Professional Documents
Culture Documents
SOS
Prepared by Vincent Chan
v2chan@math.uwaterloo.ca
April 8, 2010
I include numbers after each section heading that correspond to the textbook section headings for your
convenience. Proofs of theorems are included in your textbook, and will be omitted. Finally, I strongly rec-
ommend going through the examples and exercises in your textbook, as they are very helpful in preparing
for your exams and understanding the material.
1 Review
1
WATERLOO SOS E XAM -AID: MATH138 F INAL
Solution. This is suitable for a partial fraction decomposition. First notice the degree of the numerator is
less than that of the denominator, so we do not need to use polynomial division, and the denominator is
already factored. Then suppose
𝑥+5 𝐴 𝐵
= + .
(𝑥 + 1)(𝑥 + 3) 𝑥+1 𝑥+3
Clearing the denominator gives
𝑥 + 5 = 𝐴(𝑥 + 3) + 𝐵(𝑥 + 1).
comparing denominators or substituting values 𝑥 = −1, −3, we get 𝐴 = 2 and 𝐵 = −1. Then
1 1 1
−1
∫ ∫ ∫
𝑥+5 2
𝑑𝑥 = 𝑑𝑥 + 𝑑𝑥
0 (𝑥 + 1)(𝑥 + 3) 0 𝑥+1 0 𝑥+3
1 1
= 2 ln ∣𝑥 + 1∣∣0 − ln ∣𝑥 + 3∣∣0 = 2 ln(2) − ln(4) + ln(3) = ln(3).
Solution. Notice
1 1
0≤ 2𝑥
≤ 2𝑥 = 𝑒−2𝑥 ,
𝑥+𝑒 𝑒
and 𝑏
∞
𝑒−2
∫
−2𝑥 1 −2𝑥
𝑒 𝑑𝑥 = lim − 𝑒 = < ∞.
1 𝑏→∞ 2
1 2
∫∞ 𝑑𝑥
Then by the Comparison Test, 1 𝑥+𝑒2𝑥 is convergent.
√ 1.4. Find the volume of the solid obtained by rotating the region bounded by the curves 𝑦 = 𝑥 and
Example
𝑦 = 𝑥 about the line 𝑥 = 2.
√
Solution. First, we calculate the intersection points. We want 𝑥 = 𝑥, so that 𝑥2 = 𝑥, and so 𝑥 = 0 (𝑦 = 0)or
𝑥 = 1 (𝑦 = 1).
Washer method:
2
WATERLOO SOS E XAM -AID: MATH138 F INAL
The outer function is 𝑥 = 𝑦 2 and the inner function is 𝑥 = 𝑦. However, the outer radius is given by 2 − 𝑦 2
and the inner radius is given by 2 − 𝑦. Then
∫ 1 ∫ 1
𝑉 =𝜋 (2 − 𝑦 2 )2 − (2 − 𝑦)2 𝑑𝑦 = 𝜋 4 − 4𝑦 2 + 𝑦 4 − 4 + 4𝑦 − 𝑦 2 𝑑𝑦
0 0
∫ 1 ()1
2 4 5 3 1 5 2
=𝜋 4𝑦 − 5𝑦 + 𝑦 𝑑𝑦 = 𝜋 2𝑦 − 𝑦 + 𝑦
0 3 5 0
( ) ( )
5 1 1 1 8𝜋
=𝜋 2− + =𝜋 + = .
3 5 3 5 15
Shell method: √
The top function is 𝑦 = 𝑥 and the bottom function is 𝑦 = 𝑥. However, the radius is given by 2 − 𝑥.
Then
∫ 1 ∫ 1
√
𝑉 = 2𝜋 (2 − 𝑥)( 𝑥 − 𝑥) 𝑑𝑥 = 2𝜋 2𝑥1/2 − 2𝑥 − 𝑥3/2 + 𝑥2 𝑑𝑥
0 0
( )1
4 3/2 2 1
𝑥 − 𝑥2 − 𝑥5/2 + 𝑥3
= 2𝜋
3 5 3 0
( ) ( )
4 2 1 5 7 8𝜋
= 2𝜋 −1− + = 2𝜋 − = .
3 5 3 3 5 15
Example 1.5. A tank contains 120 litres of water in which 200 grams of salt is dissolved. Brine containing 2
grams of salt per litre is then pumped into the tank at a rate of 3 litres per minute; the well-mixed solution
is pumped out at the same rate. Find the concentration of salt in the tank at time 𝑡.
Solution. The change in mass 𝑚(𝑡) is given by the difference of the rate of inflow and the rate of outflow.
The rate is given by the product of the flow rate and the concentration, so that
𝑚(𝑡) 1
rate in = (2)(3) = 6 and rate out = (3) = 𝑚.
120 40
Then we have the differential equation
𝑑𝑚 1
= 6 − 𝑚.
𝑑𝑡 40
We can solve this in two ways, as a separable or linear differential equation. For the former, we rearrange
as
𝑑𝑚 1
= − (𝑚 − 240),
𝑑𝑡 40
then
∫ ∫
𝑑𝑚 𝑑𝑡
=−
𝑚 − 240 40
1
ln ∣𝑚 − 240∣ = − 𝑡 + 𝐶
40
1
𝑚 − 240 = 𝐴𝑒− 40 𝑡
1
𝑚 = 𝐴𝑒− 40 𝑡 + 240.
3
WATERLOO SOS E XAM -AID: MATH138 F INAL
The initial conditions give 200 = 𝑚(0) = 𝐴 + 240 for separable equations, so that 𝐴 = −40 and
1
𝑚(𝑡) = 240 − 40𝑒− 40 𝑡 .
𝑑 𝑑𝑚 1
(𝑚 − 240) = = − (𝑚 − 240).
𝑑𝑡 𝑑𝑡 40
This is the exponential model, which has solution
1
𝑚 − 240 = 𝐴𝑒− 40 𝑡 ,
and we get the same answer as before after applying the initial condition.
Finally, we can treat the differential equation as a linear differential equation, writing
𝑑𝑚 1
+ 𝑚 = 6.
𝑑𝑡 40
The integrating factor is
1 1
∫
𝑑𝑡
𝐼(𝑡) = 𝑒 40 = 𝑒 40 𝑡 ,
so that
( )
1 1 1 1
(𝑒 40 𝑡 𝑚)′ = 𝑒 40 𝑡 𝑚′ + 𝑚 = 𝑒 40 𝑡 (6)
40
∫
1 1 1
𝑒 40 𝑡 𝑚 = 6𝑒 40 𝑡 𝑑𝑡 = 240𝑒 40 𝑡 + 𝐶
1
𝑚 = 240 + 𝐶𝑒− 40 𝑡 ,
and we get the same answer as before after applying the initial condition.
Example 1.6. Determine whether the following sequence is convergent or divergent. If it is convergent, find
its limit.
𝑎𝑛 = ln(2𝑛 + 3𝑛 ) − 𝑛 ln(3).
(Winter 2006, Q1)
𝑎𝑛 = ln(𝑏𝑛 + 𝑐𝑛 ) − 𝑛 ln(𝑐),
On one hand,
ln(𝑏𝑛 + 𝑐𝑛 ) − 𝑛 ln(𝑐) ≥ ln(𝑐𝑛 ) − 𝑛 ln(𝑐) = 𝑛 ln(𝑐) − 𝑛 ln(𝑐) = 0,
4
WATERLOO SOS E XAM -AID: MATH138 M IDTERM
9 Limits of Sequences (11.1)
D EFINITION 9.1. A sequence is an ordered list of objects (numbers in our case), which we may denote as
𝑎1 , 𝑎2 , 𝑎3 , . . . , 𝑎𝑛 , . . . or {𝑎𝑛 }∞
𝑛=1 or {𝑎𝑛 }.
D EFINITION 9.2. A sequence {𝑎𝑛 } has the limit 𝐿 if for every 𝜀 > 0 there exists an integer 𝑁 such that if
𝑛 > 𝑁 , then
∣𝑎𝑛 − 𝐿∣ < 𝜀.
In this case we write
lim 𝑎𝑛 = 𝐿 or 𝑎𝑛 → 𝐿 as 𝑛 → ∞,
𝑛→∞
and say {𝑎𝑛 } is convergent. Otherwise, we say that the sequence {𝑎𝑛 } is divergent. A special case of this is if
for every 𝑀 > 0, there exists an integer 𝑁 such that if 𝑛 > 𝑁 , then
𝑎𝑛 > 𝑀.
T IP. Here is a useful sequence to remember: {𝑟𝑛 } is convergent if −1 < 𝑟 ≤ 1, divergent for 𝑟 ≤ −1, and
divergent to infinity for 𝑟 > 1. Moreover, lim𝑛→∞ 𝑟𝑛 = 0 if −1 < 𝑟 < 1 and lim𝑛→∞ 𝑟𝑛 = 1 if 𝑟 = 1.
