dubitable

Theme:

Theme Picker

Choose a custom theme!

Pick from a selection of colours for a pre-defined colour scheme!

Theme:

Farey Sequences

Adapted from "Chapter III: Farey Series and a theorem of Minkowski" in the book "An Introduction to the Theory of Numbers, Sixth Edition" by G.H. Hardy and E.M.Wright.

Note that the book calls it a series but Wikipedia says "A Farey sequence is sometimes called a Farey series, which is not strictly correct, because the terms are not summed." so I will call it a sequence.

What is a Farey Sequence?

A Farey Sequence Fn is the ascending irreducible fractions between 0 and 1 where the denominators are less than or equal to n. For example to construct F4 we can get all the fractions with denominator ≤ 4:

0 4,1 4,2 4,3 4,4 4;0 3,1 3,2 3,3 3;0 2,1 2,2 2;0 1,1 1

Obviously there's a bunch of duplicates here: all the fractions equal to zero or one are the same as each other, there's also 2 4=1 2. We also need them to be in ascending order, so therefore the proper Farey Sequence for F4 is the following:

F 4=0 1,1 4,1 3,1 2,2 3,3 4,1 1

Graphing this, using the y-axis as the denominator and the x-axis as the numerator and then reflecting it across the diagonal you can form little shapes like so:

F4

F 4=0 1,1 4,1 3,1 2,2 3,3 4,1 1

Some properties of a Farey Sequence:

If a/b and c/d are successive terms of Fn then b+d>n. The mediant of the fractions is a+c b+d and we know it will always be in between the other two like so: a c<a+b c+d<b d this is because(from the Wikipedia page for mediant) a+b c+d-a c=bc-ad c( c+d )=d( c+d )( b d-a c ) and since b/d > a/c this will always be positive so we know the mediant is greater than a/c and b d-a+b c+d=bc-ad d( c+d )=c c+d( b d-a c ) which will also be positive for the same reason so the mediant is less than b/d. It might be more intuitive if we look at an example, using 1/2 and 2/3. 1+2 2+3=3 5 which is in between the two numbers. You can kind of think of it as smooshing the 2 numbers together and getting something in between. The significance of this property for us is that the denominator of the mediant is b+d and if it was less than or equal to n then it would be in Fn but since these are successive terms it can't be, so b+d must be greater than n.

If n > 1, then all successive terms of Fn have differing denominators. Obviously for n = 1, we would have 0/1 and 1/1 in sequence, but for any other n we can't have a/b and c/b as successive terms since a+1≤c<b and a b<a b-1<a+1 bc b. So e.g. for 1/3 and 2/3 for F4 they can't be successive since 1/2 comes in between.

If a/b and c/d are successive terms of Fn then we have bc-ad=1

e.g. for F4 we have 1/2 followed by 2/3 so we get 2×2-3×1=1

Additionally, if a/b, c/d and e/f are three successive terms of Fn then we have c d=a+e b+f

e.g. for F4 we have 1/2, 2/3 and 3/4 successively so 2 3=1+3 2+4

To prove these last two we first show that the two statements are equivalent to each other. If we assume bc-ad=1 then for three successive equations we can form de-cf=1 as well. Now we want to get equations for c and d without each other so we do the following:
d=1+bc a
Putting d equal to the rest of an equation to substitute it in
If we do a similar thing for d
c=ad-1 b
Putting c equal to the rest of an equation to substitute it in
And if we put our first equation over the other equation we get c d=a+e b+f as expected.

To prove they're equivalent we need to prove it goes the other way as well. i.e. a implies b and b implies a. Thus if we assume c d=a+e b+f is true then we can show this means the other equation must be true by induction. To prove something by induction means to prove something is true for a specific case then prove that if it's true for a certain case it will be true for the next case. This is useful for us because if bc-ad=1 then we can do the following:

c d=a+e b+f
Our initial equation

The base case is evident since the first two elements for any Fn will be 0/1 and 1/n and 1×1-0×n=1 Thus we have shown that the two equations are equivalent to each other but we have still yet to prove they're true.

To do this, we will also use induction; this time using Fn-1 to Fn as the inductive case. It's obviously true for n = 1 as the base case. Suppose that a/b and e/f are consecutive for Fn-1 but for Fn they're separated by c/d. If we set ad-bc=r and cf-de=s then we can do the following, remembering that af-be=1 from Fn-1

ad-bc=r
Our starting equation for r

We can do a similar thing for d to get d=bs+fr so that we can form c d=as+er bs+fr. What values of r and s can work for this? They have to be greater than zero since otherwise it would just be equal to a/b or e/f and wouldn't be separate for the Farey Sequence. To be a new value in Fn it must be the smallest possible value otherwise and so it must be r=s=1, and thus c/d is the mediant of the other two. Thus, c d=a+e b+f. As an example, for 2/3 and 1/1 in Fn, c d=2s+1r 3s+1r since this is a new value that was just added to Fn and is in between two values that were in Fn-1, if s or r were greater than 1 then the value for r=s=1 would've already been included in Fn-1 and 2/3 and 1/1 wouldn't be successive in Fn

This is a bit handwavy and I don't formalise it much but there are other proofs given in the book chapter I'm basing this off which I may return to.