The Laplace transform
A force switches on at one instant and off at another. Can an initial value problem with that forcing be solved in one pass, by algebra, instead of solving again after every switch?
Where you are 6 in Systems and transforms
Why this exists
At a swimming pool, Zoe films the tip of a springboard while Mateo sets a sandbag on the tip at t = 1 second and lifts it off at t = 3. The forcing is 0, then a constant, then 0 again. Undetermined coefficients can handle it only in pieces: solve on [0, 1], carry y(1) and y'(1) into a new initial value problem on [1, 3], solve again, carry the values to t = 3, and solve a third time. Every switch costs a new problem.
The Laplace transform replaces a function of t by a function of a new variable s, chosen so that differentiation becomes multiplication by s and a switch at t = a becomes a factor e^{-as}. A linear initial value problem with constant coefficients becomes one algebraic equation with the initial conditions already inside it, and partial fractions turn its solution back into a function of t. Throughout, y is a function of t \ge 0, time; lower-case letters are functions of t and capitals are their transforms.
Definition
A function f is piecewise continuous on [0, \infty) when each bounded interval [0, b] contains only finitely many points where f is discontinuous, and at each of them (and at 0, from the right) the one-sided limits of f exist and are finite. It is of exponential order c when there are constants K and T with
A function of exponential order c is of every order c' > c too. Polynomials, e^{at}, \cos bt and \sin bt all have some exponential order (Exercise 19); e^{t^2} has none.
The Laplace transform of f is the function
defined at each real s for which the improper integral converges. Only the values of f for t \ge 0 enter. We write \mathcal{L}\{t^2\} for the transform of the function t \mapsto t^2.
For a \ge 0, the unit step that switches on at a is
Multiplying by u(t - a) switches a function on at t = a, and for a < b the difference u(t - a) - u(t - b) is 1 on [a, b) and 0 elsewhere.
If F = \mathcal{L}\{f\} for some piecewise continuous f of exponential order, then f is an inverse transform of F, written f = \mathcal{L}^{-1}\{F\}. By Result 6 it is determined except at its jumps, and at most one inverse transform is continuous.
Intuition
The weight e^{-st} is 1 at t = 0 and dies quickly when s is large, so F(s) is a weighted total of f in which large s sees only small t. That is why F(s) \to 0 as s \to \infty.
Integrate e^{-st} f'(t) by parts: the derivative moves onto e^{-st}, which brings down a factor s, and the boundary term leaves -f(0). So \mathcal{L}\{f'\} = sF - f(0), and an equation with derivatives in t becomes an equation without derivatives in s. A delay is as simple: a function switched on at t = a is weighed from a instead of 0, which pulls out a factor e^{-as}.
Each simple fraction in s belongs to one exponential polynomial: \displaystyle \frac{1}{s - r} to e^{rt}, \displaystyle \frac{1}{(s - r)^{k+1}} to \displaystyle \frac{t^k}{k!}e^{rt}, a quadratic factor to a damped cosine or sine. The roots of the denominator are the exponents in the solution, which is why the characteristic roots of Constant-coefficient equations reappear.
Results
In Results 1 to 6, f and g are piecewise continuous on [0, \infty) and of exponential order c, with transforms F and G.
1. Existence and decay. The integral F(s) converges absolutely for every s > c, and F(s) \to 0 as s \to \infty.
2. Linearity. For constants \alpha, \beta and s > c, \mathcal{L}\{\alpha f + \beta g\} = \alpha F + \beta G.
3. Basic transforms. For an integer n \ge 0 and real a, b:
| f(t) | F(s) | valid for |
|---|---|---|
| 1 | \dfrac{1}{s} | s > 0 |
| t^n | \dfrac{n!}{s^{n+1}} | s > 0 |
| e^{at} | \dfrac{1}{s - a} | s > a |
| \cos bt | \dfrac{s}{s^2 + b^2} | s > 0 |
| \sin bt | \dfrac{b}{s^2 + b^2} | s > 0 |
| u(t - a), a \ge 0 | \dfrac{e^{-as}}{s} | s > 0 |
4. Shifts. For real a and s > c + a,
For a \ge 0 and s > c,
5. Derivatives. Let f be continuous on [0, \infty) and of exponential order c, differentiable except at finitely many points of each bounded interval, with f' piecewise continuous. Then for s > c,
If f, f', \dots, f^{(n-1)} are continuous and of exponential order c and f^{(n)} is piecewise continuous, then for s > c
6. Uniqueness (Lerch’s theorem). If F(s) = G(s) for all s greater than some s_0, then f(t) = g(t) at every t > 0 where both are continuous. Two continuous functions with the same transform are equal.