Limits of sequences behave nicely:
T HEOREM 9.5. If {𝑎𝑛 } and {𝑏𝑛 } are convergent sequences and 𝑐 is a constant, then
(i)
lim (𝑎𝑛 + 𝑏𝑛 ) = lim 𝑎𝑛 + lim 𝑏𝑛
𝑛→∞ 𝑛→∞ 𝑛→∞
(ii)
lim (𝑎𝑛 − 𝑏𝑛 ) = lim 𝑎𝑛 − lim 𝑏𝑛
𝑛→∞ 𝑛→∞ 𝑛→∞
34
WATERLOO SOS E XAM -AID: MATH138 M IDTERM
(iii)
lim 𝑐𝑎𝑛 = 𝑐 lim 𝑎𝑛
𝑛→∞ 𝑛→∞
(iv)
lim 𝑐 = 𝑐
𝑛→∞
(v)
lim (𝑎𝑛 𝑏𝑛 ) = lim 𝑎𝑛 ⋅ lim 𝑏𝑛
𝑛→∞ 𝑛→∞ 𝑛→∞
(vi)
𝑎𝑛 lim𝑛→∞ 𝑎𝑛
lim = if lim 𝑏𝑛 ∕= 0
𝑛→∞ 𝑏𝑛 lim𝑛→∞ 𝑏𝑛 𝑛→∞
(vii) [ ]𝑝
lim 𝑎𝑝𝑛 = lim 𝑎𝑛 if 𝑝 > 0 and 𝑎𝑛 > 0
𝑛→∞ 𝑛→∞
The following is a nice result that allows us to find limits without dealing with the limiting behaviour
of the sequence directly.
T HEOREM 9.6 [S QUEEZE T HEOREM ]. If 𝑎𝑛 ≤ 𝑏𝑛 ≤ 𝑐𝑛 for 𝑛 ≥ 𝑛0 and lim 𝑎𝑛 = 𝐿 = lim 𝑐𝑛 , then lim 𝑏𝑛 = 𝐿.
𝑛→∞ 𝑛→∞ 𝑛→∞
T HEOREM 9.7.
lim ∣𝑎𝑛 ∣ = 0 =⇒ lim 𝑎𝑛 = 0.
𝑛→∞ 𝑛→∞
E XAMPLE 21. Determine whether the following sequence is convergent or divergent. If it is convergent,
find its limit.
sin( 𝑛𝜋 )
𝑎𝑛 = .
𝑛
S OLUTION. Observe that ∣ sin( 𝑛𝜋 )∣ ≤ 1, so that ∣𝑎𝑛 ∣ ≤ 𝑛1 which limits to 0 by a previous observation, and so
by a theorem,
lim 𝑎𝑛 = 0.
𝑛→∞
E XAMPLE 22. Determine whether the following sequence is convergent or divergent. If it is convergent,
find its limit. √
𝑎𝑛 = 𝑛 2𝑛 + 3𝑛 .
35
WATERLOO SOS E XAM -AID: MATH138 M IDTERM
First, consider √
𝑛
1 + ( 𝑐𝑏 )𝑛 .
Then
√ √
1 + ( 𝑐𝑏 )𝑛 ≤ 1 + ( 𝑐𝑏 )𝑛 ,
𝑛 𝑛
1= 1≤
the first inequality following since 𝑐𝑏 > 0 and the second following since the function 𝑥𝑛 is increasing for
𝑥 ≥ 0. Since 𝑏 < 𝑐, we have 0 < 𝑟 = 𝑐𝑏 < 1 and so
We also have
lim 1 = 1,
𝑛→∞
Now, √
𝑛
lim 𝑐𝑛 = 𝑐,
𝑛→∞
so by limit arithmetic,
√ √ √ √
lim 𝑎𝑛 = lim 𝑛
𝑐𝑛 𝑛
1 + ( 𝑐𝑏 )𝑛 = lim 𝑛
𝑐𝑛 ⋅ lim 𝑛
1 + ( 𝑐𝑏 )𝑛 = 𝑐.
𝑛→∞ 𝑛→∞ 𝑛→∞ 𝑛→∞
E XERCISE 13. Determine whether the following sequence is convergent or divergent. If it is convergent,
find its limit.
𝑛
𝑎𝑛 = .
log 𝑛
S OLUTION. Notice both the numerator and the denominator approach infinity as 𝑛 → ∞. We apply lHos-
pital’s Rule to the related function 𝑥/ ln 𝑥 and obtain
𝑥 1
lim = lim = ∞.
𝑥→∞ ln 𝑥 𝑥→∞ 1/𝑥
Thus by a theorem,
lim 𝑎𝑛 = ∞,
𝑛→∞
36
WATERLOO SOS E XAM -AID: MATH138 F INAL
using properties of logarithm. On the other hand,
By continuity of logarithm,
( )
lim ln(1 + ( 𝑐𝑏 )𝑛 ) = ln lim 1 + ( 𝑐𝑏 )𝑛 = ln(1) = 0.
𝑛→∞ 𝑛→∞
ln(2𝑛 + 3𝑛 ) − 𝑛 ln(3) = 0.
Definition 2.2. A sequence {𝑎𝑛 } is bounded above if there is a number 𝑀 such that
𝑎𝑛 ≤ 𝑀
𝑚 ≤ 𝑎𝑛
Theorem 2.3 [M ONOTONE C ONVERGENCE T HEOREM ]. Every bounded, monotonic sequence is convergent.
5
WATERLOO SOS E XAM -AID: MATH138 F INAL
Note that this theorem is not true in the converse: certainly there are sequences which are not monotonic
that are convergent, for example 𝑎𝑛 = 𝑛1 (−1)𝑛 . However, boundedness is always a necessary condition for
the limit to exist. (Indeed, suppose the limit of a sequence {𝑎𝑛 } is 𝐿. Then for 𝜀 = 1 > 0, there is an 𝑁 such
that if 𝑛 > 𝑁 , then ∣𝑎𝑛 − 𝐿∣ < 1. That is, ∣𝑎𝑛 ∣ < 𝐿 + 1 for 𝑛 > 𝑁 . Then as there are finitely many terms we
have neglected, consider 𝑀 = max{𝑎1 , . . . , 𝑎𝑁 }, and we will have ∣𝑎𝑛 ∣ < 𝑀 + 𝐿 + 1 for every 𝑛.)
Example 2.4. Determine whether the sequence defined as follows converges or diverges. If it converges,
find its limit. √
𝑎1 = 1, 𝑎𝑛+1 = 6 + 𝑎𝑛 for 𝑛 ≥ 1.
(Spring 2002, Q3) and (Spring 2001, Q2 - modified), (Spring 2000, Q6 - modified)
√ √ √ √ √
Solution. Making some rough estimates, 𝑎2 = 6 + 1 = 7 ≈ 2.6, and 𝑎3 = 6 + 7 ≈ 8.6 ≈ 2.9. We
can see that the sequence seems to be increasing
√ to 3. In fact, if the limit does exist, say lim𝑛→∞ 𝑎𝑛 = 𝐿,
then by limit arithmetic, we have that 𝐿 = 6 + 𝐿. That is, 𝐿2 − 𝐿 − 6 = 0, and factoring, this gives
(𝐿 + 2)(𝐿 − 3) = 0, so that 𝐿 = −2 or 𝐿 = 3. Then a limit of 3 seems like a likely candidate, since we know
the sequence begins above 0 and tends to increase. However, this is not enough for a solution! We have
not yet proved the limit exists, nor have we proved that should a limit exist, it must necessarily be positive.
This is important; we can get situations where the “limit” can be calculated much as above, but it does not
even exist.
To confirm our guesses, we use mathematical
√ induction to show 1 ≤ 𝑎𝑛 ≤ 𝑎𝑛+1 ≤ 3 for all 𝑛. The base
case of 𝑛 = 1 is already done, since 1 ≤ 1 ≤ 7 ≤ 3. If we assume this inequality is true for 𝑛 = 𝑘, then
1 ≤ 𝑎𝑘 ≤ 𝑎𝑘+1 ≤ 3
7 ≤ 6 + 𝑎𝑘 ≤ 6 + 𝑎𝑘+1 ≤ 9
√ √ √ √
1 ≤ 7 ≤ 6 + 𝑎𝑘 ≤ 6 + 𝑎𝑘+1 ≤ 9 = 3
1 ≤ 𝑎𝑘+1 ≤ 𝑎𝑘+2 ≤ 3.
In the second and third steps, we have used the fact that addition by a constant and taking the square
root does not affect
√ the inequality, as they are increasing functions. The last step came from our recurrence
relation, 𝑎𝑛+1 = 6 + 𝑎𝑛 . Thus, we have proven the inequality for 𝑛 = 𝑘 + 1, and so the result holds for
all 𝑛 by mathematical induction. This shows that the sequence is increasing and bounded (above), so that
the Monotone Convergence Theorem guarantees it has a limit, and since all the terms are at least 1, we also
know the limit is positive.
