MAT137 1 1 SET NOTATION Set Notation Elements and Sets Typically, we use lowercase letters such as x, y to denote elements (things). We use uppercase letters such as A, B to denote sets (a bag of things). Sets are made up of elements. If an element x is inside set A, we write x 2 A. Set Compliment Ac stands for the set complement of A. It contains ’everything’ except for the things in A. Set Containment We use the symbol ⇢ or ✓ to denote contained in. For example, A ⇢ B means set A is contained in set B. This means every element x in A, x is also in B. Then we arrive at the property of transitivity for set containment: If A ⇢ B and B ⇢ C, Then A ⇢ C If we interpret set A, B, C as bags, then the statement basically says: If bag A is inside bag B, but bag B is inside bag C, then bag A must be inside bag C. Set Equality Two sets are equal only when A ⇢ B and B ⇢ A. Union and Intersection We use the symbol [ to denote union of sets. We say an element x 2 A [ B only when x 2 A or x 2 B. We use the symbol \ to denote intersection of sets. We say an element x 2 A \ B only when x 2 A and x 2 B. Distributivity and DeMorgan’s Laws A [ (B \ C) = (A [ B) \ (A [ C) (A [ B)c = Ac \ B c Author: Xiao A \ (B [ C) = (A \ B) [ (A \ C) (A \ B)c = Ac [ B c Page 1 MAT137 2 2 PROOF TECHNIQUES Proof Techniques Instantiation Want to prove a statement P is true for many things (everything in set S). Example statement: 8x 2 S, P is true. (where S is a set, 8 means for every) 1. Pick x 2 S randomly, or ’arbitrarily’. Typically, things in S will have some common properties. 2. Use the properties to prove P is true for this specific x you picked. 3. Conclude P is true for all things in S. Contradiction Want to prove a statement P is true. Instead of proving P directly: 1. Suppose P is false. 2. After some logical reasoning, deduce an impossible statement. Example: 1=0 3. Conclude P is true. Implication Want to prove the statement: If A is true, then B is true. ( A ) B ) 1. Suppose A is true. 2. Using the assumption, prove B is true. 3. Conclude A ) B is true. Equivalent Want to prove: A if and only if B. ( A () B) 1. Prove ( A ) B ) is true. 2. Prove ( B ) A ) is true. 3. Conclude A () B is true. Author: Xiao Page 2 MAT137 2 PROOF TECHNIQUES Existence Want to prove something exist, which makes the statement P to be true. Example statement: 9x 2 S, P is true. (where S is a set, 9 means there exists at least 1) 1. Give one example of x 2 S. Typically, x will use quantities that are from former parts of the proof. x can also be taken directly as a element in S, such as x = 3 2 S. 2. Show that P is true for the specific x you have chosen. (Thus, we need choose x so that we can show P is true.) 3. Conclude x does exist, which makes P true. Set Containment Want to prove a set S1 is contained inside S2 : S1 ⇢ S2 . Method 1: Instantiation. 1. Pick x 2 S1 randomly, or ’arbitrarily’. 2. Using the common properties of S1 , prove x 2 S2 . 3. Conclude S1 ⇢ S2 . Method 2: Use contradiction. 1. Suppose S1 6⇢ S2 . 2. Immediately conclude 9x 2 S1 , x 62 S2 . 3. After some logical reasoning, deduce an impossible statement. 4. Conclude S1 ⇢ S2 . Set Equality Want to prove 2 sets are equal: S1 = S2 . 1. Prove S1 ⇢ S2 . 2. Prove S2 ⇢ S1 . 3. Conclude S1 = S2 . Author: Xiao Page 3 MAT137 2 PROOF TECHNIQUES Negation Want to prove a statement Q is not true. Type 1: Statements with quantifiers: for every 8 and exists 9 statement Q = 8x 2 S1 , 9y 2 S2 , P is true. (where P is another statement.) 1. Write the word not in front of the statement Q with a bracket. Example: not (Q) = not (8x 2 S1 , 9y 2 S2 , P is true.) 2. Bring the word not toward the right, passing the quantifiers 8, 9. Each time not pass across 8, change 8 to 9. Similarly, each time not pass across 9, change 9 to 8. Example: not (Q) = 9x 2 S1 , not (9y 2 S2 , P is true.) not (Q) = 9x 2 S1 , 8y 2 S2 , not (P is true.) not (Q) = 9x 2 S1 , 8y 2 S2 , P is false. 3. Prove the simplified statement of not (Q) attained at the end. Type 2: Statements with implication: statement Q = (A ) B) 1. Write the word not in front of the statement Q with a bracket. Example: not (Q) = not(A ) B). 2. Logically, the statement A ) B is equivalent to: not (A) or B. Use De Morgan’s Law and⇣ bring the word ⌘ not into the bracket. Example: not (Q) = not not (A) or B ⇣ ⌘ not (Q) = not not (A) and not (B) not (Q) = (A is true) and B is false. 3. Prove the simplified statement of not (Q) attained at the end. Contrapositive Want to prove the implication statement: A ) B Instead of proving the implication directly: 1. We can choose to prove the equivalent statement: not (B) ) not (A). 2. Suppose not (B) is true. 3. Prove not (A) is true. 4. Conclude A ) B is true. Author: Xiao Page 4 MAT137 2 PROOF TECHNIQUES Proof By Cases Split into cases, and prove the statement in both (all) cases. Type 1: Want to prove the statement: (P or Q) is true. 1. Consider the following two cases. Case 1: P is true. Case 2: P is false. (We need prove (P or Q) is true in both cases.) 2. In case 1: Conclude (P or Q) is true. In case 2: Use the new assumption P is false, prove Q is true. Conclude (P or Q) is true. 3. Conclude (P or Q) is true in both cases. Important Note: Alternatively, we can do the reverse, and split into the following 2 cases: Case 1: Q is true. Case 2: Q is false. Sometimes P and Q are very di↵erent statements, and only one of the above 2 case splits can lead to a proof. Type 2: We know (P or Q) is true. Want to prove statement R is true. 1. Consider the following two cases. Case 1: P is true. Case 2: Q is true. (We need prove R is true in both cases.) 2. In case 1: Use the new assumption P is true, prove R is true. (Do not use Q is true.) In case 2: Use the new assumption Q is true, prove R is true. (Do not use P is true.) 3. Conclude R is true in both cases. Induction Want to prove: 8n 2 N, P (n) is true. (P involves the natural number n) 1. Prove P (1) is true. 2. Suppose P (n) is true. Prove P (n + 1) is true. (Replace all n in statement P with n + 1) 3. Conclude 8n 2 N, P (n) is true. Author: Xiao Page 5 MAT137 2 PROOF TECHNIQUES Question 1 Prove the following statements are Equivalent: 1. A [ B = B 2. A ⇢ B 3. A \ B = A 4. B c ⇢ Ac Strategy: To prove several statements are equivalent, we typically prove a loop of implications. In this case, let’s try to prove 1 ) 2, 2 ) 3, 3 ) 4, 4 ) 1. In general, we can take any order of number we like. Proof Part 1: Prove 1 ) 2. This means we need to prove the statement: A [ B = B ) A ⇢ B. In proving an Implication, the first step is to suppose the left side is true. So we write in the first line of the proof: Suppose . Next, the statement we need to prove changes to: A ⇢ B. In proving a set containment, we first take a random element on the left. So we write in the second line of the proof to be: Let x 2 be chosen randomly. Now the statement we need to prove changes to: x 2 . We need to use our assumption in step 1 to arrive at this conclusion. Think: Since x 2 A, what is its relationship with A [ B in the assumption. Since x 2 A, Since A [ B = B, Thus x 2 . Thus 1 ) 2 is true. Author: Xiao . . Page 6 MAT137 2 PROOF TECHNIQUES Proof Part 2: Prove 2 ) 3. The statement we need to prove is: A ⇢ B ) A \ B = A. Since it’s an implication, we need to: Suppose . Next the statement we need to prove changes to A \ B = A. Note that this is a Set equality. We need to prove 2 statements: A \ B ⇢ A and A ⇢ A \ B. First, we prove: A \ B ⇢ A. Let x 2 be chosen randomly. The statement we need to prove changes to: x 2 Using the definition of intersection, we have x 2 . Thus x 2 . and x 2 Next, we prove A ⇢ A \ B. Let x 2 be chosen randomly. The statement we need to prove changes to: x 2 A \ B. By the definition of intersection, we need to prove: x 2 By the previous assumption, we already know x 2 . By the first assumption , we also have x 2 Thus x 2 A \ B. . and x 2 . . Thus A \ B = A. Thus 2 ) 3 is true. Author: Xiao Page 7 MAT137 2 PROOF TECHNIQUES Proof Part 3: Prove 3 ) 4. The statement we need to prove is: A \ B = A ) B c ⇢ Ac . Since it’s an implication, we need to: Suppose . Next the statement we need to prove changes to B c ⇢ Ac . Note that this is a Set containment. Let x 2 be chosen randomly. By definition of complement, x 62 . (62 means not in) The statement we need to prove changes to: x 2 By definition of complement, we need to prove: x 62 We use proof by contradiction. Suppose x 2 . Using the first assumption, we have x 2 . Using the definition of intersection, we have x 2 This is a contradiction, as we already know x 62 Thus we conclude x 62 Thus 3 ) 4 is true. Author: Xiao . . and x 2 . . . Page 8 MAT137 2 PROOF TECHNIQUES Proof Part 4: Prove 4 ) 1. The statement we need to prove is: B c ⇢ Ac ) A [ B = B. Since it’s an implication, we need to: Suppose . Next the statement we need to prove changes to A [ B = B. This is again a Set equality. First, we prove A [ B ⇢ B. Let x 2 A [ B be chosen randomly. The statement we need to prove changes to: x 2 . Note that as x 2 A [ B, then it is a possibility that x 2 A or x 2 B. Thus to continue, we need to split the proof into 2 cases. Case 1: x 2 A Case 2: x 2 B. (We need prove x 2 is true in both cases.) We use a little trick: Thus x 2 . As x 2 A, we have x 62 Ac . Using the first assumption , and x 62 Ac , we have x 62 . Thus x 2 . Thus in both cases, x 2 . Next, we prove B ⇢ A [ B. Let x 2 be chosen randomly. The statement we need to prove changes to: x 2 A [ B . By definition of union, we need to prove: x 2 By the previous assumption, we have x 2 Thus x 2 . . or x 2 . Thus A [ B = B. Thus 4 ) 1 is true. Author: Xiao Page 9 MAT137 2 PROOF TECHNIQUES Question 2 Use induction to prove the following statements P (n) to be true for all n 2 N: 1. P (n) : n X k = 1 + 2 + 3 + ... + n = k=1 2. P (n) : n X n(n + 1) 2 k 2 = 12 + 22 + 32 + ... + n2 = k=1 n X n(n + 1)(2n + 1) 6 n2 (n + 1)2 3. P (n) : k = 1 + 2 + 3 + ... + n = 4 k=1 Author: Xiao 3 3 3 3 3 Page 10 MAT137 3 ✏ 3 ✏ LIMIT DEFINITION Limit Definition The standard limit is as follows: lim f (x) = L x!a However, what exactly do these symbols mean? Intuitively, it is understood as follows: We think of a function as an arrow, taking one value a in the domain X, and ’shoot’ it to the value f (a) of the codomain Y. f :X!Y a 7! f (a) (In this course, X and Y will be both R, the set of all real numbers.) We would like to ask: If instead of starting at a, we are at the point x that is really close to a (in X), can we make sure that f (x) is really close to f (a) as well (in Y )? This seems like a reasonable requirement, and is the intuition behind limit. How can we translate this sentence into mathematical language? To keep things simple, let’s say the target is L = f (a), and we want to get f (x) to be close to L. It would be great if f (x) hit exactly L, however, that is too high of a requirement. Instead, we create an error range around L. We require f (x) to hit inside this error range. We call the radius of this error range to be ✏. After picking this error range, we need to control the values of x we pick. If x is far from a, we shouldn’t expect f (x) to be in the error range. So we give a control range centered at a for the values of x we can pick. We call the radius of this control range to be . What should the control range radius satisfy? We should be able to guarantee that for any value of x taken from the control range near a, f (x) will be inside the error range, near L. i.e. We can achieve our goal of f (x) being close to L, by concentrating the value of x around a, to be inside the control range with radius . Author: Xiao Page 11 MAT137 3 ✏ LIMIT DEFINITION Now to put it into mathematical language: (Recall that the distance between any 2 points a and b, in R is d(a, b) = |b a|) For any ✏ > 0 as the error radius, there will exist > 0 as the control radius, such that for all x inside the control radius (which is |x a| < ), we can guarantee that f (x) is in the error range (which is |f (x) f (a)| < ✏). In symbols: lim f (x) = L x!a 8✏ > 0, 9 > 0, if 0 < |x a| < ) then |f (x) L| < ✏ Remember that the value a is a fixed point given to us from the start, and x is a variable value that can change inside the control range. So: |x a| < 0 < |x ,a |f (x) < x < a+ L| < ✏ , L ✏ < f (x) < L + ✏ a| is just a detail: for limits, we do not consider the possibility of x = a. Using a similar idea, we can say the limit at infinity is done by controlling x to be very large. So instead of concentrating x inside a control range with radius , we concentrate x ”at infinity”, by requiring x > N , where N is chosen to be large. lim f (x) = L x!1 8✏ > 0, 9N, if x > N ) then |f (x) L| < ✏ What about limit at a does not exist? It means no matter what value of L (for all L), lim f (x) 6= L. x!a Negating the statement would interchange 8 and 9, negating the implication would mean there is a counter-example to the implication, namely, there is an x in the control range , with f (x) outside the error range ✏. 8L, 9✏ > 0, 8 > 0, 9 x , with 0 < |x a| < and |f (x) L| ✏ One type of limit not existing is f (x) going to infinity. In this case, we do have a goal for the values of f (x), which is to be very large (larger than a large number M ), as long as x is concentrated closed enough to the point a. lim f (x) = +1 x!a 8M > 0, 9 > 0, if 0 < |x Author: Xiao a| < ) then f (x) > M Page 12 MAT137 4 ✏ 4 ✏ PROOF OF LIMIT DEFINITION Proof of Limit Definition The standard statement we need to prove is: lim f (x) = L x!a 8✏ > 0, 9 > 0, s.t. if 0 < |x a| < ) then |f (x) L| < ✏ We would need to follow the previous proof techniques: 1. Pick a random ✏ > 0. 2. Find a value of which would make the implication to be true. ( can depend on ✏ that was chosen in step 1.) 3. Proving the implication, first assume 0 < |x 4. Attempt to prove the statement |f (x) a| < L| < ✏ The problem is step 2, on how we can obtain this . Thus, we first start rough work as follows: 1. Write the quantity |f (x) L| 2. Attemp to factor this quantity into two pieces: |x a| · |S| where S is the extra stu↵ left over from the factorization. (Note: For any 2 numbers a and b, |a · b| = |a| · |b|.) 3. Hand pick a number k > 0. Usually pick k = 1, but sometimes need to pick smaller. 4. Write a k <x<a+k 5. Use step 4 and some algebra to establish the bound on |S|. Attain m < |S| < M , where m and M are positive with no x involved. When M does not exist, pick a smaller k from step 3 and repeat. 6. Write |x a| · |S| < |x a| · M < · M < ✏ ✏ }, so we have both: M ✏ |x a| < k and |x a| < . M Using this , we would write the formal proof. 7. Conclude we pick Author: Xiao = min{k, Page 13 MAT137 4 ✏ PROOF OF LIMIT DEFINITION Let’s try an example: Prove lim x2 = 9 x!3 8✏ > 0, 9 > 0, s.t. if 0 < |x 1. Write the quantity |f (x) |x2 9| 3| < L| ) then |x2 9| < ✏ 2. Attemp to factor this quantity into two pieces: |(x 3)(x + 3)| = |x 3| · |x + 3| = |x 3| · |S| 3. Hand pick a number k = 1. 4. Write 3 1<x<3+1)2<x<4 5. Use step 4 and algebra to establish the bound on |S|. 5<x+3<7 5 < |x + 3| < 7 6. Write |x 3| · 7 < · 7 < ✏ ✏ 7. Conclude we pick = min{1, }, so we have both: 7 ✏ |x 3| < 1 and |x 3| < . 7 Formal Proof : Let ✏ > 0 be chosen randomly. ✏ Pick = min{1, }. 7 ✏ Then, we have 1 and . 7 Assume 0 < |x a| < . (Copy the work of step 2, 3, 4, 5) |f (x) L| = |x2 9| = |x 3| · |x + 3| Since |x 3| < 1 ) 2<x<4 5<x+3<7 5 < |x + 3| < 7 Then |x 3| · |x + 3| < |x 3| · 7 Since |x 3| < ) |x 3| · 7 < · 7 ✏ ✏ Since ) ·7 ·7=✏ 7 7 ✏ Thus |x2 9| = |x 3| · |x + 3| < |x 3| · 7 < · 7 · 7 = ✏. 