6.11 Practice Problems

These are the tutorial questions for this section, worked in full.

Problem 6.11.1. Use the definition to find the Laplace transform, stating for which \(s\) the integral converges. \[(a)\ 9t^3-e^{-t}\hspace {0.7cm}(b)\ 3e^{4t}\hspace {0.7cm} (c)\ 2\cosh 7t+3\sinh 7t\hspace {0.7cm}(d)\ \big (3e^{-4t}+4\big )^2\]

Show solution

Solution. The two facts needed are \(\mathcal {L}\{t^n\} = \dfrac {n!}{s^{n+1}}\) for \(s>0\) and \(\mathcal {L}\{e^{at}\} = \dfrac {1}{s-a}\) for \(s>a\); both follow from the defining integral, and each converges only where the exponential decays.

(a)

\[\mathcal {L}\big \{9t^3-e^{-t}\big \} = \frac {54}{s^4} - \frac {1}{s+1}, \qquad s>0 .\]

(b)

\(\dfrac {3}{s-4}\), for \(s>4\).

(c)

Using \(\cosh 7t = \dfrac {e^{7t}+e^{-7t}}{2}\) and the corresponding form for \(\sinh \), \[\frac {2s}{s^2-49} + \frac {21}{s^2-49} = \frac {2s+21}{s^2-49}, \qquad s>7 .\]

(d)

Expand first: \(\big (3e^{-4t}+4\big )^2 = 9e^{-8t}+24e^{-4t}+16\), so \[\frac {9}{s+8} + \frac {24}{s+4} + \frac {16}{s},\qquad s>0 ,\] the binding condition coming from the constant term.

Problem 6.11.2. Find the Laplace transform of \[(a)\ \sin \Big (3t-\frac 12\Big )\hspace {0.8cm}(b)\ \cos 3t\sin 4t \hspace {0.8cm}(c)\ 5e^{-6t}+3t^4+3\cosh 4t-2\sinh 5t\]

Show solution

Solution.

(a)

This is not a shift in \(t\) of the kind the second shifting theorem handles, since there is no unit step; expand instead: \[\sin \Big (3t-\frac 12\Big ) = \sin 3t\cos \tfrac 12 - \cos 3t\sin \tfrac 12 ,\] so the transform is \[\frac {3\cos \frac 12 - s\sin \frac 12}{s^2+9}.\]

(b)

Use the product-to-sum identity \(\cos A\sin B = \dfrac {\sin (A+B)-\sin (A-B)}{2}\), giving \(\cos 3t\sin 4t = \dfrac {\sin 7t+\sin t}{2}\), so the transform is \[\frac 12\Big (\frac {7}{s^2+49} + \frac {1}{s^2+1}\Big ).\]

(c)

Term by term, \[\frac {5}{s+6} + \frac {72}{s^5} + \frac {3s}{s^2-16} - \frac {10}{s^2-25}.\]

Problem 6.11.3. Find the Laplace transform of \((a)\ (t-3)^3u(t-7)\) and \((b)\ t^3e^{-4t}u(t-2)\).

Show solution

Solution. The second shifting theorem is \(\mathcal {L}\{g(t)u(t-a)\} = e^{-as}\mathcal {L}\{g(t+a)\}\), so the function must first be rewritten in terms of \(\tau = t-a\).

(a)

Here \(a=7\), and with \(t = \tau +7\), \[(t-3)^3 = (\tau +4)^3 = \tau ^3+12\tau ^2+48\tau +64 .\] Hence \[\mathcal {L} = e^{-7s}\Big (\frac {6}{s^4} + \frac {24}{s^3} + \frac {48}{s^2} + \frac {64}{s}\Big ).\] Note the shift in the step is \(7\) while the shift inside the cube is \(3\); taking them to be the same is the usual error.