So suppose the limit is 𝐿 = lim𝑛→∞ 𝑎𝑛 . The recurrence relation gives
√ √ √
lim 𝑎𝑛+1 = lim 6 + 𝑎𝑛 = 6 + lim 𝑎𝑛 = 6 + 𝐿,
𝑛→∞ 𝑛→∞ 𝑛→∞
by limit arithmetic. Since lim𝑛→∞ 𝑎𝑛 = 𝐿, we also have lim𝑛→∞ 𝑎𝑛+1 = 𝐿. Then solving for 𝐿, we get
𝐿2 − 𝐿 − 6 = 0, or (𝐿 + 2)(𝐿 − 3) = 0, and thus 𝐿 = −2 or 𝐿 = 3. Since 𝐿 is positive, we have 𝐿 = 3, by
uniqueness of the limit.
Aside: The reason we had two possible limits was because we took a square√ when solving for 𝐿, which
ignores sign. In fact, if we had defined the recurrence relation by 𝑎𝑛+1 = − 6 + 𝑎𝑛 instead, the sequence
would still have a limit, and this time it would be −2. However we cannot use the Monotone Convergence
Theorem to solve it, so we won’t try it.
6
WATERLOO SOS E XAM -AID: MATH138 F INAL
Exercise 2.5. Determine whether the sequence defined as follows converges or diverges. If it converges,
find its limit.
𝑎1 = 1, 𝑎𝑛+1 = 1 − 𝑎𝑛 for 𝑛 ≥ 1.
Solution. If we naı̈vely say the limit is given by lim𝑛→∞ 𝑎𝑛 and use it in our recurrence relation by limit
arithmetic, we get 𝐿 = 1 − 𝐿 and hence 𝐿 = 21 . However, the sequence looks like 1, 0, 1, 0, . . ., which does
not have a limit. In this case, notice the sequence is not monotonic, so we cannot apply the Monotone
Convergence Theorem.
Exercise 2.6. Determine whether the sequence defined as follows converges or diverges. If it converges,
find its limit.
𝑎1 = 1, 𝑎𝑛+1 = 1 + 2𝑎𝑛 for 𝑛 ≥ 1.
Solution. If we naı̈vely say the limit is given by lim𝑛→∞ 𝑎𝑛 and use it in our recurrence relation by limit
arithmetic, we get 𝐿 = 1 + 2𝐿 and hence 𝐿 = −1. However, the sequence looks like 1, 2, 3, 4, . . ., which
does not have a limit. In this case, notice the sequence is not bounded, so we cannot apply the Monotone
Convergence Theorem.
3 Series (11.2)
Definition 3.1. A series is the sum of the terms of an infinite sequence {𝑎𝑛 },
𝑎1 + 𝑎2 + 𝑎3 ⋅ ⋅ ⋅ .
It is denoted by
∞
∑ ∑
𝑎𝑛 or 𝑎𝑛 .
𝑛=1
It should be noted that although these definitions start with an initial term of 𝑎1 , we may sometimes begin
with 𝑎0 if the notation is cleaner, for example, the geometric series which will come shortly.
Definition
∑ 3.2. If the sequence {𝑆𝑛 } is convergent, say lim𝑛→∞ 𝑆𝑛 = 𝑆 for a real number 𝑆, then the series
𝑎𝑛 is said to be convergent and we write
∞
∑
𝑎𝑛 = 𝑆.
𝑛=1
7
WATERLOO SOS E XAM -AID: MATH138 F INAL
Tip. Here is a useful series to remember: the geometric series
∞
∑
𝑎𝑟𝑛 = 𝑎 + 𝑎𝑟 + 𝑎𝑟2 + ⋅ ⋅ ⋅
𝑛=0
is divergent.
Some theorems about convergence of series:
∑
Theorem 3.3 [D IVERGENCE T EST ]. If the series 𝑎𝑛 is convergent, then lim𝑛→∞ 𝑎𝑛 = 0.
Notice the converse to this theorem is false, and the harmonic series is a counterexample. This theorem
gives a nice
∑ test for divergence however, since it implies that if lim𝑛→∞ 𝑎𝑛 does not exist or is not 0, then
the series 𝑎𝑛 is divergent, hence the name of the theorem.
Series, like limits of sequences, have the expected behavior when performing arithmetic on them:
∑ ∑ ∑
Theorem 3.4. If
∑ 𝑎𝑛 and 𝑏𝑛 are convergent series, then so are the series 𝑐𝑎𝑛 (where 𝑐 is a constant) and
(𝑎𝑛 + 𝑏𝑛 ). Moreover,
∑ ∑
(i) 𝑐𝑎𝑛 = 𝑐 𝑎𝑛
∑ ∑ ∑
(ii) (𝑎𝑛 + 𝑏𝑛 ) = 𝑎𝑛 + 𝑏𝑛 .
Integral Test:
∫ ∞
1
Tip. Recall the improper integral 𝑑𝑥 is convergent if 𝑝 > 1 and divergent if 𝑝 ≤ 1. Using the Integral
1 𝑥𝑝
∞
∑ 1
Test, the p-series is convergent if 𝑝 > 1 and divergent if 𝑝 ≤ 1. In particular, notice this implies the
𝑛=1
𝑛𝑝
harmonic series diverges.
8
WATERLOO SOS E XAM -AID: MATH138 F INAL
By the method of proof of the Integral Test, we also get an error estimate for certain series.
Theorem 4.2 [I NTEGRAL T EST R∑ EMAINDER E STIMATE ]. Suppose 𝑓 (𝑘) = 𝑎𝑘 , where 𝑓 is a continuous, positive,
decreasing function for 𝑥 ≥ 𝑛 and 𝑎𝑛 is convergent. If the 𝑛th remainder term is given by 𝑅𝑛 = 𝑆 − 𝑆𝑛 , then
∫ ∞ ∫ ∞
𝑓 (𝑥) 𝑑𝑥 ≤ 𝑅𝑛 ≤ 𝑓 (𝑥) 𝑑𝑥.
𝑛+1 𝑛
Comparison Test:
∑ ∑
Theorem 4.3 [C OMPARISON T EST ]. Suppose that 𝑎𝑛 and 𝑏𝑛 are series with positive terms. We have:
∑ ∑
(i) If 𝑎𝑛 ≤ 𝑏𝑛 for all 𝑛 and 𝑏𝑛 is convergent, then 𝑎𝑛 is also convergent.
∑ ∑
(ii) If 𝑎𝑛 ≥ 𝑏𝑛 for all 𝑛 and 𝑏𝑛 is divergent, then 𝑎𝑛 is also divergent.
Tip. Our two important series will be very useful when used in conjunction with the Comparison Test
(recall the geometric series and the 𝑝-series).
We can use the Integral Test combined
∑ with the Comparison Test in order to get an error estimate for
other series. In particular, suppose 𝑏𝑛 is a series which works well with the integral test, and 𝑎𝑛 ≤ 𝑏𝑛 for
every 𝑛. The 𝑛th remainder term is given by 𝑇𝑛 = 𝑏𝑛+1 + 𝑏𝑛+2 + ⋅ ⋅ ⋅ . Since 𝑎𝑛 ≤ 𝑏𝑛 for every 𝑛, we have
𝑅𝑛 ≤ 𝑇𝑛 , so getting a bound on 𝑇𝑛 using something like the Integral Test will give the same bound on 𝑅𝑛 .
∑ ∑
Theorem 4.4 [L IMIT C OMPARISON T EST ]. Suppose that 𝑎𝑛 and 𝑏𝑛 are series with positive terms. If
𝑎𝑛
lim =𝑐
𝑛→∞ 𝑏𝑛
where 𝑐 is a finite number and 𝑐 > 0, then either both series converge or both series diverge.
Tip. This theorem will be extremely useful in testing series whose terms are algebraic functions of 𝑛, so that
they are similar to 𝑝-series.
satisfies
9
WATERLOO SOS E XAM -AID: MATH138 F INAL
1. 𝑏𝑛+1 ≤ 𝑏𝑛 for all 𝑛
2. lim𝑛→∞ 𝑏𝑛 = 0
then the series converges.
Tip. Here is another useful series to remember: the alternating harmonic series
∞
∑ (−1)𝑛−1 1 1
= 1 − + − ⋅⋅⋅
𝑛=1
𝑛 2 3
is convergent.
We also get an estimate for the remainder of alternating series.
Before we begin the other tests of convergence, we should introduce a new form of convergence.
∑ ∑
Definition 4.7. A series 𝑎𝑛 is absolutely convergent if the series of absolute values ∣𝑎𝑛 ∣ is convergent. It
is conditionally convergent if the series is convergent, but not absolutely convergent.