7 Author: Xiao 3| · |x + 3| < |x Page 14 MAT137 4 ✏ PROOF OF LIMIT DEFINITION Question 3 Prove the following limits. 1. lim x2 = 4 x!2 8✏ > 0, 9 > 0, s.t. if 0 < |x Strategy: Start with the quantity |f (x) 2| < L| and factor out |x 2. lim x!9 p b) = a2 3. lim x!3 4. p 4| < ✏ 2| x=3 8✏ > 0, 9 > 0, s.t. if 0 < |x Strategy: Use the formula (a + b)(a ) then |x2 9| < p ) then | x 3| < ✏ b2 to eliminate square root. x+1=2 1 =1 x!1 x lim Strategy: Notice that k = 1 does not provide a bound for the quantity we need, so we need to pick a smaller k. 5. 1 1 = x!1 x2 + 1 2 lim 6. lim |1 x!2 3x| = 5 Strategy: In our rough work, we may assume x is be concentrated close enough to 2 (approximately 2) to determine whether the quantity 1 3x is positive or negative. However, we need to make this precise in our formal proof (with the value of k). Author: Xiao Page 15 MAT137 4 ✏ PROOF OF LIMIT DEFINITION Question 4 Given L as a constant, suppose lim f (x) = L x!0 Prove lim f (2x) = L x!0 Strategy: We need to use our assumption to prove the statement. This means we need to write the definition of our assumption and the statement which we need to prove. We need to use new names for variables in our assumption. So, we suppose 8✏2 > 0, 9 2 > 0, s.t. (for all x2 ), if 0 < |x2 0| < 2 ) then |f (x2 ) L| < ✏2 We need to prove 8✏ > 0, 9 > 0, s.t. (for all x), if 0 < |x 0| < ) then |f (2x) L| < ✏ Notice that in the second statement, we have the quantity |f (2x) L|. Author: Xiao Page 16 MAT137 4 ✏ PROOF OF LIMIT DEFINITION Question 5 Part 1: Suppose we know the following about the function f (x): if |x 2| < 1 ) then |f (x) 5| < 2 Which of the following statements are necessarily true? A statement below is necessarily true means: We can prove that statement given below using the assumption above. A statement below is not necessarily true (considered as false) means: We can give a counter-example: a function f (x) which satisfy our assumption above, and does not satisfy that statement given below.) 1. if |x 2| < 1 ) then |f (x) 2. if |x 2| < 0.5 ) then |f (x) 3. if |x 2| < 1 ) then |f (x) 4. if |x 2| < 0.5 ) then |f (x) 5. There exists Author: Xiao 5| < 3 5| < 2 5| < 1 5| < 1 > 0 such that if |x 2| < ) then |f (x) 5| < 1 Page 17 MAT137 4 ✏ PROOF OF LIMIT DEFINITION Part 2: We still have the assumption, if |x 2| < 1 ) then |f (x) 5| < 2 Now we suppose (moreover that) lim f (x) = L x!2 This means, we know the following is true: 8✏ > 0, 9 > 0, s.t. if 0 < |x 2| < ) then |f (x) L| < ✏ Question: 1. Can L = 2? (L can equal 2 means that we can give an example f (x) with L = 2. L can not equal 2 means that we can prove using our assumption that L can not (can never) equal 2.) 2. Can L = 2.5? 3. Can L = 8? Make a preliminary conclusion. 4. Can L = 5? 5. Can L = 4? 6. Can L = 3? Make a final conclusion. Author: Xiao Page 18 MAT137 5 5 LIMIT AND CONTINUITY Limit and Continuity Two Sided Limits The limit at a point a of the function f (x) is equal to the number L: 1. When you let x approach a from the left side of a, the graph of f (x) gets close to the number L. 2. When you let x approach a from the right side of a, the graph of f (x) gets close to the number L. 3. The actual value of f (a) does NOT matter at all. lim f (x) = L x!a One Sided Limits We only let x approach a from the left side of a, and the graph of f (x) get close to the number L1 : lim f (x) = L1 x!a The minus sign next to a indicates that we are approaching from the left. Similarly, we can approach a from the right to attain another limit L2 : lim f (x) = L2 x!a+ The plus sign indicates that we are approaching from the right. Only when the one sided limits exist, and the numbers L1 = L2 , we say that the limit of f (x) at a (from both sides) exist: lim f (x) = lim f (x) = L1 = lim+ f (x) = L2 x!a x!a x!a Continuity Intuitively, f (x) is continuous in an interval if the graph is in one piece. f (x) is continuous at the point a if the limit at a is equal to f (a): lim f (x) = f (a) x!a f (x) is continuous in an interval means f (x) is continuous at every point in the interval. Most functions are continuous, most limits can be obtained by putting x = a in the formula of f (x). Every standard function we know are continuous in their domains. Any function transformations (such as +, , ⇥, /, composition) of continuous functions, are also continuous. (except division by 0) Author: Xiao Page 19 MAT137 5 LIMIT AND CONTINUITY Limits of x approaching infinity We allow a to be infinite as well. This is understood intuitively as: As x becomes larger and larger, f (x) gets closer to a number L. lim f (x) = L x!1 We can also have a being minus infinity: lim f (x) = L x! 1 Limit Properties First, we need (to know) the following limits to exist, lim f (x) = L, x!a lim g(x) = M x!a Then we can attain lim [f (x) + g(x)] = L + M x!a lim [f (x) x!a g(x)] = L M lim c · f (x) = c · L x!a lim [f (x) · g(x)] = L · M x!a f (x) L = if M 6= 0 x!a g(x) M If f (x) is continuous, then we can move the limit inside f : lim lim f (g(x)) = f [lim g(x)] x!a x!a In fact, we only need to assume f (x) is continuous at lim g(x). x!a Infinite Limits We also allow one sided limits to be infinite: When you let x approach a from one side of a, the graph of f (x) gets larger and larger. We write: lim f (x) = 1, x!a lim f (x) = 1 x!a+ We can also have limits being minus infinity as well, where the graph goes lower and lower. lim f (x) = x!a 1, lim f (x) = x!a+ 1 If the limit is infinity, it is regarded as the limit does not exist. So we can not use the limit properties above. Author: Xiao Page 20 MAT137 5 LIMIT AND CONTINUITY Verticle Asymptotes When f (x) is given by a fraction, we always require that the denominator is not zero. For example: 1 f (x) = x 1 Clearly, x = 1 is not allowed. As x gets close to 1, or any other number with division by 0, f (x) will usually become infinite. This is called a verticle asymptote at the point x = 1. Because x = 1 is not allowed, the graph can not cross the verticle asymptote. In terms of limits, we write lim x!1 1 x 1 = ±1 where lim x!1 1 x 1 = 1 and lim+ x!1 1 x 1 = +1 1 Note that the quantity does not exist, and is not equal to 1. 0 Only when we take the limit, we can say it is either +1 or 1. Horizontal Asymptotes As x approaches ±1, sometimes f (x) will attain a limit. This is called the horizontal asymptote. The graph is allowed to cross the horizontal asymptote. For example: 1 f (x) = x 1 As x becomes large in either positive or negative direction, the bottom is 1 large. So it can be thought of as ±1 lim x!1 1 x 1 ⇡ 1 =0 1 lim x! 1 1 x 1 ⇡ 1 =0 1 1 is ’equal’ to 0, since that is the unique answer ±1 we obtain after taking the limit. In this case, the value of Author: Xiao Page 21 MAT137 5 LIMIT AND CONTINUITY Question 6 Evaluate lim (x2 + 2x ). x!1 Question 7 Find conditions on A and B so that 8 > <Ax B x 1 f (x) = 3x 1<x<2 > : 2 Bx A 2x is continuous at x = 1 and discontinuous at x = 2. Question 8 Are the following statements true or false? Justify your answers. (Prove the statement is true, or give a counterexample if it is false.) 1. Suppose lim [f (x) + g(x)] exists and lim f (x) exists, x!a x!a then lim g(x) also exists. x!a 2. Suppose lim [f (x) + g(x)] exists but lim f (x) does not exist, x!a x!a then lim g(x) also does not exist. x!a 3. Suppose lim f (x) exists but lim g(x) does not exist, x!a x!a then lim [f (x) + g(x)] also does not exist. x!a 4. Suppose lim f (x) does not exist and lim g(x) does not exist, x!a x!a then lim [f (x) + g(x)] also does not exist. x!a 5. Suppose lim f (x) = L 6= 0 and lim [f (x)g(x)] = 1, x!a x!a then lim g(x) also exists. x!a 6. Suppose lim f (x) = 0 and lim [f (x)g(x)] = 1, x!a x!a then lim g(x) does not exist. x!a Author: Xiao Page 22 MAT137 5 LIMIT AND CONTINUITY Question 9 Prove the ”Continuity law for composition”: If f (x) is continuous at lim g(x), then we can move limit inside: x!a lim f (g(x)) = f [lim g(x)] x!a x!a Deduce this implies: Composition of continuous functions is continuous. Author: Xiao Page 23 MAT137 6 6 LIMIT EVALUATION TECHNIQUES Limit Evaluation Techniques Rational Functions Rational functions is a polynomial dividing a polynomial. Try to factor the numerator and the denominator. When cancellation occurs, one can try to plug in x = a to attain limit. f (x) = x2 + 3x x(x + 3) x = = x2 x 12 (x + 3)(x 4) x 4 ) lim f (x) = x! 3 3 3 4 = Rational Functions with x Approaching Infinity Only the term with the highest exponent in the numerator and denominator matters when computing limits. lim f (x) = x!1 24x6 + 1 24x6 ⇡ = 12 2x6 + 30x5 + 1 2x6 ) lim f (x) = 12 x!1 To be precise, we divide by the highest exponent on the numerator and denominator. In this example, x6 is the highest exponent. 6 lim f (x) = x!1 6 (24x + 1)/x = (2x6 + 30x5 + 1)/x6 24 + 1 x6 30 1 2+ + 6 x x = 24 + 0 24 = = 12 2+0+0 2 In general, am xm + ... + a0 bn xn + ... + b0 with degree m in the numerator and degree n in the denominator. The highest exponent of a polynomial is called the degree. f (x) = 1. m < n The denominator grows faster, the limit is 0 as horizontal asymptote. 2. m = n The numerator and denominator grows at the same pace. am The limit is f (x) ⇡ as horizontal asymptote. bn 3. m > n The numerator grows faster, and the limit is ±1. am x m The sign depends on f (x) ⇡ bn x n Author: Xiao Page 24 3 7 MAT137 6 LIMIT EVALUATION TECHNIQUES Square Roots Try to use the relation a2 b2 = (a + b)(a b). Set a to be the square root, then we multiply and divide by the other factor, and try to simplify. This is called multiply by conjugate. p p p 4x + 1 3 4x + 1 3 4x + 1 + 3 4x + 1 32 p f (x) = = ·p = x 2 x 2 4x + 1 + 3 (x 2)( 4x + 1 + 3) 4(x 2) 4 4 2 p =p ) lim f (x) = = x!2 6 3 (x 2)( 4x + 1 + 3) 4x + 1 + 3 Small Angle Approximation If the limit is given as x ! 0, then we can make the substitution = sin(x) =1 x!0 x sin(x) ⇡ x ⇡ tan(x) i.e. lim tan(x) =1 x!0 x lim Note: This also works for arcsin(x) and arctan(x). Absolute Values When absolute value is in f (x), approach from the left side and right side separately. Determine whether the expression inside the absolute value is positive or negative, and remove the absolute value sign. lim+ ( x!0 1 x 1 1 )= |x| x 1 =0 x lim ( x!0 1 x 1 1 )= |x| x 1 2 = = x x 1 Squeeze Theorem To attain lim f (x), can try to find g(x) and h(x) so that: x!a 1. g(x) f (x) h(x) near the point a. 2. lim g(x) = L = lim h(x) x!a x!a Then we conclude lim f (x) = L. x!a In practice, we want to show lim f (x) = 0. Start at |f (x)|, create a chain x!a of inequality, simplify |f (x)| to attain |g(x)|, which has limit 0 : |f (x)| ... |g(x)| ! 0 i.e. lim |g(x)| = 0 x!a Then we conclude lim f (x) = 0. x!a Author: Xiao Page 25 MAT137 6 LIMIT EVALUATION TECHNIQUES Question 10 x2 1. lim x!3 x h3 h!2 h3 2. lim 2 x!3 x 3. lim 9 3 13. lim 5h2 + 3h + 6 h2 3h + 2 14. lim x 3 2 x x!3 (x 3)2 p x+1 1 5. lim x!0 x 4. lim p y!1 y+4 p x!0 17. lim e1/x x!0 t t!0 sin(2t) 19. lim sin(2x) x!0 sin(3x) 20. lim sin(2z 2 ) z!0 cos(3z)sin2 (5z) 21. lim tan(x 3) x!3 2x 6 22. lim 2ex x!0 sin(2ex ) 23. lim x2 sin 1 x 24. lim 2x sin 1 x 8. lim 9. lim 10. lim 11. lim 12. lim 1 t!0 cos(3t) t2 y 16. lim+ e1/x 18. lim 7. lim 4y + 1 1 sin(1 cos(x)) x!0 xtan(⇡x) r 1 u+2 15. lim ( 2) u!2 2 u u 1 sin(x) x!1 x 6. lim p e3x 2e x!1 4e3x + e x! 3x 3x e3x 2e 1 4e3x + e 3x 3x 3x 5x x!1 32x + 24x x! 3x 5x 1 32x + 24x x! 6 2x + 12 |x + 6| x!0 x!0 cos 1 x Question 11 Define the function at x = 5 so that f (x) becomes continuous. p 2x 1 3 f (x) = x 5 Author: Xiao Page 26 MAT137 7 7 TYPES OF DISCONTINUITY Types of Discontinuity Recall, f (x) is continuous at point a if lim f (x) = f (a) x!a Note that this equality contains 2 independent requirements: 1. The limit lim f (x) = L would exist (as a 2-sided limit). x!a Recall that the existence of this limit does not depend on f (a). 2. The value of L must be equal to f (a), provided that f (a) exist: i.e. a is in the domain of f (x). Thus for f (x) to be not continuous at a, it could fail either 1 or 2. Removable Discontinuity In this scenario, condition 1 holds: lim f (x) = L x!a exists However, condition 2 does not hold: f (a) 6= L or f (a) does not exist In any case, we can redefine the function at a, by setting f (a) = L. Then the function becomes continuous at point a. We have ”removed” the discontinuity. Removable discontinuity typically arises when an cancellation occurs in a fraction (rational function). For example: (x 1)(x + 1) f (x) = x 6= 1 x 1 Note that f (1) does not exist. However, it is clear that lim f (x) = 2. x!1 Thus we may redefine f (1) = 2. Then f (x) becomes: ( (x 1)(x+1) x 6= 1 x 1 f (x) = 2 x=1 is now continuous at a = 1. Note that f (x) = x + 1 everywhere. Author: Xiao Page 27 MAT137 7 TYPES OF DISCONTINUITY Jump Discontinuity In this scenario, condition 1 does not hold: lim f (x) does not exist as 2 -sided limit x!a However, both 1-sided limits exist but are not equal: lim f (x) = L1 lim f (x) = L2 x!a+ x!a L1 6= L2 Note that these limits again does not depend on f (a). Intuitively, f (x) ”jumps” from L1 to L2 as x goes acroos a. Jump discontinuity typically arises when a function is piecewise defined. For example: ( 1 x>0 |x| = f (x) = x 1 x<0 We have lim f (x) = x!0 1 and lim+ f (x) = 1, as a jump at x = 0. x!0 Infinite Discontinuity In this scenario, one or both of the 1-sided limits are infinite: lim f (x) = ±1 x!a or lim f (x) = ±1 x!a+ or both Infinite discontinuity typically arises when a fraction is dividing by 0. However, we need to require that the numerator to be non-zero. For example: x 1 f (x) = x 6= 1, 1 (x 1)(x + 1) Note that lim + f (x) = x! 1 2 1 1 [+] lim + = lim + = =1 2 x! 1 x + 1 x! 1 x + 1 [+] where a non-zero number dividing by 0 gives infinity. lim+ f (x) = x!1 1 x lim+ 2 x!1 x 1 1 1 = lim + · 1 = 1 x! 1 2 2 where 0 dividing by 0 does not gives infinity, and is called an indeterminant form, which requires further manipulations. In this case, we ”cancel” (x 1) on the numerator and denominator to attain the limit. Author: Xiao Page 28 MAT137 7 TYPES OF DISCONTINUITY Essential Discontinuity The above 3 scenarios covers most instances of discontinuity. Any discontinuity that is not the above 3 scenarios is called an essential discontinuity. Essential discontinuity typically arises with limits similar to 1 lim sin x!0 x 1 approaches infinity at a faster and faster pace. x The function sin(z) oscillate 1 period whenever the input increases by 2⇡. 