a repository of mathematical know-how

To prove that two real numbers are equal, prove that neither is greater than the other

Quick description

Two real numbers x and y are equal if it is not the case that x<y and it is not the case that y<x. This is often the best way to prove that x=y when it is not obvious.

Example 1 (and general discussion)

This basic idea, phrased very differently, goes back at least as far as the ancient Greeks. If you want to prove that the area of a circle of radius r is \pi r^2, then one way to do it is to inscribe a regular 2n-gon. One then cuts the 2n-gon into ``slices'' by joining its centre to the 2n vertices, and then rearranges the slices into a parallelogram by alternating upwards-pointing slices and downwards-pointing slices. Each slice is an isosceles triangle, and the parallelogram has height the altitude of these isosceles triangles and base half the perimeter of the 2n-gon. Now let s<r. Then if n is large enough, the altitude of each slice will be greater than s and the perimeter of the 2n-gon will be greater than \pi s. These assertions are not trivial but are not too hard to prove geometrically. It follows that for every s you can find inside the circle a polygon of area greater than \pi s^2. Therefore, the area of the circle cannot be less than \pi r^2. A similar argument shows that it cannot be greater than \pi r^2, and therefore it is equal to \pi r^2.

The point here is that we did not do some calculation that ended up with \pi r^2. Rather, we proved that the area could not be greater than this and could not be less than this.

A similar idea is often used to prove that a non-negative number is zero: if x\geq 0 and x<\epsilon for every positive \epsilon then x=0. With the help of the Archimedean principle we can also state a useful variant of this: if x\geq 0 and x<1/n for every positive integer n, then x=0.

Example 2

A nice example of this is the standard proof of Liouville's theorem in complex analysis, which states that a bounded analytic function on the whole complex plane is constant. To prove this, one uses Cauchy's integral formula, which states that if z is a complex number and C is a simple closed contour that contains z in its interior, then f(z)=\frac 1{2\pi i}\int_C\frac{f(w)dw}{w-z}.

Now let z and z' be two points and let C be a simple closed contour that contains both z and z' in its interior. If we now calculate f(z)-f(z') using Cauchy's integral formula, we obtain \frac 1{2\pi i}\int_Cf(w)dw(\frac 1{w-z}-\frac 1{w-z'}, which equals \frac 1{2\pi i}\int_Cf(w)dw\frac{z-z'}{(w-z)(w-z')}. If f is bounded above in modulus by M, and C is a circle of radius R with R at least twice the maximum of |z| and |z'|, then this comes out to be at most \frac 1{2\pi}2\pi RM\frac{|z-z'|}{(R/2)^2}, which tends to 0 as R tends to infinity. Therefore, |f(z)-f(z')| is less than any positive real number, which forces it to be zero.

Example 3

One of the standard proofs of the intermediate value theorem uses the law of trichotomy in a fairly explicit way. Suppose that f is a continuous function, that a<b, and that f(a)<z<f(b). To prove that there exists c between a and b such that f(c)=z, one defines c to be the supremum of the set f(x)<z\}. Then instead of proving directly that f(c)=z, one obtains a contradiction from the assumption that f(c)>z and another contradiction from the assumption that f(c)<z.


Post new comment

(Note: commenting is not possible on this snapshot.)

Before posting from this form, please consider whether it would be more appropriate to make an inline comment using the Turn commenting on link near the bottom of the window. (Simply click the link, move the cursor over the article, and click on the piece of text on which you want to comment.)