MATH 2410 — Differential Equations
Lecture Notes. This is NOT meant to be a substitute for coming to lecture or consulting the textbook directly, but rather a helpful resource to review for exams/ review what was discussed in lecture.
What Is a Transform?
A transform is an operation that converts one type of function into a different kind of function. We have already seen several examples of this idea: differentiation transforms \(x^2\) into \(2x\) and \(\ln(x)\) into \(1/x\); integration transforms \(e^{kx}\) into \(\tfrac1ke^{kx}\).
Integral transforms. Suppose we have a function \(f(x,y)\) and hold \(y\) constant. Then
\[ \int_a^b f(x,y)\,dx = F(b,y)-F(a,y), \qquad \text{where } F_x(x,y)=f(x,y), \]turns a function of two variables into a function of one variable \(y\).
We now generalize this idea in two ways: (1) integrate \(f(t)\) against some fixed "kernel" function \(K(s,t)\), and (2) integrate over the unbounded interval \([0,\infty)\):
\[ \int_0^\infty K(s,t)f(t)\,dt = \lim_{b\to\infty}\int_0^b K(s,t)f(t)\,dt, \]where \(s\) is a real variable. Choosing \(K(s,t)=e^{-st}\) gives the Laplace transform.
The Laplace Transform
This follows immediately from linearity of the integral:
\[ \mathcal{L}\{c_1f+c_2g\} = \int_0^\infty e^{-st}(c_1f+c_2g)\,dt = c_1\int_0^\infty e^{-st}f(t)\,dt + c_2\int_0^\infty e^{-st}g(t)\,dt = c_1\mathcal{L}\{f(t)\}+c_2\mathcal{L}\{g(t)\}. \]By linearity, \(\mathcal{L}\{k\}=k\,\mathcal{L}\{1\}\), so it suffices to find
\[ \mathcal{L}\{1\} = \int_0^\infty e^{-st}\,dt = \lim_{b\to\infty}\int_0^b e^{-st}\,dt = \lim_{b\to\infty}\left(-\frac1se^{-st}\right)\Big|_0^b = \lim_{b\to\infty}\left(\frac1s - \frac{e^{-sb}}{s}\right). \]This limit is \(\tfrac1s\) if \(s>0\) (and diverges if \(s\le 0\)). So
\[ \boxed{\mathcal{L}\{k\} = \frac{k}{s}, \qquad s>0.} \]By linearity, \(\mathcal{L}\{mt+k\} = m\,\mathcal{L}\{t\}+\mathcal{L}\{k\}\), so we need
\[ \mathcal{L}\{t\} = \int_0^\infty te^{-st}\,dt = \lim_{b\to\infty}\int_0^b te^{-st}\,dt. \]Using integration by parts (\(u=t\), \(dv=e^{-st}dt\), so \(du=dt\), \(v=-\tfrac1se^{-st}\)),
\[ \int_0^b te^{-st}\,dt = \left(-\frac ts e^{-st} - \frac{1}{s^2}e^{-st}\right)\Big|_0^b = -\frac bs e^{-sb} - \frac{1}{s^2}e^{-sb} + \frac{1}{s^2}. \]As \(b\to\infty\), the first two terms vanish for \(s>0\), leaving \(\mathcal{L}\{t\} = \tfrac{1}{s^2}\) for \(s>0\). Thus
\[ \boxed{\mathcal{L}\{mt+k\} = \frac{m}{s^2}+\frac{k}{s}, \qquad s>0.} \]Integrating by parts twice (first with \(u=e^{-st}\), \(dv=\sin(t)dt\), then again on the resulting integral),
\[ \int_0^\infty e^{-st}\sin(t)\,dt = \Big[-e^{-st}\cos(t) - se^{-st}\sin(t)\Big]_0^\infty - s^2\int_0^\infty e^{-st}\sin(t)\,dt. \]Moving the last term to the left-hand side,
\[ (1+s^2)\,\mathcal{L}\{\sin t\} = \lim_{b\to\infty}\Big(-e^{-sb}\cos(b)-se^{-sb}\sin(b)\Big) - (-1-0) = 1 \qquad (s>0), \]since the boundary terms vanish as \(b\to\infty\) for \(s>0\). Thus
\[ \boxed{\mathcal{L}\{\sin t\} = \frac{1}{1+s^2}, \qquad s>0.} \]| \(f(t)\) | \(\mathcal{L}\{f(t)\}\) |
|---|---|
| \(1\) | \(\dfrac1s\) |
| \(t^n\) | \(\dfrac{n!}{s^{n+1}}\) |
| \(e^{at}\) | \(\dfrac{1}{s-a}\) |
| \(\sin(kt)\) | \(\dfrac{k}{s^2+k^2}\) |
| \(\cos(kt)\) | \(\dfrac{s}{s^2+k^2}\) |
Common Laplace transforms (all valid for \(s\) sufficiently large).
