Limits

We now discuss the "long-term behaviour" of sequences. Consider the sequence

(an)=(a1,a2,...)(a_n)=(a_1, a_2, ...)

We assume that we have a recursive or explicit method to determine each term ana_n of the sequence. We are interested in knowing what happens to ana_n if we make nn larger and larger (hence, when we move to the right in the sequence). We have already discussed some examples at the beginning of the section. Let's define it in more detail.

Definition 1: Convergent sequence

We say that the sequence (an)(a_n) is convergent if the values ana_n get closer and closer (but not necessary touch) a single real number as we make nn larger and larger. This single number is often denoted by aa_\infty, and is called the limit of the sequence. This is then written as follows:

limnan=a\lim_{n\rightarrow \infty} a_n = a_\infty

or as

annaa_n \xrightarrow[]{n\rightarrow \infty} a_\infty

(say: "the limit of the sequence (an)(a_n) is aa_\infty as nn approaches infinity", or "the sequence (an)(a_n) converges to aa_\infty if nn approaches infinity").

Definition 2: Divergent sequence

If the sequence (an)(a_n) does not converge, that is, if it does not approach a certain value, we call it divergent. There are two possible ways a sequence can be divergent.

  • Either the sequence does not approach a single value aa_\infty (e.g. by alternating forever between two values). Thus no (single) limit exists: limnan does not exist\lim_{n\rightarrow \infty} a_n \text{ does not exist}
  • or the sequence increases without upper bound, in which case we write limnan=\lim_{n\rightarrow \infty} a_n = \infty or anna_n \xrightarrow[]{n\rightarrow \infty} \infty (that is, the limit is a=a_\infty = \infty)
  • or the sequence decreases without lower bound: limnan=\lim_{n\rightarrow \infty} a_n = -\infty or anna_n \xrightarrow[]{n\rightarrow \infty} -\infty (that is, the limit is a=a_\infty = -\infty).
Example 1

Again the examples from the beginning of the chapter. Determine whether the sequences are convergent or divergent. If convergent, determine the limit of the sequence.

  1. (an)=(1,3,5,...)(a_n)=(1,3,5,...)
  2. (bn)=(1,0.5,0.25,0.125,...)(b_n)=(1, 0.5, 0.25, 0.125, ...)
  3. (cn)=(1,1,1,1,...)(c_n)=(1,-1,1,-1,...)
  4. (dn)=(1,2,4,8,...)(d_n)=(1,2,4,8,...)

Also, give an example of a sequence that converges to 1.71.7.

Solution
  1. Divergent, with limnan=\lim_{n\rightarrow \infty} a_n = \infty.
  2. Convergent, with limnan=0\lim_{n\rightarrow \infty} a_n = 0.
  3. Divergent, the sequence does not approach a single number.
  4. Divergent, with limnan=\lim_{n\rightarrow \infty} a_n = \infty.

A sequence that converges to 1.71.7 is, for example, an=1.71na_n=1.7-\frac{1}{n} (n=1,2,3,...)(n=1,2,3,...).

As a series (sn)(s_n) is also a sequence, it can also be convergent or divergent. The limit ss_{\infty} of a series, if it exists, has an interesting interpretation. It is an infinite sum, that is, the sum of infinitely many numbers:

Definition 3: Infinite sum a_1+a_2+...

Let (sn)(s_n) be the series of a sequences (an)=a1,a2,...(a_n)=a_1, a_2, ...\,. We thus have

s1=a1s2=a1+a2s3=a1+a2+a3...\begin{array}{lll} s_1 &=& a_1\\ s_2 &=& a_1+a_2\\ s_3&=& a_1+a_2+a_3\\ ...\\ \end{array}

We can think of the limit limnsn=s\lim_{n\rightarrow \infty} s_n= s_\infty as

s=a1+a2+a3+...infinitely many termss_\infty = \underbrace{a_1+a_2+a_3+...}_{\text{infinitely many terms}}

Of course, as for any sequence, the limit or infinite sum can not exist, may be ±\pm\infty or actually be a real number.

