Professional Documents
Culture Documents
TAYLOR'S
SERIES
s.i *
HISTORICAL SURVEY
f(x
h) == f(x)
hf'(x)
h2
+ 2!f"(x) +
which bears his name was first enunciated by Brook Taylor (1685-1731) as early as
1712 in a. letter to John Machin, Its first formal appearance was in his text 1v[ethodus
incrementorum directa et inversa which was published in London in the period 1715-1717.
This text also contains the easy consequence now known as Maclaurin's series, but
Taylor's proof of these expansions did not consider convergence and is worthless. The
importance of these expansions was not appreciated by analysts for over a half century
until Lagrange pointed out their applicability, and no rigorous proof of Taylor's
theorem was offered until Cauchy included a remainder term and tested for convergence
in 1821.
Colin Maclaurin (1698-1746), though an able mathematician, is improperly credited
with authorship of the expansion
x2
f(x) == f(O) + xf'(O) + ,f"(O) + ...
2.
which was contained in his Treatise of Fluxions published in Edinburgh in 1742. This
expansion is obviously a special case of Taylor's theorem, a point which was indicated
by Taylor 25 years earlier. Additionally, Maclaurin's expansion was apparently discovered independently by James Stirling and is contained in his paper M ethodus
differentialis sive Tractaius de summoiione et interpolatione serierum infiniiarum published in London in 1730. The greater fame of Maclaurin and the wider circulation of
his Treatise are accountable for this miscredit.
* This section may be omitted without loss in continuity of the technical presentation.
558
5.2
Taylor's Series
A discussion of Taylor's series builds on the base of several mean value theorems which
serve as lemmas. The first of these is the well-known
Let f(x) be a function of the real ariable x which possesses a conXl ~ X ~ X2. Let a and b be two points within the
intervalt for which f(a) = f(b) == O. Then at least one value of x can befound between a and
b, say x == t, for which f' (t) == O.
ROLLE'S THEOREl'vI:
Proof: The truth of this theorem is almost self-evident from a geometric display of the
function such as shown in Figure S.l. If the function is to be zero at a and at b, it cannot
f(x)
J------;---4--------4L.----+-~-o#__~__+_--x
FIGURE
S.l
Rolle's theorem.
Let f(x) and g(x) be two functions of the real variable x which
possess continuous first derivatives ihrouqhoui the interval Xl S X ~ X2. Let a and b be any
two points within this in ierool such that g(a) ~ g(b). If g' (z) is nowhere zero in the interval,
then for some value x == t between a and b,
THEOREIVI OF l'vIEAN VAL DE:
f(b) - f(a) _
f'(~)
(S.l)
~~~
=:~:~
t In this and all subsequent theorems of this supplement, b can be either larger or smaller than a.
SECTION
8.2
It can be observed that hex) is a function which satisfies all the requirements of Rolle's
theorem. It has a continuous first derivative in the interval and h(a) == h(b) = O. 8ince
h'(O
= 0 =
feb) - f(a)
g(b) - g(a)
g'm - 1'(0
feb) - ita)
b- a
I'm
(8.2)
b-a
(b-a)2
f(b) == f(a)
+ -,-f'(a)
+
1.
in which
~n
')'
..,.
f"(a)
f'(~o) =
feb) - f(a) - (b -, a)
1.
Define a constant K 2 by the equation
~o between a and
(x - a)
1!
f'(a) -
b for which
= 0
(x - a)2
2!
(8.4)
K2
The function hex) has a continuous first derivative in the interval, given by
h" (x)
==
f" (x)
K2
560
Taylor's Series
and since h'(a) == h'(~l) == 0, there must be a point x == ~2 between a and ~1 for which
f(b)
= f(a)
(b - a) f'(a)
1!
(b - a)2f"(~')
2!
2
(8.5)
(S.6)
from which it follows by the above procedure that !(3 = f'''(~3), where ~3lies between
a and b. Continuing this process out to the nth derivative yields the result (8.3). The
last term of this series, namely
is known as the remainder after n terms. For the important case in which f(x) is a
function with continuous derivatives of all orders, (8.3) becomes an infinite series as
n ~ co , If the remainder goes to zero in this process, the series converges to the value
f(b) and one may write
(8.7)
EXAl\1PLE
If fex)
lets a
S.l
= sin
=
x, the remainder does go to zero and the expansion (8.7) is applicable. If one
= 1/ y!2 and feb) = t. (8.7) gives
1
V2
1r -"21 (1r)2
1 (1r)3
]
12 + 6 12 + ...