1 The input of sin(z) in this case is z = , which increases faster and faster x toward infinity, passing through multiples of 2⇡ at an increasingly fast pace. 1 Thus sin oscillates faster and faster as x approach 0. x As x approach 0, An ”intuitive” limit at x = 0 would be ”all numbers between 1 and 1”. However, a limit must equal to 1 number only, not infinitely many numbers. Thus 1 lim sin does not exist x!0 x with no further classifications. Similarly, lim cos x!0 Author: Xiao 1 x does not exist Page 29 MAT137 8 8 IDEA OF DERIVATIVE Idea of Derivative The derivative of f (x) at the point a is f (a + h) f (a) h!0 h We take the limit of h approaching 0, so h 6= 0. As h is on the denominator, we can not simply substitute h = 0 into the expression. We need to perform algebraic manipulations to the fraction in order to cancel the h on the denominator. Then we may substitue h = 0 to attain the answer to f 0 (a). f 0 (a) = lim Note that f 0 (a) is typically di↵erent with di↵erent point a. This can be understood as a function: f (x + h) f (x) f 0 (x) = lim h!0 h where x is now also thought of as the variable to the function f 0 . There are several ways to intuitively understand the derivative f 0 (a): 1. Graph - Slope of f (x) On the graph of y = f (x), m = f 0 (a) corresponds to the ⇣slope of ⌘ the tangent line at the point a. The 2 points on the graph a, f (a) and ⇣ ⌘ a + h, f (a + h) define a line, where x = a + h is either slightly left or right of the point a. The slope of this line is the di↵erenece quotient f (a + h) f (a) f (a + h) = (a + h) a h f (a) which appears in the definition of the derivative. As we take the limit of h approaching 0, this line becomes a good approximation of the graph near point a. Thus the slope of this line approaches the slope of the f (x) at the point a. The tangent line at x = a is defined to be the⇣line with ⌘ the slope as 0 the derivative f (a) passing through the point a, f (a) . Using pointslope form of the line: y Author: Xiao f (a) = f 0 (a)(x a) Page 30 MAT137 8 IDEA OF DERIVATIVE 2. Rate of Change - Sensitivity to Input Changes Consider an example f (x) = 2x where f is linear. First set x = 1, we have f (1) = 2. Now if we change the input by 1 unit, with x = 2, we have f (2) = 4. We see that in this example, if we change the input by h units, the output is changed by 2h units. Typically if we change the input by a lot, we would expect the output to change a lot. However, the ratio of the changes between the output and the input is the more meaningful information. This ratio is the di↵erenece quotient f (a + h) f (a) = Average Rate of Change h The derivative measures this ratio when a very small change is done to the input (h ! 0). We would expect a small change to the input would result in a small change to the output. However, the ratio of the changes between the output and the input may in fact be small or large. A large ratio (positive or negative) indicates that the function f (x) is very sensitive to changes to its input near the point a, where a small change to the input will result in a (relatively) large change in output. A ratio of 0 indicates that the function is not sensitive to changes to its input near the point a, where a small change to the input will result in an (extremely) small change in output. By taking the limit of h approaching 0, while the change in input becomes smaller and smaller, the ratio of the changes between the output and the input may not become small. In fact, the value which this ratio approaches to, is the derivative f 0 (a), the instantaneous rate of change at point a, and it is the ’most accurate’ number to describe the sensitivity of input for f (x) at the point x = a. Because we take the limit of h approaching 0, the answer we attain only meausres the sensitivity near the point x = a. The derivative is a local phenomenon. f 0 (a) does not give any information of f (x) at any other points. Author: Xiao Page 31 MAT137 8 IDEA OF DERIVATIVE 3. Velocity Let’s use t in place of x for the variable. Let f (t) = s(t) be the position of a particle at time t. The average velocity between times t = a and t = b is s(b) s(a) di↵erence in position = b a time used This correspond ⇣ ⌘ ⇣ to the ⌘ slope of the secant line defined by 2 points a, s(a) and b, s(b) on the graph of s(t). vavg = The instantaneous velocity, or just simply velocity, at time t = a is the derivative: s(a + h) s(a) s(t) s(a) vinst = v(a) = s0 (a) = lim = lim t!a h!a h t a ⇣ ⌘ as the slope of the tangent line at a, s(a) . The velocity can again be thought of as a function of time: v(t) = s0 (t) Question 12 Find the derivative f 0 (a) by definition. 1. f (x) = x3 1 2. f (x) = p x Question 13 Find the equation of the tangent line at the point given. f (x) = x3 at (1, 1) Question 14 1. Let f 0 (x) = 0 for all x and f (0) = 0. What can you say about f (2)? 2. Let f 0 (0) = 0 and f (0) = 0. What can you say about f (2)? Question 15 Let s(t) = t2 6t. Find the average velocity for t between [4, 8] and the instantaneous velocity at t = 8. Author: Xiao Page 32 MAT137 9 9 THREE IMPORTANT THEOREMS Three Important Theorems Intermediate Value Theorem (IVT) If f (x) is continuous on [a, b], and f (a) < 0, f (b) > 0, then exists c 2 (a, b) with f (c) = 0. Intuition: To draw a continuous curve from below the x-axis to above the x-axis, the curve must cross the x-axis. Intermediate Value Theorem: Generalized Version If f (x) is continuous on [a, b], then for any K between f (a) and f (b), there exists c 2 (a, b) with f (c) = K. Intuition: To draw a continuous curve from y = f (a) to y = f (b), the curve must cross all possible horizontal lines y = K between y = f (a) and y = f (b). The theorem guarantees the existence of c, but gives no way of finding c. Extreme Value Theorem (EVT) If f (x) is continuous on [a, b], then f (x) has maximum and minimum on the interval [a, b]: 1. There exists xM , with f (xM ) f (x) for all x 2 [a, b], as a maximum. 2. There exists xm , with f (xm ) f (x) for all x 2 [a, b], as a minimum. Intuition: On a closed interval [a, b], a continuous curve would have a highest point and a lowest point. Note that the assumption of closed interval is crucial: If we only assume f (x) to be continuous on (a, b], then it is possible that f (x) becomes infinite at a, in which case there would not exist a maximum. The theorem guarantees f (x) would reach maximum and minimum, but the maximum or minimum may be reached at more than 1 point. The theorem again gives no way of finding the maximum and minimum. Author: Xiao Page 33 MAT137 9 THREE IMPORTANT THEOREMS Mean Value Theorem (MVT) If f (x) is continuous on [a, b], di↵erentiable on (a, b), then exists c 2 (a, b) with f (b) f (a) f 0 (c) = b a Intuition: Consider a car driving forward on a highway. Suppose the car travels 100km forward in an hour. Can I conclude that at some time during the trip, the velocity of the car is exactly 100km/h? 1. If the car always travels faster than 100km/h, then after 1 hour, it would have travelled more than 100km. 2. Similarly, if the car always travels slower than 100km/h, then after 1 hour, it would have travelled less than 100km. 3. Thus there must be a time t1 where the velocity is below 100km/h, and another time t2 where the velocity is above 100km/h. To change the velocity from below 100km/h to above 100km/h, the car must pass through the velocity of 100km/h at some time c between t1 , t2 . Let f (x) = s(t) as the position of a car driving forward on a highway. Then f 0 (x) = s0 (t) = v(t) is the velocity of the car. Then Mean Value Theorem states (with the new notation): There must exists c 2 (a, b) with v(c) = s0 (c) = s(b) b s(a) Total distance = = average velocity a Total time It is important to require f (x) to be di↵erentiable. In the above example, the velocity of the car must change in a smooth and continuous manner. The theorem guarantees the existence of c, but gives no way of finding c. Question 16 Show that there are at least 3 solutions in Estimate their values. Question 17 Give the region of x with (x ⇣ x ⌘3 ⇡ 2 sin x + 1 = 0? 2 x > 0, by finding all zeroes and 1)(x + 1) discontinuities of f . Author: Xiao Page 34 MAT137 10 10 DERIVATIVE Derivative f (a + h) f (a) h!0 h The function is di↵erentiable at point a if this limit exists as the derivative at a. f 0 (a) = lim f is di↵erentiable implies f is continuous. Note that continuous does not imply di↵erentiable: f (x) = |x| is continuous, but not di↵erentiable at x = 0. Tangent Line The point-slope form of line: y y0 = m(x x0 ) where (x0 , y0 ) is on the line. The tangent line at (x0 , y0 ) = a, f (a) has slope m = f 0 (a). Derivative Rules dy d = (y) (c)0 = 0 (f +g)0 = f 0 +g 0 (cf )0 = cf 0 dx dx f f 0g g0f (f g)0 = f 0 g + g 0 f ( )0 = [f (g(x))]0 = f 0 (g(x)) · g 0 (x) 2 g g 1 (xn )0 = n · xn 1 (ex )0 = ex (ln x)0 = x = eln(x) = ln(ex ) x (sin(x))0 = cos(x) (cos(x))0 = sin(x) sin(x) 1 1 cos(x) tan(x) = csc(x) = sec(x) = cot(x) = cos(x) sin(x) cos(x) sin(x) y = f (x) f 0 (x) = (tan(x))0 = sec2 (x) (sec(x))0 = sec(x)tan(x) Inverse Function Theorem (f 1 0 ) (x) = 1 f 0 (f 1 (x)) ⇡ ⇡ , ] ) sin(✓) = x 2 [ 1, 1] 2 2 arccos(x) = ✓ 2 [0, ⇡] ) cos(✓) = x 2 [ 1, 1] 1 1 arcsin0 (x) = p arccos0 (x) = p 2 1 x 1 x2 ⇡ ⇡ 1 arctan(x) = ✓ 2 [ , ] ) tan(✓) = x 2 R arctan0 (x) = 2 2 1 + x2 arcsin(x) = ✓ 2 [ Author: Xiao Page 35 MAT137 10 DERIVATIVE Higher Order Derivatives f 0 (x) is the first order derivative function of f (x), and we may take the derivative of f 0 (x) a second time. 0 f 0 (x) = f 00 (x) = f (2) (x) f 00 (x) is the second order derivative function of f (x). Similarly, we may take even more derivatives on top of f 00 (x). The n-th order derivative function of f (x), f (n) (x) is the result of taking the derivative n times onto the original function f (x). Typically, there are no short-cuts or nice formulas for calculating n-th order derivatives, one must proceed by taking one derivative at a time. Position, Velocity, Acceleration f (t) = s(t) is the position of a particle at time t s0 (t) = v(t) is the velocity of a particle at time t Then the second derivative of position s00 (t) = v 0 (t) = a(t) is the acceleration of a particle at time t Question 18 Find the derivative. 1. f (x) = x2 + 1 x2 1 3 2. f (x) = x tan(3x + 1) r x+2 3. f (x) = x 1 Question 19 Find the n-th order derivative. 1. f (x) = ln(x) Author: Xiao 4. f (x) = ex ln(ln(x)) 5. f (x) = sin2 x + sin x2 + sin2 x2 6. f (x) = ex e 7. f (x) = xx 2. f (x) = p x+1 Page 36 MAT137 10 DERIVATIVE Question 20 1. Is there a continuous function on (0, 1) that takes on only 2 distinct values? 2. Is there a continuous function on (0, 1) that does not have a maximum? Question 21 Find conditions on A and B such that ( x3 f (x) = Ax + B x1 x>1 f (x) is di↵erentiable at x = 1. Question 22 Suppose f (0) = 1, f 0 (0) = 2, f (1) = 0, f 0 (1) = 1, f (2) = 1, f 0 (2) = 1 g(0) = 2, g 0 (0) = 1, g(1) = 1, g 0 (1) = 0, g(2) = 1, g 0 (2) = 1 h(0) = 1, h0 (0) = 2, h(1) = 2, h0 (1) = 1, h(2) = 0, h0 (2) = 2 Find 1. (g f )0 (1) 2. (g f )0 (2) 3. (f h g)0 (1) 4. (g f h)0 (2) Question 23 ⇡ Find the equation of the tangent line of f (x) = cot(x) at a = . 6 Author: Xiao Page 37 MAT137 10 DERIVATIVE Exponential and Logarithms We know the derivative of the natural exponential is (ex )0 = ex . What if the base is not e but some other number a > 0? We use the relation a = eln(a) ⇣ ⌘x ax = eln(a) = ex·ln(a) Thus any other exponential is just a horizontal stretch/shrink of ex by a factor of ln(a). (Also a horizontal reflection if ln(a) < 0 with a < 1.) By chain rule, and the above equality, (ax )0 = ex·ln(a) · ln(a) = ax · ln(a) 1 We know the derivative of the natural logarithm is (ln x)0 = . x What if the base is not e but some other number a > 0? We know the logarithm base a: loga (x) is the inverse function of ax . y = loga (x) , x = ay Using the relation a = eln(a) , y ⇣ x=a = e ln(a) ⌘y = ey·ln(a) Taking ln on both sides, and using the fact that ln and e are inverse functions, ⇣ ⌘ ln(x) = ln ey·ln(a) = y · ln(a) Dividing by the constant ln(a), y = loga (x) = ln(x) ln(a) Thus any other logarithm is just a verticle stretch/shrink of ln(x) by a factor 1 of . (Also a verticle reflection if ln(a) < 0 with a < 1.) ln(a) 1 1 Using (ln x)0 = , with the extra constant of in front, x ln(a) ⇣ ⌘0 1 loga (x) = x ln(a) Author: Xiao Page 38 MAT137 10 DERIVATIVE Logarithmic Di↵erentiation We know product rule as: (f g)0 = f 0 g + f g 0 What if there are more than 2 functions multiplied together? We could use the ’generalized product rule’ where: (f gh)0 = f 0 gh + f g 0 h + f gh0 However, this rule becomes complicated if there are more than 3 functions. There is an alternative method called logarithmic di↵erentiation, where we use the facts: x ln(xy) = ln(x) + ln(y) ln( ) = ln(x) ln(y) ln(xy ) = y ln(x) y 1. Given a product or quotient of functions, such as y(x) = f (x)2 g(x) h(x) Take ln of both sides to get ln y(x) = ln f (x)2 g(x) h(x) 2. Apply the above rules, we get ln y(x) = 2 ln f (x) + ln g(x) ln h(x) 3. Take derivative against x on both sides using chain rule, we get 1 1 1 · y 0 (x) = 2 · f 0 (x) + · g 0 (x) y(x) f (x) g(x) 1 · h0 (x) h(x) 4. Solve for y 0 (x) by multiplying y(x) to the right side, h 1 i 1 1 y 0 (x) = y(x) 2 · f 0 (x) + · g 0 (x) · h0 (x) f (x) g(x) h(x) Question 24 Find the derivative y 0 (x) using logarithmic di↵erentiation. p (x2 + 1)3 (2x 5)2 x2 + 3 p a) y(x) = b) y(x) = 3 (x2 + 2)2 x2 + 2x Author: Xiao Page 39 MAT137 11 11 LINEAR APPROXIMATION AND DIFFERENTIAL Linear Approximation and Di↵erential ⇣ ⌘ Recall the tangent line is a good approximation of y = f (x) near a, f (a) : y f (a) = f 0 (a)(x a) Thus we may say, for x ”near” a, (meaning (x a) should be small,) y = f (x) ⇡ f (a) + f 0 (a)(x a) This is called the linear approximation of f (x) near a. It is also called first order Taylor approximation of f (x) near a. Di↵erential We may rewrite the tangent line with a di↵erent notation. The di↵erence in y = y = f (x) The di↵erence in x = x=x f (a) a ( is the greek letter ”delta”, used to mean ”change in”.) Then the above equation reads: y ⇡ f 0 (a) · x If we let the di↵erence to ”approach zero”, the approximation becomes better. We say y becomes the di↵erential dy, x becomes the di↵erential dx: dy = f 0 (a)dx Now that if we ”switch notation” and write a instead of x, then dy = f 0 (x)dx dy = f 0 (x) dx dy Note that while both equations are similar in spirit, is NOT a fraction, dx and can not be treated like one. Question 25 p Approximate 3.99. Author: Xiao Page 40 MAT137 12 12 INVERSE FUNCTION Inverse Function f :A!B A is the domain of f , B is the codomain of f . f is one to one if: Any 2 di↵erent points in the domain will get sent (by f ) to 2 di↵erent points in the codomain. Equivalently, f is one to one if: There can not exist 2 di↵erent points that get sent (by f ) to the same point. The simplest logical statement for one to one is: f is one to one , If f (a) = f (b) then a = b If it seems that there are 2 di↵erent points, a and b, which when put into f , result in the same value: f(a) = f(b) , it actually means the 2 points a and b are the same point to begin with: a = b Example f (x) = x+3. Suppose f (a) = f (b), a + 3 = b + 3, so a = b ) f is one to one. The image is all possible values that f can take. image(f ) = {f (x) | x 2 A} f is onto if: image = codomain. Thus, we can easily make f onto, by simply defining B to be image(f ). Example Consider the (one to one) function, f :R!R f (x) = arctan(x) Now at first, we can consider A = R (on the left), and B = R (on the right). However, as x can be any real number in A = R, arctan(x) can take on ⇡ ⇡ every value in ( , ). (We can see this with the graph). 2 2 ⇡ ⇡ Thus, image(f ) = ( , ). 2 2 Now if we set ⇡ ⇡ f :R!( , ) 2 2 Then f is one to one and onto. Author: Xiao Page 41 MAT137 12 INVERSE FUNCTION Horizontal Line Test Given a function f : f is one-to-one if and only if every possible horizontal line in the xy-plane intersect the curve at most once. A function f is one to one and onto, f :A!B 1 if and only if there exists an inverse function f 1 f , :B!A The function f takes a 2 A as input, and gives b 2 B as output. f (a) = b The function f 1 takes the b 2 B as input (which is the output of f ), and gives back a 2 A, as the output of f 1 . f 1 Thus the domain of f The inverse of f 1 1 (b) = a is B, and the image of f is f , written as (f 1 ) 1 1 is A. = f. This gives a one to one correspondence between the points in A and B. Any point in A is sent into B using f , and f 1 sends the result back into A. f 1 f f f (a) = a 1 (b) = b for all a 2 A for all b 2 B In the previous example, f (x) = arctan(x), so f 1 (y) = tan(y), where ⇡ ⇡ f :R!