Lecture Notes 17 Calculus in the Eighteenth Century

You might also like

Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 25

Lecture Notes 17: Calculus in the Eighteenth Century

A. Britain

1. De Moivre

Abraham de Moivre (1667-1754) was a French Protestant who, after suffering persecution for his
religious beliefs in France, found refuge in England in 1688 where he remained for the rest of his life.
Unable as a foreigner to obtain a university appointment, he earned a meagre living from private
tuition and by advising gamblers and underwriters in the coffee houses of London on matters of
probability. He became a good friend of both Edmund Halley, the Secretary of the Royal Society, and
with Isaac Newton, who used to collect him from Slaughter’s Coffee House most evenings and invite
him to his house for philosophical discourse. In 1697 on Newton’s recommendation he was elected a
fellow of the Royal Society. 1n 1710 he was appointed to the Royal Society Commission which
adjudicated in favour of Newton against Leibniz over priority in the discovery of the calculus, but
despite his involvement in this dispute he managed to maintain a good relationship with Johann
Bernoulli, Leibniz’ main cheerleader on the Continent.

De Moivre made a particular study of Newton’s Principia. He purchased a copy and cut out the pages
so that he could carry a few of them at all times and read them as he walked around London, going
from one pupil to the next. He was quickly able to master this difficult work. In later life when people
used to pose Newton questions about the Principia, he would refer them to De Moivre, saying, ‘ He
understands these things better than I do.’ That at least is the story. De Moivre’s first academic
paper, presented to the Royal Society in 1695, arose from his study of fluxions in the Principia.

His most important mathematical work was The Doctrine of Chances, first published in 1718 with
two further editions in 1738 and 1756. In this text on probability he not only stated general rules but
applied them to various games of chance common in his time.

For example he solved the problem of de Méré which asks how many throws of two dice are
necessary to ensure an even chance of throwing at least one double six. He began by noting that if a
denotes the number of chances of a successful outcome while b denotes the number of chances of
an unsuccessful outcome on any one throw, then bx/(a + b)x represents the probability of x
consecutive failures. Since there is to be an even chance of success, x must satisfy the equation
bx/(a + b)x = ½. Dividing both the numerator and denominator of the left hand side of this equation
by bx and writing a/b as 1/q, he obtained (1 + 1/q)x = 2. He solved for x by taking logarithms,
obtaining x = log 2/ log (1 + 1/q). After expanding log (1 + 1/q) in a power series de Moivre
concluded that ‘if q is infinite or pretty large in respect to unity’, the first term 1/q of the series is
sufficient and the solution can be written x = q log 2  0.7 q. Since the odds of throwing a double six
are 35 : 1 against, it follows that q = 35, which implies that the required number of throws in
de Méré ‘s problem is roughly 24.
2

But even more important than these calculations was his discussion of the problem of summing the
terms of the binomial (a + b)n in which the normal approximation to the binomial distribution made
its first appearance. Initially de Moivre restricted himself to the case where the chances of success
and failure are equally likely i.e. where a = b = 1. He sought to find the probability of n/2 successes in
n trials, that is, the ratio of the middle term of (1 + 1)n to the sum of all the terms, namely 2n, for
n large and even. He determined, having derived a formula for n factorial, that this ratio approaches
2/(2n). Next he showed that if Q is a term lying at a distance t from the middle term M, then for
large n, log (Q/M)   2t2/n i.e. Q  M exp ( 2t2/n). In modern notation this means that

P (X = n/2 + t)  2/(2n) exp ( 2t2/n)

De Moivre understood that the expression on the right hand side represents a curve with two
inflection points, each at a distance n/2 from the maximum point. Thus de Moivre had in effect
found what is called today the normal probability density function

f (x) = [1/(2) ] exp  (x  )2 / 2 2,

as can be seen if n/2 is substituted for  , n/2 is substituted for  and t is substituted for x   .

k
In order to find the sum of the probabilities for the different values of t, he approximated ∑P
t =0
(X = n/2 + t), by an integral which can be expressed in modern notation by 2/(2n)
k

∫ exp ( 2t /n) dt. He evaluated this integral by expanding the integrand in a power series and
2

0
integrating term by term. For k = n/2 he calculated that the value of the integral is 0.341344. He
concluded that

‘ if it was possible to take an infinite number of experiments, the probability that an event
which has an equal number of chances to happen or fail shall neither appear more frequently than
n/2 + n/2 nor more rarely than n/2  n/2 will be expressed by the double sum of the number
exhibited … that is, 0.682688.’

This was the first glimpse of what has come to be known as the central limit theorem, which implies
that when n is large, a normal distribution provides a good approximation to a binomial distribution.
In this case, where the chances of success and failure are equal, it can be seen that De Moivre’s
values of n/2 and n/2 represent, respectively, the expected value and standard deviation of a
binomial distribution with parameters n, p, where p = 1/2. De Moivre has approximated this
distribution by a normal distribution and has computed the probability that the number of successes
will fall within what is now called one standard deviation either side of the expected value. His figure
of 0.682688 only differs from the modern value in the sixth decimal place. However de Moivre
himself attached no importance to his function other than as an approximation to the binomial
distribution.