7. Initial value problems. Let a \ne 0, b, c be real, let p(s) = as^2 + bs + c be the characteristic polynomial, and let f be piecewise continuous and of exponential order. The initial value problem
has exactly one solution on [0, \infty): a function y with y and y' continuous, y'' existing and the equation holding except at the jumps of f. Its transform is, for all large s,
With y_0 = y_1 = 0 this is Y = H F, where \displaystyle H(s) = \frac{1}{p(s)} is the transfer function. A first-order equation b y' + c y = f, y(0) = y_0, with b \ne 0, gives \displaystyle Y = \frac{F + b y_0}{bs + c} in the same way.
Why it is true
Result 1. Take K, T with |f(t)| \le Ke^{ct} for t \ge T. On [0, T], f is bounded, say by B: each of its finitely many pieces extends continuously to a closed interval, where a continuous function is bounded. For s > c and t \ge T, |e^{-st}f(t)| \le Ke^{-(s-c)t}, and \displaystyle \int_T^\infty Ke^{-(s-c)t}\,dt = \frac{K}{s-c}e^{-(s-c)T} is finite. By the comparison test of Improper integrals, \int_T^\infty |e^{-st}f|\,dt converges. Absolute convergence gives convergence: with h = e^{-st}f, 0 \le |h| - h \le 2|h|, so \int_T^\infty (|h| - h)\,dt converges by comparison, and h = |h| - (|h| - h). The integral over [0, T] is an ordinary one. For s > \max(c, 0),
which tends to 0.
Result 2. The integral over [0, b] is linear; let b \to \infty, both limits existing by Result 1.
Result 5. Fix b > 0 and list the points 0 = t_0 < t_1 < \dots < t_k = b where f' fails to exist or is discontinuous, together with 0 and b. On each [t_{j-1}, t_j], e^{-st}f is continuous with a derivative that extends continuously to the closed interval, so integration by parts (Integration by parts) applies:
Add over j. Because f is continuous, the boundary terms telescope to e^{-sb}f(b) - f(0). For s > c, |e^{-sb}f(b)| \le Ke^{-(s-c)b} \to 0 and the last integral tends to F(s), so \int_0^b e^{-st}f'\,dt \to sF(s) - f(0). The general formula follows by induction: apply the case n = 1 to f^{(n-1)} and substitute the formula for \mathcal{L}\{f^{(n-1)}\}.
Result 3. For real a, \displaystyle \int_0^b e^{-(s-a)t}\,dt = \frac{1 - e^{-(s-a)b}}{s - a} for s \ne a, which tends to \displaystyle \frac{1}{s-a} as b \to \infty exactly when s > a (for s = a the integral is b). The case a = 0 is the transform of 1. For the step, \displaystyle \int_a^\infty e^{-st}\,dt = \frac{e^{-as}}{s} for s > 0. The line for t^n is Exercise 18. For f = \cos bt: f and f' are bounded, so of order 0, and f'' = -b^2 f, f(0) = 1, f'(0) = 0. Result 5 gives s^2F - s = -b^2F for s > 0, so \displaystyle F = \frac{s}{s^2+b^2}. For \sin bt, f(0) = 0 and f'(0) = b give s^2F - b = -b^2F.
