8.1 POLYNOMIAL APPROXIMAT|
TAYLOR’S FORMULA ONS By
While values of polynomial functions can be found b:
number of additions and multiplications, other fy
rithmic, exponential, and trigonometric functio
easily. We show in this section that many functions can be a ahiateg x
polynomials and that the polynomial, instead of the cieaae ale be
be used for computations when the difference between the cee &
value and the polynomial approximation is sufficiently smal]. s Ue
Various methods can be employed to approximate a Biven f
polynomials. One of the most widely used involves Taylor's one b
in honor of the English mathematician Brook Taylor (1685-4944
following theorem, which can be considered as b
A 4 generalization yy
mean-value theorem, gives Taylor's formula. Of thy
f Let f be a function such that f and its first derivatiy
re
continuous on the closed interval [a, 6). Furthermore, Wet fray
|, exist for all x in the open interval (a, 6). Then there is a number z
in the open interval (a, b) such that
Perform;
Mctions, such 4. 89a,
a
NS. Cannot be they
90) = f+ FO — + Me aps
x $A) yay g LED
i +a (b - ay tae peo
k Equation (1) also holds if b < a; in such a case [a, b) is replaced
by, [b, a}, and (a, b) is replaced by (b, a).
Note that when n = 0, (1) becomes
f(b) = fla) + f'(2)(b — a)
where z is between a and b. This is the mean-value theorem.
We defer the proof of Theorem 8.1.1 until later in this section, lint!
we replace b by x, we obtain Taylor's formula:
fo) = fy + LOG - y+ LO uate.
1!
a
(a) oalea} art ny
+ - ap +
where z is between a and x:
80
i its first deo
The condition under which (2) holds is that f and its ei vane
must be continuous on a closed interval containing « ern
(n + 1)st derivative of f must exist at all points of the corres!ga)
Aan ~ » Le ‘i “ua: FOB Rees oe 45
PO) = s@ +L, 4), Fo
interval. Formula (2) May be written as
FG) = Px) + RY) (3)
where
Y 2 (e- #43, {
and
4 = ai ay! where zis between a and (5)
P(x) is culled the nth-depree ‘T ylor polynomial of the function f at the
number 4, and R,(x) is called the remainder. The term R,(x) as given in (5)
is called the Lagrange form of the remainder, named in honor of the French
mathematician Joseph L. Lagrange (1736-1813).
The special case of Taytor’s
formula obtained by taking a = 0 in (2) is
gfe) = 70) + LO, LO, LO. LOM aay
E { 2} ni (n + 1)
where z is between 0 and x. This formu},
named in honor of the Scottish mathe
1746). However, the formula was obtained earlier by Taylor and by another
British mathematician, James Stiri
ng (1692-1770). The nth degree
Maclaurin polynomial for a function f, obtained from (4) with a = 0, is
5 :
(2) = f(0) + T+ LO 4 ye
ais called Maclaurin’s formula,
matician Colin Maclaurin (1698—
(6)
We may approximate a function by means of a Taylor polynomial at a
number @ or by a Maclaurin polynomial.
> ILLUSTRATION 1 We compute the nth degree Maclaurin poly-
nomial for the natural exponential function. If f(x) = e*, all the derivatives
of f at x are e* and the derivatives at zero are 1. Therefore, from (6),
2 3 t
ee ee
Pax) = 14+ x45 algae air 7)
Thus the first four Maclaurin polynomials of the natural exponential func-
tion are
P(x) = 1
P(x = 14x
Pz) = 1+ xt hx?
PG) Meta canta[-3. 3] by [0,4]
fiayeet
Poel
FIGURE 1
{-3. 31 by [0,4]
fl ‘
Pye l+x
FIGURE 2
[-3. 3] by 10. 4]
foyee
Pays itat ye
FIGURE 3
Figures | through 4 show the graph of f(x) = e* along w
of Pyl-x), Pyl), P2205 and P,(x), respectively, plotted in epee the atop
window, In Figure 5 the graphs of the four Maclaurin polyno, 3) by (03)
graph of f(x) = ¢' are sketched on one coordinate system, Obs and ]
polynomials approximate e* for values of x near zero, and aes how
increases, the approximation improves. Tables | and 2 give we that a,
P,(x) (when 1 is 0, 1, 2. and 3) and e* — P,(x) for x = 0.4 wets Of es
respectively. From these two values of x it appears that the close al =O)
the better is the approximation for a specific P,(x), Serv Sto zeny,
4
n
Table 1 Table 2
wt pan lo
wt Pod) et — POA) ne? PaO.) gta
ae.