(b)

Here \(a=2\), and with \(t = \tau +2\), \[t^3e^{-4t} = (\tau +2)^3e^{-4\tau -8} = e^{-8}\big (\tau ^3+6\tau ^2+12\tau +8\big )e^{-4\tau }.\] Using the first shifting theorem on each term, \[\mathcal {L} = e^{-2s-8}\Big (\frac {6}{(s+4)^4} + \frac {12}{(s+4)^3} + \frac {12}{(s+4)^2} + \frac {8}{s+4}\Big ).\]

Problem 6.11.4. Use \(\mathcal {L}\{f^{(n)}\} = s^nF(s)-s^{n-1}f(0)-\cdots -f^{(n-1)}(0)\) to find the transforms of \((a)\ \sinh ^2\omega t\) and \((b)\ t\sin \frac {\pi t}{2}\).

Show solution

Solution.

(a)

Let \(f = \sinh ^2\omega t\), so \(f(0)=0\) and \(f' = 2\omega \sinh \omega t\cosh \omega t = \omega \sinh 2\omega t\). Then \[\mathcal {L}\{f'\} = sF(s) - 0 \implies \omega \cdot \frac {2\omega }{s^2-4\omega ^2} = sF(s),\] so \[F(s) = \frac {2\omega ^2}{s\big (s^2-4\omega ^2\big )}.\] The identity \(\sinh ^2\omega t = \dfrac {\cosh 2\omega t-1}{2}\) gives the same answer directly, and agreement is worth checking.

(b)

Here the quicker route is the derivative-of-transform rule \(\mathcal {L}\{tf(t)\} = -F'(s)\) with \(f = \sin \frac {\pi t}{2}\), whose transform is \(\dfrac {\pi /2}{s^2+\pi ^2/4}\). Differentiating, \[\mathcal {L}\Big \{t\sin \frac {\pi t}{2}\Big \} = \frac {\pi s}{\big (s^2+\frac {\pi ^2}{4}\big )^2}.\]

Problem 6.11.5. Find \(\displaystyle {\mathcal {L}\Big \{\int _0^t\big (x^2-\cosh 3x\big )dx\Big \}}\) (i) by evaluating the integral first, and (ii) by the formula \(\mathcal {L}\big \{\int _0^tf\big \} = \dfrac {F(s)}{s}\).

Show solution

Solution.

(i)

\[\int _0^t\big (x^2-\cosh 3x\big )dx = \frac {t^3}{3} - \frac {\sinh 3t}{3},\] whose transform is \[\frac 13\cdot \frac {6}{s^4} - \frac 13\cdot \frac {3}{s^2-9} = \frac {2}{s^4} - \frac {1}{s^2-9}.\]

(ii)

\(F(s) = \mathcal {L}\{t^2-\cosh 3t\} = \dfrac {2}{s^3} - \dfrac {s}{s^2-9}\), so \[\frac {F(s)}{s} = \frac {2}{s^4} - \frac {1}{s^2-9},\] the same. The second route avoids the integration entirely, which is the point of setting the question both ways.

Problem 6.11.6. Use \(\mathcal {L}\{t^nf(t)\} = (-1)^n\dfrac {d^nF}{ds^n}\) to find the transforms of \[(a)\ t^ne^{-\alpha t}\hspace {0.7cm}(b)\ 3t^4e^{5t}+t\hspace {0.7cm} (c)\ (t-4)^2e^{-3t}\hspace {0.7cm} (d)\ \int _0^{\infty }te^{-2t}\cos 2t\,dt\]

Show solution

Solution.

(a)

With \(F = \dfrac {1}{s+\alpha }\), differentiating \(n\) times gives \(\dfrac {(-1)^nn!}{(s+\alpha )^{n+1}}\), so \[\mathcal {L}\big \{t^ne^{-\alpha t}\big \} = \frac {n!}{(s+\alpha )^{n+1}}.\]

(b)

By (a) with \(\alpha = -5\) and \(n=4\), \[\frac {3\cdot 4!}{(s-5)^5} + \frac {1}{s^2} = \frac {72}{(s-5)^5} + \frac {1}{s^2}.\]

(c)

