You are on page 1of 3

Convergence using the definition

10
1. Guess the limit of an = 1 + √
3 n . Prove your guess.
√ 10
Proof Since 3 n approaches infinity with n, we guess √
3n approaches 0, and so an approaches
1. Let’s prove our guess
In order to show a0 converges to 1, we the estimate the distance from an to 1 and show it is
eventually arbitrarily small, using the definition of convergence.

an −→ a iff (∀ϵ > 0)(∃N )(n > N =⇒ |an − a| < ϵ

Start with the distance from an to the limit 1


10
|an − 1| = |1 + √
3
− 1|
n
10
= √3
n
10
Working backwards, we need √
3n < ϵ, for a pre-fixed ϵ. Solving the inequality for n, we get
 3
10 1000
n> ϵ = ϵ3
.
ϵ3
Choose N > 1000
ϵ3
(this choice implies 1
N < 1000 ). Note that n > N implies 1
n < 1
N and so
1 1
3n < √
√ 3 . For n > N ,
N

10
|an − 1| = |1 + √3
− 1|
n
10
= √3
n
10
< √3
N
10
< q 3 ( why )?
3 ϵ
1000
10
= ϵ
10
= ϵ
1000
We have shown that, for a given ϵ > 0, {an } is within ϵ of 1 when n > ϵ3
.
For instance, if we wish for an to be within ϵ = 0.001 of 1, we choose n > 1000/0.0013 =
1000/0.000000001 = 1012 .
Consider the two sequences an = 1/n and bn = 1/n3 . They both converge to zero. Now the
definition of convergence

(∀ϵ > 0)(∃N )(n > N =⇒ |xn − a| < ϵ

says that, roughly speaking, an and bn are eventually within ϵ of zero, for any given ϵ > 0.
For instance,

ˆ For ϵ = 0.01, bn = 1/n < 0.001, when n > 10.


ˆ What about an , which converges to 0 too? Well, for an to be less than 0.001, n needs to
be larger than 1000 (n > 100).
ˆ For an and bn to be simultaneously less than 0.001, n has to be larger than 10 and larger
than 1000. Thus for this to happen we choose N = max(10, 1000) = 1000 and when
n > N = 1000, both sequences are within the desired distance from 0.

We will keep this simple example in mind when proving the following.

2. Show that an −→ a and bn −→ b =⇒ (an + bn ) −→ a + b.


Proof Fix ϵ > 0. By definition of convergence

ˆ There exists N1 such that n > N1 =⇒ |an − a| < ϵ/2


ˆ There exists N2 such that n > N2 =⇒ |bn − b| < ϵ/2
Notice that: i) we chose two different N1 and N2 , just as we saw in the example for
1/n and 1/n3 above.
ii) The choice ϵ/2 instead of ϵ will become clear below.
ˆ Let N = max(N1 , N2 ). Now for n larger than N , both sequences are within ϵ of their
respective limits. Thus n > N =⇒

|(an + bn ) − (a + b)| = |an − a + bn − b|


< |an − a| + |bn − b| (by the triangular inequality)
< ϵ/2 + ϵ/2
= ϵ

Here is a careless version of proof writing By definition,

ˆ There exists N such that n > N =⇒ |an − a| < ϵ/2


ˆ There exists N such that n > N =⇒ |bn − b| < ϵ/2
ˆ |an + bn − (a + b)| < |an − a| + |bn − b| < ϵ

Comments

ˆ Is N the same in both cases?


ˆ Is N the same for the sequence (an + bn ) too?

In the proof we gave for 1/ 3 n −→ 0, it was easy to solve for n in terms of ϵ when figuring
out at which point the sequence becomes within ϵ of 0. In general, it might be a difficult to
solve for n. The next examples illustrates this case.

3. Show an = √ 1 −→ 0
n2 −4

Proof Fix ϵ > 0 and write down the distance from an to its limit 0: |an − 0| = √n12 −4 .


ˆ We need to figure out for what value of n is √n12 −4 < ϵ.


ˆ Instead of solving for n in terms of ϵ, we will use an estimate on n2 − 4 to make the
problem easier. Recall that for A > B > 0 =⇒ 1/A < 1/B
ˆ It would’ve been nice if n2 − 4 > n2 , but it is not. However, n2 − 4 > n2 /4, for large
enough n. Thus
√ √
ˆ n − 4 > n2 /4 =⇒ n2 − 4 > n/2 =⇒ 1/ n2 − 4 < 2/n
2

ˆ Now it is easy to solve 2/n < ϵ. In fact, the solution is n > 2/ϵ.
Now we write down the proof.
Fix ϵ > 0. Choose N > 2/ϵ. For n > N > 2/ϵ,

√ 1 1

− 0 = √
2n −4 n2
−4
1
< p
n2 /4
1
=
n/2
2
=
n
ϵ
< 2 why?
2
= ϵ

4. If limn−→∞ an = a, then limn−→∞ sn = limn−→∞ a1 +a2 +an3 +···+an = a (the arithmetic average
of the terms of the sequence approaches the same limit as the sequence itself).
Proof Fix ϵ > 0. Since an −→ a, there exists N1 such that |an − a| < ϵ, for all n > N1 . With
that in mind, let’s estimate the difference between sn and the limit a.
a1 + a2 + a3 + · · · + an a1 + a2 + a3 + · · · + an − na


− a =
n n
a1 − a + a2 − a + a3 − a + · · · + an − a

=
n
|a1 − a| + |a2 − a| + · · · + |aN1 − a| |aN1 +1 − a| + · · · + |an − a|
≤ +
n n
|a1 − a| + |a2 − a| + · · · + |aN1 − a| ϵ + ϵ + · · · + ϵ
≤ + (why?)
n n
|a1 − a| + |a2 − a| + · · · + |aN1 − a| | (n − N1 ) ϵ|
≤ + (1)
| n
{z } | n
{z }
= I + II

We need to show that (1) = I + II < ϵ. The numerator c in I is a constant number. It


does not depend on n. Since we have shown that k/n −→ 0 as n −→ ∞ for any constant k,
I −→ 0 as n −→ ∞. Thus there exists N2 such that for n > N2 , I < ϵ.
What about the second term II? Can we make it smaller than ϵ for large enough n? This is
simple. Notice that n − N1 < n and so (n−N
n
1)
< 1. It follows that (n−N
n
1)
ϵ < ϵ.
Thus, for n > max (N1 , N2 ), (1) = I + II < ϵ + ϵ = 2ϵ. Proof is complete.

Note 1 We have shown above that |sn − a| < 2ϵ not < ϵ. Does this matter? Absolutely not.
As long as you can show |sn − s| < kϵ for a constant k (like 2 or 3 or 100, or even a billion),
it is enough. We stress a constant k (independent o n) times ϵ, not a variable n times ϵ.
Note 2 You √ may use a similar argument to show that if limn−→∞ bn = b > 0, then
limn−→∞ n b1 · b2 · b3 · · · bn = b. Just take the log of both sides to convert the problem to
what we have just shown. The difference is, instead of having a1 +a2 +···+a
n
n
, we now have
ln s1 +ln s2 +···+ln sn
n .

You might also like