Anda di halaman 1dari 6

Solutions to Quiz 5

Single Variable Calculus, Jan - Apr 2013

Introduction Dear Class, Here you can nd detailed solutions to the problems from the quiz on Chapter 5: Discretization. In this quiz, there was only one variation for each problem. The Final Exam is already live. Use these solutions, and everything else to study, study, study!! As usual, I have added a tiny preamble to each problem which outlines the concepts that the problem tested. -Vidit Problem 1 This problem is just a limit computation! Basically, if you remembered how to show
n

lim

1 1+ n

= e,

then you should have had no trouble with this problem at all. Question: Compute the limit 2n1 2n + 1 lim n 2n 1 2 Answer: e . n+1 2n1 Set yn = 2 , so the problem is just asking for the value of limn yn . Since computing 2n1 limits of powers is hard in general, we make our lives easier by considering natural logs. That is, ln(yn ) = (2n 1) ln 2n + 1 2n 1 .

Now, we modify the input to the ln function like this: 2n + 1 2n 1 + 2 2 = =1+ . 2n 1 2n 1 2n 1 The reason we wrote things like this is because of Taylor series! Remember, ln(1 + x) = x + O x 0, so we have 2 2 1 ln 1 + = +O as n 2n 1 2n 1 (2n 1)2 Using this in our formula for ln(yn ) above gives ln(yn ) = 2 + O
1
1 x2

as

1 2n 1

SINGLE VARIABLE CALCULUS, JAN - APR 2013

and so, limn ln(yn ) = 2. From this, it is easy to see that our original limit is e2 : just take exponentials of both sides to get rid of that natural log. Problem 2 This problem tests your ability to manipulate falling powers. Remember, xk = x(x 1)(x 2) . . . (x k + 1). The question gives you the dierentiation formula xk = kxk1 . Question: Consider the sequence b = (3n2 n). Which of the following sequences a = (an ) satises a = b? Answer: a = (n3 2n2 + n). As suggested by the hint in the question, lets try to write each of the possible answers in terms of falling powers. To do this, rst compute n3 = n(n 1)(n 2) = n3 3n2 + 2n. Similarly, we have n2 = n(n 1) = n2 n. Look at the coecients of the n3 and n2 terms in an . Since we want to express an in terms of falling powers, we should compute the sum n3 + n2 . Lets do that: n3 + n2 = n3 2n2 + n = an . Using the dierentiation formula for falling powers, we see an = (n3 + n2 ) = 3n2 + 2n1 . Now to check that this expression actually equals bn , you have to go back from falling powers to normal exponents. This is easy: 3n2 + 2n1 = 3n(n 1) + 2n = 3n2 3n + 2n = 3n2 n = bn . So, we have shown a = b for this choice of a. Problem 3 This problem tests Eulers method for approximating solutions to a dierential equation dx = f(x, t). dt The basic idea is to discretize the time variable into a sequence tn and use the approximation xn+1 = xn + (t) f(xn , tn ). along with an initial condition x(t0 ). Question: Use Eulers method to compute the rst three iterates, (x1 , x2 , x3 ), of the dierential equation dx = t 2x dt starting with t0 = 0, x0 = x(t0 ) = 3 and using a constant step size h = t = 0.1. Answer: (2.4, 1.93, 1.564). The best way to keep track of everything is to build a table with all the necessary information. I like to start out like this, with all the initial data the easy things lled in: n tn xn f(xn , tn ) 0 0 3 1 0.1 2 0.2 3 0.3

SOLUTIONS TO QUIZ 5

Now the way to proceed is straightforward: the missing entry in the rst row of our table is just f(x0 , t0 ) = f(3, 0) = 6. Once you know this entry, you can compute x1 = x0 + (t) f(x0 , t0 ) = 3 + (0.1)(6) = 3 0.6 = 2.4, so now we can put all this new information in our table: n tn xn f(xn , tn ) 0 0 3 -6 1 0.1 2.4 2 0.2 3 0.3 Next, we compute f(x1 , t1 ) = f(2.4, 0.1) = 0.1 4.8 = 4.7. Immediately, this lets us compute x2 using the Euler method formula. We get x2 = x1 + (t) f(x1 , t1 ) = 2.4 0.47 = 1.93 Lets update the table: n 0 1 2 3 tn xn f(xn , tn ) 0 3 -6 0.1 2.4 -4.7 0.2 1.93 0.3