( , ) 2 2 ⇡ ⇡ f 1:( , )!R 2 2 Although f 1 (y) is defined by a formula to be tan(y), it has a di↵erent domain than the standard tan(y) function. In particular, f (x) = arctan(x) and f 1 (y) = tan(y) are only inverses to each other in the domain and image as defined above. ⇡ ⇡ For y 62 ( , ), we have arctan(tan(y)) 6= y. 2 2 Author: Xiao Page 42 MAT137 12 INVERSE FUNCTION Inverse Trignometric Functions We know that sin(✓), cos(✓), tan(✓) are periodic with period 2⇡. (We can even say tan(✓) has period ⇡.) Consider f (✓) = sin(✓) for example. Let f :A=R!B=R We have image(f ) = [ 1, 1], so we set f : A = R ! B = [ 1, 1] Then f is onto. However, there are too many elements in the domain A of f , which get sent to the same value in the codomain B of f . For example, we have f (0) = 0 2 B, and f (2⇡) = 0 2 B, and f (4⇡) = 0 2 B. This is called not one-to-one. In order to make f one-to-one, we can restrict the domain, by removing points in the domain that cause the above problem: f (0) = f (2⇡) = f (4⇡). We (usually) choose an interval as the new domain A of f , to attain one-to-one. f :A!B The most common way for f (✓) = sin(✓) is to choose A = [ Then f is one to one and onto, and f 1 (x) = arcsin(x), ⇡ ⇡ , ]. 2 2 ⇡ ⇡ , ] ! [ 1, 1] 2 2 ⇡ ⇡ f 1 (x) = arcsin(x) : [ 1, 1] ! [ , ] 2 2 f (✓) = sin(✓) : [ Similarly, f (✓) = cos(✓) : [0, ⇡] ! [ 1, 1] 1 (x) = arccos(x) : [ 1, 1] ! [0, ⇡] ⇡ ⇡ f (✓) = tan(✓) : ( , )!R 2 2 ⇡ ⇡ f 1 (x) = arctan(x) : R ! ( , ) 2 2 f Author: Xiao Page 43 MAT137 12 INVERSE FUNCTION Question 26 a) Find the following values. 2⇡ 2⇡ 2⇡ x = cos( ), y = sin( ), z = tan( ). 3 3 3 b) Verify whether the following quantities exists, and evaluate them. arccos(x), arcsin(y), arctan(z). c) Verify whether the following equations hold. cos(arccos(x)) = x, sin(arcsin(y)) = y, tan(arctan(z)) = z. d) Conclude that cos(arccos(x)) = x for all x in the domain of arccos, which is [ 1, 1]. Similarly, sin(arcsin(x)) = x tan(arctan(x)) = x for all x in the domain of arcsin (which is [ 1, 1]), and arctan (which is R). e) Conclude that arccos(cos(✓)) = ✓ for ✓ 2 [0, ⇡], the restricted domain of cos(✓). Similarly, arcsin(sin(✓)) = ✓ ⇡ ⇡ for ✓ 2 [ , ], the restricted domain of sin(✓). 2 2 arctan(tan(✓)) = ✓ ⇡ ⇡ for ✓ 2 ( , ), the restricted domain of tan(✓). 2 2 5⇡ ⇡ 7⇡ f) Evaluate arcsin(sin( )), arccos(cos( )), arccos(cos( )). 6 3 6 Recall the sum of angles formula: sin(x + y) = sin(x)cos(y) + cos(x)sin(y) cos(x + y) = cos(x)cos(y) p sin(x)sin(y) p 3 ), arccos( 3), 1 g) Evaluate: arccos(1), arccos( ), arccos( 2 p2 p 1 3 arcsin(0), arcsin(1), arcsin( ), arcsin( ), arctan(1), arctan( 3) 2 2 Author: Xiao Page 44 MAT137 12 INVERSE FUNCTION Question 27 Sketch the graph of the following functions. a) f (x) = sin(arcsin(x)), where x is a ratio. b) f (x) = arcsin(sin(x)), where x = ✓ is an angle. c) f (x) = tan(arctan(x)), where x is a ratio. d) f (x) = arctan(tan(x)), where x = ✓ is an angle. Question 28 Simplify the following expressions which are given as a composition of trignometric function and inverse trignometric function. ⇡ 3⇡ a) arccos(sin(✓)), where ✓ 2 [ , ] is an angle. 2 2 ⇡ ⇡ Repeat with ✓ 2 [ , ]. 2 2 b) sec(arccos(x)), where x is a positive ratio. c) sin(arccos(x)), where x is a positive ratio. d) sin(arctan(x)), where x is a positive ratio. e) tan(arcsin(x)), where x is a positive ratio. Author: Xiao Page 45 MAT137 12 INVERSE FUNCTION Solve for inverse function f 1 1. Set y = f (x) 2. Isolate x on one side: x = g(y), where g(y) is the right side with only y. 1 3. Write f (y) = g(y), then switch all y to x, so f 1 (x) = g(x). Note: The last step of ’switching y to x’ is optional. The x after the switch is not the same x as we started out with. This is used when we would like to write f 1 (x), with x being the variable. Inverse Function Theorem (f 1 0 ) (x) = 1 f 0 (f 1 (x)) Question 29 f (x) = ( x3 x2 1 x<0 x 0 a) Sketch the graph of f and verify using the graph that f is one to one. b) Find f 1 , and then find (f 1 )0 (4) and (f 1 )0 ( 2). Question 30 Find f 1 (x). 1. f (x) = x + 3 p 2. f (x) = 2 3x 3. f (x) = 2x + 1 x 1 4. f (x) = 5. f (x) = 1+e 1 e ex e x x x 2 Question 31 1 2 Evaluate ln(e2x ), ln(2ex ), e2ln(x) , ln( 3 ), eln(ln(e )) . e Question 32 Solve for x. 1. e2 2. e2x 3x =6 4. ln(x) + ln(x 1) = 1 3ex + 2 = 0 3. ln(ln(x)) = 1 Author: Xiao 5. 1 < e2x+1 < 2 Page 46 MAT137 13 13 IMPLICIT DIFFERENTIATION Implicit Di↵erentiation 1. Replace all y with y(x) 2. Take derivative against x on both sides. 3. Isolate y 0 (x) = dy dx Remember to use product rule and chain rule, when taking derivatives. Example: cos(xy) = 1 + sin(y) 1. cos(x · y(x)) = 1 + sin(y(x)) 2. d d cos(x · y(x)) = [1 + sin(y(x))] dx dx 3. Using chain rule: d dy [x · y(x)] = 0 + cos(y(x)) · dx dx Using produict rule on the left side: dy dy sin(x · y(x)) · [1 · y(x) + x · ] = cos(y(x)) · dx dx dy Isolate : dx dy dy sin(x · y(x)) · y(x) x · sin(x · y(x)) · cos(y(x)) · =0 dx dx dy sin(x · y(x)) · y(x) + [ x · sin(x · y(x)) cos(y(x))] · =0 dx dy [ x · sin(x · y(x)) cos(y(x))] · = sin(x · y(x)) · y(x) dx dy sin(x · y(x)) · y(x) = dx x · sin(x · y(x)) cos(y(x)) sin(x · y(x)) · Question 33 dy d2 y Find and 2 of dx dx x2 + 4xy + y 3 + 5 = 0 at (x, y) = (2, 1). Author: Xiao Page 47 MAT137 14 14 RELATED RATES PROBLEM Related Rates Problem 1. Assign symbols to all variables in the problem. 2. Write an equation that relate all the variables. 3. Identify which variables change with time. If some quantity x changes as time passes, replace x with x(t). 4. Take derivative against t on both sides using implicit di↵erentiation. 5. Plug in the values given for the variables in the problem. Example: The volume of a cone is increasing at 30 m3 /min. The diameter and the height are always equal. How fast is the height increasing when it is 10 meters? 1. v = volume, r = radius, h = height 2. 1 v = ⇡r2 · h 3 Using the assumption, diameter = 2r = h ! r = h 2 1 h 1 v = ⇡( )2 · h = ⇡ · h3 3 2 12 3. As time passes, both volume and height will change. v(t) = 4. d d 1 v(t) = [ ⇡ · (h(t))3 ] dt dt 12 1 d dv = ⇡ · (h(t))3 dt 12 dt dv 1 dh = ⇡ · 3 · (h(t))2 · dt 12 dt 5. 30 = Author: Xiao 1 ⇡ · (h(t))3 12 1 dh dh ⇡ · 3 · (10)2 · = 25⇡ · 12 dt dt ) dh 6 = dt 5⇡ Page 48 MAT137 14 RELATED RATES PROBLEM Question 34 The radius of sphere is increasing at the rate of 4 cm/s. How fast is the volume increasing when the diameter is 80 cm? Question 35 Two cars start at the same point. One travels west at 25 km/h, the other travels south at 60 km/h. What is the rate of the distance between them changing after 2 hours? Question 36 A policeman is at 200m from a straight highway. Down the highway, 200m from the point on the highway closest to the policeman, is an emergency call box. The officer points the radar gun at the call box. A car passes the call box, and the radar indicates the distance between the policeman and the car is increasing at 72km/h. The speed limit is 90km/h. Is the car above the speed limit? Question 37 The water level in a hemispherical tank of radius 2 m is dropping at a rate of 5 cm/sec when the depth of the water in the tank is 1 m. At what rate is the radius of surface area of the water decreasing, when the depth of the water in the tank is 1 m? Question 38 A camera 3000 m from the launch pad is to be always aimed at a rocket as it rises. How fast is the angle of elevation of the camera changing at the instant the rocket is 4000 m above the launch pad, and rising at a rate of 880 m per sec? Question 39 A 2-m tall man walks along the side of an 8-m wide street, on which the only lamppost is on the other side of the street. The man walks at a rate of 2 m/sec. The lamppost is 10 m tall. When he is 6 m from a point directly opposite the lamppost, how fast is the length of his shadow changing? Author: Xiao Page 49 MAT137 15 15 MEAN VALUE THEOREM Mean Value Theorem Mean Value Theorem If f is continuous on [a, b], di↵erentiable on (a, b), then there exists c between (a, b) so that f 0 (c) = f (b) b f (a) a Corollary: Rolle’s Theorem If furthermore, f (b) = f (a), then there exists c between (a, b) so that f 0 (c) = 0 Increasing and Decreasing Function f (x) is increasing on an interval [a, b], if for any x1 < x2 , we have f (x1 ) < f (x2 ). f (x) is decreasing on an interval [a, b], if for any x1 < x2 , we have f (x1 ) > f (x2 ). Results from Mean Value Theorem If f 0 (x) = 0 for all x in an interval, then f (x) is a constant function. i.e. There exists some constant c such that f (x) = c for all x in the interval. (If f 0 (x) = g 0 (x), then f (x) g(x) = c is constant.) If f 0 (x) 6= 0 for all x in an interval, then f (x) is one-to-one. If f 0 (x) > 0 for all x in an interval, then f (x) is increasing. If f 0 (x) < 0 for all x in an interval, then f (x) is decreasing. Local Extrema A point c is a local maximum if there exists an interval (c l, c + l) such that for all x in the interval, f (x) f (c) (when f (x) is defined). A point c is a local minimum if there exists an interval (c l, c + l) such that for all x in the interval, f (x) f (c) (when f (x) is defined). If c is a local extrema in the interior of an interval [a, b], then c is a critical point: f 0 (c) = 0 or f 0 (c) d.n.e. First Derivative Test for Local Extrema At a critical point c ( f 0 (c) = 0 or f 0 (c) d.n.e.) : f 0 (x) < 0 on the left and f 0 (x) > 0 on the right ) c is a local minimum. f 0 (x) > 0 on the left and f 0 (x) < 0 on the right ) c is a local maximum. Author: Xiao Page 50 MAT137 15 MEAN VALUE THEOREM Question 40 Prove that f (x) = arcsin p 1 x + 2arctan x 1+x is constant on some interval. What is the constant? What’s the largest interval where this is true? Question 41 Find all functions defined on R that has f 0 (x) = 3. Furthermore, show that no other function (other than those that you have found) can satisfy f 0 (x) = 3. Question 42 a) Suppose f 0 (x) has only 1 zero on [a, b]. How many zeroes can f have at most on [a, b]? b) Suppose f 0 (x) has only 2 zero on [a, b]. How many zeroes can f have at most on [a, b]? c) Suppose f 0 (x) has only k zero on [a, b]. How many zeroes can f have at most on [a, b]? d) Suppose f 00 (x) has only 1 zero on [a, b]. How many zeroes can f have at most on [a, b]? e) Suppose f (n) (x) has only 1 zero on [a, b]. How many zeroes can f have at most on [a, b]? Question 43 Find the critical points, and classify them as local extremas. 1. f (x) = x4 2. f (x) = |x| ( x2 + 1 x < 1 3. f (x) = x+1 x 1 Author: Xiao Page 51 MAT137 15 MEAN VALUE THEOREM Question 44 Sketch the graph of f (x) = ( 2x + 2 x x3 x x > 1 1 Find f 0 (x). Determine whether f satisfies the conditions of the mean value theorem on [ 3, 2]. Find the value c from the conclusion of the mean value theorem. Question 45 Suppose f 0 (x) M for all x, where M is a constant. let a < b. Show the rate of increase of f (x) is bounded by M : f (b) f (a) + M (b Author: Xiao a) Page 52 MAT137 16 16 MAXIMUM AND MINIMUM Maximum and Minimum Extreme Value Theorem (EVT) If f (x) is continuous on [a, b], then f (x) has (absolute) maximum and minimum on the interval [a, b]: 1. There exists xM , with f (xM ) f (x) for all x 2 [a, b], as a maximum. 2. There exists xm , with f (xm ) f (x) for all x 2 [a, b], as a minimum. To find the maximum and minimum of f (x) on [a, b] when f is di↵erentiable, we can use the second derivative test: 1. Compute f 0 (x). Set f 0 (x) = 0. The values of x that satisfy this equation are called critical points. (Technically, if f 0 (x) doesn’t exist, then x is also a critical point.) 2. Compute f 00 (x). For each critical point a, If f 00 (a) > 0, a is a local minimum. If f 00 (a) < 0, a is a local maximum. If f 00 (a) = 0, then it is unclear. (We can also use the first derivative test for local max/min.) 3. Compare all values of f (a) for all critical points and endpoints of interval. The largest value is absolute (global) maximum. The smallest value is absolute (global) minimum. We can carefully extend the above results to f (x) with domain being R. 1. For lim f (x) = +1 = lim f (x), then there exists absolute minimum. x!1 2. For lim f (x) = x!1 x! 1 1 = lim f (x), then there exists absolute maximum. x! 1 3. For lim f (x) = L = lim f (x), where L 6= ±1: x!1 x! 1 If there exists x1 with f (x1 ) > L, then there exists absolute maximum. If there exists x2 with f (x2 ) < L, then there exists absolute minimum. 4. For lim f (x) = +1, lim f (x) = x!1 x! 1 1 (infinities of opposite signs), there are no absolute maximum and no absolute minimum. Note: it is important to assume f (x) has domain being R, that is to say f (x) can not be approaching infinity anywhere on R. Author: Xiao Page 53 MAT137 16 MAXIMUM AND MINIMUM Example: f (x) = 1. x2 4 x2 + 4 2x · (x2 + 4) 2x · (x2 4) 2x3 + 8x 2x3 + 8x = (x2 + 4)2 (x2 + 4)2 16x = 2 =0 (x + 4)2 The denominator is always positive, so the numerator is 0. The only critical point is x = 0 f 0 (x) = 2. f 00 (x) = 16 · (x2 + 4)2 16x · 2(x2 + 4) · 2x (x2 + 4)4 16 · (x2 + 4) 64x2 64 48x2 = (x2 + 4)3 (x2 + 4)3 At the critical point x = 0, = f 00 (0) = 64 >0 43 Thus x = 0 is a local minimum. 3. Compare the ”endpoints”, in this case: lim f (x) = 1 x!±1 We do have f (0) = 1 < 1. However, notice that f (x) < 1 since the numerator is always smaller. So, there are no absolute maximum, but there exists absolute minimum. Thus x = 0 is the absolute minimum. Question 46 Find all critical points, local extrema, global extrama, interval of increase and decrease. a) f (x) = (x 1)2 (x 2)2 on the interval [0, 3] b) f (x) = ( |x2 1| on R 2 2 2x x 2x0 c) f (x) = |x 2| 0<x3 Author: Xiao Page 54 MAT137 17 17 OPTIMIZATION Optimization 1. Identify the quantity that needs to be maximized/minimized. Set as y. 2. Identify the quantity that can be changed. Set as x. 3. Write an equation that relate all the variables. Draw a diagram if needed. 4. Calculate dy , and set to 0. dx 5. Solve for x, and the quantities needed. Examples: There are 1200cm2 of material to construct a rectangular container with square base and with open top. What is the largest volume? 1. Set v = volume 2. Set b = base, h = height 3. v = b2 · h and 1200 = b2 + 4hb Solving for h: 1200 b2 h= 4b 2 4 1200b b v= = 300b 4b 4. 