oO 14918 t 04918 0 1.2214 5 (02)
1 14918 14 0.0918 1 1.221412 $2214
2 14918 148 Ook 2 1.2214 1.22 rae
3. L918 1.4907 0.001) 3 (1.2214 1.2213 ae
n
The Lagrange form of the remainder when P,(x) is the nth deg
Maclaurin polynomial for the natural exponential function is from (5
e
= aneh 3h
are nit where z is between 0 and x 8)
R(x)
In particular, if Ps(x) is used to approximate e*
Ro) = a" where z is between 0 and x
and
et = Py(x) + Rix)
_ eee
PD EXAMPLE! Usea Maclaurin polynomial to find the valueofVe
accurate to four decimal places.
Solution Iff(x) = ¢, the nth degree Maclaurin polynomial of fis
by (7) and the Lagrange form of the remainder is given by (8). If.
(8)
a!) = aes) hereO <2<$
9) ~ (n+ II\2 where?
Thus
ee aan'=000005 and y = -0.00008
FIGURE 6
Because e¢ < 4, then e? <
1
(3)
ret be he o ‘ vitae be accurate to four decimal places we want
a3 ess than 0.00005; |{R,(4)| wi e 2 i
1/2n + 1)! < 0.00005, When m \ | will be less than 0.00005 if
1 I
2(n + (720)
= 0.00004
2. and we have
—2
eine IY FOr
Because 0.00004 ~ 0.00005, we take P\(4) as the approximation of Ve
accurate to four decimal places. Because from (7)
PUY = VEER eh ot
we obtain Ve ~ 1.6487 <
> _ EXAMPLE 2 Estimate ona graphics calculator the values of x for
which P,(x) approximates e* accurate to four decimal places.
Solution If P.(.x) is to approximate e' accurate to four decimal places
Je’ — Pax)| < 0.00005
Figure 6 shows the graph of y = e' — P.(x}, that is
y set (Lh + et 02/21 + x7/3t + xt/at + x°/51)
and the lines y = +0.00005 plotted in the [—1, 1] by [—0.0001, 0.0001]
window. Using the intersect capability (or trace and zoom-in) of our graph-
ics calculator, we determine that the curve and the line y = 0.00005 inter-
sect when x = —0.5824 and x = 0.5667. Thus we conclude that when
0.5824 < x < 0.5667, P.(x) approximates e' accurate to four decimal
places.
This answer supports our conclusion in Example | that Px(y) approxi-
iE ; ; =
mates Ve accurate to four decimal places.
(ILLUSTRATION 2 We now determine the ath degree Maclaurin
polynomial for the sine function. If f(x) = sin x. then
f'(x) = cos x f(x) = msinx f(x) = —cos x
Posse [eco Pi = sin x
and so on. Thus f(0) = 0,
POS) FO=0 POSS PW=0 PMs)1-6, 6) by [-4. 4]
fla) = sine
Pisa
FIGURE 7
{-6. 6} by [-4, 4]
f(x) = sine
3
Px) 5.x =
2 6
FIGURE 8
[-6, 6} by [-4. 4]
flo ssn
5
c
Phdar- +
. 6
FIGURE 9
oe
120
a
and so on. From (6),
eyo 4 on 0
ee ee nie
Pix) = x 31 di 5! 7! ay) Qn = Ti
Thus Pylx) = 0.
Pil) = x as
_, 8 Pix) = x —
FA tie ela
Pe) ie ‘
ey ok P(x) =x - 4 4
PG) ae
eo 2 roa
eye ey SH ny ee '
Pd=x— e+ i597 soa Plt) = 6” ip a
500
and so on,
Figures 7 through 10 show the graph of the sine function along with
graphs of its Maclaurin polynomials of degrees 1, 3, 5, and 7, respecinge
plotted in the [~6, 6] by [—4, 4] window. Figure 11 shows tegen
these four Maclaurin polynomials and the graph of f(x) = sin x Sketched ¢
the same coordinate system. Notice that the polynomial APPrOX:mation,
improve as n increases. 4
> EXAMPLE 3 (a) Determine the accuracy e
c hen the seventh de.
gree Maclaurin polynomial of the sine function, P,(x), is used to appron.
mate sin 0.5. (b) Support the answer in part (a) graphically. (c). Compute
P,(0.5) to approximate sin 0.5 accurate to the number of decimal places
allowed by the answer in part (a).
