Taylor Polynomials The tangent line approximation L(x) is the best first-degree (linear) approximation to f(x) near x=a

because L(x) and f(x) have the same rate of change (derivative) at a. For a better approximation than a linear one, lets try a second-degree (quadratic) approximation P(x). In otherwords, we approximate a curve by a parabola instead of by a straight line. To make sure that the approximation is a good one, we stipulate the following: P a = f a ( and should have the same value at a.) i. ii. iii. P ' a= f ' a ( and should have the same rate of change at a.) P ' ' a = f ' ' a (The slopes of P and f should change at the same rate.) P x =ABxCx
2

1. Find the quadratic approximation

to the function

that satisfies conditions (i), (ii), and (iii) with a = 0. Graph P, f, and the linear approximation L(x) on a common screen. Comment on how well the functions P and L approximate f. P x =ABxCx2 i. P a = f a A B0C 02=cos 0 A=1 ii. P ' a= f ' a B2C0=sin 0
B=0

f x =cos x

P(x) must satisfy conditions i, ii and iii with a = 0.

iii.

P ' ' a = f ' ' a 2C=cos 0 C= C= 1 cos 0 2 1 2

1 2 Therefore: P x =1 x 2

2. Determine the values of for which the quadratic approximation f x =P x in Problem 1 is accurate to within 0.1. [Hint:Graph y=P x , y=cos x0.1 f x =cos x , and y=cos x0.1 on a common screen.] f x =cos x0.1
y 1

Graph of Quadratic Approximation in Problem 1, along with graphs for f x =cos x0.1 , and

(x,y) = (-1.26124242785830,0.20463376908505) (x,y) = (1.26124242785830,0.20463376908505) y = 1 y = 1-(xx)/2 y = cos(x) y = cos(x)+0.1 y = cos(x)-0.1

x 1 1

The approximation is accurate from (-1.26, 1.26) as shown in the graph.

3. To approximate a function f by a quadratic function P near a number , it is best to write P in the form P x =AB x a C xa2 : Show that the quadratic function that satisfies conditions (i), (ii), and (iii) is P x = f a f ' a x a 1 f ' ' a xa2 2

By differentiating repeatedly: i. P a = f a P a =A BaaC aa 2= f a A= f a ii. P ' a= f ' a P ' a= B2Caa = f ' a B= f ' a iii. P ' ' a = f ' ' a P ' ' a =2C= f ' ' a C= 1 f ' ' a 2 1 f ' ' a xa2 is the formula for 2

Therefore: P x = f a f ' a x a Quadratic Approximation.

4. Find the quadratic approximation to

f x = x3 near a=1 . Graph f,

the quadratic approximation, and the linear approximation from Example 3 in Section 3.11 on a common screen. What do you conclude? Linear Approximation: P x = f a f ' a x a = 13 Quadratic Approximation: P x = f a f ' a x a P x = 13 1 f ' ' a xa2 2 1 1 x1=2 x1 4 2 13

1 1 1 1 x1 x12=2 x1 x12 3 4 32 2 13 4 13
3 y

y = (1/4)(x-1) +2 y = sqrt(x+3) y = 2+((x-1)/4)-((x-1)(x-1)/32)

x 2 1 1 2

The Quadratic Approximation (in green) is accurate over a larger interval than that of the Linear Approximation.

5. Instead of being satisfied with a linear or quadratic approximation to f(x) near x=a, lets try to find better approximations with higher-degree polynomials. We look for an nth-degree polynomial

T n x=c 0c 1 xac 2 xa2c 3 xa3. . .c n xan

such that

T n and its first n derivatives have the same values at x = a as f and

its first n derivatives. By differentiating repeatedly and setting x = a, show that these conditions are

satisfied if where

c 0= f a , c 1= f ' a , c 2=

1 f a f ' a and in general c k = 2 k!

k !=1234....k

. The resulting polynomial

T n x= f a f ' a xa

f ' ' a f n a x a2 ... xa n is called the nth2! n!

degree Taylor polynomial of f centered at a.

By differentiating repeatedly: i.

T k x =c 0c 1 aac2 aa c 3 aa . . .cn aa = f a c 0= f a T ' k x=c12 c 2 aa3c 3 aa 2. . .k c n aak 1= f ' a c 1= f ' a T ' ' k x =2 c 26c3 aa. . .k k 1c n aa n2= f ' a 2c 2= f ' ' a c 2= 1 f ' ' a 2
k

ii.

iii.

And thus:

T x =k k 1 k2 ...1 c k = f a k ! c k = f a
k

ck =

f k a k!
2 3 n

By plugging these values into the original equation:

T n x=c0 c 1 xac 2 xa c 3 xa . . .c n xa

we get

T n x= f a f ' a xa
looking for.

f ' ' a f n a x a2 ... xa n which is what we were 2! n!

6. Find the 8th-degree Taylor polynomial centered at a = 0 for the function f x =cos x . Graph f together with the Taylor polynomials in the viewing rectangle [5, 5] by [1.4, 1.4] and T 2, T 4, T 6, and T 8

comment on how well they approximate f . f ' ' a f n a 2 T n x= f a f ' a xa x a ... xa n 2! n! Solving for i. T 2, T 4, T 6, and T 8 where f x =cos x and a = 0:
f ' ' a xa 2 2!

T 2= f a f ' a xa

T 2=cos 0sin 0 xa 1 T 2=1 x 2 2 ii.

cos 0 x a2 2

f ' ' a f iii a f iv a 2 3 T 4 = f a f ' a xa xa xa xa4 2 3! 4! T 4 =cos 0sin 0 xa 1 x4 T 4 =1 x 2 2 24 cos 0 sin 0 cos 0 xa 2 xa 3 xa4 2 3! 4!

iii.

T 6= f a f ' a xa

f ' ' a f vi a x a2 ... xa 6 2! 6! cos 0 sin 0 cos 0 xa2 xa 3 xa4 2 3! 4!

T 6=cos0 sin 0 xa

sin 0 cos 0 xa5 xa6 5! 6!

1 x4 x6 T 6=1 x 2 2 24 720 iv. T 8= f a f ' a xa f ' ' a f viii a xa2... xa8 2! 8!

cos 0 sin 0 cos 0 xa2 x a3 xa 4... 2 3! 4!

T 8=cos 0sin 0 xa

...

sin 0 cos 0 sin0 cos 0 xa5 xa6 xa7 xa 8 5! 6! 7! 8!

1 2 x4 x6 x8 T 6=1 x 2 24 720 40320

Graphs of T 2, T 4, T 6, T 8 and f(x) where

y y y y y y = = = = = 1-(xx/2) cos(x) 1-(xx/2)+(xxxx/24) 1-(xx/2)+(xxxx/24)-(xxxxxx/720) 1-(xx/2)+(xxxx/24)-(xxxxxx/720)+(xxxxxxxx/40230)

f x =cos x

x 4 3 2 1 1 2 3 4 5