Inverse Laplace Transforms
Reading the transform table from Section 7.1 in reverse gives us a small table of inverse transforms:
\[ \mathcal{L}^{-1}\left\{\frac1s\right\}=1, \qquad \mathcal{L}^{-1}\left\{\frac{1}{s^2}\right\}=t, \qquad \mathcal{L}^{-1}\left\{\frac{1}{s+3}\right\}=e^{-3t}. \]By linearity,
\[ \mathcal{L}^{-1}\left\{\frac{-2s+6}{s^2+4}\right\} = -2\,\mathcal{L}^{-1}\left\{\frac{s}{s^2+4}\right\} + 6\,\mathcal{L}^{-1}\left\{\frac{1}{s^2+4}\right\}. \]Using \(\mathcal{L}\{\cos(2t)\}=\dfrac{s}{s^2+4}\) and \(\mathcal{L}\{\sin(2t)\}=\dfrac{2}{s^2+4}\),
\[ \boxed{\mathcal{L}^{-1}\left\{\frac{-2s+6}{s^2+4}\right\} = -2\cos(2t)+6\sin(2t).} \]When the denominator doesn't match a known form directly, we turn to partial fractions.
Since the inverse transform works with sums and differences but not with products or quotients, we first factor and decompose:
\[ \frac{6s+2}{(s-1)(s+1)} = \frac{A}{s-1}+\frac{B}{s+1}. \]Then \(6s+2 = A(s+1)+B(s-1)\). Setting \(s=1\): \(8=2A \Rightarrow A=4\). Setting \(s=-1\): \(-4=-2B \Rightarrow B=2\). So
\[ \mathcal{L}^{-1}\left\{\frac{6s+2}{(s-1)(s+1)}\right\} = \mathcal{L}^{-1}\left\{\frac{4}{s-1}\right\} + \mathcal{L}^{-1}\left\{\frac{2}{s+1}\right\} = \boxed{4e^{t}+2e^{-t}.} \]Solving Initial Value Problems with the Laplace Transform
- Apply the Laplace transform to both sides of the differential equation.
- Use the theorem above to rewrite each derivative term, and solve the resulting algebraic equation for \(F(s)=\mathcal{L}\{y(t)\}\).
- Apply the inverse Laplace transform (using partial fractions if needed) to recover \(y(t) = \mathcal{L}^{-1}\{F(s)\}\).
Taking the Laplace transform of both sides,
\[ \mathcal{L}\{y'\}+3\mathcal{L}\{y\} = \big(sF(s)-y(0)\big)+3F(s) = \frac{16}{s-1}. \]Substituting \(y(0)=1\) and solving for \(F(s)\),
\[ F(s)(s+3) = \frac{16}{s-1}+1 = \frac{s+15}{s-1} \;\Longrightarrow\; F(s) = \frac{s+15}{(s+3)(s-1)}. \]Using partial fractions, \(\dfrac{s+15}{(s+3)(s-1)} = \dfrac{A}{s+3} +\dfrac{B}{s-1}\), so \(s+15=A(s-1)+B(s+3)\). Setting \(s=1\): \(16=4B \Rightarrow B=4\). Setting \(s=-3\): \(12=-4A \Rightarrow A=-3\). So
\[ y = \mathcal{L}^{-1}\{F(s)\} = \mathcal{L}^{-1}\left\{\frac{-3}{s+3}\right\} + \mathcal{L}^{-1}\left\{\frac{4}{s-1}\right\} = \boxed{-3e^{-3t}+4e^{t}.} \]Taking the Laplace transform of both sides,
\[ \mathcal{L}\{y''\}+5\mathcal{L}\{y'\}+4\mathcal{L}\{y\} = \big(s^2F(s)-sy(0)-y'(0)\big) + 5\big(sF(s)-y(0)\big) + 4F(s) = 0. \]Substituting \(y(0)=1\), \(y'(0)=0\),