5. dv = 300 db b3 4 3b2 =0 4 1200 = 400 ) b = 20 3 1200 400 h= = 10 80 v = b2 · h = 400 · 10 = 4000 b2 = Author: Xiao Page 55 MAT137 17 OPTIMIZATION Question 47 How much fencing is needed to define two adjacent rectangular playgrounds of the same width and total area 15, 000m2 ? Question 48 Find the point(s) on the parabola y = x2 closest to the point (3, 0). Question 49 You want to sell a square open box: the bottom is a square, the four sides are identical rectangles, and there isn’t anything on the top. The box should have a volume of 1000 cm3 . The material for the bottom costs $2 per cm2 and the material for each of the four sides costs $3 per cm2 . You know you will be able to sell each box for $1,250. Is this operation profitable? Question 50 A farmer wants to hire workers to pick 1600 bags of beans. Each worker can pick 10 bags per hour and is paid $1.00 per bag. The farmer must also pay a supervisor $20 per hour while the picking is in progress. She has additional miscellaneous expenses of $8 per worker (but not for the supersivor). How many workers should she hire to minimize the total cost? What will the cost per bag picked be? Question 51 A power line is needed to connect a power station on the shore of a river to an island 4 kilometers downstream and 1 kilometer o↵shore. Find the minimum cost for such a line given that it costs $50,000 per kilometer to lay wire under water and $30,000 per kilometer to lay wire under ground. Question 52 A bus company runs a bus at a fare of $37 per person if 16 to 35 passengers sign up for the trip. The company does not charter trips for fewer than 16 passengers. The bus has 48 seats. If more than 35 passengers sign up, then the fare for every passenger is reduced by 50 cents for each passenger in excess of 35 that signs up. Determine the number of passengers that generates the greatest revenue for the bus company. Author: Xiao Page 56 MAT137 18 18 CONCAVITY AND POINT OF INFLECTION Concavity and Point of Inflection When f 0 (x) > 0, f (x) is increasing at x. When f 0 (x) < 0, f (x) is decreasing at x. When f 00 (x) > 0, f (x) is concave up at x. When f 00 (x) < 0, f (x) is concave down at x. Concave up looks like a bowl right-side up sitting on the table. Concave down looks like a bowl that’s put upside down. A point x is a point of inflection if f 00 (x) = 0 and f 00 changes sign at x. (If f 00 still changes sign at x, x is a point of inflection even if f 00 (x) d.n.e.) Solve f 00 (x) = 0, and check whether f 00 (x) changes sign. Example: x2 4 x2 + 4 16x f 0 (x) = 2 (x + 4)2 The denominator of f 0 (x) is always positive, so the sign of f 0 (x) depends only on the numerator. Thus f (x) is increasing for x > 0, and decreasing for x < 0. f (x) = f 00 (x) = 64 48x2 =0 (x2 + 4)3 The denominator of f 00 (x) is always positive, so set the numerator is 0. 64 4 2 = ! x = ±p 64 48x2 = 0 ! x2 = 48 3 3 00 Note that f (x) has no verticle asymptotes, so f (x) always exists. We need 2 to check the 3 regions set out by x = ± p . 3 p p 00 We have f (0) > 0, so f (x) is concave up between 2/ 3 and -2/ 3. As x approaches +1 and 1, 64 48xp2 approaches p1. Thus f (x) is concave down for x < 2/ 3 and x > 2/ 3 p So x = 2/ 3 and x = Author: Xiao p 2/ 3 are points of inflection. Page 57 MAT137 19 19 GRAPHING Graphing We can try to accurately sketch a graph by finding the following: 1. Domain 2. Verticle asymptotes: lim f (x) = ±1 x!a (typically happens for x not in the domain) 3. Horizontal asymptotes: lim f (x) and lim f (x) x!1 x! 1 Possible slant asymptote for rational functions: am xm + ... + a0 f (x) = bn xn + ... + b0 m=n+1 lim f (x) = ±1, but f (x) has a slant asymptote attained by long division. x!±1 f (x) = kx + b + r(x) where kx + b is the slant asymptote. q(x) Find intersections to horizontal/slant asymptotes if possible. 4. Points of intersection with axis: Set x = 0, and y = 0 5. Take derivative: Intervals of increase/decrease, critical points f 0 (a) = 0. 6. Take 2nd derivative: Intervals of concavity, points of inflection. Classify critical points as maximum, minimum, or neither. 7. Other properties: order of zeroes (for rational functions), discontinuity, endpoint behaviour (for intervals), verticle tangent lines, corners... Question 53 Sketch the graph. 1. f (x) = x2 (x + 2)3 2. f (x) = x2 x2 1 3 3. f (x) = Author: Xiao x +1 x2 4. f (x) = ex ex + 1 5. f (x) = x1/3 p 6. f (x) = x 1 7. f (x) = |2x x2 x2 | Page 58 MAT137 20 20 L’HOPITAL’S RULE L’Hopital’s Rule When lim f (x) 0 = lim g(x) 0 lim OR ± f (x) f 0 (x) = lim 0 g(x) g (x) 1 1 Usually, we plug in the value of x and get 00 , so we take derivative and plug in x again. Example: x3 0 lim 2 = x!0 x + x 0 Taking derivative of the numerator and denominator, 3x2 0 = =0 x!0 2x + 1 1 = lim Case of 0 · 1 lim[f (x) · g(x)] = 0 · 1 We can do one of two things: lim[f (x) · g(x)] = lim lim[f (x) · g(x)] = lim f (x) 0 = 1 0 ( g(x) ) g(x) 1 = 1 1 ( f (x) ) Typically, the simpler fraction out of the two should be used to get the answer. Example: 1 sin(1/x) 0 lim x · sin( ) = lim = 1 x!1 x!1 x 0 (x) = lim x!1 Author: Xiao cos(1/x) · 1 x2 1 x2 1 = lim cos( ) = cos(0) = 1 x!1 x Page 59 MAT137 20 L’HOPITAL’S RULE Case of 01 , 10 , 11 , ... When the base of an exponent includes x, one must utilize the formula: f (x)g(x) = (eln(f (x)) )g(x) = eg(x)·ln(f (x)) Example: lim (tan(2x))x = lim+ ex·ln(tan(2x)) x!0+ x!0 Using property of limits, we can move the limit onto the exponent. h i lim x · ln(tan(2x)) = e x!0+ So we only need to compute lim+ x · ln(tan(2x)) = lim+ x!0 = lim+ ln(tan(2x)) x!0 1 tan(2x) · sec2 (2x) · 2 1 x2 x!0 = lim+ x!0 1 x cos(2x) 1 sin(2x) cos2 (2x) 1 x2 = ·2 ln(0) 1 = 1 1 = lim+ x!0 2x2 sin(2x)cos(2x) The limit of cos(2x) is 1, so we can split up the limit to simplify calculation. = lim+ x!0 2x2 1 2x2 0 · lim+ = lim+ ·1= sin(2x) x!0 cos(2x) x!0 sin(2x) 0 Using L’hopital: 4x 0 = =0 x!0 cos(2x) · 2 2 Thus the limit in the exponent is 0. = lim+ lim ex·ln(tan(2x)) = elim x · ln(tan(2x)) = e0 = 1 x!0+ L’hopital rule should not be used when the limit can be solved by other techniques, as derivative of function typically is more complicated than the original function. Question 54 An important limit - Compount interest (with 100% interest per year): lim n!1 Author: Xiao ⇣ 1+ 1 ⌘n =e n Page 60 MAT137 20 L’HOPITAL’S RULE Question 55 tan(3t) t!0 ln(1 + 2t) 1. lim h3 h!2 h3 5h2 + 3h + 6 h2 3h + 2 2. lim 1 cos(3t) t!0 t · ln(1 + t) 3. lim 4. lim (x7 5x3 + 2)e x!1 5. lim (x x!1 6. lim p x!0 x 1)sin(x) ex cos(x) x+1 x 1 3 7. lim x · tan( ) x!1 x 1 1 ) x!0 xsin(x) xtan(x) p x4 + 5x3 x2 9. lim x!1 x 8. lim ( sin(x) x!1 x 10. lim 11. lim (cos(x))1/x 2 x!0 ln(2) 12. lim x 1 + ln(x) x!1 13. lim+ x p x x!0 14. lim (2 x!1 Author: Xiao x)tan(⇡x/2) Page 61 MAT137 21 21 IDEA OF INTEGRATION Idea of Integration Let’s first assume f (x) > 0. We want the area under the curve of f (x) on [a, b] (and above the x-axis). We split the area into many rectangles, and add up their total area. Area of Rectangle = base · height P P The total area of the rectangles is base · height = bi · h i where i runs over all rectangles in the partition P . i The way we pick the bases of the rectangles on [a, b] is called partition. P : t0 < t1 < ... < tn t0 = a tn = b Once we have a partition P , we can split the rectangles into even smaller rectangles (by adding some additional points in the list of ti ). This is called a refinement of P . But, how do we choose the height of the rectangles? There are many methods to choose the height of the rectangles. The following 2 methods for choosing height are important for theoretical purposes: For Upper Sum, pick the height of a rectangle to be the maximum value (actually the supremum) of f in that rectangle. The total area of the rectangles computed with this method is called Uf (P ). For Lower Sum, pick the minimum value (the infimum) of f instead. The total area of the rectangles computed with this method is called Lf (P ). However, finding maximum (supremum) and minimum (infimum) of the functions in many small rectangles are typically very difficult. The following 2 methods are easy for computation: For Right-point method, pick the height of a rectangle to be the function value of the right most endpoint of the rectangle. For Left-point method, pick the height of a rectangle to be the function value of the left most endpoint of the rectangle. Author: Xiao Page 62 MAT137 21 IDEA OF INTEGRATION Given a partition of [a, b], where t0 = a and tn = b P : t0 < t1 < ... < tn We may compute the total area explicitly. In the i-th rectangle with base [ti 1 , ti ], the right most endpoint is ti and the left most endpoint is ti 1 . The total area computed by the Right-point method is: Total area = n X i=1 bi · h i = n X (ti i=1 ti 1 ) · f (ti ) The total area computed by the Left-point method is: Total area = n X i=1 bi · h i = n X (ti i=1 ti 1 ) · f (ti 1 ) Notice that the formula for the Right-point method is slightly easier. If we furthermore assume that we split partition the interval [a, b] into n equal length intervals, then we attain a formula for ti : b a Each rectangle would have base length bi = . n b a Since t0 = a, we must have t1 = a + bi = a + . n b a Then we must have t2 = t1 + bi = a + 2 . n Thus the pattern continues and we attain: ti = a + i b a n The formula for the total area with Right-point method becomes: n n X X b a ⇣ b a⌘ Total area = bi · h i = ·f a+i n n i=1 i=1 The formula for the total area with Left-point method becomes: n n X X b a ⇣ b a⌘ Total area = bi · h i = · f a + (i 1) n n i=1 i=1 Author: Xiao Page 63 MAT137 21 IDEA OF INTEGRATION Intuitively, as we use more rectangles, the total area of the rectangles would become a better approximation to the true area under the curve of f (x). Thus an informal definition of the integral is defined to be the true area under the curve of f (x), approximated using more and more rectangles: True Area = Z b f (x)dx = lim a n!1 n X i=1 bi · h i where hi corresponds to some (potentially unknown/undeterministic) way of choosing the height of the rectangles. However, the only requirement is that hi needs to be the function value of a point xi in the i-th rectangle. That is to say hi must equal to f (xi ) for some xi 2 [ti 1 , ti ]. These types of sum (where we don’t specify the exact way of choosing base and height) is called a Riemann Sum. We may intuitively think that as n becomes large, bi becomes very narrow, so it is centered around a point x. Then we can change notation bi ! dx. If dx is narrow enough, the function values in the small rectangle are almost identical, so we may set hi ! f (x). Then informally: The ’sum’ becomes the ’integral’ sign. This definition extends to the case where f (x) may also be negative. In this case, we may still use the Riemann Sum formula, but we would be approximating the ’Net area’: Net area = Z b f (x)dx = lim a n!1 n X i=1 bi · h i where 1. When f (x) > 0, we count the area above the x-axis and below f (x) to be ’positive area’. 2. When f (x) < 0, we count the area below the x-axis and above f (x) to be ’negative area’. The ’Net area’ is the sum of both the ’positive’ and ’negative’ areas. Author: Xiao Page 64 MAT137 21 IDEA OF INTEGRATION When computing the total area of the rectangles using Riemman Sums when f (x) are simple polynomials, the following formulas are helpful: n X i=1 n X n(n + 1) i= 2 n X i=1 c · s(i) = c · i=1 n X i=1 n(n + 1)(2n + 1) i = 6 s(i) 2 n X i=1 ⇥ n X i3 = i=1 ⇤ s(i) + t(i) = n X i=1 n2 (n + 1)2 4 s(i) + n X t(i) i=1 where s(i), t(i) are expressions involving i, and c is a constant not involving i. Question 56 Compute the following integral by using the right-point method with the partition being n equal length intervals. Z 1 Z 2 1. 2x dx 2. 2x2 x dx 0 1 Question 57 Let f (x) = x2 2 and x 2 [0, 3] be partitioned into 3 equal length intervals. Compute the Riemann Sum explicitly, with the midpoint method, where we choose the height to be the function value of the midpoint in each interval. Distance and Displacement The velocity v can be positive or negative. The speed |v| is the absolute value of velocity and is always positive. If speed is constant, then total distance travelled = speed · time. If speed is not constant, we can split the time interval [a, b] into many small time intervals, each with speed |v(t)| changing with time. We use the integral to add up the distance travelled in each small time interval to attain the total distance travelled, Z b Total distance travelled with t 2 [a, b] = |v(t)|dt a If we integrate v, then we are also considering the interval when v is negative. When v > 0, we are moving forward. When v < 0, we are moving backward. The displacement is the change in position, Z b Total displacement with t 2 [a, b] = v(t)dt a Author: Xiao Page 65 MAT137 22 22 1D INTEGRATION 1D Integration We want the area under the curve of f (x) on [a, b]. We split the area into many rectangles, and add up their total area. The way we pick the base of the rectangles on [a, b] is called partition. Once we have a partition P , we can split the rectangles into more smaller rectangles. This is called a refinement of P . For theoretical purposes, the following 2 methods of choosing height are important: For Upper Sum, pick the height of a rectangle to be the maximum value (actually the supremum) of f in each rectangle. The total area of the rectangles computed with this method is called Uf (P ). For Lower Sum, pick the minimum value (the infimum) of f instead. The total area of the rectangles computed with this method is called Lf (P ). Area of Rectangle = base · height P P The total area of the rectangles is base · height = bi · h i i where the index i runs over all rectangles in the partition P. The total area of the rectangles using Upper Sum will always be larger than (or equal to) the true area. Similarly, the total area using Lower Sum will always be smaller (or equal) to the true area. Thus, the true area is always sandwiched between Upper Sum and Lower Sum. In fact, even if 2 di↵erent partitions are used, this is still true: Result 1: For any partitions P1 and P2 , Lf (P1 ) ”T rue Area” Uf (P2 ) By picking more rectangles, we can make the upper sum and lower sum closer and closer together: Result 2: If P2 is a refinement of P1 (P2 has all the rectangles of P1 and some more), then Lf (P1 ) Lf (P2 ) ”T rue Area” Uf (P2 ) Uf (P1 ) Author: Xiao Page 66 MAT137 22 1D INTEGRATION Definition 1: A function is integrable if the Upper Sum and Lower Sum can made as close as you want (by picking more rectangles): 8✏ > 0, 9 P such that Uf (P ) Lf (P ) < ✏ Notice that the same partition P is used here in upper and lower sum. What is the ”True Area”? If we really can squeeze the upper sum and lower sum together, then there is one intuitive answer. The ”True Area” must be the ”minimum” of all possible upper sum, where we try every possible way of picking the bases, and calculate the area using the maximum (supremum) value. It should also be equal to the ”maximum” of all possible lower sum. To get around the fact that minimum or maximum might not exist: Definition 2: R b ”True Area” = a f = infimum of all possible {Uf (P )} = I ba = supremum of all possible {Lf (P )} = I ba Supremum and Infimum A number u is the supremum of a set S if it is the least upper bound of S. i.e. u is the smallest number which can satisfy: For all s 2 S, s u. (u is larger than everything in S) A number l is the infimum of a set S if it is the greatest lower bound of S. i.e. l is the largest number which can satisfy: For all s 2 S, l s. (l is smaller than everything in S) The supremum is intuitively understood as the maximum of the set S. The infimum is intuitively understood as the minimum of the set S. Result 3: ✏ characterization of supremum and infimum For all ✏ > 0, there exists s 2 S such that u ✏ < s u. (As u is the ’maximum’ of S, moving down even a tiny amount will bump into an element of S.) For all ✏ > 0, there exists s 2 S such that l s < l + ✏. (As l is the ’minimum’ of S, moving up even a tiny amount will bump into an element of S.) Author: Xiao Page 67 MAT137 22 1D INTEGRATION Question 58 Show whether the following functions f are integrable. 1. f (x) = 2 for x 2 [0, 2]. ( 2 x 6= 1 2. f (x) = for x 2 [0, 2]. 4 x=1 ( 2 x<1 3. f (x) = for x 2 [0, 2]. 