Expand and use the first shifting theorem: \((t-4)^2e^{-3t} = \big (t^2-8t+16\big )e^{-3t}\), so \[\frac {2}{(s+3)^3} - \frac {8}{(s+3)^2} + \frac {16}{s+3}.\]

(d)

This is not a transform but a number: it is \(\mathcal {L}\{t\cos 2t\}\) evaluated at \(s=2\). Now \[\mathcal {L}\{t\cos 2t\} = -\frac {d}{ds}\frac {s}{s^2+4} = \frac {s^2-4}{\big (s^2+4\big )^2},\] and at \(s=2\) the numerator vanishes, so \[\int _0^{\infty }te^{-2t}\cos 2t\,dt = 0 .\]

Problem 6.11.7. Find the Laplace transforms of \((a)\ \dfrac {\cosh ^2t}{t}\) and \((b)\ \dfrac {3-3e^{-3t}}{t^3}\).

Show solution

Solution. Neither transform exists. The rule \(\mathcal {L}\Big \{\dfrac {f(t)}{t}\Big \} = \int _s^{\infty }F(u)\,du\) requires \(\dfrac {f(t)}{t}\) to remain bounded as \(t\rightarrow 0\), and in both cases it does not.

(a)

\(\cosh ^2 0 = 1\), so near the origin \(\dfrac {\cosh ^2t}{t}\sim \dfrac 1t\) and \[\int _0 \frac {\cosh ^2t}{t}e^{-st}dt\] diverges logarithmically at the lower limit. Numerically, cutting the integral off at \(t=\varepsilon \) with \(s=5\) gives \(4.77\), \(7.07\) and \(9.37\) for \(\varepsilon = 10^{-3}, 10^{-4}, 10^{-5}\) — growing without bound, as \(\ln \frac 1\varepsilon \) does.

(b)

Here \(3-3e^{-3t} = 9t - \frac {27}{2}t^2+\cdots \), so the quotient behaves like \(\dfrac {9}{t^2}\) near the origin, and the divergence is worse.

The versions that do converge

Each becomes a standard exercise if the power of \(t\) matches the order of vanishing of the numerator. Since \(\sinh ^2 t\) vanishes to second order, \[\mathcal {L}\Big \{\frac {\sinh ^2t}{t}\Big \} = \int _s^{\infty }\frac {2\,du}{u\big (u^2-4\big )} = \frac 14\ln \frac {s^2}{s^2-4},\] and with a single power of \(t\), \[\mathcal {L}\Big \{\frac {3-3e^{-3t}}{t}\Big \} = 3\int _s^{\infty }\Big (\frac 1u-\frac {1}{u+3}\Big )du = 3\ln \frac {s+3}{s}.\] Both were confirmed numerically. The first of these is exactly what the next question needs.

Problem 6.11.8. Find the inverse Laplace transforms of \[(a)\ \frac {3}{s^{5/2}}\hspace {0.6cm} (b)\ \frac {1}{s^2}-\frac {e^{-s}}{s^2}+\frac {e^{-2s}}{s}\hspace {0.6cm} (c)\ \frac {24}{s^4+20s^2+64}\hspace {0.6cm} (d)\ \ln \frac {s^2}{s^2-4}\hspace {0.6cm} (f)\ \frac {2s+3}{s^2+6s+13}\]

Show solution

Solution.

(a)

Using \(\mathcal {L}\{t^a\} = \dfrac {\Gamma (a+1)}{s^{a+1}}\) with \(a = \dfrac 32\), and \(\Gamma \big (\tfrac 52\big ) = \dfrac {3\sqrt \pi }{4}\), \[\mathcal {L}^{-1}\Big \{\frac {3}{s^{5/2}}\Big \} = \frac {3t^{3/2}}{\Gamma (5/2)} = \frac {4t^{3/2}}{\sqrt \pi }.\]

(b)

Each term is read with the second shifting theorem: \[t - (t-1)u(t-1) + u(t-2).\]

(c)