To illustrate de Moivre’s result with a simple example expressed in modern notation, suppose a fair
coin is tossed 100 times. What is the probability of obtaining between 45 and 55 heads? The
expected value of this binomial experiment, where the probability p of obtaining heads on any toss
is ½, is given by np = 50 and the standard deviation is given by [np (1  p)]= n/2 = 5. This binomial
distribution can be approximated by a normal distribution with expected value  = 50 and standard
deviation  = 5. So the probability of obtaining a number of heads within 5 (i.e. within one standard
deviation) of the expected value of 50 is approximately 0.68.
3

2. Taylor

In 1715 Brook Taylor (1685-1731), for a time Secretary of the Royal Society, published a slim volume
entitled The Method of Increments. The first part describes his method of finite differences and its
relationship to Newton’s fluxional calculus while the second part contains applications of these ideas
to problems in mathematics and mechanics. Several of these problems, such as the determination of
the period of a vibrating string, were treated for the first time by Taylor, but others, like the
catenary, had been considered earlier by Continental mathematicians, in particular Huygens, Leibniz
and Johann Bernoulli. Taylor did not help himself by failing to acknowledge any of his predecessors
except Newton. Bernoulli accused Taylor of passing off as his own ‘things which belong to others’
while Leibniz said that such a man is ‘not fit to carry out the office of Secretary of the Royal Society’.
(see Appendix 1). Nevertheless there is evidence that Taylor derived many of his important
theorems independently. His treatise, despite its obscure style and confusing notation, subsequently
influenced mathematicians both in Britain and on the Continent, notably Maclaurin, Daniel Bernoulli,
D’Alembert, Euler and Lagrange.

The series bearing his name appears as Proposition 7 of his treatise, though he had been anticipated
in obtaining power series expansions by the Scottish mathematician James Gregory (1638-1675) and
also by Newton, who had given an explicit statement of the series in an early draft of On Quadrature
but which was omitted from the version actually published in 1704. The theorem states in modern
notation that

f (x + v) = f (x) + v f ’ (x) + v2 f ’’ (x) / 2! + v3 f ’’’ (x) / 3! + …

Taylor’s proof, slightly modernised, was as follows. Starting with a function f (x) = z, he computed
the increments in z produced by small increments in x, at each stage adding the fluxion of the
previous expression:
4

f (x) = z

f x + ẋ ) = z + ż

f [x + ẋ + ( ẋ + ẍ )] = f (x + 2 ẋ + ẍ ) = z + 2 ż + z̈

f (x + 3 ẋ + 3 ẍ + ⃛x ) = z + 3 ż + 3 z̈ + ⃛z

… …

Taylor then simplified the left hand side of his table by assuming that ẋ is a constant, so that ẍ =
⃛x = ….. = 0. Furthermore he noted that the coefficients on the right hand side are simply the
coefficients of Newton’s binomial expansion. Therefore after n iterations

f (x + n ẋ ) = z + n ż + [n (n  1)/1.2] z̈ + [n (n  1) (n  2)/1.2.3] ⃛z + ……

At this point he (implicitly) allowed n to become infinitely great, so that (approximately)


n = (n  1) = (n  2) = ….. He also let v = n ẋ or n = v/ ẋ , obtaining

f (x + v) = z + v ż / ẋ + v2 z̈ /1.2 ẋ 2 + v3 ⃛z /1.2.3 ẋ 3 + … ,

which is his theorem.

If now ż / ẋ is written as dz/dx, z̈ / ẋ 2 as d2z/dx2 and so on, the previous equation becomes

f (x + v) = z + v dz/dx + (v2/2!) d2z/dx2 + (v3/3!) d3z/dx3 + …,

or, given that z = f (x),

f (x + v) = f (x) + v f’(x) + v2 f ‘‘ (x)/2! + v3 f ’’’ (x)/3! + … ,

which is the modern form of Taylor’s theorem.

It is important to note that in this theorem Taylor interpreted the fluxion ẋ , not as a velocity, but as
an infinitesimal distance, since in his view ‘the fluxions of quantities are proportional to the nascent
increments of the quantities’.

Taylor’s was the first proof of this theorem and he was the first to apply it to the solution of
problems such as the extraction of roots. If an approximation x0 has been found for an equation z
(x) = 0, where the true root is x0 + v, then the correction v can be determined by solving

0 = z (x0 + v) = z (x0) + z’ (x0) v + z’’ (x0) v2/2! + z’’’ (x0) v3/3! + …

Taylor explained that for sufficiently small v one can ignore terms containing v3 and higher powers
and solve the resulting quadratic equation 0 = z (x0) + z’ (x0) v + z’’ (x0) v2/2! for v. The solution v0 can
be added to x0 to produce a closer approximation x1 = x0 + v0 to the root. The same process can be
repeated for x1 and so on, so that the root can be approximated as closely as desired.
5

3. Simpson

Thomas Simpson (1710-1761), a weaver by trade, was a self-taught mathematician who initially
earned a living lecturing in London coffee houses. But after publishing several textbooks he obtained
a position in 1743 as Professor of Mathematics at the Royal Military Academy at Woolwich,
following which he was elected to the Royal Society.

In his New Treatise of Fluxions (1737) and The Doctrine and Application of Fluxions (1750) he
provided a comprehensive account of Newton’s method of fluxions and its applications, containing
numerous worked examples.

