Oliver Heaviside (1850--1925).
Brief History of Laplace Transform
The Laplace transform is named after the French mathematician and astronomer Pierre-Simon Laplace (1749--1827). However, he did not actually invent what we now call the Laplace transform. Indeed, Laplace himself, a notoriously vain and selfish
person in spite of his scientific genius, was careful to credit Leonhard Euler (1707--1783) with the basic formula. Well before the
work of Laplace, however, mathematical genius Leonhard Euler had studied differential equations. One of his many
noteworthy contributions in this field was the idea of transforming a function X(x) into a new function z via
the equation
\[
z = \int e^{ax}\, X(x)\,{\text d} x ,
\]
which looks fairly similar to the modern Laplace transform, only with an indefinite rather than a definite integral.
In a 1753 paper (entitled
Methodus aequationes differentiales altiorum graduum integrandi ulterius promota --
it’s a good thing mathematicians don’t use Latin any more…), Euler used methods based on this transform to give a
systematic method of solving second order linear differential equations. Later in his career, he further clarified the
method and introduced the definite integral form
\[
y(u) = \int_a^b e^{K(u)\, Q(x)}\, P(x)\,{\text d} x .
\]
In particular, this expression appeared in Euler’s 1768
Institutiones Calculi Integralis, vol. II, where he used it to solve the equation
\[
L\, \frac{{\text d}^2 y(u)}{{\text d} u^2} + M\, \frac{{\text d} y(u)}{{\text d}u} + N\,y(u) = U(u) ,
\]
if
\[
U(u) = R(a)\, e^{K(u)\, Q(a)} .
\]
However, Euler did not pursue this topic very far. Joseph Louis
Lagrange (1736--1813), born as Giuseppe Lodovico Lagrangia in
Turin, Italy, who succeeded Euler (since Leonhard returned to Russia) as the
director of mathematics at the Prussian Academy of Sciences in
Berlin, began to study integrals in the form \(
\int_0^{\infty} f(t)\,e^{-at}\,\mathrm{d}t \)
in connection with his work on integrating probability density
functions. Laplace was the next person to seriously work on this
topic, and took a critical step forward by applying the idea of a
"transformation" rather than just looking for a solution in the form
of an integral. He looked for solutions with the following equation:
\( \int x^{s}\,\phi (x)\,\mathrm{d}x = F(s) .\) In 1809, Laplace extended his transform to
find solutions that diffused indefinitely -- giving us our popular Laplace transformation. Laplace appeared to have
quickly understood the importance of his discovery, as he went on to use Laplace transforms numerous times in his
later work and generalize his integrals to create Fourier and Mellin transforms as well. So while Laplace may not
have invented his transforms, he certainly deserves credit for producing a systematic body of theory that went far
beyond anything created by his predecessors.
Although the results had been published for at least 70 years, the
transformation was not given a true physical and mathematical meaning until Oliver Heaviside (1850--1925) came up with completely new ideas on his own
in the 1880s. His predecessors used (what we call now the Laplace)
integral as an analytical tool and never even tried to establish the
inverse Laplace transform---without this tool the theory is not
complete at all. Oliver Heaviside did not actually
use (and most likely was unfamiliar with) the Laplace integral in his derivations because it was not needed in his pioneering work
in establishing the operator methods.
Let us pause explanations of Heaviside's achievements and focus on his
biography and historical circumstances that surrounded his science breakthrough.
Oliver Heaviside was born on May 18, 1850 in Camden Town, London,
England. He caught scarlet fever when he was a young child and this
affected his hearing. This was to have a major effect on his life,
making his childhood unhappy, with relations between himself and
other children difficult. However, his school results were rather
good, and in 1865 he was placed fifth from five hundred pupils. Academic
subjects seemed to hold little attraction for Heaviside, however, and
at age 16 he left school. Perhaps he was more disillusioned with
school than with learning since he continued to study (Oliver was completely
self-taught) after leaving
school, in particular he learned Morse code, studied electricity, and
studied further languages, in particular Danish and German. He was
aiming at a career as a telegrapher, in this he was advised and
helped by his uncle Charles Wheatstone (the piece of electrical
apparatus known as the Wheatstone bridge is named after him).
In 1868, Heaviside went to Denmark and became a telegrapher. He
progressed quickly in his profession and returned to England in 1871
to take up a post in Newcastle upon Tyne in the office of Great
Northern Telegraph Company, which dealt with overseas
traffic. Heaviside became increasingly deaf but he worked on his own
research into electricity. While still working as chief operator
in Newcastle he began to publish papers on electricity, the first in
1872, and then the second in 1873, which was of sufficient interest to
James Clerk Maxwell (1831--1879) that he mentioned the results in the
second edition of his
Treatise on Electricity and Magnetism. Despite this hatred of
rigor, Heaviside was able to greatly simplify Maxwell's twenty
equations with twenty variables, replacing them by four equations with two
vector variables (the electric field E and the magnetic
field B). Today we call these 'Maxwell's equations' forgetting that
they are in fact 'Heaviside's equations.'