∑
Theorem 4.8. If a series 𝑎𝑛 is absolutely convergent, then it is convergent.
Notice the converse is not true, that is, it is possible to be conditionally convergent (and so our definition
is not wasted). Indeed, the Alternating Series Test shows that the alternating harmonic series is convergent,
while the series of its absolute values is the harmonic series, which is not convergent.
Ratio Test:
𝑎𝑛+1
Theorem 4.9 [R ATIO T EST ]. (i) If lim = 𝐿 < 1, then the series ∑ 𝑎𝑛 is absolutely convergent (and
𝑛→∞ 𝑎𝑛
hence convergent).
𝑎𝑛+1
(ii) If lim = 𝐿 > 1 or lim 𝑎𝑛+1 = ∞, then the series ∑ 𝑎𝑛 is divergent.
𝑛→∞ 𝑎𝑛 𝑛→∞ 𝑎𝑛
𝑎𝑛+1
(iii) If lim = 1, then the Ratio Test is inconclusive, that is, the series ∑ 𝑎𝑛 can be convergent or divergent.
𝑛→∞ 𝑎𝑛
10
WATERLOO SOS E XAM -AID: MATH138 F INAL
The canonical example of (iii) is comparing the 𝑝-series when 𝑝 = 1 and 𝑝 = 2. In both cases, the limit of
the ratio of terms is 1, but the case 𝑝 = 1 is divergent and the case 𝑝 = 2 is convergent.
It should be noted that the case when the limit does not exist will not be covered, but for interest’s sake,
there is a version of the Ratio Test that uses lim sup rather than lim, which always exists (and is a term you
do not need to know).
Tip. If 𝑎𝑛 are rational or algebraic functions of 𝑛, then ∣𝑎𝑛+1 /𝑎𝑛 ∣ → 1, and so the Ratio Test will be incon-
clusive.
Root Test:
√
𝑛
∑
Theorem 4.10 [R OOT T EST ]. (i) If lim ∣𝑎𝑛 ∣ = 𝐿 < 1, then the series 𝑎𝑛 is absolutely convergent (and
𝑛→∞
hence convergent).
√ 𝑎𝑛+1 ∑
(ii) If lim 𝑛 ∣𝑎𝑛 ∣ = 𝐿 > 1 or lim = ∞, then the series 𝑎𝑛 is divergent.
𝑛→∞ 𝑛→∞ 𝑎𝑛
√ ∑
(iii) If lim 𝑛 ∣𝑎𝑛 ∣ = 1, then the Root Test is inconclusive, that is, the series 𝑎𝑛 can be convergent or divergent.
𝑛→∞
Tip. It turns out that if 𝐿 = 1 in the Ratio Test, then 𝐿 = 1 in the Root Test as well. Conversely, if 𝐿 = 1 in
the Root Test, then 𝐿 = 1 in the Ratio Test as well. Thus, if one of these tests gives an inconclusive result
(i.e. 𝐿 = 1), then don’t attempt the other test.
As in the Ratio Test, it should be noted that the case when the limit does not exist will not be covered,
but for interest’s sake, there is a version of the Root Test that uses lim sup rather than lim, which always
exists (and is a term you do not need to know).
4. Series that involve products, such as factorials and raising to the 𝑛th power, may be approachable with
Ratio Test. A notable exception: if 𝑎𝑛 are rational or algebraic functions of 𝑛, then ∣𝑎𝑛+1 /𝑎𝑛 ∣ → 1, and so
the Ratio Test will be inconclusive.
5. If 𝑎𝑛 is of the form (𝑏𝑛 )𝑛 , then the Root Test may yield some useful information.
11
WATERLOO SOS E XAM -AID: MATH138 F INAL
∫∞
6. If 𝑎𝑛 = 𝑓 (𝑛) where 1 𝑓 (𝑥) 𝑑𝑥 is easily evaluated and 𝑓 is a continuous, positive, decreasing function,
then the Integral Test will be of use.
1 1
Solution. Notice lim𝑛→∞ 5+2−𝑛 = 5 ∕= 0, so the Divergence Test tells us that the series diverges.
√
𝑛 1
Solution. Notice lim𝑛→∞ 𝑛3 +1 = lim𝑛→∞ 𝑛5/2 +𝑛 −1/2 = 0, so the Divergence Test tells us nothing. However,
the terms are algebraic functions of 𝑛, so we try the Comparison Test with a 𝑝-series. Notice all of these
terms are positive, and as 𝑛−1/2 > 0, we have
√
𝑛 1 1
= 5/2 ≤ 5/2 .
𝑛3 + 1 𝑛 + 𝑛−1/2 𝑛
∑ 1 5
∑ √𝑛
Now, 𝑛5/2 converges by the 𝑝-series with 𝑝 = 2 > 1, so by the Comparison Test, 𝑛3 +1 converges. To
∑∞ 1
be precise, we only have 𝑛=1 𝑛5/2 converges, whereas our sum starts at 𝑛 = 10. However, when dealing
with series, the first finitely many terms never matter.
Alternatively, we can try the Limit Comparison Test. We have
√
𝑛
𝑛3 +1 𝑛5/2 1
lim 1 = lim −1/2
= lim = 1 > 0.
𝑛→∞ 𝑛→∞ 𝑛 5/2 +𝑛 𝑛→∞ 1 + 𝑛−6/2
𝑛5/2
√
∑ 1 5
∑ 𝑛
Now, 𝑛5/2
converges by the 𝑝-series with 𝑝 = 2 > 1, so by the Limit Comparison Test, 𝑛3 +1 converges.
12
WATERLOO SOS E XAM -AID: MATH138 F INAL
Solution. As in the previous example, terms are algebraic functions of 𝑛, so we try the Comparison Test
with a 𝑝-series. However, the presence of the negative sign will not allow us to compare each term with
1
𝑛5/2
as before. Instead, notice 𝑛−1/2 ≤ 12 𝑛5/2 for 𝑛 ≥ 2. Then
√
𝑛 1 1 1
= 5/2 ≤ 5/2 1 5/2 = 2 5/2 .
𝑛3 − 1 𝑛 − 𝑛−1/2 𝑛 − 2𝑛 𝑛
∑ 1 ∑ √𝑛
Now, 2 𝑛5/2
converges by the 𝑝-series with 𝑝 = 52 > 1, so by the Comparison Test, 𝑛3 +1 converges. To
∑∞ 1
be precise, we only have 𝑛=1 𝑛5/2 converges, whereas our sum starts at 𝑛 = 10. However, when dealing
with series, the first finitely many terms never matter.
Alternatively, we can try the Limit Comparison Test. We have
√
𝑛
𝑛3 +1 𝑛5/2 1
lim 1 = lim = lim = 1 > 0.
𝑛→∞ 𝑛→∞ 𝑛5/2 − 𝑛−1/2 𝑛→∞ 1 − 𝑛−6/2
𝑛5/2
√
∑ 1 5
∑ 𝑛
Now, 𝑛5/2
converges by the 𝑝-series with 𝑝 = 2 > 1, so by the Limit Comparison Test, 𝑛3 −1 converges.
Example 5.4. Determine whether the following series converges absolutely, converges conditionally, or di-
verges.
∞
∑ 1
(−1)𝑛 .
𝑛=2
𝑛 ln 𝑛
1
Solution. It is clear that the terms limit to 0, so the Divergence Test will tell us nothing. The function 𝑥 ln 𝑥
is positive, decreasing, and continuous, so the Integral Test may be applied. Using the substitution 𝑢 = ln 𝑥
or by inspection, ∫ ∞
1 ∞
𝑑𝑥 = [ln(ln 𝑥)]∣2 = lim ln(ln 𝑥) − ln(ln(2)) = ∞,
2 𝑥 ln 𝑥 𝑥→∞
so this series does not converge absolutely. However, it may still be conditionally convergent; the presence
1
of the (−1)𝑛 factor indicates we should try the Alternating Series Test. The terms 𝑛 ln 𝑛 are certainly
∑∞ positive,
1 1
lim𝑛→∞ 𝑛 ln 𝑛 = 0, and the terms are decreasing. Then by the Alternating Series Test, 𝑛=2 (−1)𝑛 𝑛 ln 𝑛
converges. Therefore, this series converges conditionally.
Example 5.5. Determine whether the following series converges absolutely, converges conditionally, or di-
verges.
∞
∑ cos(𝑛𝜋)𝑛2
.
𝑛=1
𝑛!
Solution. The presence of the cos 𝑛𝜋 is simply there to trick you. Notice that for 𝑛 even, cos 𝑛𝜋 = 1 and
for 𝑛 odd, cos 𝑛𝜋 = −1. That is, cos 𝑛𝜋 = (−1)𝑛 , and we have a usual alternating series. It is not trivial to
13
WATERLOO SOS E XAM -AID: MATH138 F INAL
see why the terms decrease to 0, but the fact that we are dealing with products (in the form of a factorial)
implies we might be able to try the Ratio Test.