His derivation of the product rule for fluxions illustrates his skill. Suppose it is required to obtain the
fluxion of xy. Simpson let z = x + y. Then ż = ẋ + ẏ and z2 = x2 + y2 + 2xy, whence xy = ½ (z2 x2 y2).
Taking the fluxion of both sides of the last equation, Simpson obtained

˙ ) = z ż  x ẋ  y ẏ
( xy

= (x + y) ( ẋ + ẏ )  x ẋ  y ẏ

= x ẏ + y ẋ

His derivation of the quotient rule is equally ingenious. Suppose it is required to obtain the fluxion of
u/z. Simpson set x = u/z, so that xz = u and so by the product rule

x ż + z ẋ = u̇,

whence

ẋ = u̇/z  x ż /z

= u̇/z  u ż /z2

= (zu̇  u ż )/z2
6

Simpson was also the first mathematician to publish for the rule for differentiating the sine, using a
method discovered twenty years earlier by Roger Cotes (1682-1716), the editor of the second
edition of Newton’s Principia. Simpson stated that ‘the fluxion of any circular arc is to the fluxion of
its sine as is the radius (of the circle) to the cosine’:

In the figure adapted from Simpson’s original diagram, z denotes the arc of a circle of radius An = 1,
while x = Ab represents sin z and bn represents cos z. In the differential triangle nrm, the hypotenuse
nr represents ż , the fluxion of the arc, while mr represents ẋ , the fluxion of the sine. (Simpson
understood the fluxion of a variable at a point to be equal to the distance it would have moved had
it continued uniformly at its instantaneous velocity within a minimal unit time period). The
similarity of triangles Abn, nrm then implies that mr : nr = bn : An or ẋ : ż = cos x : 1. Thus if x = sin z,
then ẋ : ż = dx/dz = cos z.
7

4. Maclaurin

Colin Maclaurin (1698-1746) was made a Professor of Mathematics at the University of Aberdeen at
the age of 19 and was later appointed, upon Newton’s recommendation, to the Chair of
Mathematics at Edinburgh, where he remained for the rest of his life. In 1742 he published
A Treatise of Fluxions in which he discussed the entire range of problems to which calculus is
applied, including extremum points, points of inflection, asymptotes, curvature and the volume and
surface area of solids of revolution. He also extended Simpson’s work on trigonometry, showing for
example that ’the fluxion of an arc is to its tangent in the duplicate ratio of its cosine’
i.e. d/dz (tan z) = 1/ cos2 (z).

The series bearing his name appears in this work. Suppose that y can be expressed as a series in x,
⃛ , … represent the values of y ẏ , ÿ , ⃛y , … when x =
namely y = A + Bx + Cx2 + Dx3 + … , while E, Ė , Ë , E
0. Then

y = A + Bx + Cx2 + Dx3 + … , so E = A

ẏ = B + 2Cx + 3Dx2 + … , so Ė = B

ÿ = 2C + 3.2Dx + … , so Ë = 2C

y⃛ = 3.2 D + … , ⃛ = 3.2 D
so E

⃛ / 3!, etc, yielding


Therefore A = E, B = Ė , C = Ë /2!, D = E

⃛ x3/1x2x3 + …
y = E + Ė x + Ë x2/1x2 + E

In modern terms the theorem states that

f (x) = f (0) + f ‘(0) x + f’’ (0) x2/2! + f’’’ (0) x3/ 3! + … ,


usually called the Maclaurin series. Maclaurin acknowledged that this theorem had already been
published by Brook Taylor but in fact his method of proof was similar to Newton’s derivation.

Maclaurin used his version of the Taylor series to derive the theorem now known as the Euler
Maclaurin summation formula. This theorem, which was independently discovered by Euler at about
the same time, is considered to be one of the most important in mathematics. The aim of the
theorem is to express the sum of a progression in terms of an integral and derivatives, which are
usually simple to compute. Maclaurin’s version of his formula can be written in modern notation as

n −1 n

∑ f (i) = ∫ f ( z ) dz  1/2 [f(n)  f(0)] + 1/12 f 1(n)  f 1(0)]  1/720 [f 3(n)  f 3(0)]
i=0 0

+ 1/30240 [f 5(n)  f 5(0)]  …..

Maclaurin provided an iterative method to generate the coefficients  1/2, 1/12, 0, 1/720, 0,
1/30240 etc. He pointed out that the formula gives an approximate value for the sum, whose
accuracy can be improved by taking more terms, though neither he nor Euler gave an expression for
the remainder term when the series is truncated. Maclaurin proceeded to apply the summation
formula to a number of problems, verifying results obtained by Bernoulli, De Moivre and Stirling
among others. For example, he verified Bernoulli’s summation of the squares of the first n natural
numbers:

n n

∑ k =∫k
2 2
dk + (n2 – 0)/2 + [2n  0]/12 = n3/3 + n2/2 + n/6
0 0

It is worth noting that both he and Euler applied the formula to the Basel problem, to find the sum of
the squares of the reciprocals of the natural numbers. Maclaurin computed that

∑ 1 /k 2
= 1.64493. Euler extended the calculation to obtain a closer approximation,
1

1.64493400684822643647… Euler no doubt recognised this number as 2/6, a fact which he


famously proved in 1735.

It is clear that British and Continental mathematicians were aware of each other’s work on the
calculus during the eighteenth century. Indeed Euler and Maclaurin learned of each other’s
discovery of the summation formula before either had published. Nevertheless it is probably true to
say that British mathematicians, in remaining faithful to Newton’s dot notation as opposed to
Leibniz’ more flexible notation, cut themselves off from the mainstream of progress in the calculus
during the eighteenth century. Leibniz’ calculus only became accepted in England in the 1820’s
following the efforts of the Analytical Society.
9

B. The Continent

1. Fatio de Duiller

Nicolas Fatio de Duiller (1664-1753) was a Swiss mathematician and scientist who, though twenty
years younger, became a close friend of Isaac Newton. The two men shared an interest in biblical
prophecy and alchemy, though their relationship came to an abrupt end in 1694 for reasons which
are unclear. In 1699 Fatio published a mathematical pamphlet outlining his solution to the
brachistochrone problem (see Johann Bernoulli below) in which he proclaimed Newton as the first
inventor of the calculus, implying that Leibniz had borrowed some of his ideas. This pamphlet
unfortunately ignited the controversy between the respective followers of Newton and Leibniz over
priority in the discovery of the calculus.