Result 4. e^{-st}e^{at}f(t) = e^{-(s-a)t}f(t), whose integral is F(s - a), convergent when s - a > c. For the second shift the integrand vanishes for t < a; substituting \tau = t - a,
Result 6. We prove it for continuous f and g. Let h = f - g, continuous and of exponential order c, with transform 0 for s > s_0. Choose an integer N > \max(s_0, c). For each integer n \ge 0, s = N + n + 1 gives \int_0^\infty e^{-(N+n+1)t}h(t)\,dt = 0. Substitute x = e^{-t}, so \displaystyle dt = -\frac{dx}{x}:
As x \to 0^+, t \to \infty and |\varphi| \le Ke^{-(N - c)t} \to 0, so \varphi(0) = 0 makes \varphi continuous on [0, 1]. By linearity \int_0^1 P\varphi\,dx = 0 for every polynomial P. By the Weierstrass approximation theorem (proved in real analysis, alongside Pointwise and uniform convergence) there are polynomials P_k with \max_{[0,1]}|\varphi - P_k| \to 0. Then
so \int_0^1 \varphi^2 = 0, and a continuous \varphi with this property is 0. Hence h = 0. For piecewise continuous f and g the theorem is stated without proof here; the same substitution works with \varphi approximated in the mean, as in D. V. Widder’s The Laplace Transform.
Result 7. We prove existence, uniqueness and the exponential order of y when f is an exponential polynomial on each of finitely many intervals, which covers every example here. On the first interval, Result 5 of Undetermined coefficients gives exactly one solution, an exponential polynomial. Its values y and y' at the switch are the initial values of the next problem, which again has exactly one solution, and so on. Each piece is of exponential order (Exercise 19), hence so is y, and so are y' and \displaystyle y'' = \frac{1}{a}(f - by' - cy). For general f these facts follow from the formula of Variation of parameters and are stated without proof here. Now Result 5 with n = 2, and Result 2, transform the equation into
and solving for Y gives the formula. By Result 6, y is the one continuous inverse transform of Y. The first-order case is the same with n = 1.
Worked examples
A ramp that levels off, from the definition
Find the transform of f(t) = t for 0 \le t < 3, f(t) = 3 for t \ge 3.
Solutionf is continuous and bounded, so of order 0, and F(s) exists for s > 0. Split the integral at 3. By parts, with \displaystyle -\frac{1}{s}e^{-st} as antiderivative of e^{-st},
The tail is \displaystyle 3\int_3^\infty e^{-st}\,dt = \frac{3e^{-3s}}{s}. The terms in \displaystyle \frac{e^{-3s}}{s} cancel:
Check with Result 4: f(t) = t - (t - 3)\,u(t - 3), so \displaystyle F = \frac{1}{s^2} - e^{-3s}\frac{1}{s^2}.
Transforms and inverses from the table
Find \mathcal{L}\{4t^3 - 6e^{-2t} + e^{t}\cos 3t\}, and find \mathcal{L}^{-1}\left\{\dfrac{2s + 5}{s^2 + 4s + 13}\right\}.
SolutionBy linearity and Result 3, with the first shift (a = 1) for the last term,
valid where all three are, for s > 1.
For the inverse, the denominator has no real roots, so complete the square: s^2 + 4s + 13 = (s + 2)^2 + 9. Write the numerator in powers of s + 2: 2s + 5 = 2(s + 2) + 1. Then
Each piece is a table entry shifted by a = -2, so the inverse is \displaystyle e^{-2t}\left(2\cos 3t + \frac{1}{3}\sin 3t\right).
An initial value problem with a repeated factor
Solve y'' - 4y' + 3y = 4e^{3t}, y(0) = 1, y'(0) = 0.
SolutionHere p(s) = s^2 - 4s + 3 = (s - 1)(s - 3) and \displaystyle F = \frac{4}{s-3}. Result 7 with a = 1, b = -4 gives numerator F + s - 4:
Since 3 is a root of p and the forcing’s exponent, s - 3 is repeated. Write \displaystyle Y = \frac{A}{s-1} + \frac{B}{s-3} + \frac{C}{(s-3)^2}. Covering s - 1 and setting s = 1 gives \displaystyle A = \frac{10}{4} = \frac{5}{2}; multiplying by (s - 3)^2 and setting s = 3 gives \displaystyle C = \frac{4}{2} = 2; the coefficient of s^2 in the numerator gives A + B = 1, so \displaystyle B = -\frac{3}{2}. By Result 3 and the first shift, \displaystyle \mathcal{L}^{-1}\{\frac{1}{(s-3)^2}\} = te^{3t}, so
Check: \displaystyle y(0) = \frac{5}{2} - \frac{3}{2} = 1 and \displaystyle y'(0) = \frac{5}{2} - \frac{9}{2} + 2 = 0. The term 2te^{3t} is the particular solution with the factor t^m, m = 1, of Undetermined coefficients: there the multiplicity of a root, here a repeated factor.