( )
(𝑛+1)2 2
𝑎𝑛+1
= lim ( 2 ) = lim 𝑛 + 2𝑛 + 1 𝑛!
(𝑛+1)!
lim 𝑛 𝑛→∞ (𝑛 + 1)𝑛! 𝑛2
𝑛→∞ 𝑎𝑛 𝑛→∞ 𝑛!
( )
2 1 1
= lim 1 + + 2 = 0 < 1,
𝑛→∞ 𝑛 𝑛 𝑛+1
so the series converges absolutely. Notice in this case, we do not need to check conditional convergence.
(Had we started with using Alternating Series Test to check for convergence instead of absolute conver-
gence, we would have wasted time since absolute convergence implies convergence.)
Example 5.6. How many terms of the following series must you sum so that the partial sum differs from
the actual sum by no more than 0.01?
∞
∑ 𝑛2
(−1)𝑛 .
𝑛=1
𝑛!
Solution. The previous example shows that this series satisfies the Alternating Series Test, so by the Alter-
nating Series Remainder Estimate Theorem, the error is given by
𝑛2
∣𝑅𝑛 ∣ = ∣𝑆 − 𝑆𝑛 ∣ ≤ 𝑏𝑛+1 = .
𝑛!
1
Then to get an error of at most 0.01, we simply need to find 𝑛 such that 𝑛𝑛!2 ≥ 0.01 = 100. Notice the
(𝑛−1)!
expression on the left simplifies to 𝑛 , which is approximately (𝑛 − 2)!. Then as 5! = 120, we would
expect 𝑛 around 7 to work, and simple calculation gives 𝑛 = 7 works (since 6! = 720 > 100(7)).
𝑥
Solution. We will illustrate several ways of solving this. First, notice ln( 𝑥+1 ) is positive, decreasing, and
continuous, so the integral test applies. We can compute
∫ the integral in two ways: directly by parts, or
using ln(𝑎𝑏) = ln(𝑎) + ln(𝑏) and the known integral ln 𝑥 𝑑𝑥 = 𝑥 ln 𝑥 − 𝑥. For the former method, we try
the “1” trick:
( )
𝑥
𝑢 = ln 𝑑𝑣 = 𝑑𝑥
𝑥+1
𝑥 + 1 (𝑥 + 1) − 𝑥 1
𝑑𝑢 = 𝑑𝑥 = 𝑑𝑥 𝑣=𝑥
𝑥 (𝑥 + 1)2 𝑥(𝑥 + 1)
14
WATERLOO SOS E XAM -AID: MATH138 F INAL
Then
∫ ∞ ( ) ( )∞ ∫ ∞
𝑥 𝑥 1
ln 𝑑𝑥 = 𝑥 ln + 𝑑𝑥
1 𝑥+1 𝑥 + 1 1 1 𝑥 + 1
( )∞
𝑥 + ln(𝑥 + 1)∣∞ ,
= 𝑥 ln 1
𝑥+1 1
and this second term does not converge (this is a little sloppy here, since we should really take a limit
to solve an improper integral). Alternatively, we could manipulate the integrand prior to integrating. To
illustrate, the indefinite integral can be computed using:
∫ ( ) ∫
𝑥
ln 𝑑𝑥 = ln(𝑥) − ln(𝑥 + 1) 𝑑𝑥
𝑥+1
= (𝑥 ln(𝑥) − 𝑥) − ((𝑥 + 1) ln(𝑥 + 1) − (𝑥 + 1)) + 𝐶
= 𝑥(ln(𝑥) − ln(𝑥 + 1)) − ln(𝑥 + 1) + 1 + 𝐶,
which yields the same result as before (the 1 gets absorbed into the constant anyways).
The second method involves noticing that if we exponentiate, the logarithms will disappear and the sum
becomes a product. To make this precise, consider the partial sums 𝑆𝑛 . Since 𝑒𝑥 is a continuous function,
𝑒𝑆𝑛 will converge to the same value as 𝑒lim𝑛→∞ 𝑆𝑛 . Calculating,
( 𝑛 ) 𝑛
∑ ( 𝑖 ) ∏ 𝑖 123 𝑛 1
𝑆𝑛
𝑒 = exp ln = = ⋅⋅⋅ = → 0,
𝑖=1
𝑖 + 1 𝑖=1
𝑖 + 1 2 3 4 𝑛 + 1 𝑛 + 1
∑𝑛 ( )
𝑖
so 𝑖=1 ln 𝑖+1 → −∞.
The easiest method is to use the logarithm property outlined in the second Integral Test method, but
directly on the series. That is, the partial sums 𝑆𝑛 give:
𝑛 ( ) 𝑛
∑ 𝑖 ∑
ln = ln(𝑖) − ln(𝑖 + 1) = ln(1) − ln(2) + ln(2) − ln(3) + ⋅ ⋅ ⋅ − ln(𝑛 + 1)
𝑖=1
𝑖+1 𝑖=1
= − ln(𝑛 + 1) → −∞.
Solution. It is not trivial to see why the terms decrease to 0, but the fact that we are dealing with products
(in the form of a factorial) implies we might be able to try the Ratio Test.
(𝑛+1)! 𝑛
𝑎𝑛+1
= lim (𝑛 + 1)𝑛! 𝑛
(𝑛+1)𝑛+1
lim = lim
𝑛! 𝑛+1
𝑛→∞ 𝑎𝑛 𝑛→∞
𝑛𝑛 𝑛→∞ (𝑛 + 1)
𝑛!
(𝑛 + 1)𝑛𝑛
= lim .
𝑛→∞ (𝑛 + 1)𝑛+1
15
WATERLOO SOS E XAM -AID: MATH138 F INAL
𝑛𝑛
To handle the (𝑛+1)𝑛+1 term, notice binomial expansion gives
(𝑛 + 1)𝑛 𝑛−1
(𝑛 + 1)𝑛+1 = 𝑛𝑛+1 + (𝑛 + 1)𝑛𝑛 + 𝑛 + ⋅ ⋅ ⋅ + (𝑛 + 1)𝑛 + 1 ≥ 𝑛𝑛+1 + (𝑛 + 1)𝑛𝑛 .
2
Then
(𝑛 + 1)𝑛𝑛
𝑎𝑛+1 𝑛+1 1
lim
≤ lim = lim = < 1,
𝑛→∞ 𝑎𝑛 𝑛→∞ 𝑛 𝑛+1 + (𝑛 + 1)𝑛 𝑛 𝑛→∞ 2𝑛 + 1 2
so the series converges absolutely. Notice in this case, we do not need to check conditional convergence.
(Had we started with using Alternating Series Test to check for convergence instead of absolute conver-
gence, we would have wasted time since absolute convergence implies convergence.)
Exercise 5.9. Determine whether the following series converges absolutely, converges conditionally, or di-
verges.
∞
∑ ln 𝑛
(−1)𝑛 .
𝑛=2
𝑛
(Spring 2002, Q3 - modified)
Solution. Let us check if the terms go to 0. Since lim𝑥→∞ ln 𝑥 = ∞ = lim𝑥→∞ 𝑥, l’Hôspital’s Rule gives
ln 𝑥 (1)
lim = lim 𝑥 = 0,
𝑥→∞ 𝑥 𝑥→∞ 1
𝑛
so by a theorem about sequences, lim𝑛→∞ ln𝑛𝑛 = 0, and thus lim𝑛→∞ (−1)𝑛 ln 𝑛 = 0 by another theorem.
Then the Divergence Theorem will be inconclusive, and we should begin checking for both convergence
and absolute convergence. Of course, if we are lucky enough for it to be absolutely convergent, it is auto-
matically convergent by a theorem. For 𝑛 ≥ 3, we have
ln 𝑛 1
≥ ,
𝑛 𝑛
∑∞
and 𝑛=2 𝑛1 diverges (this is the harmonic series starting from the second term). Then by the Comparison
∑∞ ∑∞
Test, 𝑛=2 ln𝑛𝑛 diverges, and hence so does 𝑛=1 ln𝑛𝑛 . Thus, this series does not converge absolutely.
However, it may still be conditionally convergent; the presence of the (−1)𝑛 factor indicates we should try
the Alternating Series Test. The terms ln𝑛𝑛 are certainly positive and lim𝑛→∞ ln𝑛𝑛 = 0 as proved before, so
we need only check that the terms are decreasing, so consider again the function ln𝑥𝑥 . Recall a function is
decreasing if its derivative is negative, so we calculate
𝑑 ln 𝑥 ( 1 )(1) − (ln 𝑥)(1) 1 − ln 𝑥
= 𝑥 = .
