Lecture 24 - Intro to Functions

We want to find the limits of functions. What is:

limx→cf(x)=L

Let A⊆R and f:A→R where L∈R. We want c to be a Limit Point of A.

Limit of a function, ε−δ Version

limx→cf(x)=L
means that ∀ε>0∃δ>0(0<|x−c|<δ) and x∈A implies that:
|f(x)−L|<ε

Example: Limit of a function

Consider f(x)=2x2+5 where A⊆R. We want to show:
limx→2f(x)=13
The scratch is:
|f(x)−13|<ε⟺|2x2+5−13|<ε⟺|2x2−8|<ε⟺2|x2−4|<ε⟺2|x−2||x+2|<ε
Now this will work if 0<|x−2|<1 (we can make the left side as small as any number) so that implies 0<|x+2|<5 so then the LHS will have a factor of 10, so that implies δ=min{1,ε10}. The proof highlights this.

Proof

Let ε>0.

Let ε=min{1,ε10}. Suppose 0<|x−2|<δ. Note that since |x−2|<δ≤1 then |x+2|<5. Then:
|f(x)−13|=|2x2−8|=2|x−2||x+2|<2⋅δ⋅5≤2⋅ε10⋅5=ε

☐

Man this seems like a pain to do right? Do we have to do all of our Convergence of a Sequence definitions again but now for functions? Not quite. By the following theorem:

Sequential Criteria for Functional Limits

Let A⊆R and f:A→R and L∈R. Let c be a Limit Point of A.
The following are equivalent:

  1. limx→cf(x)=L
  2. For all sequences (xn)⊆A−{c} with xn→c, we have f(xn)→L.

Proof

(⇒): Suppose limx→cf(x)=L. Then ∀ε>0∃δ>0(|x−c|<δ→|f(x)−L|<ε). Now let (xn)⊆A−{c} where xn→c be arbitrary. We want to show f(xn)→L. Let's use the definition of convergence of a sequence.

Let ε>0. By our given then δ>0 such that 0<|x−c|<δ→|f(x)−L|<ε (where x∈A). Since xn→c then ∃N∈N where ∀n≥N(|xn−c|<δ). Now let's choose any n≥N, so then 0<|xn−c|<δ (the xn≠c by construction). This gives us that |f(x)−L|<ε thus completing the proof for convergence.

(⇐): We'll prove the contra-positive; namely that not (1) implies not (2). Now suppose that ¬[limx→cf(x)=L]. That implies that ∃ε>0∀δ>0∃x∈A(0<|x−c|<δ∧|f(x)−L|≥ε). We need to prove the opposite of (2); namely we need to construct this sequence such that f(xn) doesn't converge to L.

In particular, for n∈N we can find some xn∈A−{c} such that 0<|xn−c|<1n=δ such that then |f(xn)−L|≥ε (see our given above). But then that implies that f(xn)↛L (it's the opposite of the convergence definition).

☐

This leads to some very useful corollaries:

Let f,g:A→R where A⊆R and c is a Limit Point of A. Let L,M∈R. If:
limx→cf(x)=L,limx→cg(x)=M
then:

  • limx→c[αf(x)]=αL for all α∈R.
  • limx→c[f(x)+g(x)]=L+M
  • limx→c[f(x)g(x)]=LM
  • limx→cf(x)g(x)=LM provided M≠0.

Proof

Use the Sequential Criterion For Functional Limits, but now using the Limit Laws (Algebraic Limit Theorem) for sequences for each.

☐

This is a corollary from the Sequential Criterion For Functional Limits:

Existence of Functional Limits

Let f:A→R where A⊆R and c is a Limit Point of A.
If ∃ sequences (xn)⊆A−{c} and (yn)⊆A−{c} with xn→c,yn→c but (f(xn)) and (f(yn)) do not converge to the same value. Then:
limx→cf(x)
does not exist.

Namely this is the opposite of the Sequential Criterion For Functional Limits (2).

Example of Finding Functional Limits

Let f(x)=sin⁡(1x) where A=R−{0}.

---
title: 
xLabel: 
yLabel: 
bounds: [-2, 2, -1, 1]
disableZoom: false
grid: true
---
f(x) = sin(1/x)

A reasonable question to ask is if/what limx→0f(x) is here (notice 0 here is a limit point of the domain set). But notice that the sequence of points that intersect the graph at 0 on the right hand side make up a sequence where f(xn)→0. But we can do the same thing with the points that intersect the line y=1. Then f(yn)→1, so we have two sequences that converge to different limits, so then the limit of the original function must not exist.

More explicitly, let n∈N. Then choose:
xn=1nπ,yn=1π2+2nπ
Then notice that xn,yn→0 individually, but:
f(xn)=sin⁡(11nπ)=sin⁡(nπ)=0
f(yn)=sin⁡(11π2+2nπ)=sin⁡(π2+2nπ)=1
So since f(xn)→0 and f(yn)→1 then by our corrolary above then we get the limit does not exist.