The springboard
The tip of Zoe’s springboard, y decimeters below rest at time t seconds, obeys
with y(0) = y'(0) = 0: the sandbag alone would hold the tip 1 decimeter down. Find y and its largest value.
SolutionBy Result 3, \displaystyle F = \frac{5(e^{-s} - e^{-3s})}{s}, so Result 7 gives Y = (e^{-s} - e^{-3s})\,G(s) with \displaystyle G(s) = \frac{5}{s(s^2 + 2s + 5)}. Partial fractions: \displaystyle G = \frac{A}{s} + \frac{Bs + C}{s^2 + 2s + 5} with \displaystyle A = \frac{5}{5} = 1; matching 5 = (s^2 + 2s + 5) + (Bs + C)s gives B = -1, C = -2. With s^2 + 2s + 5 = (s + 1)^2 + 4 and s + 2 = (s + 1) + 1,
so \displaystyle g(t) = 1 - e^{-t}\left(\cos 2t + \frac{1}{2}\sin 2t\right). By the second shift,
Since g(0) = g'(0) = 0, y and y' are continuous at 1 and 3, while y'' jumps by 5 and -5, as the forcing does. On [1, 3], y = g(t - 1) and \displaystyle g'(t) = \frac{5}{2}e^{-t}\sin 2t, positive on \displaystyle (0, \frac{\pi}{2}) and negative on \displaystyle (\frac{\pi}{2}, 2]. So the largest value there is at \displaystyle t = 1 + \frac{\pi}{2} \approx 2.57: y = 1 + e^{-\pi/2} \approx 1.21 decimeters, an overshoot past the level 1. The critical points of g are \displaystyle \tau = \frac{k\pi}{2}, where g = 1 - (-1)^k e^{-k\pi/2}, so g \le 1 + e^{-\pi/2}, and g > 0 for \tau > 0. For t > 3, y = g(t - 1) - g(t - 3) < g(t - 1): no later value is larger. Both terms tend to 1, so y \to 0, with swings shrinking like e^{-t}. The figure shows the whole motion, and one formula carried it across both switches.
A forced system by Cramer’s rule
Solve x_1' = -2x_1 + x_2 + 3, x_2' = x_1 - 2x_2 with x_1(0) = x_2(0) = 0.
SolutionEach x_i is an exponential polynomial (Linear systems of differential equations with a constant particular solution), so Result 5 applies to both. Transforming each equation,
The determinant is (s + 2)^2 - 1 = (s + 1)(s + 3), and Cramer's rule gives
Three distinct linear factors, each found by covering: \displaystyle X_1 = \frac{2}{s} - \frac{3/2}{s + 1} - \frac{1/2}{s + 3} and \displaystyle X_2 = \frac{1}{s} - \frac{3/2}{s + 1} + \frac{1/2}{s + 3}. So
Both start at 0. The eigenvalues -1, -3 of the matrix appear as the roots of the determinant, and (2, 1) is the equilibrium the forcing creates.
A short proof: periodic functions
Prove that if f is piecewise continuous on [0, \infty) and P-periodic, then for s > 0
and find \mathcal{L}\{|\sin t|\}.
Solutionf is bounded on [0, P], hence everywhere, so it is of order 0 and F(s) converges for s > 0 by Result 1. Since the improper integral converges, it is the limit along b = nP:
Substitute t = \tau + kP and use f(\tau + kP) = f(\tau): the kth integral is e^{-kPs}I(s), where I(s) = \int_0^P e^{-s\tau}f(\tau)\,d\tau. The geometric series with ratio e^{-Ps} < 1 sums to \displaystyle \frac{I(s)}{1 - e^{-Ps}}.