𝑑𝑥 𝑥 𝑥2 𝑥2
This derivative is negative when 1 − ln 𝑥 < 0, that is, when 𝑥 > 𝑒. Thus, ln𝑛𝑛 is decreasing for 𝑛 ≥ 3. By the
∑∞ ∑∞
Alternating Series Test, 𝑛=3 (−1)𝑛 ln𝑛𝑛 converges, and hence so does 𝑛=2 (−1)𝑛 ln𝑛𝑛 . Therefore, this series
converges conditionally.
Exercise 5.10. Determine whether the following series converges absolutely, converges conditionally, or
diverges.
∞ ( )𝑛
∑ 3
(−1)𝑛 √ √ .
𝑛=0
2+ 5
16
WATERLOO SOS E XAM -AID: MATH138 F INAL
(Spring 2002, Q3)
√ √ √ √ √ √
Solution. Notice 2 + 5 > 3. Indeed, squaring gives ( 2 + 5)2 = 2 + 2 10 + 5 > 7 + 2 1 = 9. Thus,
3√
this series converges absolutely by the Geometric Series with ∣𝑟∣ = ∣ √2+ 5
∣ < 1.
Exercise 5.11. Determine whether the following series converges absolutely, converges conditionally, or
diverges.
∞
∑ cos 𝑛𝜋
.
𝑛=0
𝑛2/3
(Winter 2002, Q3)
Solution. The presence of the cos 𝑛𝜋 is simply there to trick you. Notice that for 𝑛 even, cos 𝑛𝜋 = 1 and for
𝑛 odd, cos 𝑛𝜋 = −1. That is, cos 𝑛𝜋 = (−1)𝑛 , and we have a usual alternating series. The series does not
converge absolutely, since this is a 𝑝-series with 𝑝 = 23 ≤ 1, but we do have that the terms 𝑛2/3
1
are positive,
1
decreasing, and lim𝑛→∞ 𝑛2/3 = 0. Then by the Alternating Series Test, this series converges conditionally.
Solution. This looks almost like a 𝑝-series, except the exponent of 𝑛 is not a constant. Instead of directly
1
using a 𝑝-series, we might try using the Comparison Test, except that all we can say is that 𝑛1+1/𝑛 ≤ 𝑛1𝑝 for
1 1
𝑝 ≤ 1, or 𝑛1+1/𝑛 ≥ 𝑛𝑝 for 𝑝 > 1, which does not help. We could try using the Limit Comparison Test:
1
lim 𝑛
1 = lim 𝑛1/𝑛 = 1.
𝑛→∞ 𝑛→∞
𝑛1+1/𝑛
This limit might not be familiar to you, so we will show this (this question was also on a past Math 138
exam). First, continuity of the exponential gives
( )
1
lim 𝑥1/𝑥 = lim exp(ln(𝑥1/𝑥 )) = exp lim ln 𝑥 .
𝑥→∞ 𝑥→∞ 𝑥→∞ 𝑥
1
Thus, lim𝑥→∞ 𝑥1/𝑥 = 𝑒0 = 1, and so the corresponding sequence
∑
∑∞result1holds as well. Now, 𝑛 diverges
(this is the harmonic series), so by the Limit Comparison Test, 𝑛=1 𝑛1+1/𝑛 diverges.
17
WATERLOO SOS E XAM -AID: MATH138 F INAL
6 Power Series (11.8)
Definition 6.1. A power series is a series of the form
∞
∑
𝑐𝑛 𝑥𝑛 = 𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥2 + ⋅ ⋅ ⋅
𝑛=0
where 𝑥 is a variable and 𝑐𝑛 are constants called the coefficients of the series. For each fixed 𝑥, this power
series is a series of constants which we can test for convergence.
More generally, a series of the form
∞
∑
𝑐𝑛 (𝑥 − 𝑎)𝑛 = 𝑐0 + 𝑐1 (𝑥 − 𝑎) + 𝑐2 (𝑥 − 𝑎)2 + ⋅ ⋅ ⋅
𝑛=0
∑∞
Theorem 6.2. For a power series 𝑛=0 𝑐𝑛 (𝑥 − 𝑎)𝑛 , there are only three possibilities:
(i) The series converges only when 𝑥 = 𝑎.
(ii) The series converges for all 𝑥.
(iii) There is a positive number 𝑅 such that the series converges if ∣𝑥 − 𝑎∣ < 𝑅 and diverges if ∣𝑥 − 𝑎∣ > 𝑅.
Definition 6.3. The number 𝑅 in (iii) is called the radius of convergence. By convention, 𝑅 = 0 in (i) and
𝑅 = ∞ in (ii). The interval of convergence of a power series is the interval consisting of all points 𝑥 for which
the series converges. In (i), this interval is the singleton {𝑎}. In (ii), the interval is (−∞, ∞). Case (iii) is more
complicated, as convergence may or may not hold at the endpoints 𝑎 ± 𝑅. You must check the endpoints
separately, and your final interval of convergence will be one of four possibilities:
(𝑎 − 𝑅, 𝑎 + 𝑅), [𝑎 − 𝑅, 𝑎 + 𝑅), (𝑎 − 𝑅, 𝑎 + 𝑅], [𝑎 − 𝑅, 𝑎 + 𝑅].
To determine radii and intervals of convergence, you should appeal to the Ratio or Root Tests to get a
radius of convergence, then test the endpoints (if necessary) using other methods.
Example 6.4. Find the radius and interval of convergence of the following series.
∞
∑ 𝑥𝑛
.
𝑛=1
𝑛!
By the Ratio Test, the series converges if 0 < 1, and diverges if 0 > 1. That is, we have convergence for all
𝑥, so that the radius of convergence is 𝑅 = ∞ and the interval of convergence is (−∞, ∞).
18
WATERLOO SOS E XAM -AID: MATH138 F INAL
(1 + 𝑛1 )3 1
= lim ∣𝑥∣ = ∣𝑥∣.
𝑛→∞ (3 + 3 )(3 + 2 )(3 + 1 ) 27
𝑛 𝑛 𝑛
1 1
By the Ratio Test, the series converges if 27 ∣𝑥∣ < 1, and diverges if 27 ∣𝑥∣ > 1. That is, we have convergence
for ∣𝑥∣ < 27 and divergence for ∣𝑥∣ > 27, so that the radius of convergence is 𝑅 = 27.
Example 6.6. Find the radius and interval of convergence of the following series.
∞
∑ (𝑥 + 3)𝑛
(−1)𝑛 .
𝑛=2
𝑛2𝑛
By the Ratio Test, the series converges if 12 ∣𝑥 + 3∣ < 1, and diverges if 21 ∣𝑥 + 3∣ > 1. That is, we have
convergence for ∣𝑥 + 3∣ < 2 and divergence for ∣𝑥 + 3∣ > 2, so that the radius of convergence is 𝑅 = 2.
To find the interval of convergence, we must check what happens at the endpoints 𝑥 = −5 and 𝑥 = −1.
At 𝑥 = −5, the series becomes
∞ ∞
∑ (−2)𝑛 ∑ 1
(−1)𝑛 𝑛
= ,
𝑛=2
𝑛2 𝑛=2
𝑛
which diverges (the harmonic series). At 𝑥 = −1, the series becomes
∞ ∞
∑ 2𝑛 ∑ 1
(−1)𝑛 𝑛
= (−1)𝑛 ,
𝑛=2
𝑛2 𝑛=2
𝑛
19
WATERLOO SOS E XAM -AID: MATH138 F INAL
Exercise 6.7. Find the radius and interval of convergence of the following series.
∞
∑ (𝑥 + 3)𝑛+1
(−2)𝑛 .
𝑛=1
𝑛3/2 + 𝑛1/2
By the Ratio Test, the series converges if 2∣𝑥 + 3∣ < 1, and diverges if 2∣𝑥 + 3∣ > 1. That is, we have
convergence for ∣𝑥 + 3∣ < 12 and divergence for ∣𝑥 + 3∣ > 12 , so that the radius of convergence is 𝑅 = 12 .
To find the interval of convergence, we must check what happens at the endpoints 𝑥 = − 27 and 𝑥 = − 25 .
At 𝑥 = − 72 , the series becomes
∞ ∞
∑ (− 21 )𝑛 ∑ 1
(−2)𝑛 3/2 1/2
= ,
𝑛=1
𝑛 +𝑛 𝑛=1
𝑛 + 𝑛1/2
3/2
3
which converges upon using the Limit Comparison Test with the 𝑝-series where 𝑝 = 2. At 𝑥 = − 52 , the
series becomes
∞ ∞
∑
𝑛 ( 21 )𝑛 ∑ 1
(−2) 3/2 1/2
= (−1)𝑛 3/2 ,
𝑛=1
𝑛 + 𝑛 𝑛=1
𝑛 + 𝑛1/2
which converges since the previous case shows this series converges absolutely.
Thus, the interval of convergence is [− 27 , − 52 ].
Exercise 6.8. Find the radius and interval of convergence of the following series.
∞
∑ (𝑥 − 1)𝑛
√ .