\[ F(s)(s^2+5s+4) - s - 5 = 0 \;\Longrightarrow\; F(s) = \frac{s+5}{(s+4)(s+1)}. \]Using partial fractions, \(\dfrac{s+5}{(s+4)(s+1)} = \dfrac{A}{s+4} +\dfrac{B}{s+1}\), so \(s+5=A(s+1)+B(s+4)\). Setting \(s=-4\): \(1=-3A \Rightarrow A=-\tfrac13\). Setting \(s=-1\): \(4=3B \Rightarrow B=\tfrac43\). So
\[ y = \mathcal{L}^{-1}\left\{\frac{-1/3}{s+4}\right\} + \mathcal{L}^{-1}\left\{\frac{4/3}{s+1}\right\} = \boxed{\frac43e^{-t}-\frac13e^{-4t}.} \]The First Translation Theorem
To see why, substitute directly into the definition:
\[ \mathcal{L}\{e^{at}f(t)\} = \int_0^\infty e^{-st}e^{at}f(t)\,dt = \int_0^\infty e^{-(s-a)t}f(t)\,dt = F(s-a). \]Taking the Laplace transform of both sides,
\[ \mathcal{L}\{y'\}-\mathcal{L}\{y\} = \big(sF(s)-y(0)\big)-F(s) = F(s)(s-1). \]On the right, \(\mathcal{L}\{1\}=\dfrac1s\), and by the First Translation Theorem, \(\mathcal{L}\{te^t\} = \dfrac{1}{(s-1)^2}\) (since \(\mathcal{L}\{t\}=1/s^2\)). So
\[ F(s)(s-1) = \frac1s+\frac{1}{(s-1)^2} \;\Longrightarrow\; F(s) = \frac{1}{s(s-1)}+\frac{1}{(s-1)^3}. \]Using partial fractions, \(\dfrac{1}{s(s-1)} = -\dfrac1s+\dfrac{1}{s-1}\), so
\[ F(s) = -\frac1s+\frac{1}{s-1}+\frac{1}{(s-1)^3}. \]Since \(\mathcal{L}^{-1}\{1/s^3\}=t^2/2\), the First Translation Theorem gives \(\mathcal{L}^{-1}\{1/(s-1)^3\} = \tfrac12t^2e^t\). Thus
\[ \boxed{y = -1+e^t+\frac12t^2e^t.} \]The Unit Step Function and the Second Translation Theorem
Since \(f(t)=0\) for \(t\lt 1\),
\[ \mathcal{L}\{f(t)\} = \int_1^\infty e^{-st}(2t-3)\,dt = 2\int_1^\infty te^{-st}\,dt - 3\int_1^\infty e^{-st}\,dt. \]Evaluating each piece (using integration by parts on the first),
\[ \int_1^\infty e^{-st}\,dt = \frac{e^{-s}}{s}, \qquad \int_1^\infty te^{-st}\,dt = \frac{e^{-s}}{s}+\frac{e^{-s}}{s^2}. \]So
\[ \mathcal{L}\{f(t)\} = 2\left(\frac{e^{-s}}{s}+\frac{e^{-s}}{s^2}\right) - 3\frac{e^{-s}}{s} = \boxed{\frac{2}{s^2}e^{-s} - \frac1s e^{-s}.} \]More generally, for \(\mathcal{L}\{f(t-a)\mathscr{U}(t-a)\}\), split the integral at \(t=a\) (the piece below \(a\) vanishes) and substitute \(v=t-a\):
\[ \mathcal{L}\{f(t-a)\mathscr{U}(t-a)\} = \int_a^\infty e^{-st}f(t-a)\,dt = \int_0^\infty e^{-s(v+a)}f(v)\,dv = e^{-as}\int_0^\infty e^{-sv}f(v)\,dv = e^{-as}F(s). \]By linearity and the Second Translation Theorem,
\[ \boxed{\mathcal{L}\{f(t)\} = \mathcal{L}\{2\} - 4\mathcal{L}\{\mathscr{U}(t-3)\} = \frac2s - \frac4s e^{-3s}.} \]Taking the Laplace transform of both sides,
\[ \big(sF(s)-y(0)\big)+F(s) = F(s)(s+1) = 5\mathcal{L}\{\mathscr{U}(t-1)\} = \frac{5}{s}e^{-s}. \]So
\[ F(s) = \frac{5e^{-s}}{s(s+1)} = 5e^{-s}\left(\frac1s-\frac{1}{s+1}\right). \]Applying the Second Translation Theorem to each piece,