For |\sin t|, P = \pi and |\sin t| = \sin t on [0, \pi]. An antiderivative of e^{-st}\sin t is \displaystyle -\frac{e^{-st}(s\sin t + \cos t)}{s^2 + 1}, so \displaystyle I(s) = \frac{1 + e^{-\pi s}}{s^2 + 1} and
Boundaries
The hypotheses of Result 1 are sufficient, not necessary. t^{-1/2} is unbounded near 0, so not piecewise continuous, yet its transform is \displaystyle \sqrt{\pi/s}. \displaystyle \frac{1}{t} has no transform: the integral diverges at 0 for every s. e^{t^2} has none either (Exercise 20).
The method needs linear equations with constant coefficients. A coefficient t turns \mathcal{L}\{ty'\} into a derivative of sY in s (by differentiating under the integral sign, not proved here), so a variable-coefficient equation becomes another differential equation. For a nonlinear term such as y^2 there is no rule at all: \mathcal{L}\{y^2\} is not a function of Y.
The transform of a product is not the product of transforms: \displaystyle \mathcal{L}\{t \cdot t\} = \frac{2}{s^3}, while \displaystyle \left(\frac{1}{s^2}\right)^2 = \frac{1}{s^4}. The function whose transform is FG is the convolution \int_0^t f(\tau)g(t - \tau)\,d\tau, beyond this lesson.
Not every function of s is a transform: by Result 1 a transform tends to 0, so 1, s and \displaystyle \frac{s}{s + 1} are not (Exercise 21). The input that would have transform 1 is an instantaneous impulse, which is no function.
An inverse transform is unique only up to values at jumps: the step with \displaystyle u(0) = \frac{1}{2} has the same transform as u(t). And the second shift needs both factors: f(t - a)\,u(t - a) is f delayed, while f(t)\,u(t - a) is f cut off, with a different transform (Exercise 17).
Where this goes next
Looking back at Springs, damping and resonance, a spring has transfer function \displaystyle \frac{1}{m s^2 + \gamma s + k}: its poles are the characteristic roots, and with \gamma = 0 forcing at the natural frequency squares the factor s^2 + \omega_0^2, whose inverse carries the growing t\sin\omega_0 t of resonance. Convolution and the impulse, which this lesson only names, complete the method for arbitrary forcing. For a random variable X \ge 0 with density f, F(s) = E[e^{-sX}] is the moment-generating function at -s. Equations with variable coefficients, where the transform gives no algebra, are the subject of Power series solutions, and Fourier series treats periodic forcing frequency by frequency.
Exercises
Transforms
- From the definition, find the transform of f(t) = 2 for 0 \le t < 4, f(t) = 0 for t \ge 4, for s > 0.
- Find \mathcal{L}\{5 - 3t^2 + 4e^{-2t}\} and where it is valid.
- Find \mathcal{L}\{e^{3t}\sin 2t - t^2e^{-t}\} and where it is valid.
- Find \mathcal{L}\{t\,u(t - 2)\}.
- Let f be 2-periodic with f(t) = 1 on [0, 1) and f(t) = -1 on [1, 2). Find F(s) for s > 0.
Inverse transforms
- \mathcal{L}^{-1}\left\{\dfrac{s + 7}{s^2 - s - 6}\right\}
- \mathcal{L}^{-1}\left\{\dfrac{3s - 4}{s^2 + 6s + 25}\right\}
- \mathcal{L}^{-1}\left\{\dfrac{4}{s^2(s + 2)}\right\}
- \mathcal{L}^{-1}\left\{\dfrac{e^{-2s}}{s^2 + 9}\right\}
- \mathcal{L}^{-1}\left\{\dfrac{10}{(s - 1)(s^2 + 4)}\right\}
Initial value problems
- y' + 3y = 6, y(0) = 1.
- y'' + 4y' + 3y = 0, y(0) = 2, y'(0) = 0.
- y'' + 2y' + y = 4e^{-t}, y(0) = 0, y'(0) = 1.
- y'' + 4y = 8u(t - 1), y(0) = y'(0) = 0.