The denominator factors as \(s^4+20s^2+64 = \big (s^2+4\big )\big (s^2+16\big )\), and \[\frac {24}{\big (s^2+4\big )\big (s^2+16\big )} = 2\Big (\frac {1}{s^2+4} - \frac {1}{s^2+16}\Big ),\] so the inverse is \[\sin 2t - \frac {\sin 4t}{2}.\]

(d)

From the previous question, \(\mathcal {L}\Big \{\dfrac {\sinh ^2t}{t}\Big \} = \dfrac 14\ln \dfrac {s^2}{s^2-4}\), so \[\mathcal {L}^{-1}\Big \{\ln \frac {s^2}{s^2-4}\Big \} = \frac {4\sinh ^2t}{t} = \frac {2\big (\cosh 2t-1\big )}{t},\] confirmed numerically at \(s=5\).

(f)

Complete the square: \(s^2+6s+13 = (s+3)^2+4\), and write the numerator to match: \[\frac {2s+3}{(s+3)^2+4} = \frac {2(s+3)-3}{(s+3)^2+4},\] so the inverse is \[e^{-3t}\Big (2\cos 2t - \frac 32\sin 2t\Big ).\]

Problem 6.11.9. Solve the initial value problems.

(a)
\(y''+4y'+4y = e^{-4t}\), \(y(0)=y'(0)=1\)
(b)
\(y'''+2y''+y' = e^{3t}\), \(y(0)=y'(0)=y''(0)=0\)

Show solution

Solution.

(a)

Transforming and using the initial values, \[\big (s^2Y-s-1\big ) + 4\big (sY-1\big ) + 4Y = \frac {1}{s+4},\] so \[(s+2)^2Y = \frac {1}{s+4} + s + 5 .\] Hence \[Y = \frac {1}{(s+4)(s+2)^2} + \frac {s+5}{(s+2)^2}.\] Partial fractions give \[Y = \frac {1}{4(s+4)} + \frac {3}{4(s+2)} + \frac {7}{2(s+2)^2},\] and inverting, \[y = \frac {e^{-4t}}{4} + \Big (\frac 34 + \frac {7t}{2}\Big )e^{-2t}.\]

(b)

With all initial values zero, \[\big (s^3+2s^2+s\big )Y = \frac {1}{s-3} \implies Y = \frac {1}{s(s+1)^2(s-3)}.\] Partial fractions give \[Y = -\frac {1}{3s} + \frac {1}{48(s-3)} + \frac {5}{16(s+1)} + \frac {1}{4(s+1)^2},\] so \[y = -\frac 13 + \frac {e^{3t}}{48} + \Big (\frac {5}{16} + \frac {t}{4}\Big )e^{-t}.\]

Problem 6.11.10. Solve \(t^2y''+ty'-y = 0\).

Show solution

Solution. This is an Euler–Cauchy equation, and the Laplace transform is the wrong tool for it — transforming would produce a differential equation in \(s\) rather than an algebraic one, since the coefficients are not constant.

Trying \(y = t^m\) gives \[m(m-1) + m - 1 = m^2-1 = 0 ,\] so \(m = \pm 1\) and \[y = At + \frac {B}{t}.\]

Problem 6.11.11. Solve the integral equation \(\displaystyle {y(t) = t^2 + 6\int _0^t(t-x)^4y(x)\,dx}\).

Show solution

Solution. The integral is a convolution, \(\big (t^4 * y\big )(t)\), so transforming turns it into a product: \[Y = \frac {2}{s^3} + 6\cdot \frac {4!}{s^5}\,Y = \frac {2}{s^3} + \frac {144}{s^5}Y .\] Solving, \[Y\Big (1-\frac {144}{s^5}\Big ) = \frac {2}{s^3} \implies Y = \frac {2s^2}{s^5-144}.\] That the convolution becomes multiplication is exactly why integral equations of this kind are approached by transform rather than directly.

Questions on this section

Stuck on something here? Ask below and it stays attached to this topic.