I will leave it to you to compute f(x2 , t2 ) and then x3 to get the nal answer. There is no need to compute f(x3 , t3 ) since the problem doesnt ask for the value of x4 . Problem 4 This problem relates to convergence tests for series. Question: Which of the following statements are true about the series

3 n2 (ln n)2

n=2

Choose all that apply.


(1) The integral x=2 x2 (3 dx converges, so the integral test says that the series converges. ln x)2 3 3 (2) n2 (ln n)2 < n2 for n > 2, so the comparison test says that the series converges. (3) The integral x=2 x2 (3 dx diverges, so the integral test says that the series diverges. ln x)2 3 3 (4) n2 (ln n)2 > n for n > 2, so the comparison test says that the series diverges. 3 (5) n2 (3 <n 4 for n > 2, so the comparison test says that the series converges. ln n)2 3 (6) limn n2 (ln n)2 = 0, so the n-th term test says that the series converges. + +

Answer: Only the rst two statements are true, the rest are false. To see that (1) is true, note that the integral does converge by the p-test for integrals. And the 3 comparison in (2) is valid, since the second series n 2 certainly has a smaller denominator for n > 2 than the one in our question. Of course, this second series converges by the p-test for series! Since the integral in (1) converges, you immediately know that (3) is false. Now (3) is trickier: all it is doing is comparing our series with the divergent series 3n . Being smaller than a divergent series doesnt automatically imply divergence: you have to be larger for that to work! With (5) one has to 3 be careful: the comparison doesnt work, since the denominator of n 4 is larger than our denominator, so that other series is actually smaller than our series term-by-term. Finally, the last statement is completely bogus: the n-term test only tells us when things diverge, not when things converge.

SINGLE VARIABLE CALCULUS, JAN - APR 2013

Problem 5 This problem tests the limit comparison test Question: Determine the asymptotic behavior of the sequence a = (an ) = n +. Using this information, determine whether the series 1 1 Answer: an = n +O n as n +, so the series diverges. 3 Lets use some basic Taylor series and big-O: e1/n = 1 +
2

n e1/n n

as

n=1

an converges or diverges.

1 1 , so +O 2 n n4 1 1 2 ne1/n = n + + O , and n n4 1 1 2 . ne1/n n = + O n n4 From this, and the fact that the sum
1 n=1 n

diverges, we get our answer!

Problem 6 This problem is about absolute vs. conditional convergence vs. divergence of series. Question: Which of the following statements are true? Select all that apply. n 2n (1) n=1 (1) 1+3n converges absolutely. n 2n (2) n=1 (1) 1+3n converges conditionally. n 2n (3) n=1 (1) 1+3n diverges. (1)n (4) n=1 2n+3 converges conditionally. (1)n (5) n=1 2n+3 converges absolutely. (1)n (6) n=1 2n+3 diverges. 3n 1000n3 diverges. (7) n=1 3n +1 3n 1000n3 (8) n=1 3n +1 converges absolutely. 3n 1000n3 (9) diverges. n=1 3n +1 Answer: Only (1), (4) and (7) are true, the other statements are false. Note that if (1) is true, then automatically (2) and (3) are false. Similarly, if (4) is true then (5) and (6) are false, and if (7) is true then (8) and (9) are false. After all, a series can do exactly one of three things: converge absolutely, converge conditionally or diverge. No two are possible at the same time! 2n To see that (1) is true, compare the absolute value 1+ of the summand to the geometric series 3n 2 n n=1 ( 3 ) . To see that (4) is true, rst note that the series does not converge absolutely by performing a 1 limit-comparison test with of the absolute value 2n1+3 to the divergent series n=1 2n . But note that n 1) these terms ( are going to 0 as n goes to , so certainly we have conditional convergence at least. 2n+3 n 1000n3 To see that (7) is true, use the n-th term test: the terms 3 do not go to 0 as n goes to : 3n +1 we can just use LHopitals rule four times to get rid of the 1000n3 from the numerator and the 1 from the denominator. This leaves us with a limit of 1. Problem 7 This problem tests your ability to compute the radius of convergence using the ratio test. (1)n n! n Question: For which values of x does the power series n=1 4812(4n) (x 1) converge? Answer: 3 < x < 5