4 x 1 8 > <2 x < 1 4. f (x) = 4 1 x < 2 for x 2 [0, 2]. > : 6 x=2 5. f (x) = x for x 2 [0, 2]. ( x x<1 6. f (x) = for x 2 [0, 2]. x+2 x 1 7. Suppose(g(x) is integrable and bounded (by M) on [a, b]. g(x) x 6= c f (x) = where f (c) 6= g(c). f (c) x = c 8. Suppose g(x) is integrable and bounded (by M) on [a, b]. Let f (x) = g(x) except at finitely many points {c1 , ..., cn } where f (ci ) 6= g(ci ). ( 1 x2Q 9. f (x) = for x 2 [0, 1]. 0 x 62 Q 10. Bonus:8 <1 f (x) = q :0 Author: Xiao p written as an irreducible fraction q for x 2 [0, 1]. x 62 Q, x = 0, x = 1 x= Page 68 MAT137 22 1D INTEGRATION Question 59 Suppose f is increasing on [a, b]. Prove f is integrable. (Note that we do not assume f is continuous.) Now, suppose f is decreasing on [a, b]. Prove f is integrable. A function f which is only increasing or only decreasing is called monotonic. Question 60 R Rb b Let a f (x)dx < a g(x)dx. Let P be a partition of [a, b]. Are the following statements true or false? Justify your answers. (Prove the statement is true, or give a counterexample if it is false.) 1. Lf (P ) < Ug (P ) 2. Lf (P ) < Lg (P ) Rb 3. Lf (P ) < a f (x)dx 4. Uf (P ) < Ug (P ) Rb 5. a f (x)dx < Ug (P ) Rb 6. Uf (P ) < a g(x)dx Question 61 Prove inf f (x) · (b x2[a,b] a) Z b a f (x) sup f (x) · (b a) x2[a,b] Question 62 Suppose f is integrable, Prove |f | is integrable. Is it true that: If |f | is integrable, then f is integrable? Question 63 Rb Suppose f is continuous, f 0, a f = 0. Prove f (x) = 0 for all x 2 [a, b]. Is this still true if we do not assume f is continuous? Author: Xiao Page 69 MAT137 23 23 FUNDAMENTAL THEOREM OF CALCULUS Fundamental Theorem of Calculus Fundamental Theorem of Calculus I Let f (x) be continuous, constant a. Consider the function: Z x F (x) = f (t) dt a (Notice the integrand is the function f , except with variable changed to t.) Then F 0 (x) = f (x) (The derivative is the integrand, replace all values of t with x). Combining with Chain Rule: Z g(x) F (x) = f (t) dt F 0 (x) = f (g(x)) · g 0 (x) a (Replace all t with g(x), then multiply g 0 (x).) Fundamental Theorem of Calculus II For f continuous, let F 0 (x) = f (x): Z b f (x) = F (b) F (a) a To compute an integral, we only need to find an anti-derivative. i.e. Find a function F (x) such that F 0 (x) = f (x). So we ’define’ the indefinite-integral to be one of the anti-derivative. Z f (x) = F (x) There are infinitely many anti-derivative of f (x), they are di↵erent up to an additive constant. Common anti-derivatives Z 1 xn = xn+1 n+1 Author: Xiao Z x e =e x Z 1 = ln |x| x Page 70 MAT137 23 FUNDAMENTAL THEOREM OF CALCULUS Properties of Integrals For f is integrable: Z b Z b Z b (f + g) = f+ g a Z b f= a Z a a Z a f b Z b f a Z Z c f= a b b cf = c a Z b f+ a Z Z b f a c f b |f | a Since f (x) is integrable, f (x) must be bounded: There exists m and M such that m f (x) M for all x 2 [a, b]. (Intuitively, bounded means f (x) is not approaching infinity.) Z b m(b a) f M (b a) a If f g on all of [a, b], then If f is odd: f ( x) = f (x), If f is even: f ( x) = f (x), Z b f a Z Z Z g a a f =0 a a f =2 a b Z a f 0 Mean Value Theorem for Integrals For f (x) continuous, there exists c 2 [a, b] such that Z b f (x) = f (c)(b a) a Some Conditions that guarantee f is integrable on [a, b] 1. bounded and monotone (only increasing or decreasing on [a, b]) 2. bounded and continuous 3. bounded and discontinuous only at finitely many points If f (x) is approaching infinity, we are not able to integrate f (x) (at this time). Author: Xiao Page 71 MAT137 23 FUNDAMENTAL THEOREM OF CALCULUS Function Definition using Integral p x2 arccos x = x 1 ln x = Question 64 Find the derivative of F (x). 1. Z F (x) = 2. F (x) = 3. F (x) = 4. F (x) = Question 65 Consider Z Z Z +2 x 1 x Z p 1 t2 dt x 1 dt t x2 +1 a 4 ex 2 dt 1 + et 5 dt 1 + cos(t) x3 +x2 x2 x 1 1 dt 1 + sin2 (t) 5 Z Z 1 dt 1 + cos(t3 ) 2 x+1 0 Find c 2 [0, 2] in the conclusion of the mean value theorem for integral. Question 66 Let g be an unknown di↵erentiable function. Define Z x G(x) = g(t)dt 0 Find lim x!0 Author: Xiao R G(x) 0 arctan(s + 2s2 )ds x2 Page 72 MAT137 23 FUNDAMENTAL THEOREM OF CALCULUS Question 67 Evaluate the integral with FTC II (by finding anti-derivative). Z 1 1. 2x 3 0 2. 3. 4. 5. 6. 7. 8. 9. Z Z Z 8 Z Z Z x 0 5 1 ⌘ d ⇣p 1 + x2 dx 4 f (x) where f (x) = 0 Z Z p 3 2 2 1 x2 3 1 x 1 1 3 3 3 2x 0 x 1 x 1<x4 1 x2 3 1 ( 1 x 1 x Question 68 Find f (x) where f 0 (x) = x2 and f (2) = 1. Question 69 Find the area bounded by curves x + y 2 Author: Xiao 4 = 0 and x + y = 2. Page 73 MAT137 24 24 INTEGRATION TECHNIQUES Integration Techniques We would like to compute the integral (anti-derivative) of a function Z f (x)dx Substitution (for indefinite integral: finding anti-derivative) 1. Set a portion of f (x) to be equal to u = u(x) 2. Compute du = u0 (x) dx 3. Rearrange for dx regarding the left side as a fraction dx = du u0 (x) 4. Replace dx in integral with the du expression above. Replace the portion in f (x) you set to be u. Hopefully, there is no more quantities involving x. 5. If there still are quantities involving x, solve x in terms of u x = x(u) Replace all remaining x in integral with x(u). 6. Once all quantities in integral are in terms of u, carry out the new integral with u being the integration variable. 7. Substitute x back into u, to get the final answer in terms of x. When to use substitution Given f (x), try to set the most troublesome piece of f (x) to be u. Hopefully, u0 (x) will be on the outside of the function f (x), directly multiplied to dx. In this case, the quantity u0 (x) will cancel in the substitution process. Note: When the bounds of the integral is given (definite integral), one can also compute the new bounds for u, from the bounds given in x. However, it is not necessary to compute the new bounds for u. Simply find the antiderivative in x, and plug in the bounds of x at the end. Author: Xiao Page 74 MAT137 24 INTEGRATION TECHNIQUES Integration by Parts 1. Given f (x), attempt to split the function into 2 pieces, u and dv Z Z f (x) = u · dv Note: After choosing the quantity u, dv must be the rest of the function f (x) 2. Compute the derivative and the anti-derivative of u and dv Z du 0 0 = u (x) ) du = u (x) · dx v = dv dx (dx here can be ommited.) 3. Apply formula Z Z u · dv = u · v du · v 4. Carry out the new integral on the right When to use integration by parts When f (x) is of the form xn sin(x), xn cos(x), xn ex , set u = xn , then du would have one power less than before. Continue the same process until there is no more powers of x left. When f (x) contains functions with a lot of letters such as ln(x), arctan(x), arccos(x), arcsin(x), set u to be this quantity. Then du would be only a fraction, with no letters involved. Example u=x Z du = 1 · dx x · cos(x) = u · v Author: Xiao Z x · cos(x) v= Z dv = cos(x) cos(x) = sin(x) Z Z du · v = x · sin(x) 1 · sin(x) dx = x · sin(x) + cos(x) Page 75 MAT137 24 INTEGRATION TECHNIQUES Question 70 Z 1. e 5x+1 dx 2. 3. 4. 5. 6. 7. Z Z Z Z Z Z x 2 p ⇡/4 + 1 dx sin x dx cos3 x 0 2 x3 p x 1(x + 1) dx 2. 3. 4. 5. 6. 7. Z Z Z Z Z 9. 10. 11. 1 p x2 2 + x dx 1/2 1/4 1 dx x ln x sin x cos x dx sin x + cos x Question 71 Z 1. xex dx Z 8. x sin(2x) dx x2 cos x dx 12. 13. Z Z Z Z 14. 8. Z 9. 10. 11. ln x dx 12. 2 x ln x dx Author: Xiao Z Z ex cos x dx sin(ln x) dx Z 13. Z Z Z Z Z p cos x p dx x p x x2 4 x2 dx 2x 7 dx 7x + 10 x3 dx (x + 1)2 x2 x3 2x 3x2 5 dx x+2 dx x2 + 1 2x2 x dx x+2 arcsin x dx x arctan x dx arctan p x dx x sec2 x dx sec x dx sec3 x dx Page 76 MAT137 24 INTEGRATION TECHNIQUES Question 72 Let v(t) = 2t 2. Find the total displacement and total distance travelled, in the time interval [0, 3]. Question 73 Evaluate the limit by viewing it as a right endpoint Riemann Sum on [0, 1]. n i2 P 1. lim n!1 i=1 n3 2. lim n P n!1 i=1 i2 i + n2 Question 74 Find the area under the curve. p 1. y = x x2 + 1 on 0 x 1 2. y = 2 sin x + 2 on 0 x ⇡ Author: Xiao Page 77 MAT137 25 25 AREA Area Given y = f (x) > 0 on x in [a, b], the area under the curve y = f (x) (and above the x axis) is given by Z b Z b Area = f (x) dx = y dx a a If y = f (x) > g(x), then the area between the 2 curves is given by the larger area minus the smaller area. Z b Z b Area Di↵erence = f (x) dx g(x) dx a a This can be extended to the case when the functions are negative as well. Integration with Respect to y Sometimes, we are given a curve implicitly, such as y 5 + y + 1 = x. It may be difficult to solve for y = f (x), and sometimes we can not solve y = f (x). In these scenarios, we can still compute area by taking integral against y. Given x = f (y) > 0 on y in [c, d], the area between the curve x = f (y) and the y-axis is given by Z d Z d Area = f (y) dy = x dy c c If x = f (y) > g(y) which means the curve f (y) is to the right of g(y) , then the area between the 2 curves is given by the larger area minus the smaller area. Z d Z d Area Di↵erence = f (y) dy g(y) dy c c This can be extended to the case when the functions are negative as well. Question 75 1. Find the total area between y = x3 2. Find the area between y = x 3. Find the area between y 2 + 3y Author: Xiao 2x + 2 and y = x2 + 2. 1 and y 2 = 2x + 6. x = 16 and |3y| x = 0. Page 78 MAT137 26 26 VOLUME Volume Volume by revolution There are 2 quantities, radius r and height h. Depending on the rotation axis, they will correspond to x and y. The relation between x and y transform into relation between r and h. Rotation axis as y-axis r=x h=y Given x = f (y) ) r = f (y). We can compute the volume bounded between x = f (y) and the y-axis, by adding up many horizontal circles on di↵erent values of y in (c, d). Z Z d 2 V = ⇡r dy = ⇡[f (y)]2 dy c Given y = f (x) ) h = f (x). We can compute the volume bounded between y = f (x) and the x-axis, by adding up many thin cylinders on di↵erent values of x in (a, b). Z Z b V = 2⇡r · h dx = 2⇡x · f (x) dx a Rotation axis as x-axis h=x r=y Given y = f (x) ) r = f (x). We can compute the volume bounded between y = f (x) and the x-axis, by adding up many verticle circles on di↵erent values of x in (a, b). Z Z b 2 V = ⇡r dx = ⇡[f (x)]2 dx a Given x = f (y) ) h = f (y). We can compute the volume bounded between x = f (y) and the y-axis, by adding up many thin cylinders on di↵erent values of y in (c, d). Z Z d V = 2⇡r · h dy = 2⇡y · f (y) dy c In both cases, we can take the di↵erence in volume between 2 curves. Author: Xiao Page 79 MAT137 26 VOLUME Question 76 1. Consider the region bounded by x = y 3 , x = 8, y = 0. Find the volume after revolving this region around y-axis. 2. Consider the region bounded by y = x2 , y = x1/3 . Find the volume after revolving this region around y-axis. p 3. Consider the region bounded by y = x, x + y = 6, y = 1. Find the volume after revolving this region around x-axis. 4. Consider the region bounded by y = x2 , y = 9. Find the volume after revolving this region around x-axis. 5. Find the volume of a cylinder, and a cone of the same dimensions. 6. Find the volume of a sphere. 7. Consider the region bounded by y = x2 + 4, y = 6x x2 . Find the volume after revolving this region around x-axis, and y-axis. p p 8. Consider the region bounded by x = 0, y = 0, x + y = 1. Find the volume after revolving this region around x = 3. Do this in 2 ways: Integrate with respect to x, and with respect to y. 9. Consider a hemispherical bowl of radius R. We place a small iron ball with radius r inside the bowl, where 2r R (so it is completely inside the bowl). Let the water level be h. Find the total volume of water. Note, there are 2 cases: If the ball is completely submerged, we subtract the ball’s total volume. If the ball is only partially submerged, then we need to subtract that portion of its volume. 10. Find the volume of a square based pyramid. (Add area of squares instead of circles.) 11. Two cylinders with radius r intersect with their axis at right angle. Find the volume of the intersection (that is inside both cylinders). Strategy: Take the 2 cylinders to be x2 + z 2 = r2 and y 2 + z 2 = r2 . Each horizontal section at a constant value of z is a square. Add up all areas of the square at all values of z to attain the volume. Author: Xiao Page 80 MAT137 27 27 TRIGNOMETRIC INTEGRALS Trignometric Integrals Multiples of sin(x) and cos(x) Let’s start with the basics: Z sin(x) = cos(x) Z cos(x) = sin(x) Only 1 power of sin(x) or cos(x) Use subtitution, set u to be the one with more powers. Z sin(x) · cos100 (x)dx Z u = cos(x) sin(x) · cos100 (x)dx = Z du = sin(x)dx u101 = 101 u100 du = cos101 (x) 101 One of sin(x) or cos(x) have odd power Use the formula sin2 (x) + cos2 (x) = 1 to interchange cos2 (x) and sin2 (x) of the odd power so that one power remain. Z Z 5 100 cos (x)sin (x) = cos(x)cos4 (x)sin100 (x) = = = Z Z Z cos(x) 1 cos(x) 1 cos(x)sin100 (x) sin2 (x) 2 sin100 (x) 2sin2 (x) + sin4 (x) sin100 (x) Z Z 102 2 cos(x)sin (x) + cos(x)sin104 (x) Solve each integral with the method of 1 power of sin(x) or cos(x). Note: This method also includes when one of sin(x) or cos(x) is missing. Z Z Z 2 5 4 sin (x) = sin(x)sin (x) = sin(x) 1 cos2 (x) = Author: Xiao Z sin(x) 2 Z 2 sin(x)cos (x) + Z sin(x)cos4 (x) Page 81 MAT137 27 TRIGNOMETRIC INTEGRALS Only sin2 (x) or cos2 (x) Use the sum of angles formula: cos(2x) = cos(x + x) = cos(x)cos(x) sin(x)sin(x) = cos2 (x) sin2 (x) Interchange the cos2 (x) and sin2 (x) as before, keep the one you need. sin2 (x) sin2 (x) cos(2x) = cos2 (x) 1 cos(2x) = 1 2sin2 (x) cos(2x) = 2cos2 (x) 1 cos(2x) = 1 1 sin2 (x) = Z Z 1 2 sin (x) = cos(2x) 2 cos(2x) 2 Both of these integral are easy. cos2 (x) 1 + cos(2x) cos2 (x) = 2 Z Z 1 + cos(2x) cos2 (x) = 2 Only even powers of sin(x) and cos(x) Use the above formula to interchange all even powers into powers of cos(2x) Z 2 4 sin (x)cos (x) = 1 = 8 Z (1 Z ✓ 1 cos(2x) 2 ◆✓ 1 + cos(2x) 2 ◆2 cos(2x)) 1 + 2cos(2x) + cos2 (2x)) Z 1 = 1 + 2cos(2x) + cos2 (2x) cos(2x) 2cos2 (2x) cos3 (2x) 8 Collect like terms, and replace the even powers again with the above formula note that the term is now cos2 (2x). Integrate the odd powers as before. Z 1 = 1 + cos(2x) cos2 (2x) cos3 (2x) 8 ✓ ◆ Z 1 1 + cos(4x) = 1 + cos(2x) cos(2x) 1 sin2 (2x) 8 2 For the last integral, use the substitution u = sin(2x) as before. Author: Xiao Page 82 MAT137 27 TRIGNOMETRIC INTEGRALS Multiples of sec(x) and tan(x) Let’s start with the basics: d sec(x) = tan(x)sec(x) dx Z sec(x) = tan(x)sec(x) d tan(x) = sec2 (x) dx Z tan(x) = sec2 (x) Starting with the formula and dividing by cos2 (x): sin2 (x) + cos2 (x) = 1 ) tan2 (x) + 1 = sec2 (x) This allows us to interchange tan2 (x) and sec2 (x). Only 2 powers of sec(x) Use substitution, set u = tan(x). Z sec2 (x)tan100 (x)dx u = tan(x) Z 2 sec (x)tan 100 (x)dx = Z du = sec2 (x)dx u100 du = u101 tan101 (x) = 101 101 Only 1 power of tan(x) and at least 2 powers of sec(x) Note that for 1 power of sec(x), the answer is above. Use substitution, set u = sec(x). Z tan(x)sec100 (x)dx Z u = sec(x) du = tan(x)sec(x)dx Z 100 tan(x)sec (x)dx = tan(x)sec(x)sec99 (x)dx = Author: Xiao Z u99 du = u100 sec100 (x) = 100 100 Page 83 MAT137 27 TRIGNOMETRIC INTEGRALS Even powers of sec(x) Use the formula tan2 (x) + 1 = sec2 (x) to interchange sec2 (x) with tan2 (x) so that only 2 powers of sec(x) remain. Z Z 4 100 sec (x)tan (x) = sec2 (x)sec2 (x)tan100 (x) Z = sec2 (x)(tan2 (x) + 1)tan100 (x) Z Z 2 102 = sec (x)tan (x) + sec2 (x)tan100 (x) Both are done with the substitution u = tan(x). Odd powers of tan(x) Replace so that only 1 power of tan(x) remain. Z Z 5 100 tan (x)sec (x) = tan(x)tan4 (x)sec100 (x) Z 2 = tan(x) sec2 (x) 1 sec100 (x) Z = tan(x) sec4 (x) 2sec2 (x) + 1 sec100 (x) Z Z Z 104 102 = tan(x)sec (x) 2 tan(x)sec (x) + tan(x)sec100 (x) All are done with the substitution u = sec(x). Note: Odd powers of sec(x) and even powers of tan(x) are very difficult. Z sec(x) = ln |sec(x) + tan(x)| Multiples of csc(x) and cot(x) Use 1+cot2 (x) = csc2 (x) and csc0 (x) = cot(x)csc(x) and cot0 (x) = csc2 (x) Di↵erent frequencies of sin(mx) and cos(nx) 1 sinAcosB = [sin(A B) + sin(A + B)] 2 1 sinAsinB = [cos(A B) cos(A + B)] 2 1 cosAcosB = [cos(A B) + cos(A + B)] 2 Author: Xiao Page 84 MAT137 27 TRIGNOMETRIC INTEGRALS Question 77 Z 1. sin4 x cos x dx 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. Z Z Z Z Z Z Z Z Z Z 5 3 3 2 sin x cos x dx sin x cos x dx 3 2 cos x dx 14. 