Heaviside went on to achieve further advances in knowledge, again
receiving less than his just desserts. In a 1887 paper Heaviside gave,
for the first time, the conditions necessary to transmit a signal
without distortion. In Electromagnetic Theory (1893--1912), he
postulated that an electric charge would increase in mass as its
velocity increases, an anticipation of an aspect of Einstein’s special
theory of relativity. Heaviside was elected a Fellow of the Royal
Society in 1891, perhaps the greatest honor he ever received. In 1902
Heaviside predicted that there was a conducting layer in the
atmosphere which allowed radio waves to follow the Earth's
curvature. This layer in the atmosphere, the Heaviside layer, is named
after him. Its existence was proved in 1923 when radio pulses were
transmitted vertically upward and the returning pulses from the
reflecting layer were received.
■
On the eve of the twentieth century, new physical ideas required new mathematical tools to utilize these theories. It is not an accident that many new mathematical notations and definitions were introduced by physicists, but not mathematicians. We cannot neglect to mention the famous Dirac's delta-function as well as Einstein summation
abbreviation. The latest example
gave us the Internet, including the HTML protocol. Therefore,
Heaviside's invention of operational calculus has naturally evolved to quantum mechanics and then as a byproduct, a useful
technique for solving initial value problems for differential
equations.
The Laplace transform is an integral transformation because it
maps a function of a real variable t (as a rule time) to a function of a complex variable λ (complex frequency):
\[
f^L (\lambda ) \equiv {\cal L} (f) = \int_0^{\infty} f(t)\,e^{-\lambda t} \,{\text d} t ,
\]
which mathematicians credited to Laplace. However, to determine the
inverse
Laplace transform it took about 20 years until a group of
people came up with the answer. The first publication was made
by the English mathematician
Thomas John I'Anson Bromwich (1875--1929)
who showed that the inverse Laplace transform can be expressed as
the
contour integral
\[
{\cal L}^{-1} (F) = \mbox{V.P. }\frac{1}{2\pi {\bf j}} \,
\int_{s-{\bf j}\infty}^{s+{\bf j}\infty} F(\lambda )\,e^{\lambda t}
\,{\text d} \lambda = \frac{1}{2\pi {\bf j}} \, \lim_{\omega \to \infty}
\int_{s-{\bf j}\omega}^{s+{\bf j}\omega} F(\lambda )\,e^{\lambda t}
\,{\text d} \lambda \qquad ({\bf j}^2 = -1),
\]
where abbreviation V.P. stands for the
Cauchy principal value, which indicates that the symmetrical
improper integration is done along the vertical line Reλ
=
s in the
complex plane such that
s is greater than the
real part of all singularities of
F(λ)
and
F(λ) is bounded on the line of integration.
This integral is usually referred to as the
Bromwich integral. The problem of the recovery of a real-valued function
f(
t) for
t ≥ 0, given its Laplace transform
\[
\int_0^{\infty} f(t)\,e^{-s\, t} \,{\text d} t = g(s)
\]
for real values of
s ∈ ℝ is actually solving the
integral equation of the first kind. It is an example of ill-posed problem in the sense of
J. Hadamard, who was the first to introduce the concept of a
well-posed problem in 1902. Ill-posed problems may suffer from
numerical instability when solved with finite precision, or with errors in the data. A typical example of an ill-posed problem is numerical differentiation. The Laplace transformation can be transformed into the convolution problem with the following substitution:
\[
s = a^x \qquad\mbox{and} \qquad t = a^{-y} , \quad\mbox{ where }\ a > 1.
\]
Then
\[
g \left( a^x \right) = \int_{-\infty}^{\infty} \ln a \, e^{-a^{x-y}} f \left( a^{-y} \right) a^{-y} {\text d} y .
\]
Multiplying both sides by 𝑎
x, we obtain the convolution equation
\[
\int_{-\infty}^{\infty} K(x-y)\,F(y)\,{\text d}y = G(x) ,
\]
where
\begin{align*}
G(x) &= a^x g \left( a^x \right) = s\, g(s) , \\
K(x) &= \ln a\, a^x e^{-a^s} = \ln a\,s\,e^{-s} , \\
F(y) &= f \left( a^{-y} \right) = f(t).
\end{align*}
A convolution equation occurs widely in the applied sciences. Such equations are usually solved with the aid of filtered approximation, for example with
Tikhonov regularization.
To this day, the Inverse Laplace Transform is the most difficult process to understand when solving differential
equations in this method. In this tutorial, we will not use the above formal definition of the inverse Laplace
transform; instead, we apply the residue method based on the novel approach by Vladimir Dobrushkin.
The revolutionary idea of Heaviside's operational calculus consisted in establishing the spectral
decomposition of the (unbounded) derivative operator \(
\texttt{D} = {\text d}/{\text d} t \) acting in the space
of functions on half line \( [0, \infty )
. \) This means that the operator was replaced by simple
multiplication. We will meet this approach in the second part of this
tutorial when spectral decomposition will be applied to square
matrices.