Fatio had mastered the differential and integral calculus by 1687 with the encouragement of
Huygens, a remarkable achievement given that so little had been published on the subject by that
date. He seems to have been the first mathematician to discover the technique of introducing an
integrating factor to solve differential equations. His idea was to multiply an equation by a factor
xmyn so as to render it, where possible, in integrable form. He communicated this idea to Huygens
who in turn reported it to Leibniz. Later Fatio revealed his method to Newton who included it, with
due acknowledgement, in the manuscript version of his treatise On Quadrature.

One of the examples illustrating his method reported by Huygens to Leibniz is the equation

 2xy dx + 4x2 dy  y2 dy = 0,

which, after being multiplied through by the integrating factor y 5, becomes

 2xy- 4dx + 4x2y 5 dy  y 3 dy = 0.

This equation can be expressed as


d ( x2y 4) + d (y 2/2) = 0,

whence, integrating both sides, the solution is

 x2y 4 + y 2/2 = C.

(See Appendix 3 for a modern interpretation of this technique).

10

2. Johann Bernoulli

The Bernoulli brothers, Jakob (1654-1705) and Johann (1667-1748) were among the first to
understand and apply Leibniz’ calculus. Johann started his mathematical career as a pupil of Jakob
but eventually became a jealous rival. Upon his brother’s death Johann succeeded to the Chair of
Mathematics at the University of Basel, where he remained for the rest of his life, numbering Euler
among his pupils. He was an aggressive supporter of Leibniz in the dispute with Newton over priority
in the invention of the calculus (see Appendix 2).

Johann made important contributions to the calculus in the early years. In 1694 he derived an
integral formula to solve differential equations, using this to obtain the series expansions of the sine
and the logarithm. He wrote this formula, which he called his series universalissima, as

Integr. n dz = nz  (1/1.2) zz dn/dz + (1/1.2.3) zzz ddn/dx2  (1/1.2.3.4) zzzz ddd/dz3 + … ,

but in modern notation this would be written

 f ‘(t) dt = f’ (t).t  f’’(t).t2/2! + f’’’(t).t3/3!  …

Bernoulli’s formula, which can also be derived by repeated integration by parts, is an integrated
form of the Taylor series. So Bernoulli was indignant when Taylor published his series in 1715
without citing Bernoulli’s earlier version, though the evidence suggests that Taylor arrived at his
result independently. But in fact Bernoulli himself had been anticipated by Gregory, Newton and
Leibniz. Together with Leibniz Bernoulli also developed a technique of partial differentiation, a
discovery which they kept secret for twenty years for reasons which are not clear. Another discovery
was his integration of the function y = xx :
 10 xx dx = 1  1/22 + 1/33  1/44 + …

Bernoulli also contributed to L’Hospital’s Analysis of Infinitely Small Quantities, the first textbook on
the differential calculus. He was almost certainly responsible for L’Hospital’s famous rule (see
below).

11

In mechanics, along with Leibniz and Huygens, he obtained the differential equation of the catenary,
the curve made by a hanging chain. And most famously he solved the brachistochrone problem
which he had issued as a challenge to the mathematicians of Europe and to Newton in particular.
This was to find the quickest path of descent under gravity between two points lying in a vertical
plane where one point is not directly above the other. His solution was particularly interesting as it
was based on an apparently unrelated problem in optics. He realised that the behaviour of a body
falling in a field of varying gravitational attraction is analogous to that of a light ray propagating
through a medium of varying density. Both cases adhere to the principle of least time and so in both
cases Snell’s law applies. Using Snell’s law he succeeded in deriving the differential equation of the
curve which he recognised as the equation of a cycloid (the path of a fixed point on the rim of a
wheel as it rolls without slipping). Bernoulli noted that the brachistochrone is the same curve as the
tautochrone discovered by Huygens, namely the path along which the time taken by an object
sliding to its lowest point under gravity is independent of the point from which it starts. He remarked
on the simplicity of Nature which uses one curve, the cycloid, to serve two different functions. This
problem gave rise to an important branch of mathematics called the calculus of variations by Euler
whose aim is to determine the curve for which a given integral is a maximum or a minimum.

3. L’Hospital

Guillaume François Antoine de L’Hospital (1661-1704), born into a wealthy noble family, was an able
amateur mathematician with a gift for clear exposition. He paid Johann Bernoulli a monthly salary in
return for instruction on the calculus and for exclusive publishing rights over any mathematical
discoveries Bernoulli might make. By 1696 he decided that that he understood the differential
calculus well enough to write a treatise of his own. In that year he published the first calculus text,
Analysis of Infinitely Small Quantities for the Understanding of Curves, which proved extremely
successful, running into several editions over the course of the eighteenth century. Despite the
rivalry between British and Continental mathematicians it was translated into English in 1730
(though Leibniz’ d notation was changed into Newton’s dot notation – see Appendix 6).