𝑛=1
3𝑛 𝑛
By the Ratio Test, the series converges if 13 ∣𝑥 − 1∣ < 1, and diverges if 31 ∣𝑥 − 1∣ > 1. That is, we have
convergence for ∣𝑥 − 1∣ < 3 and divergence for ∣𝑥 − 1∣ > 3, so that the radius of convergence is 𝑅 = 3.
To find the interval of convergence, we must check what happens at the endpoints 𝑥 = −2 and 𝑥 = 4.
At 𝑥 = −2, the series becomes
∞ ∞
∑ (−3)𝑛 ∑ 1
𝑛
√ = (−1)𝑛 √ ,
𝑛=1
3 𝑛 𝑛=1
𝑛
20
WATERLOO SOS E XAM -AID: MATH138 F INAL
which converges upon using the Alternating Series Test. At 𝑥 = 4, the series becomes
∞ ∞
∑ 3𝑛 ∑ 1
√
𝑛 𝑛
= √ ,
𝑛=1
3 𝑛=1
𝑛
21
WATERLOO SOS E XAM -AID: MATH138 F INAL
for ∣𝑥 − 𝑎∣ < 𝑅, then its coefficients are given by
𝑓 (𝑛) (𝑎)
𝑐𝑛 = .
𝑛!
Then this series
∞
∑ 𝑓 (𝑛) (𝑎)
𝑓 (𝑥) = (𝑥 − 𝑎)𝑛
𝑛=0
𝑛!
is called the Taylor series of f at a (or about a or centered at a). For the special case 𝑎 = 0, the series is known as a
Maclaurin series.
To go back to the first question as to whether a function has a power series representation, we use the
following two theorems, which relies on the following definition.
Theorem 7.4. If 𝑓 (𝑥) = 𝑇𝑛 (𝑥) + 𝑅𝑛 (𝑥) where 𝑇𝑛 is the 𝑛th-degree Taylor polynomial of 𝑓 at 𝑎 and lim 𝑅𝑛 (𝑥) = 0
𝑛→∞
for ∣𝑥 − 𝑎∣ < 𝑅, then 𝑓 is equal to the sum of its Taylor series for ∣𝑥 − 𝑎∣ < 𝑅.
Theorem 7.5 [TAYLOR ’ S I NEQUALITY ]. If ∣𝑓 (𝑛+1) (𝑥)∣ ≤ 𝑀 for ∣𝑥 − 𝑎∣ ≤ 𝑑, then the remainder 𝑅𝑛 (𝑥) of the
Taylor series satisfies the inequality
𝑀
∣𝑅𝑛 (𝑥)∣ ≤ ∣𝑥 − 𝑎∣𝑛+1
(𝑛 + 1)!
for ∣𝑥 − 𝑎∣ ≤ 𝑑.
Using these, we can develop some useful Maclaurin series.
∞
1 ∑
= 𝑥𝑛 𝑅=1
1 − 𝑥 𝑛=0
∞
∑ 𝑥𝑛
𝑒𝑥 = 𝑅=∞
𝑛=0
𝑛!
∞
∑ 𝑥2𝑛+1
sin 𝑥 = (−1)𝑛 𝑅=∞
𝑛=0
(2𝑛 + 1)!
∞
∑ 𝑥2𝑛
cos 𝑥 = (−1)𝑛 𝑅=∞
𝑛=0
(2𝑛)!
∞
∑ 𝑥2𝑛+1
tan−1 𝑥 = (−1)𝑛 𝑅=1
𝑛=0
2𝑛 + 1
∞
∑ 𝑘 ( )
(1 + 𝑥)𝑘 = 𝑥𝑛 𝑅 = 1.
𝑛
𝑛=0
22
WATERLOO SOS E XAM -AID: MATH138 F INAL
Example 7.6. Find the Taylor series for ln(2 + 𝑥2 ) around 𝑎 = 0 and determine its radius of convergence.
To determine 𝐶, we have the condition that at 𝑥 = 0, ln(2 + 𝑥2 ) = ln(2), and the series disappears at 𝑥 = 0,
so 𝐶 = ln(2). Thus, the Taylor series about 𝑎 = 0 is given by
∞
∑ 𝑥2𝑛+2
ln(2 + 𝑥2 ) = ln(2) + (−1)𝑛 .
𝑛=0
2𝑛 (2𝑛
+ 2)
(This may not look like a usual power series due to the constant floating outside of the sum, but think of
it as a part of the 𝑐0 term. That is, 𝑐0 = ln(2) + 21 .) To find its radius of convergence, we could go through
1
the argument of using the Ratio Test again, or remark that the power series representation of 1−𝑢 that we
applied had a radius of convergence of 1, and as we are integrating, the radius of convergence does not
2 √ √
change. That is, we have ∣ 𝑥2 ∣ < 1, so that ∣𝑥∣ < 2. Thus, the radius of convergence is 2.
Example 7.7. If 𝑓 (𝑥) = ln(2 + 𝑥2 ), find the values of 𝑓 19 (0) and 𝑓 20 (0).
𝑐𝑛 𝑥𝑛 , we have that
∑
Solution. Recall that since 𝑓 has a power series about 0, say
𝑓 (𝑛) (0)
𝑐𝑛 = .
𝑛!
Then
𝑓 19 (0) 𝑓 20 (0)
= 𝑐19 = 𝑐20
19! 20!
By the previous example,
∞
∑ 𝑥2𝑛+2
ln(2 + 𝑥2 ) = ln(2) + (−1)𝑛 .
𝑛=0
2𝑛 (2𝑛
+ 2)
Note that 𝑐19 corresponds to the term 𝑥 , not the term 𝑛 = 19. However, the coefficient of the 𝑥19 term is 0,
19
23
WATERLOO SOS E XAM -AID: MATH138 F INAL
2
𝑒−𝑥 𝑑𝑥 and determine its radius of convergence. (Spring 2006,
∫
Example 7.8. Find the Maclaurin series for
Q11)
𝑢𝑛
∑∞
Solution. The power series representation of 𝑒𝑢 is given by 𝑛=0 𝑛! with convergence everywhere, so that
∞ ∞
−𝑥2
∑ (−𝑥2 )𝑛 ∑ 𝑥2𝑛
𝑒 = = (−1)𝑛 .
𝑛=0
𝑛! 𝑛=0
𝑛!
Then
∞ ∫ ∞
𝑥2𝑛 𝑥2𝑛+1
∫
−𝑥2
∑ ∑
𝑒 𝑑𝑥 = (−1)𝑛 𝑑𝑥 = (−1)𝑛 + 𝐶,
𝑛=0
𝑛! 𝑛=0
𝑛!(2𝑛 + 1)
for a constant 𝐶. Integrating does not change the radius of convergence, so this series converges for −𝑥2 ∈
ℝ, that is, for 𝑥 ∈ ℝ.
∫1 2
Example 7.9. Write the number 0
𝑒−𝑥 𝑑𝑥 as the sum of a convergent series. (Spring 2006, Q11)
2
Example 7.10. Let 𝑇9 (𝑥) be the 9th-degree Taylor polynomial of 𝑒−𝑥 𝑑𝑥 at 𝑥 = 0. Estimate the bound
∫
∫1 2
involved in approximating 0 𝑒−𝑥 𝑑𝑥 by 𝑇9 (1). (Spring 2006, Q11)
2
Solution. We may be tempted to use Taylor’s inequality, but the derivatives of 𝑒−𝑥 𝑑𝑥 are not well-
∫
bounded. Instead, notice that the previous example showed that this number has a power series
∫ 1 ∞
2 ∑ 1
𝑒−𝑥 𝑑𝑥 = (−1)𝑛 .
0 𝑛=0
𝑛!(2𝑛 + 1)
which satisfies the alternating series test, and so by the remainder estimate, the error is bounded by:
1 1
∣𝑅9 (1)∣ ≤ 𝑏10 = = .
10!(2(10) + 1) 10! ⋅ 21
(. . .which is about 1.3 × 10−8 , but on the exam the above representation should suffice.)
24
WATERLOO SOS E XAM -AID: MATH138 F INAL
Example 7.11. Find the Taylor series for 𝑒3𝑥+1 and state its radius of convergence.
∑ 𝑢𝑛We have not specified the center of the series here, so we will consider two possibilities. Since
Solution.
𝑒𝑢 = 𝑛! , we can use 𝑢 = 3𝑥 + 1 directly to get
∞ ∞
∑ (3𝑥 + 1)𝑛 ∑ 3𝑛
𝑒3𝑥+1 = = (𝑥 + 13 )𝑛 .
𝑛=0
𝑛! 𝑛=0
𝑛!
