Proof Writing for Real Analysis

TagsGuides

The fundamentals

Start a proof by getting a general intuition for why it’s true. then, abstract it away and talk about informal terms. Lastly, put down the formal proof

Contradiction is your friend!

General form: show B from A. Immediately, you should write out the definitions of A, and the formal definitions of B. Try to see if you can massage the definitions of A to B.

General form: show that there exists some X and satisfies A and B. This is an existential proof, so the best bet is to construct something iteratively such that we satisfy both A and B. Alternatively, we can show that all A belong to B, or vice versa (although this is not generally true) .

General form: show that f(x)=f(y)f(x) = f(y), where f is some function, x, y are inputs, where the property of y is known. Mainly, you should write out the definition of what f(y)f(y) means. This k=f(y)k = f(y) satisfies all these properties. Now, you show that this kk satisfies the properties of f(x)f(x) as well.

Example: show that min(ST)=min(S)\min(S \cup T) = \min(S) if all S<TS < T. Well, if x=min(S)x = \min(S), then x<Tx < T for all TT because S<TS < T, which means that x=min(ST)x = \min(S \cup T) as well.

General form: show that f(x) = f(y) but neither properties of x, y are known. In this case, you will need to massage x,yx, y back to something and use original truths.

General form: show that x is unique. Start by assuming that there are two x, y that satisfy a property. Then, show that x=yx = y.

Induction does NOT work for infinite cases!

Using universals and the “algorithm” view

If you want to show a universal statement for some xx, you essentially want to develop an algorithm that shows that something is true for one arbitrary xx.

If you want to use a universal statement for some xx, you can assume that one such function exists.

So, to show a universal statement from another one, you can say that you pick some xx’. You might know that a universal works for xx. Just massage this xx’ into the input xx such that your final result shows that this xx’ works for the universal.

Here’s a concrete example. Say that you wanted to show that all multiples of 66 are even. You can express this as 6n6n'. We know, as a universal, that all 2n2n are even. In this case, we can let n=3nn = 3n’. This is the mapping from the desired, arbitrary input to a the input that works for our assumed universal. This yields 2(3n)2(3n’), and we can apply the universal to show that 6n6n’ is even.

This is also why you can start with some anl<ϵ|a_n - l| < \epsilon’ and use some anb<ϵ|a_n - b| < \epsilon, and then at the end, you might show that you can write ϵ\epsilon as a function of ϵ\epsilon’. We might have a shorthand where we start with ϵ\epsilon’ directly and plug in the function of ϵ\epsilon’ in the original definition of ϵ\epsilon.

When can you turn << into \leq?

Generally, if you have some f(x)<kf(x) < k, as xx → \infty, we can say f(x)kf(x) \leq k. It might not be that limf(x)=k\lim f(x) = k, but it is still correct to switch the bounds to an inclusive.

Bounding

Contradiction

If you want to show (or dissapprove) that something can be written in a form

Generally, the logic flow of a contradiction starts with two coflicting definitions. You work with one definition and massage it until you see that it clearly violates the other definition. The key here is that you only go with one definition.

Tools

A great table

Set Facts