12

In his textbook L’Hospital derived the basic differential formulas for sums, products, quotients and
roots in the manner of Leibniz. He obtained the tangent to parabolas, hyperbolas and several
transcendental curves such as the quadratrix and the spiral. He showed how to obtain maxima and
minima and points of inflection. He also gave the formula for the radius of curvature.

But his work is probably most famous as the source of ‘L’Hospital’s rule’ (even though it was
probably devised by Johann Bernoulli) for calculating the value of a function in fractional form at the
point where both numerator and denominator approach zero:

The figure shown above, where A is the origin and AB the x axis, shows three curves, all of which
(despite appearances) represent positive functions of x. AMD is the curve y (x) with ordinate
PM = y, ANB is the curve p (x) with ordinate PN = p and COB is the curve q (x) with ordinate PO = q.
The three curves are related by the relation y(x) = p(x)/q(x). At the point B, where x = a, suppose that
p = q = 0. It is required to evaluate the length of the ordinate BD at this point, in other words to
evaluate y where its value is given by 0/0.

L’Hospital took a point b on the axis infinitely close to B and drew the ordinate bd meeting the curve
ANB at f, the curve COB at g and the curve AMD at d. Thus the segment Bb can be regarded as dx,
bf can be regarded as ‘dp’ and the segment bg as ‘dq’. The vertical distance between D and d (not
shown) by the same token can be considered as ‘dy’. Given that y = p/q, the value of the ordinate bd
will be

y + dy = (p + dp) / (q + dq).
(where the d in the equation signifies Leibniz’ differential, not the point d in the diagram). Now as b
approaches B, the ordinate bd approaches BD = y, so that dy vanishes. But at the same time p
and q are equal to zero at B. So L’Hospital concluded that on the left hand side of the previous
equation only y will remain, while on the right hand side only dp/dq will remain and hence y = dp/dq.

He explained, ‘if the differential of the numerator be found and that be divided by the differential of
the denominator, after having made x = a, we shall have the value of the ordinate sought’.

13

The modern version of L’Hospital’s rule states that if f(x) and g(x) are differentiable functions on an
open interval containing a point a such that f(a) = g(a) = 0 but g‘ (x)  0 (except possibly at x = a),
then

lim x  a f(x) / g(x) = lim x  a f ‘ (x) / g‘ (x),

providing the limit on the right hand side exists.

This rule makes it possible to evaluate indeterminate quotients such as 0/0 as x approaches a limit.
For example, the value of sin x / x as x  0 is

lim x  0 sin x / x = lim x 0 (d (sin x)) / (d(x)) = lim x 0 cos x / 1 = 1.

Unfortunately L’Hospital died before he could write a volume on the integral calculus, so Bernoulli,
effectively released from his contract, eventually published his own lectures on that subject.

4. Euler

Leonhard Euler (1707-1783), the son of a pastor, studied at the University of Basel, where he
received extra-curricular tuition on Saturday afternoons from Johann Bernoulli. After gaining his
doctorate at Basel he obtained a teaching post in mathematics in 1727 at the University of St
Petersburg and in 1731 was promoted to Professor of Physics. He gained a worldwide reputation in
1735 when he solved the so-called Basel problem, which was to find the sum of the squares of the
reciprocals of the natural numbers. In 1741 he took up a post at the Berlin Academy at the invitation
of Frederick the Great of Prussia. He lived for over 25 years in Berlin where he wrote over 380
articles. He also wrote over 200 letters to Frederick’s niece on mathematics and physics which were
later collected into a best-selling popular science volume entitled Letters to a German Princess. In
1766 he returned to St Petersburg where he remained for the rest of his life. At that time he became
almost totally blind but this did not impair his astonishing productivity. For example in 1775 he
wrote on average one mathematical paper per week. Over his lifetime he wrote over 800 books and
papers amounting to over 25 000 pages. On the day he died he had spent the morning calculating
the orbit of the newly discovered planet Uranus.

14

a. Differentiation

Euler’s approach to the calculus was based on a study of functions rather than curves. In his
Institutiones Calculi Differentialis (1755) he provided a comprehensive account of differentiation
with numerous examples. To obtain the differential of a function his starting point was Leibniz’
equation dy = f (x + dx)  f (x). In respect of transcendental functions he expanded the right hand
side in an infinite series and deleted all infinitesimals of higher degree such as (dx)2, (dx)3 etc.

For example he computed the differential of the function y = log x as follows:

dy = log (x + dx)  log x

= log [(x + dx)/ x]

= log (1 + dx/x)

= dx/x  (dx/x)2/2 + (dx/x)3/3  … (using Nicholas Mercator’s expansion of log (1 + x))

= dx/x.

To differentiate sin x, Euler proceeded as follows:

d(sin x) = sin (x + dx)  sin x

= sin x cos dx + cos x sin dx  sin x (using the formula for sin (A + B))

= (sin x) (1  dx2/2! + dx4/4!  …) + (cos x) ( dx  dx3/3! + …)  sin x

(using the power series expansions of cos x and sin x)

= cos x dx.

Given the absence of a sound definition of higher order differentials, Euler attempted to circumvent
them altogether. He argued as follows. If t is a variable with a constant differential dt, then
d2t = d3t = … = 0. Now suppose that x is a function of t. Let dx = p dt, dp = q dt, dq = r dt … Then the
higher differentials of x become

d2x = d(dx) = d(p dt) = dp dt + p d2t = dp dt = q dt dt = q (dt)2.


Similarly d3x = r (dt)3 and so on. Thus the higher order differentials can be defined in terms of the first
order differential dt.