Due to his research, Heaviside dealt with multi-degree differential
equations for electrical systems in 1880s. To solve corresponding
initial value problems, Heaviside substituted the derivative operator by a
letter p (we use contemporary notation \(
\texttt{D} \) instead), which yields algebraic
equations. After that, he quickly solved the algebraic equation and proceeded with a solution to the
original differential equation. These
kinds of equations (usually with 10 or more derivatives of dependent
variable) would usually take days or weeks for most people to solve.
Heaviside was able to solve these kinds of equations within hours.
Between 1880--1887, he invented operational calculus
-- a new method for solving the differential equations.
Unfortunately, Oliver Heaviside failed to explain how he had derived the
solutions to the initial value problems, and as often happens the work of geniuses is not often understood. It took many years for
academia to accept his results because of a lack of a rigorous proof for
his methods. He replied to this criticism
with the famous statement "Mathematics is an experimental science, and
definitions do not come first, but later on," claiming that "I do not
refuse my dinner simply because I do not understand the process of
digestion." As a result, mathematicians did not accept his technique
and tried to explain what is going on behind his manipulations. After
many years of work, it became clear that Heaviside used the integral
transformation.
The Laplace transformation, denoted by ℒ, is the mapping of a set
X of integrable functions on half line [0, ∞) that grow no faster than exponentials into another set
Y of holomorphic complex-valued factions in half plane Re
z ≤
h, so ℒ:
X ↦
Y. However, when the Laplace transform is used in differential equations, it is usually applied to subset of
X that contains piecewise smooth functions having almost everywhere on half line [0, ∞) derivatives.
Let
\( \texttt{D} = {\text d}/{\text d}t \) be the derivative operator. Then the Laplace transform provides the spectral decomposition to the derivative operator, that is,
\[
\left( {\cal L}\texttt{D} f\right) (\lambda ) = \lambda\,{\cal L}f (\lambda ) .
\]
This means that the Laplace transform converts differentiation into multiplication operation. In general, when
\( P[\texttt{D}] = p_0 \texttt{D}^0 + p_1 \texttt{D} + \cdots + p_n \texttt{D}^n \) is a differential operator with constant coefficients, then the Laplace transform converts it into multiplication by a function:
\[
\left( {\cal L}\, M\left[ \texttt{D}\right] f\right) (\lambda ) = M\left[ \lambda\right] {\cal L}f (\lambda ) .
\]
Essentially, we will be doing almost the same thing that Heaviside had done-- we will take a differential equation,
subject to the initial conditions
\[
{\it L} (y) \equiv a_n \frac{{\text d}^n y}{{\text d} t^n} + a_{n-1} \frac{{\text d}^{n-1} y}{{\text d} t^{n-1}} +
\cdots + a_0 y(t) = f(t) , \qquad y(0) = y_0 , \ \dot{y} (0) = y_1 , \ldots ,\left. \frac{{\text d}^{n-1} y}{{\text d} t^{n-1}}
\right\vert_{t=0} = y_{n-1} ,
\]
use differential operator notation
\(
\texttt{D} = {\text d}/{\text d} t \) to create an algebraic equation
\[
a_n \texttt{D}^n y + a_{n-1} \texttt{D}^{n-1} y +
\cdots + a_0 y(t) \equiv \left( a_n \texttt{D}^n + a_{n-1} \texttt{D}^{n-1} +
\cdots + a_0 \right) y = f(t) ,
\]
that can be solved for the Laplace Transform of "y" (our dependent variable), which we will denote by
\( y^L (\lambda ) , \) apply the Laplace transform to the given initial value problem
to reduce our problem to an algebraic equation
\[
\left( a_n \lambda^n + a_{n-1} \lambda^{n-1} + \cdots + a_0 \right) y^L = f^L + a_n \left( y(0) + \lambda \dot{y} (0) + \cdots + y_{n-1} \right) + \cdots ,
\]
solve this algebraic equation to represent the Laplace transform of unknown function
\( y^L (\lambda ) \)
as a ratio of two polynomials in λ:
\[
y^L (\lambda ) = f^L \, \frac{1}{{\it L} (\lambda )} + \frac{P(\lambda )}{{\it L} (\lambda )} ,
\]
where
\( {\it L}(\lambda ) = a_n \lambda^n + a_{n-1} \lambda^{n-1} + \cdots + a_0 \) is the characteristic polynomial for the differential operator
\( {\it L}(\texttt{D}) \) of order
n and
P(λ) is the polynomial
of degree less than
n in λ depending on the initial conditions;
finally, we apply the inverse Laplace transform to obtain the
required solution. We again will consider Initial Value Problems (IVP) with specified initial conditions that have
unique solutions. Thus, our answer will also be unique when using the Laplace Transform method -- we will
have no arbitrary constants. Something to note, however, is that this method is ONLY valid for "
t" (our independent
variable) when "
t" is > 0. We will be using a Heaviside function that will allow our answer to maintain these
conditions (the Heaviside function will be more well defined in the section labeled "
Heaviside Function")
-
Hadamard, Jaques, Sur les problèmes aux dérivées partielles et leur signification physique. Princeton University Bulletin, 1902, pp. 49--52.
-
Halmos, P.R., What does the spectral theorem say?, The American Mathematical Monthly, 1963, Vol. 70, No. 3, pp. 241--247.
-