22.09.2016 Views

5128_Ch03_pp098-184

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Section 3.2 Differentiability 111<br />

y<br />

m 1 =<br />

f(a + h) – f(a – h)<br />

2h<br />

m 2 =<br />

f(a + h) – f(a)<br />

h<br />

tangent line<br />

Derivatives on a Calculator<br />

Many graphing utilities can approximate derivatives numerically with good accuracy at<br />

most points of their domains.<br />

For small values of h, the difference quotient<br />

f a h f a<br />

<br />

h<br />

a – h a a + h<br />

Figure 3.16 The symmetric difference<br />

quotient (slope m 1 ) usually gives a better<br />

approximation of the derivative for a given<br />

value of h than does the regular difference<br />

quotient (slope m 2 ), which is why the<br />

symmetric difference quotient is used in<br />

the numerical derivative.<br />

x<br />

is often a good numerical approximation of f a. However, as suggested by Figure 3.16,<br />

the same value of h will usually yield a better approximation if we use the symmetric<br />

difference quotient<br />

f a h f a h<br />

2h ,<br />

which is what our graphing calculator uses to calculate NDER f a, the numerical<br />

derivative of f at a point a. The numerical derivative of f as a function is denoted by<br />

NDER f x. Sometimes we will use NDER f x, a for NDER f a when we want to<br />

emphasize both the function and the point.<br />

Although the symmetric difference quotient is not the quotient used in the definition of<br />

f a, it can be proven that<br />

lim<br />

h→0<br />

f a h f a h<br />

2h<br />

equals f a wherever f a exists.<br />

You might think that an extremely small value of h would be required to give an accurate<br />

approximation of f a, but in most cases h 0.001 is more than adequate. In fact,<br />

your calculator probably assumes such a value for h unless you choose to specify otherwise<br />

(consult your Owner’s Manual). The numerical derivatives we compute in this book<br />

will use h 0.001; that is,<br />

f a 0.001 f a 0.001<br />

NDER f a <br />

<br />

0.002 .<br />

EXAMPLE 2<br />

Computing a Numerical Derivative<br />

Compute NDER x 3 ,2, the numerical derivative of x 3 at x 2.<br />

SOLUTION<br />

Using h 0.001,<br />

NDER x 3 ,2 <br />

2.001 3 1.999 3<br />

<br />

0.002<br />

12.000001.<br />

Now try Exercise 17.<br />

In Example 1 of Section 3.1, we found the derivative of x 3 to be 3x 2 , whose value<br />

at x 2 is 32 2 12. The numerical derivative is accurate to 5 decimal places. Not bad<br />

for the push of a button.<br />

Example 2 gives dramatic evidence that NDER is very accurate when h 0.001. Such<br />

accuracy is usually the case, although it is also possible for NDER to produce some surprisingly<br />

inaccurate results, as in Example 3.<br />

EXAMPLE 3<br />

Fooling the Symmetric Difference Quotient<br />

Compute NDER x, 0, the numerical derivative of x at x 0.<br />

continued

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!