15

b. Integration

In the years 1768-1770 Euler published a three volume treatise on integral calculus, the Institutiones
Calculi Integralis in which he discussed many of the techniques familiar today, such as integration by
parts, integration by partial fractions, integration by substitution and so on. The bulk of the text,
however, deals with the solution of first, second and higher order ordinary differential equations,
concluding with a discussion of partial differential equations.

21`In another paper published in 1769 Euler gave the first detailed explanation of the use of double
integrals. As his first example he calculated the volume of the first octant of a sphere whose
equation is z = √ (a2 x2  y2).

Taking an element of area dx dy in the first quadrant of the circle in the xy plane, he noted that the
volume V of the solid column above this infinitesimal rectangle is given by V = √ (a2 x2 y2) dx dy,
which he then integrated, first with respect to y and then with respect to x, to obtain the solution
V = πa3/6 .

In the same paper he showed how to transform the integral  Z dx dy into a new integral with area
element dt dv. His result, in modern notation, is

  f(x, y) dx dy =   f(x(t, v), y(t, v)) x y / t v  x y / v t dt dv,

(where the scaling termx/t . y/v  x/v . y/t is now known as the Jacobian.)

Euler also used his mastery of complex analysis to evaluate non-standard integrals. Towards the end
of his life, for example, he showed that 0 sin x /x dx = /2,an integral of considerable importance
in mathematics and mathematical physics. Around the same time he also showed that 0
sin (s2) ds = ½ (/2)

16

c. Differential Equations

Euler made notable advances in the field of differential equations, many of which would have arisen
in his work in mathematical physics. For example, Euler (independently of Clairaut) discovered that
the first order linear differential equation

dz / dx + P(x) z = Q (x)

can be solved by multiplying through by the integrating factor e  P(x) dx


.

He also found the general solution to the higher order linear differential equation

y + a1 dy/dx + a2 d2y/dx2 + a3 d3y/dx3 + …. + an dny/dxn = 0,

which involves solving the auxiliary equation

1 + a1 p + a2 p2 + a3 p3 + …. + an pn = 0

Euler noted that for each linear factor (1  α p) there is a solution of the form y = A e  x/α , while for
each irreducible quadratic factor (1 + α p + β p2) there is a solution

y = e α x/ 2β [C sin ( 4β  α2 ) x / 2β + D cos ( 4β  α2) x / 2β ]

where the sine term corresponds to the imaginary part of the roots of the quadratic factor and the
cosine term corresponds to the real part. The general solution is then the sum of the solutions
corresponding to each factor.

Euler solved the equation proposed by Daniel Bernoulli

y  k4 d4y/dx4 = 0.

The corresponding auxiliary equation

1  k4 p4 = 0

has factors (1  k p), (1 + k p) and (1 + k2 p2), so the general solution is


y = A e x/ k + B e x / k + C sin x / k + D cos x / k.

17

d. Calculus of Variations

In a paper of 1744 entitled Method of Finding Curves with Extremal Properties Euler developed an
important branch of mathematics now known as the calculus of variations. The aim of this calculus
is to find the curve y = y(x) for which an integral such as I (y) =  Z (x, y, p) dx (where p = dy/dx) is a
maximum or a minimum.

He showed that the curve y = y(x) must satisfy the equation

Z / y  d/dx ( Z / y’ ) = 0.

As one of the first of many examples in his treatise, Euler used this equation to show that the
brachistochrone is a cycloid (see Appendix 5).

e. Euler-Maclaurin Summation Formula

In contrast to Maclaurin’s geometrical derivation of this formula, Euler obtained a version by purely
analytic means (see Appendix 4):

Sn = n0 t dn + t/2 + 1/12 dt/dn – 1/720 d3t/dn3 + … ,

where t is the function and Sn is the sum to n terms.

Euler illustrated his formula with an example, t = n2 + 2n. It is clear that

 t dn = n3/3 + n2

t = n2 + 2n

dt/dn = 2n + 2.

Therefore

Sn = (n3/3 + n2) + (n2 + 2n) / 2 + (2n +2) / 12

= (2n3 + 9n2 + 7n) / 6,


where fractions can safely be ignored. (The third and further derivatives are zero).
4
Thus ∑ k 2 + 2k = [2(4)3 + 9(4)2 +7(4)] /6 = [128 + 144 + 28] /6 = 300/6 = 50.
1

18

5. D’ Alembert

Jean D’Alembert (1717-1783) was left on the steps of a church by his mother soon after his birth and
was placed in an orphanage for foundlings. But he was later found by his father, an artillery officer,
who placed him with a glazier’s wife with whom he lived for almost fifty years. His father, who did
not want to recognise the boy as his son, secretly paid for his education.

D’ Alembert (an assumed name) was a man of wide intellectual interests, including music,
philosophy and literature in addition to science. For his achievements in mathematics, principally in
mechanics and hydromechanics, he was elected to the French Academy of Sciences, the Berlin
Academy, the Royal Society and the American Academy of Arts and Sciences.

In a paper of 1747 entitled Investigation of the Curve formed by a Vibrating String, D’ Alembert
provided a solution to the (one dimensional) wave equation,

2y/x2 (x, t) = c2 2y/t2 (x, t)

His solution is given by

y (x, t) = f(x + ct) + g(x  ct),


where c is the speed of the waves and f and g are arbitrary functions.