16. 2 sin x dx 17. 2 sin x cos x dx 4 sin x dx 18. 19. 2 sec x dx 20. tan x sec x dx 5 13. 15. cos x dx 2 12. 2 tan x sec x dx 21. 22. Z Z Z Z Z Z Z Z Z Z Z tan x sec4 x, dx sec4 x tan4 x dx tan3 x sec x dx sec4 x dx tan3 x dx sec5 x dx tan2 x dx tan2 x sec x dx sin 2x sin 3x dx cos 2x sin 5x dx cos2 3x sin 2x dx Question 78 Verify the Orthogonal Relation for Fourier Sine and Cosine Series: ( ( Z ⇡ Z ⇡ 0 if m 6= n 0 if m 6= n sin mx sin nx dx = cos mx cos nx dx = ⇡/2 if m = n ⇡/2 if m = n 0 0 Z ⇡ sin mx cos nx dx = 0 (for any n, m) where n, m are integers. ⇡ Author: Xiao Page 85 MAT137 28 28 TRIGNOMETRIC SUBSTITUTION Trignometric Substitution p Integral involving pa x2 This looks similar to 1 sin2p (✓) = cos(✓) Try to turn the expression into 1 z 2 for some new z. Use substitution, set z = sin(✓). Z x2 p dx 4 x2 s ✓ r r ◆ 2 2 p 4x x 4 x2 = 4 = 4 1 =2 1 4 4 x = sin(✓) 2 dx = 2cos(✓) d✓ Z Z x2 x2 p q dx = 4 x2 2 1 ) ⇣ x ⌘2 2 x = 2sin(✓) ) x 2 dx = 2cos(✓)d✓ Z x2 p dx = · 2cos(✓)d✓ 2 2 1 sin(✓)2 Notice we still have x in the expression, so we need to change them to ✓. Z Z 4sin2 (✓) p = · 2cos(✓)d✓ = 4sin2 (✓)d✓ = 2✓ sin(2✓) 2 2 1 sin(✓) Now we need to express ✓ in terms of x. For the first term, we can do ✓ = arcsin(x/2). However, the second term involves a sin(2✓). While substituting the above ✓ expression gives the right answer, we need to simplify the answer. The substitution relates the triangle with the p opposite side length x and hypotenuse length 2. The adjacent length is 4 x2 . Using the double angle formula, p sin(✓ + ✓) = sin(2✓) = 2sin(✓)cos(✓). Using the triangle, cos(✓) = 4 x2 /2. p ⇣x⌘ ⇣x⌘ 4 x2 = 2arcsin 2sin(✓)cos(✓) = 2arcsin x 2 2 2 Author: Xiao Page 86 MAT137 28 TRIGNOMETRIC SUBSTITUTION p Integral involving pa + x2 This looks similar to 1 + tan2p (✓) = sec(✓). Try to turn the expression into 1 + z 2 for some new z. Use subtitution, set z = tan(✓). p Integral involving px2 a This looks similar to sec2 (✓) p1 = tan(✓). Try to turn the expression into z 2 1 for some new z. Use substitution, set z = sec(✓). p Integral involving ax2 + bx + c Complete the square inside the square root. Then, depending on the situation, do a trig substitution from above. Z 1 p dx (x + 1) x2 + 2x 3 Recall the procedure for completing the square: add and subtract (b/2)2 . x2 + 2x Z 3 = x2 + 2x 3 + 1 1 = (x + 1)2 4 Z 1 p dx = dx 3 (x + 1) (x + 1)2 4 1 (x + 1) x2 + 2x p This looks similar to sec2 (✓) 1. Changing the 4 to 1: Z Z 1 1 q q dx = 2 (x + 1)2 ( x+1 )2 (x + 1) 4( (x+1) 1) 2 4 p x+1 = sec(✓) 2 ) x = 2sec(✓) dx 1 1 dx = 2sec(✓)tan(✓) ) dx = 2sec(✓)tan(✓)d✓ d✓ ✓ ◆ Z Z 1 sec(✓) ✓ 1 x+1 = ·2sec(✓)tan(✓)d✓ = d✓ = = arcsec (x + 1)2tan(✓) 2sec(✓) 2 2 2 Question Z p 792 Z Z 9 x 1 1 p 1. dx 2. dx 3. x4 (x2 + 2)3/2 ex e2x Author: Xiao 4 dx 4. Z p x2 1 2x Page 87 3 dx MAT137 29 29 PARTIAL FRACTION Partial Fraction We want to integrate any rational function, f (x) = p(x) q(x) where p(x), q(x) are polynomials. We use a process that together is typically referred to as partial fractions. Long Division Long division is the first step in the process of integrating rational function. We compare the degrees of the numerator p(x) and the denominator q(x). (Degree of a polynomial p(x) is the value of its highest power.) If deg p(x) < deg q(x), then we skip long division and proceed to the next step. If deg p(x) deg q(x), we perform long division on p(x) , and produce q(x) p(x) r(x) = g(x) + q(x) q(x) where g(x) is the quotient polynomial, r(x) is the remainder polynomial. p(x) , we integrate the 2 terms on the right. q(x) The polynomial g(x) is easy to integrate. r(x) We only need to integrate the rational function . q(x) It is guaranteed that deg r(x) < deg q(x), and thus we have made sure the numerator has degree less than the denominator. To integrate Thus using the technique of long division, we can reduce any integral of a rational function into a rational function with degree of the top less than the degree of the bottom. From now on, we will assume deg p(x) < deg q(x). Author: Xiao Page 88 MAT137 29 PARTIAL FRACTION Partial Fraction 1. Completely factor the denominator q(x) into irreducible factors. The irreducible factors must have either degree 1 or degree 2. q(x) = q1 (x) . . . qn (x) Let’s first assume each qi (x) are distinct (di↵erent from each other). 2. Write the original fraction into a sum of fractions of polynomials, with qi (x) as denominators. p(x) p1 (x) pn (x) = +·+ q(x) q1 (x) qn (x) If qi (x) is degree 1, we set pi (x) to be a constant, such as pi (x) = A. If qi (x) is degree 2, we set pi (x) to be degree 1, such as pi (x) = Ax + B. Example x3 + x2 x + 1 A B Cx + D = + + 2 2 x(x 1)(x + 1) x x 1 x +1 3. Take a common denominator on the right side. Solve for the constants in pi (x) based on p(x) as a system of linear equations. Example x3 + x2 x + 1 A (x = · 2 x(x 1)(x + 1) x (x 1)(x2 + 1) B x(x2 + 1) (Cx + D) x(x + · + 2 · 1)(x2 + 1) x 1 x(x2 + 1) x + 1 x(x x3 + x2 x + 1 = A(x 1)(x2 + 1) + Bx(x2 + 1) + (Cx + D)x(x 1) Expand and factor out xn : x3 +x2 x+1 = (A+B +C)x3 +( A C +D)x2 +(A+B D)x A·1 Compare the coefficients in front of the powers of xn , we see that 1=A+B+C 1= A C +D 1=A+B D 1= A Solving this system, we see that A = 1, B = 1, C = 1, D = 1. So we attain x3 + x2 x + 1 1 1 x+1 = + + 2 2 x(x 1)(x + 1) x x 1 x +1 Author: Xiao Page 89 1) 1) MAT137 29 PARTIAL FRACTION 4. In the event that a factor has multiplicity n, that is to say q(x) have n a repeated irreducible factor, such as q1 (x) , n q(x) = q1 (x) q2 (x) . . . qn (x) then to form a sum of fractions for this factor, we must repeatedly form k new fractions in the sum, with denominator q1 (x) with k increasing from 1 up to n. The numerator is still determined by whether q1 (x) is linear or quadratic. Example p(x) 3x2 + x + 1 A B C D = = + + + 3 2 q(x) x(2x 1) x 2x 1 (2x 1) (2x 1)3 Notice that we have (2x 1)k as the denominator, with k = 1, 2, 3. In each of the instances, we have a constant as numerator, since (2x 1) as the repeated factor, is linear. Example p(x) 3x2 + x + 1 A B Cx + D Ex + F Gx + H = 2 2 = + 2+ 2 + 2 + 2 3 2 q(x) x (x + 1) x x (x + 1) (x + 1) (x + 1)3 Notice that we have xk as denominator, with k = 1, 2. The numerators are set to be constants, since the irreducible repeated factor x is linear. That is to say, we view x2 = (x)2 . However, x2 + 1 is an irreducible deg 2 repeated factor, so we have (x2 + 1)k as denominator, with k = 1, 2, 3. The numerators are set to be linear, since the irreducible repeated factor (x2 + 1) is quadratic. Thus, the partial fraction method writes f (x) as a sum of fractions with deg 1 or deg 2 denominator, and each fraction can be integrated separately. We are left with solving integrals of the following forms: Z Z A Ax + B k 2 (ax + b) (ax + bx + c)k where ax2 + bx + c is an irreducible quadratic. Author: Xiao Page 90 MAT137 29 PARTIAL FRACTION Integration 1. Linear term (degree 1) on denominator. Z A (ax + b)k take substitution u = ax + b, to result in an integral of the form Z 1 uk 2. Quadratic term (degree 2) on denominator. Linear term on numerator. Z Ax + B 2 (ax + bx + c)k We can split the fraction using +. We first focus on the term with x in the numerator. Z Ax dx 2 (ax + bx + c)k We must set u = ax2 + bx + c from the denominator. Then du = (2ax + b)dx as a linear term. So we try to make the numerator to be du = (2ax + b)dx exactly. Z Z Z ⌘ A 2ax A⇣ 2ax + b b = = dx+ 2a (ax2 + bx + c)k 2a (ax2 + bx + c)k (ax2 + bx + c)k The first integral after the substitution u = ax2 + bx + c result in Z 1 uk The second integral is of the form Z Z B 1 =B 2 k 2 (ax + bx + c) (ax + bx + c)k which is the form of the 2nd term in the fraction. Author: Xiao Page 91 MAT137 29 PARTIAL FRACTION 3. Quadratic term (degree 2) on denominator. Constant term (equivalent to the number 1) on the numerator. Z 1 2 (ax + bx + c)k This case is difficult, we focus on the case of k = 1. Z 1 2 ax + bx + c By the partial fraction procedure, the quadratic must be irreducible. We complete the square on the denominator, and try to attain the form Z 1 ⇣ ⌘2 cx + d + 1 We take u = cx + d. This gives the integral of the form Z 1 = arctan(u) 2 u +1 The case of k > 1 requires integration by part to reduce to the case of k = 1. In conclusion, the result of the integral Z p(x) q(x) only involve terms of the form: um , ln(u), arctan(u). (where m can be positive or negative) Question 80 1. 2. Z Z 2x3 + x2 x (x 1)2 4 dx x3 + x2 + x + 2 dx x4 + 3x2 + 2 Author: Xiao 3. 4. Z Z x3 x+1 dx 2x2 + 2x (x2 1 dx + 1)2 Page 92 MAT137 30 30 SEQUENCES Sequences A sequence S = {an }1 n=1 is like a regular function f (n) = an , except you can only ask its limit as n ! 1. A sequence S is bounded above, if every number in the sequence is less than some big number M. If a sequence S is bounded above, there exists a quantity that is very similar to the maximum of S, it is called the supremum of S, denoted sup S. It is like the maximum, because, if you take a step to the left, from the supremum, no matter how small the step is, you will bump into something in S along the way. 8✏ > 0, 9s 2 S such that, sup S ✏ < s sup S Similarly, if S is bounded below, then there exists a quantity that is similar to the minimum of S, called infimum of S, denoted inf S. Of course, you take a step to the right, from the infimum, you will bump into something in S. 8✏ > 0, 9s 2 S such that, inf S s < inf S + ✏ The definition of limit as n ! 1 for S = {an } and the computation of limit is exactly the same as a regular function f (n) = an . For a sequence {an } , the limit L exists when the values of an get close to L. How close is close? For any error range ✏, as long as we take n large, we can always make sure an is within that error to L 8✏ > 0, 9N such that, if n > N ) |an L| < ✏ If the limit exists, then S gets close to L, so it is not going to infinity, so it’s bounded above and below. However, being bounded (above and below) is not enough for the limit to exist. Author: Xiao Page 93 MAT137 30 SEQUENCES If the limit of S = {an } does not exist, we say that it diverges. It coule be going to infinity, or it could oscillate and not approach a specific number L. If S is unbounded, S would get bigger and bigger, so it can’t converge. increasing means each term is bigger than the term before: 8n, an+1 > an . non-decreasing is just a technical way of saying 8n, an+1 an . If S is bounded above, and it’s non-decreasing, then it converges to its ’maximum’ value, or the supremum. decreasing means 8n, an+1 < an , non-increasing means 8n, an+1 an . If S is bounded below, and it’s non-increasing, then it converges to its ’minimum’ value, or the infimum. Being non-increasing or non-decreasing is called monotone. An important note: the limit of a sequence has nothing to do with how it behaves at the beginning. We only ask about how the sequence behaves as n ! 1, so we can ignore finitely many terms when computing limit. Using this, we can say the following are true ’eventually’, which means when n is large enough: ln(n) ⌧ np for any power of p > 0. (log grows slower than polynomial) n p ⌧ an for any base a > 1. (polynomial grows slower than exponential) an ⌧ n! for any base a > 1. (factorial grows very fast) n n! ⌧ n . (nn grows the fastest) We attain the chain of ascending growth rates: p 1 ⌧ ln(n) ⌧ n ⌧ n ⌧ n2 ⌧ ... ⌧ 2n ⌧ en ⌧ n! ⌧ nn where an ⌧ bn if an =0 n!1 bn Compute limits when the base of not a constant To compute limit, L’Hopital Rule is usually the method of choice. Change to base e, then bring the limit to the exponent: lim a = eln(a) Author: Xiao ak = ek·ln(a) Page 94 MAT137 30 SEQUENCES Example (n + 2)1/n = eln(n+2) 1/n = eln(n+2)/n Using the chain of growth, the exponent approach zero, thus the answer to the limit is e0 = 1. Question 81 Find the infimum and supremum of the following sets. 1. S = {x : |x 1| < 2} 2. S = {0.9, 0.99, 0.999, ....} 3. S = {( 1)n (1 1 ) : n 2 N} n Question 82 Find the limit of the following sequences. n + ( 1)n n 6. 1 2n 4n 3. p 4n2 + 2 7. n ln n 2n 3n + 5 n+1 5. n p 2 n 8. (n + 2)1/n 1. 2. (1.01)n 4. 1 2n + 3 3n+1 9. n 1 4 Question 83 Prove an ! 0 if and only if |an | ! 0. Question 84 Consider the recursively defined sequence. Determine whether it converges. If it converges, find its limit. 1 1. a1 = 1, an+1 = an + 1. 2 p p 2. a1 = 2, an+1 = 2 + an Author: Xiao Page 95 MAT137 31 31 IMPROPER INTEGRAL Improper Integral Integral is many rectangle’s area which is base times height. What happens if one of them is infinite? Type 1: base is infinite This corresponds to one of the limit of integration being infinity. Z 1 Z b f (x) = lim f (x) a b!1 a We integrate to some big number b as usual, then take the limit. Type 2: height is infinite This corresponds to the function going to infinity. Assume f (x) appraoches infinity at x = k, and a < k, then Z k Z b f (x) = lim f (x) a b!k a We integrate to some number b close to k, then take the limit. When multiple infinite points arise We need to split the integral into several pieces, so that in each piece, only 1 of the 2 bound is approaching an infinite point. When one piece results in infinity, the entire integral diverges. Example Z 1 1 (x 1)3 0 The integral has infinite points at x = 1, and x = 1. We need to approach x = 1 on both sides. So we choose some number c 2 (1, 1) and split up the integral: Z 1 Z c Z 1 1 1 1 = + + 3 3 1) 1) (x 1)3 0 (x 1 (x c Z b Z c Z b 1 1 1 = lim + lim+ + lim 3 3 b!1 c (x b!1 b!1 1) 1) 1)3 0 (x b (x 1 = 1+1+ = ±1 2(c 1)2 Author: Xiao Page 96 MAT137 31 IMPROPER INTEGRAL Integrals can’t always be computed. However, we can ask whether an integral is convergent (not infinite). Comparison Test for Improper Integrals If f, g continuous, and 0 f (x) g(x) for x a, Then Z 1 Z 1 0 f (x) g(x) a a In particular: R1 R1 If the larger integral aR g(x) < 1, then the smaller integral a f (x) R 1< 1. 1 If the smaller integral a f (x) = 1, then the larger integral 1 a g(x). To prove an integral converges, compare it to a larger integral which can be computed to converge. To prove an integral diverges, compare it to a smaller integral which can be computed to diverge. Typically, the new integral will be modified from the original, by removing complicated terms. Limit Comparison Test for Improper Integrals If f, g continuous, both positive for x a, If f (x) 6= 0 6= 1 x!1 g(x) lim Then f (x) ⇡ g(x) In particular: Z 1 a f (x) converges if and only if (The rate of growth is similar) Z 1 g(x) converges a This is typically used on rational functions. When f (x) is a rational function, we take the leading powers on the top and bottom to create a new function g(x). This method guarantees the the limit is not zero or infinite. Author: Xiao Page 97 MAT137 31 IMPROPER INTEGRAL Question 85 Compute the improper integral, or show it diverges. Z 1 Z 1 1 1. 