1 - 12
Use of only the first four terms of this series gives the approximation
f(b)
S.3
TAYLOR'S SERIES
0.4999
X2,
and if a and x
(S.8)
in which
~n
SECTION
8.4
561
f(x)
~ (x - a)m
,fm(a)
==
(8.9)
n~.
m=O
is a convergent series representation for f(x), valid within the entire interval. (8.9) is
known as the Taylor's series expansion of f(x) about the point a.
The special case of this result in which a = 0 is known as Maclaurin's series, and can
be written
00
\' x
L -
f(x) =
m=O
m!
(8.10)
jm(O)
f(x
Both x and x
EXAMPLE
-t-
~x)
is expanded in a series
~ (~x)m
= L _,_jm(x)
~x)
m=O
(S.11 )
1n.
8.2
~x)
= (x
fm(x) =
+ Lix)n in which
n'.
(n - m)!
xn -
fm(x) = 0
m>n
+ ~x)n
= \'
n!
xn-m(~x)m
n(n - 1)
2!
xn-2(~x)2
+ ... +
nx(~x)n-l
(S.12)
(~x)n
S.4
TAYLOR'S SERIES
FOR SEVERAL
VARIABLES
The results of the previous section may be extended to functions of more than one
variable with little difficulty. Let j(x,y) be any function which, together with all its
partial derivatives, is continuous in the interval Xl ~ X ~ X2, YI ~ Y ~ Y2. If (a,b) and
(x,y) are any two points within this interval, then by Equation (8.9),
x ) = ~ (x - a)m amf(a,y)
j( ,y
L
,
axm
m=O
m.
(8.13)
But the functions of y appearing on the right side of (8.13) also can be expanded in a
562
Taylor's Series
so that
All the series in (8.13), (8.14), and (8.15) rnust converge for all points in the interval
in order for this to be a valid procedure. When they do, (8.15) is known as a 'I'aylor's
series expansion of f(x,Y) about the point (a,b).
A useful alternative form of (8.15) arises when f(x + LlX, Y + ~y) is expanded in a
Taylor's series about (x,y). Direct substitution in (S.15) gives
f(x
LlX, Y
/1y) =
m=O n=O
-- -- --n!
nd
axmayn
(8.16)
Next, let !(x,y,z) be any function which, together with all its partial derivatives, is
continuous in the interval Xl ~ X ~ X2, Yl S; Y :s; 1}2, ZI ~ Z ~ Z2. If (a,b,c) and (x,Y,z)
are any t\VO points within this interval, then by (S.15),
~
L L
f(x,Y,z) -
1n!
m=On=O
n!
axmayn
(8.17)
am+n.f(a,b,z) =
~o
axmayn
p!
axmaynaz p
(8.18)
_I
oo
!(x,Y,z) -
I~
I
oo
m=On=Op=O
(8.19)
When it is assumed that the necessary convergence conditions are met, (8.19) is known
as the Taylor's series expansion of !(x,Y,z) about the point (a,b,c).
In an alternative form,
j(x
LlX, Y
~Y,
<'
.'wI
LlZ) -
L L L
m=O n=O
(8.20)
The extension of these results to functions of four or more variables follows the same
procedure and can be predicted by inspection.
EXAMPLE
8.3
In a vacuum triode, the plate current ib is a function of both the plate voltage eb and the
grid voltage e.. In many applications the triode has a plate current which consists of a
time-independent, or d.c. component, and a time-varying component. 'The plate current
can then be expressed in the form
i b = Ib + i
in which Ib is the quiescent value and i p is the superimposed time-varying part. These t\VO
component currents flow in response to the voltages eb = Eb
e p and e c = E e + eg, with
SECTION
8.4
(Eb,EJ the quiescent portions and (ep,e ll ) the time-varying portions. When Equation (8.16)
is applied to this situation, one obtains
If the triode is biased to operate in the linear portion of its characteristic, then all higher
order derivatives vanish and this expansion simplifies to
(8.21)
If one defines the plate conductance gp and transconductance gm by the relations
REFERENCES
1.
Cajori, F., A History of Miuhenuiiics, 2d ed., pp. 226-229, The Macmillan Company, New
York, 1919.
2. Love, C. E., and E. D. Rainville, Differential and Integral Calculus, 6th ed., pp. 439-447,
The Macmillan Company, New York, 1962.
3.
Smith, D. E., History of Mathematics, vol. 1, pp. 449-454, Ginn and Company, New York,
1923.