D’Alembert’s paper is notable for his use of partial differential calculus which he had introduced in a
Treatise on Dynamics in 1743. While his notation is somewhat different from modern practice, his
approach laid the groundwork for the solution given in textbooks today. D’ Alembert first
differentiated y with respect to x holding t constant and then differentiated y with respect to t
holding x constant, establishing the total differential of y as

dy = y/x dx + y/t dt

He then obtained expressions for a derivative of both y/x and y/t, noting Euler’s result that
2y/x t = 2y/t x. A little algebra and two integrations provided D’Alembert with his result.

19

Lagrange

Joseph-Louis Lagrange (1716-1813) in his Theory of Analytic Functions (1797) attempted to define
the derivatives of a function in terms of the coefficients of its Taylor series (see Lecture Notes 23).
This attempt to justify the differential calculus without reference to infinitesimal quantities was
unsuccessful in that not all functions can be expressed in terms of a power series. Indeed, by
definition, only analytic functions can be so represented. Nevertheless in the course of his discussion
Lagrange derived a useful expression for the remainder term of a Taylor expansion. He showed in
effect that a function f (x) can be approximated about the point x = 0 by the Taylor polynomial

f (x) = f(0) + f ‘(0) x + f ‘’(0) x2/2! + … + f n(0) xn/n! + f n+1(c) xn+1/(n + 1)! ,

for some c between 0 and x. The last term on the right is known today as the Lagrange form of the
remainder. Its importance lies in giving bounds for the error in approximating a given function by its
Taylor polynomial. The proof that a function f (x) can be represented by its Taylor series consists in
showing that for all x and for any c  [0, x] the remainder term approaches zero as n approaches

infinity.

Appendix 1: Bernoulli on Taylor and the English

‘I have not yet seen Taylor’s book … I can easily believe that Taylor cites no one except Newton:
for it is a characteristic of the English that they begrudge everything to other (nations) and attribute
all things to themselves or to their nation.’

(Johann Bernoulli to Leibniz,1716)

Appendix 2: Johann Bernoulli on Priority in the Invention of the Calculus

‘When in England war was declared against M. Leibniz for the honour of the first invention of the
new calculus of the infinitely small, I was despite my wishes involved in it; I was pressed to take part.
After the death of M. Leibniz the contest fell to me alone. A crowd of English antagonists fell upon
my body. It was my lot to meet the attacks of Messrs. Keil, Taylor, Pemberton, Robins and others. In
short I alone like the famous Horatio Cocles kept at bay the entire army at the bridge.’

(Johann Bernoulli)

20

Appendix 3: Integrating Factors

It is not clear how Fatio determined the integrating factor needed to solve his differential equation.
He may have started with an arbitrary equation which he differentiated to obtain the corresponding
differential equation and then multiplied its terms by a power of x or y. Having obtained a new
differential equation in this way, he may then have considered the steps needed to get back to the
original equation.

An explanation of his solution using modern notation would be as follows. Given his equation

 2xy dx + 4x2 dy  y2 dy = 0, … (1)

multiplication by the integrating factor y− 5 (assuming it has been found) produces the following:

 2xy- 4dx + 4x2y 5 dy  y 3 dy = 0. … (2)

Equation (2) is of the form

M (x, y) dx + N (x, y) dy = 0,

where M =  2xy- 4 and N = 4x2y 5 y 3.

It can be seen that My = Nx (the subscripts denoting partial differentiation) since

( 2xy- 4)y = (4x2y 5 y 3)x = 8xy-5.

In these circumstances the differential equation is said to be exact. This in turn implies that there
exists a solution f (x, y) such that f (x, y)x = M and f(x, y) y = N.

Therefore

f (x, y)x =  2xy – 4.

Integration with respect to x holding y constant yields

f (x, y) = − x2y – 4 + h(y), … (3)


where h is an, as yet, undetermined function of y. Partial differentiation with respect to y of
equation (3) yields

f (x, y) y = 4x2y-4 + h’(y).

But f(x, y)y also equals N = 4x2y 5 y 3. It follows comparing the two expressions for f(x, y)y that

4x2y-4 + h’(y) = 4x2y 5 y 3,

whence h’(y) =  y 3 and so h(y) = y -2/2.

Thus f (x, y) = − x2y-4 + y -2/2 and the solution of the differential equation is

−x2y-4 + y -2/2 = C.

21

Appendix 4: Euler’s Derivation of the Summation Formula

Euler wrote two papers on his summation formula in the 1730s. His approach, in contrast to
Maclaurin’s, was analytic rather than geometrically based. Given a progression whose general term
n
is t(x), his aim was to evaluate the sum S(x), where S = ∑ t ¿ ¿). The general term t being the last
1

term of the series, it follows that t(n) = S(n) – S(n – 1). Expanding S(n – 1) in a Taylor series, he found

S(n – 1) = S(n) – S’(n) + S’’(n)/2! – S’’’(n)/3! + … ,

whence

t = S(n)  S(n  1)

= S(n)  [S(n)  dS/dn + 1/2! d2S/dn2 + 1/3! d3S/dn3 

= dS/dn + 1/2! d2S/dn2 + 1/3! d3S/dn3  … (1

Euler then inverted this series, expressing S in terms of t:

dS/dn =  t +  dt/dn +  d2t/dn2 + …

Integration of the previous equation yields


n
S =  ∫ t dn + [ t +  dt/dn + … ]n0 , … (2)
0

where the coefficients , ,  … are yet to be determined.