Exercise 7.12. Find the Maclaurin series for 𝑥3 cos 𝑥4 and state its radius of convergence. (Spring 2002, Q5)
Exercise 7.13. If 𝑓 (𝑥) = 𝑥3 cos 𝑥4 and 𝑇11 is the 11th-degree Taylor polynomial, find a constant 𝐶 such that
25
WATERLOO SOS E XAM -AID: MATH138 F INAL
remainder estimate for the Alternating Series Test, we have
(𝑥8 )2 1 1
∣𝑓 (𝑥) − 𝑇11 (𝑥)∣ = ∣𝑥3 𝑅1 (𝑥)∣ ≤ 𝑥3 𝑏2 = 𝑥3 = 𝑥 9,
(2 ⋅ 2)! 24
1
so 𝐶 = 24 works.
∫1 1
Exercise 7.14. Approximate the integral 0
𝑥3 cos 𝑥4 𝑑𝑥 with error less than 480 . (Spring 2002, Q5)
Notice that this is an alternating series and satisfies the hypothesis of the remainder estimate (clearly the
terms are positive and decrease to 0). Then the error when using the 𝑛th-degree Taylor polynomial is given
by
1
∣𝑅𝑛 ∣ ≤ 𝑏𝑛+1 = .
(2𝑛)!(8𝑛 + 4)
1
If we want the error to be less than 480 , we need 480 ≤ (2𝑛)!(8𝑛 + 4), and we notice that 𝑛 = 2 gives
(2𝑛)!(8𝑛 + 4) = 480 precisely. Then we require the 2nd-degree Taylor polynomial to approximate the
1
integral with error less than 480 , that is,
2
∑ 1 1 1 1 1 1 1 101
(−1)𝑛 = − + = − + = .
𝑛=0
(2𝑛)!(8𝑛 + 4) 4 2 ⋅ 12 4! ⋅ 20 4 24 480 480
Example 8.2. A particle travels on the path (3 sin 𝑡, 2 cos 𝑡) for 0 ≤ 𝑡 ≤ 2𝜋. Sketch this curve, indicating the
direction of motion, and points 𝑡 = 0, 𝜋2 .
(Winter 2006, Q2)
26
WATERLOO SOS E XAM -AID: MATH138 F INAL
Solution. Notice ( 𝑥3 )2 + ( 𝑦2 )2 = 1, so this is an ellipse with semi-major axis 3 (along 𝑥-axis) and semi-minor
axis 2 (along 𝑦-axis). The curve begins at (0, 2) and moves in a clockwise direction since sin begins to in-
crease and cos begins to decrease. The points 𝑡 = 0, 𝜋2 correspond to the points (0, 2) and (3, 0) respectively.
In the last example, we were fortunate enough to have a familiar identity, but this will not always be the
case. To aid in curve sketching, we consider tangent lines. The slope of a tangent to a curve in ℝ2 is given
by
𝑑𝑦
𝑑𝑦 𝑑𝑡
= 𝑑𝑥 ,
𝑑𝑥 𝑑𝑡
if 𝑑𝑥
𝑑𝑡 ∕= 0, and otherwise the slope is undefined (because the tangent is vertical). In particular, the curve is
horizontal when 𝑑𝑦 𝑑𝑥
𝑑𝑡 = 0 and vertical when 𝑑𝑡 = 0. Combining information about the slopes of the tangents
with other information such as intercepts makes sketching curves much easier.
Example 8.3. Consider the curve given by the parametric equations 𝑥(𝑡) = 𝑡(𝑡2 − 3), 𝑦(𝑡) = 3(𝑡2 − 3), for
−2 ≤ 𝑡 ≤ 2. Find the intercepts of the curve and the points where the tangent to the curve is horizontal or
vertical, and include them on a sketch of the curve.
(Winter 2002, Q7)
2
√
√ we are solving 0 = 𝑥(𝑡) = 𝑡(𝑡 − 3), so that 𝑡 = 0 or 𝑡 = ± 3. Solving for
Solution. To find the 𝑦-intercept(s),
𝑦, we have 𝑦(0) = −9 and 𝑦(± 3) = 0, so the 𝑦-intercepts
√ are (0, 0) and (0, −9). To find the 𝑥-intercept(s),
we are solving 0 = 𝑦(𝑡) = 3𝑡2 − 9, so that 𝑡 = ± 3, and we have already considered this point.
𝑑𝑦 𝑑𝑦
We have 𝑑𝑥 2
𝑑𝑡 = 3𝑡 − 3 and 𝑑𝑡 = 6𝑡. The curve is horizontal when 𝑑𝑡 = 0, that is, when 6𝑡 = 0 and hence
𝑡 = 0. We saw that this was one of the 𝑦-intercepts, at (0, −9). The curve is vertical when 𝑑𝑥 𝑑𝑡 = 0, that is,
when 3𝑡2 − 3 = 0 and hence 𝑡 = ±1. At both of these points, 𝑦 = −6 and 𝑥 = ±2, so the curve is vertical at
(±2, −6).
We should check the slope(s) at the other 𝑦-intercept (which is the 𝑥-intercept,
√ remember) to get more
𝑑𝑦
information about the curve. Then we are considering the slope 𝑑𝑥 at 𝑡 = ± 3. We have
𝑑𝑦
𝑑𝑦 𝑑𝑡 6𝑡
= 𝑑𝑥
= ,
𝑑𝑥 𝑑𝑡
3𝑡2 − 3
so √
𝑑𝑦 √ ±6 3 √
(± 3) = = ± 3.
𝑑𝑥 9−3
27
WATERLOO SOS E XAM -AID: MATH138 F INAL
This tells us that there are two tangents at (0, 0), so the curve intersects itself, with one tangent having
negative the slope of the other.
Finally, we should check the endpoints of the curve given by the bounds on 𝑡. At 𝑡 = ±2, we have
𝑥 = ±2 and 𝑦 = 3. Putting this all together, we can sketch the curve.
Recall the arc-length formula: if the curve is in the form of 𝑦 = 𝐹 (𝑥), for 𝑎 ≤ 𝑥 ≤ 𝑏 (and 𝐹 ′ is continuous),
then the length of the curve is given by
√ ( )2
∫ 𝑏
𝑑𝑦
𝐿= 1+ 𝑑𝑥.
𝑎 𝑑𝑥
Similarly, one can use the substitution rule to obtain the arc-length formula for parametric curves.
Theorem 8.4. Suppose a curve 𝐶 is described by the parametric equations 𝑥 = 𝑓 (𝑡), 𝑦 = 𝑔(𝑡) with 𝛼 ≤ 𝑡 ≤ 𝛽, where
𝑓 ′ and 𝑔 ′ are continuous on [𝛼, 𝛽] and 𝐶 is traversed exactly once as 𝑡 increases from 𝛼 to 𝛽. Then the length of 𝐶 is
given by √
∫ 𝛽 ( )2 ( )2
𝑑𝑥 𝑑𝑦
𝐿= + 𝑑𝑡.
𝛼 𝑑𝑡 𝑑𝑡
Notice this equation reduces to our previous equation if 𝑥 = 𝑡, so that 𝑦 is a function of 𝑥.
3𝜋
Example 8.5. A particle travels on the path (−3 + cos 𝑡, 1 + sin 𝑡) for 0 ≤ 𝑡 ≤ 2 . Find the length of the curve.
(Winter 2006, Q2)
𝑑𝑥 𝑑𝑦
Solution. We have 𝑑𝑡 = − sin 𝑡 and 𝑑𝑡 = cos 𝑡. Then
∫ 3𝜋/2 √
2 2
∫ 3𝜋/2 √ 3𝜋/2 3𝜋
𝐿= (− sin 𝑡) + (cos 𝑡) 𝑑𝑡 = 1 𝑑𝑡 = 𝑡 ∣0 = .
0 0 2
28
WATERLOO SOS E XAM -AID: MATH138 F INAL
Exercise 8.6. Consider the curve given by the parametric equations 𝑥(𝑡) = 𝑡(𝑡2 − 3), 𝑦(𝑡) = 3(𝑡2 − 3), for
−2 ≤ 𝑡 ≤ 2. Find the length of the curve.
(Winter 2002, Q7)
𝑑𝑥 𝑑𝑦
Solution. We have 𝑑𝑡 = 3𝑡2 − 3 and 𝑑𝑡 = 6𝑡. Then
∫ 2 √ ∫ 2 √
2 2
𝐿= (3𝑡2 − 3) + (6𝑡) 𝑑𝑡 = 9𝑡4 − 18𝑡2 + 9 + 36𝑡2 𝑑𝑡
−2 −2
∫ 2 √ ∫ 2 √
= 9(𝑡4 + 2𝑡2 + 1) 𝑑𝑡 = 9(𝑡2 + 1)2 𝑑𝑡
−2 −2
∫ 2
)2
3(𝑡2 + 1) 𝑑𝑡 = 𝑡3 + 3𝑡 −2 = 14 − (−14) = 28.
(
=
−2
29
Proofs
Comparison Test
Geometric Series
Monotone Sequence Theorem
Ratio Test