) are generally much harder, and most have no solution in terms of familiar functions. But when the coefficients are at most linear in \(t\), such as \(ty''\) or \(3ty'\), the Laplace transform can still help. It trades a second order equation in \(t\) for a first order linear equation in \(s\), which the integrating factor method from Linear equations and integrating factors can solve.

Transforming a factor of t

The tool comes from Computing Laplace transforms: \(\mathcal{L}\{tf(t)\} = -F'(s)\). Apply it with \(f = y'\), whose transform is \(sY - y(0)\). Differentiating with respect to \(s\), the constant \(y(0)\) disappears, so

\mathcal{L}\{ty'\} = -\frac{d}{ds}\big(sY - y(0)\big) = -sY' - Y

With \(f = y''\), whose transform is \(s^2Y - sy(0) - y'(0)\), the same steps give

\mathcal{L}\{ty''\} = -s^2Y' - 2sY + y(0)

Here \(Y'\) means \(\frac{dY}{ds}\). Notice that \(y'(0)\) has vanished from the second formula, because it is a constant.

Exponential order and a limit

Recall from The definition of the Laplace transform that \(f\) is of exponential order \(\alpha\) if \(|f(t)| \le Me^{\alpha t}\) for all large \(t\). For such functions, piecewise continuous on \([0, \infty)\), there is a useful fact:

\lim_{s \to \infty} F(s) = 0

Every table entry obeys it: \(\frac{1}{s}\), \(\frac{n!}{s^{n+1}}\) and \(\frac{b}{s^2 + b^2}\) all shrink to zero. So if solving for \(Y(s)\) produces a term that grows without bound as \(s \to \infty\), it cannot be the transform of a solution of exponential order, and its constant must be zero. This is how a constant of integration is removed without an initial condition.

Worked example: setting up

Solve \(y'' + ty' - y = 0\) with \(y(0) = 0\) and \(y'(0) = 1\). 1. Transform each term. The first gives \(s^2Y - 0 - 1\). The second, by the rule above, gives \(-sY' - Y\). The third gives \(-Y\). 2. So the transformed equation is \(s^2Y - 1 - sY' - 2Y = 0\). 3. Rearrange into standard linear form by dividing by \(-s\):

Y' + \Big(\frac{2}{s} - s\Big)Y = -\frac{1}{s}

Check: this is first order and linear in \(Y\), exactly as expected from a coefficient linear in \(t\). The original second order equation in \(t\) has become a first order equation in \(s\).

Worked example: solving for Y

Continue with \(Y' + \big(\frac{2}{s} - s\big)Y = -\frac{1}{s}\). 1. The integrating factor is \(\mu = e^{\int (2/s - s)\,ds} = s^2e^{-s^2/2}\). 2. Multiply through: \(\big(s^2e^{-s^2/2}Y\big)' = -se^{-s^2/2}\). 3. Integrate: \(s^2e^{-s^2/2}Y = e^{-s^2/2} + C\), so

Y(s) = \frac{1}{s^2} + \frac{Ce^{s^2/2}}{s^2}

4. As \(s \to \infty\), the second term grows without bound unless \(C = 0\), so \(C = 0\). 5. Invert: \(Y = \frac{1}{s^2}\) gives \(y = t\). Check: \(y'' + ty' - y = 0 + t - t = 0\), with \(y(0) = 0\) and \(y'(0) = 1\).

Worked example: a constant fixed by data

Solve \(ty'' + ty' - y = -3\) with \(y(0) = 3\) and \(y'(0) = 5\). 1. Transform:

(-s^2Y' - 2sY + 3) + (-sY' - Y) - Y = -\frac{3}{s}

2. Group:

-s(s + 1)Y' - 2(s + 1)Y = -\frac{3(s + 1)}{s}

3. Divide by \(-s(s + 1)\): \(Y' + \frac{2}{s}Y = \frac{3}{s^2}\). The integrating factor is \(s^2\), so \((s^2Y)' = 3\), \(s^2Y = 3s + C\) and \(Y = \frac{3}{s} + \frac{C}{s^2}\). 4. Both terms tend to 0, so the limit cannot fix \(C\). Invert instead: \(y = 3 + Ct\). 5. Now use \(y'(0) = 5\), which the transform never needed: \(C = 5\) and \(y = 3 + 5t\). Check: \(0 + 5t - (3 + 5t) = -3\).

Where the constants go

The two examples show the two ways the leftover constant is settled. If its term grows as \(s \to \infty\), as \(Ce^{s^2/2}/s^2\) did, exponential order forces it to vanish. If its term is a genuine transform, as \(\frac{C}{s^2}\) was, the constant survives into \(y(t)\), and an initial condition that dropped out during the transform fixes it at the end. Either way, every constant must be accounted for before you state the answer. A leftover \(C\) in the final solution means a step was skipped.

The limits of the method

The trick depends on each \(t\) turning into one derivative in \(s\). A coefficient of \(t^2\) brings in \(Y''(s)\), so the transformed problem is second order again and no easier than the original. Even with linear coefficients, the first order equation for \(Y\) may have no tidy solution, or its answer may not appear in any table. There is also a subtlety at \(t = 0\). In the second example the coefficient of \(y''\) is \(t\), which vanishes there, so the usual existence and uniqueness theorem does not apply at that point. Always confirm the answer by substituting it back in.

Common mistakes to avoid

Transforming \(ty'\) as \(t\) times \(sY - y(0)\), mixing the variables \(t\) and \(s\): the factor \(t\) must become \(-\frac{d}{ds}\). Forgetting the product rule when differentiating \(sY\), which gives \(Y + sY'\), not \(sY'\). Keeping \(y(0)\) inside the derivative: it is a constant, so \(\frac{d}{ds}\) removes it. Dropping a growing term without saying why: the reason is exponential order. Leaving an arbitrary constant in the answer when an unused initial condition was available to fix it.

Key takeaways

For coefficients linear in \(t\), use \(\mathcal{L}\{tf\} = -F'(s)\): \(\mathcal{L}\{ty'\} = -sY' - Y\) and \(\mathcal{L}\{ty''\} = -s^2Y' - 2sY + y(0)\). The transformed problem is a first order linear equation for \(Y(s)\), solved with an integrating factor. Transforms of functions of exponential order tend to 0 as \(s \to \infty\), which removes any growing term. A surviving constant is fixed by an initial condition after inverting. Higher powers of \(t\) give higher order equations in \(s\), so the method has narrow reach.

Flashcards

Flashcard

L{t f(t)}

-F′(s)

A factor of t becomes -d/ds.

Test your understanding

Question 1 of 5

What is the Laplace transform of t y′(t), written in terms of Y(s)?

Question 2 of 5

What is the Laplace transform of t y″(t)?

Question 3 of 5

Solving for Y gives Y(s) = 2/s³ + C e^(s²)/s³. Why must C be 0?

Question 4 of 5

Why does the Laplace transform not simplify y″ + t²y = 0 the way it simplifies y″ + ty′ - y = 0?

Question 5 of 5

For an equation with y(0) = 2 and y′(0) = -1, the transform gives Y(s) = 2/s + C/s². What is y(t)?

← Previous topicNext topic →