Solution
(a) From (3) with f(x) = sin x and n = 7, we have
sin x = P,(x) + R,(x) :
Thus
sin 0.5 = P,(0.5) + R,(0.5)
where from (5) with x = 0.5 and a = 0
1B)
R0.5) = Posy where z is between 0 and 0.5
'
= 0.0000001 sin z
Because |sin z| < |
{R,(0.5)| < 0.0000001
We therefore conclude that when P,(0.5) is used to approximate sin O05,
our value is accurate to six decimal places.(b) 2 Pe ie Our answer in part (a) we must show that when vy = 0.5,
UJ 1 5 9.0000001. Because R,(t) = sin x — P.G9), and from the
computation in Mlustration 2, for the sine function.
q Py = 4 -
6 120 5040
r, we plot the graphs of
ay ( wt x* a?
{-o. 6} by | 44] Sd 2 Se T0 sian) and y = +0,0000001
Mem in the [=1. 1] by [ -0.0000002, 0.000002] window as shown in
ene ae an Figure 12, Using the intersect capability (or trace and zoom-in) of our
6 120 50d
graphics calculator, we determine that the curve and the lines intersect
MCURETO at the points where x = +0,6921. Because -0,6921 < 0.5 < 0.6921
we have supported our answer in part (a).
Substituting 0.5 for x in the expression for P,(.x), we have
‘ .
05 - (0.5) , (0.5) _ (0.5)
6 120 5040
0.47942553
(c
u
P,(0.5)
From part (a), this computation is accurate to six decimal places.
Therefore
sin 0.5 = 0.479426 <
EXAMPLE 4 Find the third-degree Taylor polynomial of the
cosine function at 37 and the Lagrange form of the remainder. Plot the
graphs of the polynomial and the cosine function in the same window.
Solution Let f(x) = cos x. Then from (4).
pune Ae) +o 2) AEF SD
Because f(x) = cos x, f'(x) = ~sin x, f(x) = —cos xf" x = sins,
a
fim = V2 fim) = -$V2 PG) = -N9 fra) = 2
(+1, 1] by {-0.0000002, 0.000002]
fee Therefore
Nese (ee ee _ .
inx=[t— + 195” 5040 pla) = {V2 - Vala = fa) — EV 3k = bey? + AVI Lap
y = 0.0000001 and y = ~0.0000001 3
‘3 ¢) = cos x, we obtain from (5)
FIGURE 12 Because f'"\(x) = cos +:
Rylx) = $(cos x - in) where z is between $7 and x
AXP = Bae we
Because |cos =| = 1, we conclude that |Rx(x)| = 3a ~ ¢7)* for all x.Figure 13 shows the graphs of P,(x) and f().= eos x
—r, 7] by [-2. 2] window. Observe how the graph of thei
approximates the graph of the cosine function near x = in
EXAMPLE 5 Use the third-degree Taylor Polynomi
cosine function at 4 7 found in Example 4 to Compute an approxima
of cos 47°, and determine the accuracy of the result,
ea eutaty) Solution 47° A 7 ee ra in the solution of Example:
Mipiwereb v= fi mand.x — 77 = 47, and
cos 47° = }V2[1 — Aa = fhm? + hw Rm)
where
V2 G- Jas hv7 W-} RG) = keos 2(%5 7)" where }7 S 2 < Aha
FIGURE 13 Because 0 < cos z < 1,
0 < RG m) < 4h 7)" < 0.00000007
Taking ¢ 7 = 0.0349066 we obtain from (9)
cos 47° = 0.681998
which is accurate to six decimal places because of inequality
We now prove Theorem 8.1.1. There are several knoy
theorem, although none is very well motivated. The one fol
use of Cauchy's mean-value theorem (7.7.3). o&
Let F and G be two functions de
Fa) = 6) — fo) - f@b- 9 - 2% — 9 -, = FOG _ y1 ROE
2! (n— 1)! ink
and
g(a yt ‘
2) ieee -
FIG) = —f'(x) + f'G) — f(b — x) + 2b 9) 2
FAO = x)
oe i Saag |e arte
Combining terms we see that the sum
_ following even-numbered term is zero:Therefore -
, = _ f"™ G9
F'(x) a = 59} (13)
Differentiating in (12) we obtain
iy oe "
G'(x) = me? ox) (14)
Checking the hypothesis of Cauchy’s mean
@) F and Gare continuous on {a, b);
(ii) F and G are differentiable on (a, b);
(iii) for all’x in (@, 5), G(x) #0.