Then he substituted the successive derivatives of equation (2) into equation (1), obtaining

t = ( t +  dt/dn +  d2t/dn2 +.. ) – 1/2!( dt/dn +  d2t/dn2 +.. ) + 1/3!( d2t/dn2 +  d3t/dn3 +..) − …

Comparing the coefficients of t, dt/dn, d2t/dn2 etc. on either side of this last equation, he obtained
=1

 = /2

 = /2 – /6

 = /2 – /6 + /24

This gave  = 1,  = ½,  = 1/12,  = 0,  = − 1/720, …

Hence, after substituting these coefficients into equation (2), he obtained

Sn = n0 t dn + t/2 + 1/12 dt/dn – 1/720 d3t/dn3 + … ,

which is his version of the summation formula.

22

Appendix 5: Euler’s Calculus of Variations

In a paper of 1744 entitled Method of Finding Curves with Extremal Properties Euler developed a
branch of mathematics now known as the calculus of variations. The aim of this calculus is to find
the curve y = y(x) for which an integral such as I(y) =  Z (x, y, p) dx (where p = dy/dx) is a maximum
or a minimum.

Euler’s central idea was to approximate the curve by a polygonal line. He partitioned the interval
[a, b] into n equal subintervals [xi- 1, xi] of length ∆x and drew the polygonal line connecting the
points (xi, yi) where yi = y (xi):

The integral I(y) could then be approximated by


I(y) = I(y1 , y2 ,…., yn- 1 ) = Σ Z (xi , yi , pi ) ∆x.

But pi = (yi+1 – yi)/ ∆x. So the approximation for I(y) can be written as

I(y) = Σ Z [xi , yi , (yi+1 – yi)/∆x] ∆x.

Now for the curve y(x) to be an extremum, the (partial) derivative of the integral with respect to
each yi must be zero. Euler therefore needed to estimate this derivative and set it equal to zero in
order to obtain the equation of the curve.

His method of estimating the derivative was to determine the effect on the integral by increasing
yi by a small quantity h. Only Z [xi-1 , yi-1 , (yi – yi-1)/∆x] and Z [xi , yi , (yi+1 – yi)/∆x] are affected by such
a change since they alone depend on yi.

23

Accordingly he considered the differential

dZ (xi-1 , yi-1 , pi-1) = Mi-1 dxi-1 + Ni-1 dyi-1 + Pi-1 dpi-1

where Mi-1 = Zi-1 /xi-1, Ni-1 = Zi-1 /yi and Pi-1 = Zi-1 /pi-1 . Only the last term on the right hand side
is affected by a change in yi. Since pi-1 is approximately equal to (yi yi-1) / ∆x, which implies that pi-
1 / yi = 1/∆x, an increase in yi by a small quantity h produces an increase in the differential equal to
Pi-1 . h / ∆x.

Similarly he considered the differential

dZi (xi, yi, pi) = Mi dxi + Ni dyi + Pi dpi .

The first term on the right is unaffected by a change in yi. The second term increases by Ni . h. And
since pi is approximately equal to (yi+1  yi) / ∆x, which implies that pi / yi =  1/∆x, the third term
is decreased by Pi . h / ∆x.

Thus the total change in the differentials dZi-1 and dZi, and hence in the differential value of the
integral, is therefore given by

Pi-1 . h / ∆x + Ni . h  Pi . h / ∆x = h [Ni  (Pi  Pi-1) / ∆x ].

This implies that the partial derivative of the integral with respect to yi is

Ni  (Pi  Pi-1) / ∆x.

But Pi  Pi-1 = dP, and so in the limit, as ∆x  0, the curve y (x) satisfies the equation

N  dP/dx = 0,

or, in modern notation,


Z / y  d/dx ( Z / y’ ) = 0.

As one of the first of many examples in his treatise, Euler used this equation to show that the
brachistochrone, the path of quickest descent under gravity, is an inverted cycloid.

24

Appendix 6: Preface to Edmund Stone’s English Translation of L’Hospital’s Analysis of Infinitely


Small Quantities (1730)

‘Upon this latter foundation is built the Calculus Differentialis, first published by Mr Leibnitz in the
Year 1684; having been since followed by almost all the Foreigners; who represent the first
Increment, or Differential (as they call it) by the letter d, the second by dd etc; the Fluents or Flowing
Quantities being called Integral. But since this method in the practice thereof does not differ from
that of Fluxions, and an Increment or Differential may be taken for a Fluxion; out of regard to Sir
Isaac Newton, who invented the same before the year 1669, I have altered the Notation of our
Author, and instead of a d, dd, d3 etc. put his notation viz. ẋ , ẍ , ⃛x , etc. or some others of the final
Letters of the Alphabet, pointed thus, and called the infinitely small Increment, or Differential of a
Magnitude, the Fluxion of it.’

At the end of his translation Stone added an appendix of his own on the inverse method of fluxions.
This was so successful that it was translated into French and published in 1735 as an adjunct to
L’Hospital’s text.

Further Reading

*Carl Boyer, The First Calculus Textbooks, jstor.org

*Boyer and Merzbach, A History of Mathematics, Wiley (1991), Ch. 20,


atiekubaidillah.files.wordpress.com,

*C. H. Edwards, The Historical Development of the Calculus, pp. 277-281, isidore.co

*V. Katz, A History of Mathematics, Addison-Wesley (2009), pp.584-641, UCL library online

*L. Feigenbaum, Brook Taylor and the Method of Increments, jstor.org


S. Mills, The Independent Derivations by Leonhard Euler and Colin Maclaurin of the Euler-Maclaurin
Summation Formula, jstor. org

Grateful acknowledgement to V. Katz and C. H. Edwards for use of their texts in preparing these
notes.

You might also like