\[ \boxed{y = 5\mathscr{U}(t-1) - 5e^{-(t-1)}\mathscr{U}(t-1).} \]Since \(\sin(t)\) is periodic with period \(2\pi\), \(\sin(t-2\pi)=\sin(t)\), so we can write the forcing term as \(\sin(t-2\pi)\mathscr{U}(t-2\pi)\) — exactly the form needed for the Second Translation Theorem, with \(f(t)=\sin(t)\), \(F(s) = \dfrac{1}{s^2+1}\). Taking the Laplace transform of both sides,
\[ \big(s^2F(s)-s\cdot1-0\big)+4F(s) = (s^2+4)F(s)-s = e^{-2\pi s}\cdot\frac{1}{s^2+1}. \]So
\[ F(s) = \frac{s}{s^2+4} + \frac{e^{-2\pi s}}{(s^2+1)(s^2+4)}. \]Using partial fractions, \(\dfrac{1}{(s^2+1)(s^2+4)} = \dfrac{1/3}{s^2+1} - \dfrac{1/3}{s^2+4}\), so
\[ F(s) = \frac{s}{s^2+4} + \frac{e^{-2\pi s}}{3}\cdot\frac{1}{s^2+1} - \frac{e^{-2\pi s}}{3}\cdot\frac{1}{s^2+4}. \]Taking the inverse transform of each piece (applying the Second Translation Theorem to the last two, and using \(\sin(t-2\pi)=\sin(t)\), \(\sin(2(t-2\pi))=\sin(2t)\)),
\[ \boxed{y = \cos(2t) + \frac13\sin(t)\mathscr{U}(t-2\pi) - \frac16\sin(2t)\mathscr{U}(t-2\pi).} \]Differentiating a Laplace Transform
Suppose \(F(s) = \mathcal{L}\{f(t)\}\). Since \(F(s)\) is defined by an integral in \(t\) with parameter \(s\), we can differentiate under the integral sign with respect to \(s\):
\[ \frac{d}{ds}F(s) = \frac{d}{ds}\int_0^\infty e^{-st}f(t)\,dt = \int_0^\infty \frac{\partial}{\partial s}\big(e^{-st}\big)f(t)\,dt = \int_0^\infty -te^{-st}f(t)\,dt = -\mathcal{L}\{tf(t)\}. \]So \(\mathcal{L}\{tf(t)\} = -F'(s)\). Differentiating a second time,
\[ \frac{d^2}{ds^2}F(s) = \int_0^\infty \frac{\partial^2}{\partial s^2} \big(e^{-st}\big)f(t)\,dt = \int_0^\infty t^2e^{-st}f(t)\,dt = \mathcal{L}\{t^2f(t)\}. \]Continuing this pattern gives the general result.
We know \(\mathcal{L}\{\sin(4t)\} = F(s) = \dfrac{1}{s^2+16}\). By the theorem with \(n=1\),
\[ \mathcal{L}\{t\sin(4t)\} = -F'(s) = -\frac{d}{ds}\left(\frac{1}{s^2+16}\right) = -\left(\frac{-2s}{(s^2+16)^2}\right). \]Thus
\[ \boxed{\mathcal{L}\{t\sin(4t)\} = \frac{2s}{(s^2+16)^2}.} \]We know \(\mathcal{L}\{\cos(2t)\} = F(s) = \dfrac{s}{s^2+4}\). By the theorem with \(n=2\), we need \(F''(s)\). First,
\[ F'(s) = \frac{(s^2+4)-s(2s)}{(s^2+4)^2} = \frac{4-s^2}{(s^2+4)^2}. \]Differentiating again (quotient rule, with \(u=4-s^2\), \(v=(s^2+4)^{-2}\), so \(u'=-2s\), \(v'=-4s(s^2+4)^{-3}\)),
\[ F''(s) = -2s(s^2+4)^{-2} + (4-s^2)\big({-4s(s^2+4)^{-3}}\big) = \frac{-2s(s^2+4) - 4s(4-s^2)}{(s^2+4)^3} = \frac{2s^3-24s}{(s^2+4)^3}. \]So \(F''(s) = \dfrac{2s(s^2-12)}{(s^2+4)^3}\), and since \((-1)^2=1\),
\[ \boxed{\mathcal{L}\{t^2\cos(2t)\} = \frac{2s(s^2-12)}{(s^2+4)^3}.} \]