So by the conclusion of the theorem
F(b) — Fa) _ F’@
G(b) - Ga)” GD
where z is in (a, b). But F(b) = 0 and G(b) = 0. So
_ FR) (15)
F(a) = @ G(a)
for some z in (a, b).
Letting x = ain(12),x = zin(13),andx = zin(14) and substituting
into (15) we obtain
-value theorem we see that
Beg) al (b - a"!
nah = at nt se [-sesl (n+ 1)!
inh)
F(a) = £ + a (Bray (16)
If x = a in (11), we obtain
fra) fa)
fa)
{
P _ Dye nee
Fla) = f(b) ~ fla) — f'(ab — a) Gb a ~ 2. Gay 9) -
ituting 16) into the above equation, we get
Substituting from (16) oy
i 2 LO = ay + - a"
fe) = fla) + plaid — @) + LM - af +... FR a + Ty
2
i m holds if b < a because the conclu-
ich is ired result. The theorem | 1 1
ae cacy = mean-value theorem is unaffected if a and b are inter-
sion o § :
changed.
are other forms of the remainder in Taylor’s formula. Depending
dene fie one form-of the remainder may be more desirable to use
rae iS The following theorem, known as Taylor's formula with inte-
than another.
s e remai as an integral.
inder, expresses the remainder as an integ
z the remainder,
gral form of8.1.2 Theorem Taylor's Formula with Integral Form of ;},,
tg .
LTT
and the principal
EXERCISES 8.1
In Exercises 1 through 10, find the Maclaurin polynomial
of the stated degree for the function f with the Lagrange
form of the remainder. Plot the graphs of f and the poly-
nomial in the same window and observe how the graph of
the polynomial approximates the graph of f near the point
where x = 0.
el =
1. f(x)
3. f(x)
5. f(x) = cos x; degree 6
7. f(x) = sinh x; degree 4
9. f(x) = (1 + xP?; degree 3
10. f(x) = Ul — x)"; degree 4
; degree 4 2. f(x) eG
4. f(x) = tan x; degree 3
6. f(x) = cosh x; degree 4
8. f(x) = e-*; degree 3
; degree 5
= e*; degree 5
In Exercises 11 through 18, find the Taylor polynomial of
the stated degree at the given number a for the function f
with the Lagrange form of the remainder. Plot the graphs
of f and the polynomial in the same window and observe
how the graph of the polynomial approximates the graph
of f near the point where x = a.
UL f(x) = x; a = 4; degree 3
12; f(x) = Vasa = 4; degree 4
13. f(x) = sin x; a = £77; degree 3
14. f(x) = cos x; a = $7; degree 4
15, f(x) = In x; a = 1; degree 5
The proof of this theorem is left as an exercise (See Exercise
You have seen how a function can be approximated by a succes:
Taylor polynomials, The values of these polynomials for a given
can be thought of as a sequence of numbers, the subject matter: of the
section, The nth degree Taylor polynomial is the sum of n + le
as n increases without bound, the sum may or may not Approach a lim
Such considerations form the basis of infinite series, defined in Section g
40),
valntog
topic of this chapter.
16. f(x) = In(x + 2); a = —1; degree 3
17. f(x) = In cos x; a = 477; degree 3
18. f(x) = x sin x; a = 57; degree 5
19. Compute the value of e accurate to five decimal
places, and prove that your answer has the re
accuracy. Support your answer graphically.
20, Do Exercise 19 for the value of e-¥/2,
21. Compute sin 31° accurate to three decimal plac
using a Taylor polynomial and prove that yot
has the required accuracy. Support your answer
graphically. '
22. Compute cos 59° accurate to three decimal
using a Taylor polynomial and prove that
has the required accuracy. Support your answer
graphically,
23, Estimate the error that results when cos x is Te}
by 1 — 32x? if |x| < O01. :
24. Estimate the error that results when sin x!
by x = 4x3 if |x| < 0,05. ‘
25. Estimate the error that results when
placed by 1 + 5x if0