5. cos x 2 1 x +1 0 Z 1 Z 1 1 1 2. 6. 2 x+2 2 0 x Z 1 Z 1 1 p 3. ln x 7. x 0 0 Z 1 Z 1 1 4. ln x 8. 2 x 1 1 2 Question 86 Let p > 0. For which values of p would the integral converge or diverge? Z 1 1 xp 1 This integral is sometimes called the p-series. Question 87 Using the comparison tests, show whether the integral converges or diverges. Z 1 sin x + 2 cos x + 10 1. x2 1 Z 1 x 7 2. 2 x +x+5 0 p Z 1 x 6 3. 2 10 3x + 5x + 11 Z 1 arctan x 4. x1.1 0 Z 1 sin x 5. 4/3 0 x Z 1 2 6. e x 0 Author: Xiao Page 98 MAT137 32 32 SERIES Series We would like to add numbers together, but it gets tricky if we want to add infinitely many of them. We define a series as a sequence of partial sums. For a series 1 X an n=1 Consider the sequence: s1 s2 s3 s4 = a1 = a1 + a2 = a1 + a2 + a3 = a1 + a2 + a3 + a4 .. . sn = a1 + a2 + a3 + ... + an The sequence {sn } are the partial sums. We define the series converge if the sequence {sn } converge. Usually, infinite series are very difficult to calculate. However, one important result is the geometric series, which converges: 1 X xn = n=0 1 1 |x| < 1 x Only when series converge, we can perform regular algebra as follows: 1 X an + n=a 1 X bn = n=a c· 1 X n=a an = 1 X an + b n n=a 1 X n=a c · an With these, we can manipulate the geometric series to attain exact answers to some series. Author: Xiao Page 99 MAT137 32 SERIES Convergence of Series Most of the time, we can not expect to attain the exact answer to the series. Instead, we can characterize whether a series converge or not. When taking this approach, we again arrive at the previous conclusion that: we only care about how the terms behave as n ! 1, and we can ignore finitely many terms. Thus, the starting point of the series does not a↵ect convergence. However, since we can not ignore infinitely many terms, if the terms get larger and larger, the series can’t possibly converge: Test 0: Zero Test for Divergence 1 X If lim an 6= 0, then an diverges. n!1 n=a However, even when lim an = 0, it might oscillate, or diverge for other reasons. n!1 Thus test 0 is almost never useful. When the series does not oscillate, it usually will be decreasing ’eventually’, and will remain positive, then we can connect the points an smoothly to form a continuous function. The series can be thought of as area of the rectangles below the curve, so it must be the case that the integral is approximately equal to the series. Test 1: Integral Test For an positive and decreasing ’eventually’, 1 X n=a an ⇡ Z 1 an a In particular, the sum (series) converges if and only if the integral converges. We should all know by now that integrals can get really difficult. Although this approximation can almost always be made, it is almost never useful. (Decreasing ”eventually”: there exists N such that for all n > N , an+1 < an .) Author: Xiao Page 100 MAT137 32 SERIES P-series Which integrals are easy to do? Clearly powers like xp . For a power p: Z 1 Z 1 Z 1 1 X 1 1 1 ⇡ = dx = x p dx p p p n n x a a a n=a For p 6= 1, this is easily calculated: x p+1 p+1 = x=1 x=a The evaluation at x = a is just a number so it’s irrelevant, and so we only need to focus on the infinity part. The answer depends upon the exponent. If the exponent is positive, then it’s infinite. If the exponent is negative, then it’s finite. For p = 1, then answer is ln(x), which is infinite. Thus: Z 1 1 X 1 1 ⇡ dx p n xp Diverges for p 1 Converges for p > 1 Given this result, what can we say about: 1 X 1 n2 1 Well, it most likely converges because we only care about the behaviour as n becomes large, so the number 1 is irrelevant. In fact, this is true in general: When an is a rational function, we only need to take the leading powers 1 X on the top and bottom to create a new series bn . Test 2: Limit Comparison Test For an , bn positive ’eventually’, 1 1 X X an 6= 0 6= 1, then an ⇡ bn n!1 bn If lim In particular, the first sum (series) converges if and only if the second converges. Author: Xiao Page 101 MAT137 32 SERIES Now that we are able to test convergence for rational function, we next need to worry about other functions, such as ln(x), sin(x), cos(x), arctan(x). Fortunately, a function like sin(x) is at most 1, so it being inside a series would at most be multiplying the original series by 1, which shouldn’t matter if the series already converge. Test 3: Comparison Test Suppose an bn and positive ’eventually’, then If 1 X If bn Converges, then 1 X an Diverges, then 1 X 1 X 1 X an 1 X bn : an Converges bn Diverges Note that in both cases, the implication is only one direction. 1 X arctan(n) n2 So the left series converges too. 1 X ⇡ 1 Converges 2 n2 Using ln(n) < np for any p > 0: 1 X ln(n) 1 p X n < n2 So the left series converges too. 1 X 1 = Converges 2 3/2 n n Now we focus on series including exponents an and factorials n!. Test 4: Ratio Test For an positive ’eventually’, compute an+1 lim =L n!1 an If L < 1, then 1 X an converge. 1 X If L > 1, then an diverge. If L = 1, then it’s inconclusive. Author: Xiao Page 102 MAT137 32 SERIES Convergence of Series involving Positive and Negative Terms In general, this is a difficult problem. Thus it’s much easier to just get rid of all the negative signs. Test 5: Absolute Convergence Test 1 1 X X If |an | Converges, then an Converges The quantity |an | means making every term positive. Intuitively, if the positive series converge, then some terms becoming negative on the right would only make the sum smaller, and would still converge. 1 1 X X When |an | converges, we say an is absolutely convergent. Note: the positive sum diverging does not mean the original series diverge. 1 1 X X In the case when the an does converge but |an | does not converge, 1 X we say an is conditionally convergent. 1 X sin(n) n2 This series oscillate due to sin(n), but putting on absolute values: 1 1 X X sin(n) 1 | | Converges 2 n n2 Thus the original series converges absolutely by comparison test. A special case is when the positive and negative terms alternate in turn: a1 a2 + a3 a4 ... This can be written with the notation ( 1)n or ( 1)n+1 . Test 6: Alternating Series Test For an positive and decreasing ’eventually’, consider the alternating series 1 X ( 1)n an = a1 a2 + a3 a4 .... Then we only need to compute the limit of an : 1 X lim an = 0, if and only if ( 1)n an Converges n!1 The test also holds with cos(n⇡) = ( 1)n = sin (n + 1/2)⇡ , and ( 1)n+1 . Author: Xiao Page 103 MAT137 33 33 SUMMARY OF TESTS FOR SERIES Summary of Tests for Series Combining all the tests, the steps for determining convergence of a series 1. Take the absolute value of an 1 X an : 2. If there are exponents (2n ) or factorial (n!), apply Ratio Test (or Root Test): ⇣ ⌘ p an+1 lim | |=L or lim n |an | = L n!1 an n!1 1 X L < 1 conclude an absolutely converge. 1 X L > 1 conclude an diverge. 3. ln(n) in the numerator, compare 1 < ln(n) < np with p small. ln(n) in the denominator, compare 1 > 1/ln(n) > 1/np with p small. For sin and cos in numerator, compare | sin(n)| 1 ,| cos(n)| 1. For arctan in numerator, compare arctan(n) < ⇡/2. 4. For series in rational function form, take bn to be leading powers of top and bottom. Attain bn = n1p . 1 X p > 1 conclude an absolutely converge. 1 1 X X p1 |an | diverge, but an may still converge. 1 X 5. If the original series can be written in the form ( 1)n an . For an decreasing, compute lim an = L . n!1 1 X L = 0 conclude ( 1)n an conditionally converge. 1 X L 6= 0 conclude ( 1)n an diverge. 1 , there is one power of n on the bottom, use integral test. n · ln(n) Use small angle approximation on: For an = sin(1/n) ⇡ 1/n ⇡ tan(1/n) ⇡ arcsin(1/n) ⇡ arctan(1/n) 1 X If lim an 6= 0, then conclude an diverge (by Test 0). n!1 Author: Xiao Page 104 MAT137 33 SUMMARY OF TESTS FOR SERIES Question 88 Show whether the series is Absolutely convergent, Conditionally convergent, or Divergent. 1. 1 X n=1 2. 1 X n=2 3. 4. 1 2 + cos n 13. n=1 15. 1 X 1 n2 n=1 16. 17. n2 + 4n p 6. n5 + 4n4 + 2n n=1 1 p X n sin n 7. n2 n=1 ( 1)n sin n=1 1 X 1 n 1 9. ( 1) sin 2 n n=1 11. 12. n 1 X (n + 1)2n n=1 1 X cos n⇡ n! 1 X n! 23n n=1 1 X n! (2n)! n=1 Author: Xiao ln n 1 X ln n n=2 1 X 10. 1 X ( 1)n 2n nn n=1 n=2 1 X ( 1)n 5. n2 1 n=2 8. (3n)! 1 X 1 14. nn n=1 1 n ln n 1 X ( 1)n n n=1 1 X 1 X n!(2n)! n3 1 p X n ln n 18. (n + 1)2 n=2 1 X p 1 X n2 (ln n)2 1 n ln n n=2 p 1 X n+1 20. 1.4 n (ln n)2 n=2 19. 21. n=2 1 1 X (ln n)10 p 22. n n=2 1 X 1 n(ln n)3 n=2 p 1 X n4 + 2n3 + 1 p 24. 3 2n9 + 4n8 + 5n6 (ln n)5 n=2 23. Page 105 MAT137 34 34 TAYLOR SERIES Taylor Series We would like to approximate the function f (x) with a polynomial. We choose the point a as the center of expansion. We require the polynomial to have the same derivatives as the function at a. This (infinite) polynomial is called the Taylor (power) Series of f (x): 1 X f (n) (a) f (x) ⇡ P (x) = (x a)n n! n=0 Usually, the approximation only holds in a symmetric interval centered at a. This is called the interval of convergence. Half of the length of the interval is called the radius of convergence. an+1 Use Ratio Test to attain the interval of convergence, set | | < 1. an The series may converge or diverge on the boundary of the interval. For a = 0, the most important Taylor Series to know are: 1 X xn x2 x3 x e = =1+x+ + + ... x2R n! 2 3! n=0 sin(x) = 1 X ( 1)n 2n+1 x =x (2n + 1)! n=0 cos(x) = 1 1 x 1 X ( 1)n 2n x =1 (2n)! n=0 = 1 X x3 x5 + 3! 5! x2 x4 + 2! 4! x2R ... ... xn = 1 + x + x2 + x3 + ... n=0 x2R |x| < 1 (The note x 2 R means the approximation holds for any number x.) Since the formula holds for any x inside the interval of convergence, we can substitute x with any other quantity. Example e x2 = 1 X (x2 )n n=0 1 3 n! = 1 + x2 + x4 x6 + + ... 2 3! x2R i 1 1 1 X x n 1h x x 2 x 3 = · = ( ) = 1+ +( ) +( ) +... x 3 1 x/3 3 n=0 3 3 3 3 3 Author: Xiao 1 x <1 3 Page 106 MAT137 34 TAYLOR SERIES Derivatives and Integrals We can take derivatives and integrals of Taylor Series on both sides of the equation. For the sum, we take derivatives and integrals term by term. In other words, Z ⇣X ⌘ XZ X d d X an dx = an dx an = an dx dx The interval of convergence remains the same. Example: arctan(x) and ln(1-x) 1 1 X X 1 1 2 n = = ( x) = ( 1)n x2n 1 + x2 1 ( x2 ) n=0 n=0 | x2 | < 1 (We get | x2 | < 1 since we plugged in x2 , but this is equivalent to |x| < 1.) Taking integral on both sides: Z 1 Z X 1 = arctan(x) = C + ( 1)n x2n |x| < 1 2 1+x n=0 arctan(x) = C + 1 X ( 1)n n=0 To solve for C, plug in x = 0, x arctan(x) = 2n+1 1 X = 0, so arctan(0) = 0 = C + ( 1)n n=0 Similarly, ln(1 x) = 1 X xn n=1 n = x2n+1 2n + 1 x2n+1 2n + 1 x2 2 x |x| < 1 x3 3 ... P 0. |x| < 1 Limit Using Taylor Series for x ! 0 We can plug in the Taylor Series of the common functions into the limit. The higher powers of the Taylor Series would be zero after taking x ! 0, leaving the answer as the leading term.We can divide top and bottom by the lowest power to attain the leading term. 3 5 x x3! + x5! ... sin(x) x2 x4 lim = =1 + ... x!0 x x 3! 5! Plugging in x = 0, we attain the answer 1. Note: If the limit is x ! a, we need to expand the Taylor series at a. Author: Xiao Page 107 MAT137 34 TAYLOR SERIES Question 89 Find the radius of convergence and the interval of convergence for the power series. 1. 1 X n(x 2)n n=1 2. 1 X x3n 2n n=2 3. 1 X ( 1)n (x 3)2n n=3 1 X (n!)2 4. (x (2n)! n=4 3)n Question 90 Find the Taylor Series of the functions centered at a = 0. 1. 1 1+x 1 2. 1 + x2 2 x 3. x e 4. cos(x4 ) 5. sin(2x) 6. e x 7. ex e 8. 1 3x + 2 9. x2 x sinx x 10. ln(1 x) 11. ln(1 + x) Z 12. f (x) = x e t2 dt 0 Question 91 Find the Taylor Series of the functions centered at the given value of a. 1. ex at a = 1 1 2. at a = 3 1 x ⇡ 3. sin(x) at a = 4 Author: Xiao Page 108 MAT137 34 TAYLOR SERIES Question 92 Find the exact value of the series (using Taylor Series). 1. 1 X nx n 5. n=1 2. 3. 6. 2n 7. n! 1 X n=0 1 X (n + 1)xn 8. n! n=0 1 X xn (n + 2)n! ( 1)n n=0 1 X xn+1 n=0 4. n=0 1 X n2 n=0 1 X (n + 1)x2n+1 (2n + 1)! xn (n + 1)(n + 2) 1 X (4n2 + 8n + 3)2n n! n=2 Question 93 View the infinite repeating decimal expansion as a series, and compute its value. a) 0.99999... c) 0.252525... b) 0.11111... d) 0.3121212... Question 94 Find the limit using Taylor Series. 6 sin x 6x + x3 1. lim x!0 x5 2. lim x!0 ex 2 cos(2x) 3x2 x2 sin(x2 ) Question 95 A telescoping series is when adjacent positive and negative terms cancel, N X so the N-th partial sum: an would only consist of the first and the last n=1 term, and we can compute the series by taking the limit N ! 1. 1. 1 X [arctan n n=0 ⇣ n ⌘ 2. ln n+1 n=1 1 X Author: Xiao arctan(n + 1)] 3. 1 X n=1 1 n2 + 3n 1 X n+2 4. n3 n n=3 Page 109 MAT137 35 35 POWER SERIES Power Series g(x) = 1 X a)n an (x is called a power series. n=0 For any power series g(x), there exists a unique radius of convergence R 1. The power series g(x) converges absolutely for |x 2. The power series g(x) diverges for |x a| < R. (in the circle) a| > R. (outside the circle). If R > 0, then the power series g(x) (as a function of x) is analytic in |x a| < R and g 0 (x) is given by term by term di↵erentiation 0 g (x) = 1 X a)n nan (x 1 n=1 with the same radius of convergence R. Note that: 1. g(a) = a0 is always defined, and so the power series at least converge at a. Thus R 0. However, g(x) may not converge for any other x, and then we say R = 0. 2. If g(x) converges absolutely for all x, then we say R = 1, and so g(x) is analytic everywhere. 3. g(x) may or may not converges on the boundary of the interval (circle) of convergence, which needs to be checked separately. 4. The point a is called the center of expansion. 5. As g 0 (x) is also a power series, we have g(x) is in fact infinitely di↵erentiable in |x a| < R. Given an analytic function f (x), then for any a in the domain, f (x) must equal its Taylor Series in an interval (circle) near a: f (x) = 1 X f (n) (a) n=0 n! (x a)n Note that Taylor Series is a power series. This is called expanding f (x) at the point a. Author: Xiao 0: Page 110 MAT137 35 POWER SERIES The radius of convergence R can be obtained in 2 ways: Ratio test: R= 1 an+1 lim n!1 an Root Test: R= 1 p lim n |an | n!1 Taylor Polynomial If we only take finitely many terms of the Taylor Series, we get a polynomial called the (N-th order) Taylor Polynomial PN (x). The Taylor Polynomial will not be equal to f (x), but it will be an approximation. The di↵erence between the Taylor Polynomial and f (x) is called the remainder RN (x). PN (x) = N X f (n) (a) n=0 f (x) = N X f (n) (a) n=0 n! (x n! a)n + RN (x) (x a)n RN (x) = f (n+1) (c) (x (n + 1)! a)n+1 where c 2 (a, x) is undetermined by the mean value theorem. The 1st-order Taylor Polynomial is the tangent line at a: P1 (x) = f (a) + f 0 (a)(x f (x) x!a x The definition of the derivative, f 0 (a) = lim a) f (a) gives: a f (x) P1 (x) f (x) f (a) f 0 (a)(x a) = lim = f 0 (a) f 0 (a) = 0 x!a x!a x a x a As x approaches a, we know that x a is going to zero. The 1st order approximation P1 (x) = f (a) + f 0 (a)(x a), as x approaches a, the error (remainder) between the function f (x) and P1 (x) would also go to zero. However, we have that this error (remainder) between the function f (x) and the 1st order approximation is going to zero faster than 1 power of (x a). lim As N increaess, the speed for which (x a)N is going to zero increases. In general, the N-th order Taylor Polynomial would have its error go to zero faster than N powers of (x a): lim x!a Author: Xiao f (x) (x PN (x) =0 a)N Page 111