Let's try to learn together the basis of calculus!

Theoretical look at limits

Theoretical framework around limits

First we will look at the definition of limits and we will try out a few proof to see how it works. In practice we never come back to the actual definition, but we use theorems and properties derived from that. We will see in a next lesson a few of those properties, how to prove them and how to use them.

Convergence Definition

The theory

Let’s look at the case of a finite limit - we will deal with infinity later.

We say that a sequence ana_n converges to a limit LL if:

  • for any ϵ>0\epsilon > 0, there exists an integer NN such that for all n>Nn > N, we have anL<ϵ|a_n - L| < \epsilon.

(By the way, the symbol ϵ\epsilon is the Greek letter “epsilon”, and it is often used in mathematics to denote a small positive number. Epsilon is kinda the ancester of the letter “e” in the Latin alphabet, but I digress.)

Ouch. That’s quite a mouthful. Let’s try to understand this definition. The overall idea is that we need to give a meaning to the phrase “getting as close as we want to LL”.

So let’s imagine two mathematicians looking at the proof, let’s call them Alice and Bob.
Bob will be convinced that the sequence converges to LL if, each time he gives Alice a number ϵ\epsilon, Alice can find a number NN such that “the sequence is closer than Epsilon” for all n>Nn > N. In mathematical terms, that would be when we have anL<ϵ|a_n - L| < \epsilon.

Their dialogue would look like this:

  • Bob: “Can you find a rank after which the sequence is closer than 1 to LL?”
  • Alice: “Sure, here it is.”
  • Bob: “Ok, well done. Hmmm - what about closer than 0.1?”
  • Alice: “Give me a second… Here it is.”
  • Bob: “Wow - okay not bad. Can you do it for 0.000000001?”
  • Alice: “Sure, here is one”

If Alice can do this for any ϵ\epsilon, then Bob is convinced that the sequence converges to LL. This is the idea behind the definition of convergence.

👀 Hint!

Generally, when a definition calls for “there exists” a certain value, we will have to “build” one value according to the other values or parameters we have laying around.
Try to remember this recipe each time you see “there exists” in a definition or a problem.

How it goes with an example

Let’s look at the sequence an=1na_n = \frac{1}{n}. We would like to prove that the limit of this sequence is 00. Let’s look at some cases first, like in the discussion between Alice and Bob - and we will play the role of Alice.

  • Bob: “Can you find a rank after which the sequence is closer than 0.1 to 00?” Alice has to find a rank NN such that for all n>Nn > N, we have an0<1|a_n - 0| < 1.
  • Let’s try N=1N = 1.
  • a10=10=1|a_1 - 0| = |1 - 0| = 1 - not good enough.
  • Let’s try N=2N = 2.
  • a20=0.50=0.5|a_2 - 0| = |0.5 - 0| = 0.5 - not good enough but getting closer. A quick work of algebra will show us that for N=10N = 10, we have a100=0.10=0.1|a_{10} - 0| = |0.1 - 0| = 0.1. Now Alice has to prove that for every N>10N > 10, we have an0<0.1|a_n - 0| < 0.1. I will leave this aside for now as we will have to do something really similar for the general case.

Generally speaking we can try a few values of ϵ\epsilon to get a feel for the way we can “build” the value of NN. Let’s look at the general example. Bob asks us to find a rank N0N_0 such that for all n>N0n > N_0, we have an0<ϵ|a_n - 0| < \epsilon. Please note that we are looking for one of such N0N_0 - we don’t have to find the minimum or anything like that. Just find one which works! We have an=1na_n = \frac{1}{n}, so an<ϵa_n < \epsilon is equivalent to:

  • 1n<ϵ\frac{1}{n} < \epsilon. When we invert both sides of the inequality, we have to reverse the sign of the inequality:
  • n>1ϵn > \frac{1}{\epsilon}. Now we have a good candidate for N0N_0, we have to prove that the sequences stays close to the limit for every rank above N0N_0.

If N>N0N > N_0, then we have 1N<1N0\frac{1}{N} < \frac{1}{N_0}.
We know that 1N0<ϵ\frac{1}{N_0} < \epsilon, so we have 1N<ϵ\frac{1}{N} < \epsilon.

Here we go, we have proved that for every arbitrary ϵ\epsilon, we can find a rank N0N_0 such that the sequence is closer than ϵ\epsilon for all ranks above N0N_0. So we can say that the limit of the sequence an=1na_n = \frac{1}{n} is 00.

Infinite limit definition

The theory

The theory is a similar idea than the one above. But instead of proving that the sequence is “closer” to the limit, we need to prove that the sequence becomes “bigger” than any arbitrary number after a certain rank.

We say that a sequence ana_n has for limit ++\infty if:

  • for any K>0K > 0, there exists an integer N0N_0 such that for all n>N0n > N_0, we have an>Ka_n > K.

Here again we have to prove the existence of a certain number, so we will need to build it according to the value of KK.

How it goes with an example

Let’s look at the sequence an=n2a_n = n^2. We would like to prove that the limit of this sequence is ++\infty.

Let’s look at some cases first, like in the discussion between Alice and Bob. Can we go higher than, say, 100100?
Well, for N=10N=10 we have a10=100a_{10} = 100.
We know as well that if a>b>0a > b > 0, then a2>b2a^2 > b^2. So we can say that for N>10N > 10, we have aN=N2>100a_N = N^2 > 100.

Looking at the general case, let’s look at a given KK. We have to find a rank N0N_0 such that for all n>N0n > N_0, we have an>Ka_n > K. We have an=n2a_n = n^2, so n2>Kn^2 > K is equivalent to:
n>Kn > \sqrt{K}.

So we can take any N0N_0 bigger than K\sqrt{K}, and we are done. Every n>N0n > N_0 will have n2>N02n^2 > N_0^2 and we know that N02>KN_0^2 > K. So we can say that for every n>N0 n > N_0, we have an>Ka_n > K. We have proved that the limit of the sequence an=n2a_n = n^2 is ++\infty.

The negative case

Having a limit of -\infty is the same idea, but we have to prove that the sequence is smaller than any arbitrary number after a certain rank.

Previous Lesson
Next Lesson