Extra: Convergent sequences - a deeper look for the brave

Show

What exactly do we mean by ana_n converges towards a real number aa_\infty? Intuitively, it is reasonably clear: aa_\infty is that number towards which the terms of the sequence move if we move to the right in the sequence. However, note that none of the terms ana_n has to actually be the value aa_\infty (see figure below):

Let's describe this idea in mathematical terms:

Definition 4

The sequence (an)(a_n) converges to aa_\infty if for every small ϵ>0\epsilon>0 there exists a position n0n_0 in the sequence such that all subsequent terms ana_n are closer than ϵ\epsilon to aa_\infty:

(aϵ)an(a+ϵ) for all nn0(a_\infty-\epsilon) \leq a_n \leq (a_\infty+\epsilon) \text{ for all } n\geq n_0

Side note: the above inequality is also often written like this:

aanϵ for all nn0\vert a_\infty-a_n\vert \leq \epsilon \text{ for all } n\geq n_0
Example 2

Intuitively, it is clear that the sequence

(bn)=(1,0.5,0.25,0.125,...)(b_n)=(1, 0.5, 0.25, 0.125, ...)

tends towards 00:

limnbn=0\lim_{n\rightarrow \infty} b_n = 0

Therefore, for any value ϵ>0\epsilon>0, no matter how small, there must be a position n0n_0 in sequence so that for any position nn to the right of n0n_0 (that is, nn0n \geq n_0) it is

0ϵbn0+ϵ0-\epsilon \leq b_n \leq 0+\epsilon

Determine such a n0n_0 for ϵ=0.00001\epsilon=0.00001.

Solution

It is bn=(12)n1b_n=\left(\frac{1}{2}\right)^{n-1} (für n=1,2,3,...n=1,2,3,...). As bn>0b_n>0 for all nn, find an n0n_0 with

bn0+ϵ for all nn0b_n\leq 0+\epsilon \text{ for all } n \geq n_0

that is

=(12)n10.00001=\left(\frac{1}{2}\right)^{n-1}\leq 0.00001

Let's find nn such that

(12)n1=0.00001ln(..)ln((12)n1)=ln(0.00001)ln(.)(n1)ln(12)=ln(0.00001):ln(12),+1n=ln(0.00001)ln(12)+1=17.61\begin{array}{rll} \left(\frac{1}{2}\right)^{n-1} &=& 0.00001 \quad\vert \ln(..)\\ \ln\left(\left(\frac{1}{2}\right)^{n-1}\right) &=& \ln(0.00001)\quad \vert \ln(.)\\ (n-1)\ln(\frac{1}{2}) &=& \ln(0.00001) \quad \vert :\ln(\frac{1}{2}), +1\\ n&=&\frac{\ln(0.00001)}{\ln(\frac{1}{2})}+1 = 17.61\\ \end{array}

For all n18n\geq 18 ist 0.00001bn0.00001-0.00001\leq b_n\leq 0.00001 (that is n0=18n_0=\underline{18}).

Extra: Divergent sequences - a deeper look for the brave

Show

Divergent sequences do not approach a single number. This can mean that the sequence jumps back and forth between different numbers forever and ever (see figure below):

But it can also mean that the sequence tends towards \infty or towards -\infty. Again, it is intuitively clear what we mean by this: A sequence (an)(a_n) tends towards \infty if the terms become larger and larger without ever reaching a limit, that is, without an upper bound. And a sequence converges towards -\infty if the sequence terms become more and more negative without ever reaching a limit or lower bound (see Figure below).

To be more exact:

Definition 5

A sequence (an)(a_n) converges to \infty if for any arbitrarily large number NN there exists a position n0n_0 in the sequence such that all subsequent terms ana_n are greater than NN:

anN for all nn0a_n \geq N \text{ for all } n\geq n_0

A sequence (an)(a_n) converges to -\infty if for any arbitrarily large negative number NN there exists a position n0n_0 in the sequence such that all subsequent terms ana_n are smaller than NN:

anN for all nn0a_n \leq N \text{ for all } n\geq n_0
Example 3

The sequence of square numbers

(an)=(1,4,9,16,...)(a_n)=(1,4,9,16,...)

is divergent and converges to \infty. Thus, for any arbitrarily large value NN, we can find a n0n_0 such that

anN for all nn0a_n\geq N \text{ for all } n\geq n_0

Find a n0n_0 for N=1000000N=1\,000\,000.

Solution

It is an=n2a_n=n^2. We have to find an n0n_0 such that

n21000000 fu¨r alle nn0n^2 \geq 1\,000\,000 \text{ für alle } n\geq n_0

This is n0=1000000=1000n_0=\sqrt{1\,000\,000}=\underline{1\,000}.