- y' + 2y = 4\big(1 - u(t - 1)\big), y(0) = 0. Give y for t \ge 1 in closed form.
- x_1' = x_1 + 2x_2, x_2' = 2x_1 + x_2, x_1(0) = 1, x_2(0) = 0.
Proofs
- Let f be piecewise continuous and of exponential order c, and a \ge 0. Prove \mathcal{L}\{f(t)\,u(t - a)\} = e^{-as}\,\mathcal{L}\{f(t + a)\} for s > c, and use it to redo Exercise 4.
- Prove \mathcal{L}\{t^n\} = \dfrac{n!}{s^{n+1}} for s > 0 and every integer n \ge 0.
- Prove that t^k e^{\alpha t}\cos\beta t and t^k e^{\alpha t}\sin\beta t are of exponential order \alpha + 1, and that every exponential polynomial has some exponential order.
- Prove that e^{t^2} has no Laplace transform: the integral diverges for every real s.
- Prove that no piecewise continuous f of exponential order has F(s) = \dfrac{s}{s + 1}.
- Let f be piecewise continuous and of exponential order c. Prove \mathcal{L}\left\{\int_0^t f(\tau)\,d\tau\right\} = \dfrac{F(s)}{s} for s > \max(c, 0).
Answers
- \dfrac{2(1 - e^{-4s})}{s}. The integral is \displaystyle 2\int_0^4 e^{-st}\,dt = \frac{2}{s}(1 - e^{-4s}).
- \dfrac{5}{s} - \dfrac{6}{s^3} + \dfrac{4}{s + 2} for s > 0, by linearity and \displaystyle \mathcal{L}\{t^2\} = \frac{2}{s^3}.
- \dfrac{2}{(s - 3)^2 + 4} - \dfrac{2}{(s + 1)^3} for s > 3, by the first shift with a = 3 and a = -1.
- e^{-2s}\left(\dfrac{1}{s^2} + \dfrac{2}{s}\right). Write t = (t - 2) + 2 and use the second shift on \big((t - 2) + 2\big)u(t - 2).
- \dfrac{1 - e^{-s}}{s(1 + e^{-s})}, which is \dfrac{\tanh(s/2)}{s}. Here \displaystyle \int_0^2 e^{-st}f\,dt = \frac{(1 - e^{-s})^2}{s}; divide by 1 - e^{-2s} = (1 - e^{-s})(1 + e^{-s}) as in Example 6.
- 2e^{3t} - e^{-2t}. The denominator is (s - 3)(s + 2); covering gives \displaystyle \frac{10}{5} = 2 and \displaystyle \frac{5}{-5} = -1.
- e^{-3t}\left(3\cos 4t - \dfrac{13}{4}\sin 4t\right). Here s^2 + 6s + 25 = (s + 3)^2 + 16 and 3s - 4 = 3(s + 3) - 13.
- 2t - 1 + e^{-2t}. With \displaystyle \frac{A}{s} + \frac{B}{s^2} + \frac{C}{s + 2}: B = 2, C = 1, and A + C = 0.
- \dfrac{1}{3}\sin\big(3(t - 2)\big)\,u(t - 2), by the second shift applied to \displaystyle \frac{1}{3}\sin 3t.
- 2e^{t} - 2\cos 2t - \sin 2t. Covering gives \displaystyle \frac{2}{s - 1}; then 10 = 2(s^2 + 4) + (Bs + C)(s - 1) gives B = -2, C = -2.
- y = 2 - e^{-3t}. \displaystyle Y = \frac{1 + 6/s}{s + 3} = \frac{s + 6}{s(s + 3)} = \frac{2}{s} - \frac{1}{s + 3}.
- y = 3e^{-t} - e^{-3t}. \displaystyle Y = \frac{2s + 8}{(s + 1)(s + 3)}; covering gives 3 and -1.
- y = te^{-t} + 2t^2e^{-t}. \displaystyle Y = \frac{1}{(s + 1)^2} + \frac{4}{(s + 1)^3}, and \displaystyle \mathcal{L}\{t^2e^{-t}\} = \frac{2}{(s + 1)^3}.