SOLUTIONS TO QUIZ 5

Let an =

(1)n n! 4812(4n)

be our coecient. We know that the radius of convergence equals the limit R = lim |an | , n |an+1 |

whenever this limit exists. Lets compute this limit, noting that we can ignore all that (1)n stu because of the absolute value. For any n, we have the following: |an | = |an+1 | n! 4 8 12 (4n) 4 8 12 (4n + 4) (n + 1)! .

Almost everything cancels! We are left with the simple expression 4n + 4 |an | = =4 |an+1 | n+1 This sequence is constant: it doesnt even depend on n, so its limit is just 4. So, the radius of convergence is 4. Since the series is centered at 1, we have convergence whenever |x 1| < 4, which gives our answer. Problem 8 This problem tests your ability to compute the radius of convergence using the root test. 3n+1 n n Question: What is the radius of convergence R of the power series x ? n=0 2n+2 2 Answer: 3 . n+1 n . As the hint suggests, using the limit Lets write the coecients of our series as an = 3 2n+2 of a ratio like we did in Problem 7 wont work so well, since everything is raised to the n-th power! However, we have a test to solve exactly this problem. Remember, R= 1 limn |an |1/n

3n+1 Now, |an |1/n = 2 , and a single use of LHopital tells us that the limit of this sequence as n n+2 3 equals 2 . So, R which is just the reciprocal of this limit must be 2 . 3

Problem 9 This problem tests your understanding of error terms in alternating series. Remember, if you have n an alternating sum s = 0, and you want to compute how close you can get n=1 (1) an where an to s by only adding up the rst N terms, then you have the error term:
N

EN = s Question: Consider the series

(1)n an
n=1

|aN+1 |

s=

(1)n
n=1

16 n4

Using the alternating series error estimate, what is the best (lowest) estimate for N such that the error EN of the approximation
N

s= satises EN 10 ? Answer: N 200.


8

(1)n
n=1

16 + EN n4

SINGLE VARIABLE CALCULUS, JAN - APR 2013

We want to nd the lowest N so that EN 108 . By the alternating series error estimate, we just want to nd an N large enough so that |aN+1 |, which is larger than EN , is less than 108 . That is, 16 < 108 , so (N + 1)4 16 (N + 1)4 > 8 , so 10 (N + 1)4 > 16 108 . Taking 4-th roots, we get (N + 1) > 2 102 = 200. So, we have N > 199, which means N Problem 10 This question tests the remainder theorem for Taylor series. Remember, if the rst N terms of the Taylor seroes for f(x) at 0 have been chosen as an approximation, then the error term is bounded like this: n f(n) (x) n |fN+1 (c)| N+1 x < x , EN (x) = f(x) n! (N + 1)! n=0 where c is any point between 0 and x. Question: Suppose you try to approximate e1/2 using the rst three terms in the Taylor series expansion of the exponential function: x2 ex = 1 + x + + E2 (x) 2 1 What is the the best bound for E2 2 provided by Taylors theorem? 1 Answer: E2 1 2 48 0 1 One gets this answer by noting that E2 (1/2) < e ( )3 from the error formula above. After all, 3! 2 and 0 is the second derivative of ex is ex itself, and the largest value attained by ec for c between 1 2 1 1 0 at c = 0. Note that e = 1, so in fact our error term is bounded by 3!23 = 48 . 200.

Anda mungkin juga menyukai