12 Chapter One Mathematical Preliminaries ‘The meaning of the mysterious terms in this theorem is as follows: Re HH) soar ran y. and so on. Letting fi; = 8f/8x, fy = f/8y, fer = 2° f/8x", fry = °f/0x By, ‘iy = 22/892, we can write the first few terms of (5) as FO thy +R =F (hfe thy) + § (fax + Qikfey + fy) + where on the right-hand side the function f and each of the following partial deriva- tives are evaluated at (x, y). EXAMPLE 4 What are the first few terms in the Taylor formula for f(x, y) = e0s(xy)? Solution For the given function, we find that af ysinjy) 3 = —xsincey) ef , Pf __a Bray TAPER) —sinGay) Fy = =A? cosy) ‘Thus, if we let n = 1 in Taylor’s formula (5), the result is cos[(x + A)(y +4)] = costxy) ~ hy sin(xy) — kx sin(xy) + Bx(h, ‘The remainder Ey is the sum of three terms—-namely, 1 2 = GHG + Ob) cost + 41D +90] — h(x + Oh)(y + 8k) cost(x + Oh)(y + OK)] + sin[(x + Oh)(y + OK)]) —!eeetonente+6b,0-+001 : PROBLEMS 1.1 1. Show that [x?—4] < ¢ when 0 < [x —2| < e544)" and prove lim, .2 x? = 4by using these inequalities 2. Show thatthe function f(x) = xsin(I/x), with f() Aiferentiabe a0 3. Show that f(x) = 6, is continuous at 0 but not is once differentiable at 0 but not twice. ? sin(L/x), with FO)1.1 Basic Concepts and Taylor's Theorem 13 Let f(x) = x°*(x — sin) for x # 0. How should (0) be defined in order that be continuous? Will tals be differentiable? ‘5. a. Derive the Taylor series at 0 forthe function f(x) = In(x + 1). Write this series in surimation notation. Give two expressions forthe remainder when the seties is truncated. ‘i, Determine the smallest number of terms that must be taken inthe series to yield In 1.5 with an error less than 10-* © Determine the numberof terms necessary to compute In 1.6 with error 10~ at most. 6, Determine whether the following function is continuous, once differentiable, or twice ditferentiable P4x-1 ifeso fe fee 1 ifxs Pax ifxso 7. (Continuation) Repeat the preceding problem for the function itx>1 ‘8 Criticize this reasoning: The function f defined by atx ite x ifx=0 wr-{ has the properties Jim, /*(s) = lim 6 =0 lim f(x) = lim 6x =0 Therefore, J" is continuous. 9. Prove tha iff is differentiable at x, then fer 2 2 =s's) ‘Show that for some functions that are not differentiable at x, the preceding limit exists, (See Eggermont [1988] or te following problem.) 10, Prove or disprove this assertion: If is differentiable atx, then fora #4 mn SO +N = fletah) ° hah £e) 11, Show that im, .1(4x +2) = 6 by means ofan €-5 proof 12, Show tht lim,2(1/2) = 3 by means ofan -5 proof 43. For the function f(x) = 3 — 2x +x? and the interval [a,b] that occurs in the Mean- Value Theorem, 14, (Continuation) Repeat the preceding problem with the function f(x) = x64.x4 — 1 and the interval [0,1]. 15, Find the Taylor series for f(x) = cosh.x about the point ¢ = 0. 16. the series for inx is truncated after the term involving (x ~ 1)!” and is then used to ‘eompute In2, what bound on the eror ean be given? 1,3, find the number €14 Chapter One Mathematical Preliminaries 17. Fin the Taylor series for f(x) = e* about the point = 3, Then simplify the series and show how it could have been obtained directly from the series for f about ¢ =0 18, Let k be a positive integer and let < a < 1. To what classes C*(R) does the function 3° belong? 19, Prove: If f € CYR), then f' € C%'(R) and f° F() dt belong wo C™*"(B. 20, Prove Rolle’ Theorem directly (aot asa special case ofthe Mea-Value Theorem). 21. Proves IF f € C*(R) and fe) = f1) = = Fhe) =O for x0 < xy < ++ < fy then FG) =O for some & € (Z,%). Hint: Use Rolie’ Theorem times. 22, Prove thatthe function f(x) = x*is continuous everywhere. 23, For small values of x the approximation sinx * x is often used. Estimate the error in using this formula withthe aid of Taylor's Theorem. For what range of values of x wil this approximation give results correct to six decimal places? 224, For small values of x, how good isthe approximation cos x = 1 ~ 127? For what ange of values wil this approximation give correct results tounded to three decimal places? 25. Use Taylor's Theorem with n = 2 to prove that the inoquality I+. < eis valid fr all real numbers except x = 0. 5. Derive the Taylor series with remainder term for In( I +) about 1. Derive an inequality that gives the numberof terms that must be taken to yield In 4 with errr less than 2-" 227, Whats the third term in the Taylor expansion of x +x ~ 2 about the point 3? 28. Using the series fr *, how many tems are needed to compute e* correctly to four deci- smal places (rounded)? 29, Develop the Taylor series for f(x) = In about e, writing the results in summation notation and giving the remainder term. Suppose [x ~ el < 1 and accuracy § x 10"! is desied. What isthe minimum numberof terms in the series required to achieve this | ooomnsy? } 20, Determine the fist 190 tems ofthe Taylor series for <* about 1 and the remsnder ‘erm Ey 21 Detemine the Taylor polynomial of dere 2 for fs) =e” expanded about te pam. | 42. Fit dcop union in sexes of powers of ~ 1) an ten 1 appro mate 0.99999 99995 to ten decimal places. a. joosx — (1 27, ; B fine —xC1—/5} i 4, Determine function at nb termed the Hnerizaen of? — 2 a2 } | ; 23S. How many terms are required in the series S1 Da to give e with an eror of at most 6/10 unit inthe 20th decimal place? 136, Find the first two terms in the Taylor expansion of x! about the point x = 32. Approx ‘mate the fifth root of 31.999999 using these two terms inthe series. How accurate is your answer? ial / 31 Find he Taylor polynomial of degree? fr the ntion (2) = & sn expanded shot 3 | the pont x/2.1.2. Orders of Convergence and Additional Basic Concepts 15 38. Determine the Lagrange form of the remainder when Taylor's Theorem is applied tothe function f(x) = cos.x, with n = 2 and ¢ = 2/2, How small must we make [x ~ 1/2] if this remainder term is not to exceed } x 10~ in absolute value? An error term of the form (—1)"(n + 1)~§-*-*(x — 1)" was obtained in the example illustrating Taylor's formula, Compare ths to the eror term that arises from the integral form ofthe remainder. Use Taylor's Theorem with Integral Remainder and the Mean-Value ‘Theorem for Inte- als to deduce Taylor's Theorem with Lagrange Remainder. Orders of Convergence and Additional Basic Concepts In numerical calculations, especially on high-performance computers, it often hap. pens that the answer to a problem is not produced all at once. Rather, a sequence of approximate answers is produced, usually exhibiting progressively higher accuracy. Convergence of sequences is an important subject that will be taken up again later, such as in Chapter 3 (p. 73). Here we present just a few introductory concepts. Convergent Sequences Let us consider an idealized situation in which a single real number is sought as the answer to a problem. It might be, for example, a zero of a complicated equation oor the numerical value of an intractable definite integral. In such a case, a computer program may produce a sequence of real numbers x1, %2, x3, ... that are approaching the correct answer. We write lim x if there corresponds to each positive ¢ a real number r such that xy — L| < ¢ when- ever n > r. (Here n is an integer.) For example, whenever n> &~! For another example, recall the equation