- y = 2\big(1 - \cos 2(t - 1)\big)\,u(t - 1). \displaystyle Y = \frac{8e^{-s}}{s(s^2 + 4)} and \displaystyle \frac{8}{s(s^2 + 4)} = \frac{2}{s} - \frac{2s}{s^2 + 4}.
- y = 2(1 - e^{-2t}) - 2\big(1 - e^{-2(t-1)}\big)u(t - 1); for t \ge 1, y = 2(e^2 - 1)e^{-2t}. \displaystyle Y = \frac{4(1 - e^{-s})}{s(s + 2)} and \displaystyle \frac{4}{s(s + 2)} = \frac{2}{s} - \frac{2}{s + 2}.
- x_1 = \dfrac{e^{3t} + e^{-t}}{2}, x_2 = \dfrac{e^{3t} - e^{-t}}{2}. The transformed system (s - 1)X_1 - 2X_2 = 1, -2X_1 + (s - 1)X_2 = 0 has determinant (s - 3)(s + 1), so \displaystyle X_1 = \frac{s - 1}{(s - 3)(s + 1)} and \displaystyle X_2 = \frac{2}{(s - 3)(s + 1)}.
- The function t \mapsto f(t + a) is piecewise continuous, and |f(t + a)| \le Ke^{ca}e^{ct} for t \ge T, so it has order c and its transform exists for s > c. The integrand of \mathcal{L}\{f(t)u(t - a)\} vanishes for t < a, and substituting t = \tau + a, \int_a^b e^{-st}f(t)\,dt = e^{-as}\int_0^{b-a} e^{-s\tau}f(\tau + a)\,d\tau. Let b \to \infty. For Exercise 4, f(t + 2) = t + 2 has transform \displaystyle \frac{1}{s^2} + \frac{2}{s}.
- Induction on n. For n = 0 it is Result 3 for 1. For n \ge 1, f = t^n is continuous with f' = nt^{n-1} continuous and f(0) = 0, and both are of every order c > 0, since t^ne^{-ct} \to 0. Result 5 gives n\,\mathcal{L}\{t^{n-1}\} = s\,\mathcal{L}\{t^n\} for s > c, hence for every s > 0. So \displaystyle \mathcal{L}\{t^n\} = \frac{n}{s}\cdot\frac{(n-1)!}{s^n} = \frac{n!}{s^{n+1}}.
- Since t^ke^{-t} \to 0 as t \to \infty (by repeated l’Hôpital), there is T with t^k \le e^{t} for t \ge T. Then |t^ke^{\alpha t}\cos\beta t| \le t^ke^{\alpha t} \le e^{(\alpha + 1)t} for t \ge T, and the same for sine. An exponential polynomial is a finite sum \sum_j \lambda_j g_j of such terms, |g_j| \le e^{c_jt} for t \ge T_j; with c = \max c_j and T = \max T_j, |\sum_j \lambda_j g_j| \le \big(\sum_j|\lambda_j|\big)e^{ct} for t \ge T.
- Fix s and let t_1 = \max(s, 0). For t \ge t_1, e^{-st}e^{t^2} = e^{t(t - s)} \ge 1, so \int_{t_1}^b e^{-st}e^{t^2}\,dt \ge b - t_1 \to \infty. The integral over [0, t_1] is finite, so the improper integral diverges.
- By Result 1 every such transform tends to 0 as s \to \infty, but \displaystyle \frac{s}{s + 1} \to 1.
- Let g(t) = \int_0^t f(\tau)\,d\tau. Then g is continuous, g(0) = 0, and by the The fundamental theorem of calculus g' = f except at the jumps of f, so g' is piecewise continuous. Let |f| \le B on [0, T] and |f(t)| \le Ke^{c't} for t \ge T, where c' = c if c > 0 and otherwise c' is any number in (0, s). For t \ge T, \displaystyle |g(t)| \le BT + \int_T^t Ke^{c'\tau}\,d\tau \le \left(BT + \frac{K}{c'}\right)e^{c't}. So Result 5 applies for s > c': F(s) = sG(s) - g(0) = sG(s), and \displaystyle G = \frac{F}{s}.