play and relax: games for kids games
  Cut the knot: learn to enjoy mathematics
A math books store at a unique math study site. Learn to enjoy mathematics.
Google
Web CTK
Best sites for teachers
Sites for teachers
Sites for parents
Terms of use
Awards

Interactive Activities
CTK Exchange
CTK Insights - a blog

Games & Puzzles
What Is What
Arithmetic/Algebra
Geometry
Probability
Outline Mathematics
Make an Identity
Book Reviews
Eye Opener
Analog Gadgets
Inventor's Paradox
Did you know?...
Proofs
Math as Language
Things Impossible
Visual Illusions
My Logo
Math Poll
Cut The Knot!
MSET99 Talk
Other Math sites
Front Page
Movie shortcuts
Personal info
Privacy Policy

Guest book
News sites

Recommend this site

Best sites for teachers
Sites for teachers
Sites for parents

Education & Parenting

Manifesto: what CTK is about Search CTK Buying a book is a commitment to learning Table of content Things you can find on CTK Chronology of updates Email to Cut The Knot Recommend this page

Fractions on a Binary Tree

Let a/b be a fraction in lowest terms, where a and b are positive integers, a < b. Place it at the root of a binary tree. Next form its two children a/(a+b) and b/(a+b). Continue this process indefinitely to obtain the following tree diagram.

Find a and b such that the diagram starting with a/b contains all positive fractions less than 1.

Solution

References

  1. J. Cofman, What To Solve?, Oxford Science Publications, 1996.

 

 

 

 

 

 

Solution 1

Let's first show that if the diagram contains all positive fractions, then a/b must be equal to 1/2. Assume, on the contrary, that 1/2 appears somewhere inside the tree. This means that for some p and q either p/(p+q) = 1/2 or q/(p+q) = 1/2. In both cases, p = q = 1 and the direct ancestor of 1/2 - p/q or q/p, as the case may be, equals 1. But 1 can't possibly belong to the tree because the algorithm generates only fractions less than 1.

The algorithm has the property that successors of a fraction in lowest terms are automatically in lowest terms as well. If gcd(p, q) = 1, then too gcd(p, p+q) = 1, since a factor common to both p and (p+q) also divides q.

Assume now that a/b = 1/2 and let p/q be a positive fraction in lowest terms less than 1. We want to show that p/q belongs to the tree. If p/q = 1/2, there is nothing to prove. We have to consider two cases

  1. p/q < 1/2
  2. p/q > 1/2

In general, for positive fractions r/s less than 1 define the ancestor function A in the following manner: if r/s < 1/2 let A(r/s) = r/(s-r), if r/s > 1/2 let A(r/s) = (s-r)/r.

Starting with r/s, form a sequence of fractions:

(1) p/q, A(p/q), A(A(p/q)), ...

Observe that the sequence of denominators is decreasing. Therefore, (1) is a finite sequence. Let An(p/q) be the last fraction in (1). Its denominator must be 2, for, otherwise, we would be able to continue. So the fraction is 1/2. The function A is called ancestor because it reverses the algorithm that defines the tree. Going backwards, in n steps, we'll get from 1/2 to p/q. Therefore, p/q is on the tree.

Solution 2

In the discussion on the Stern-Brocot tree we defined two functions: fR(r/s) = r/s+1 = (r+s)/s and fL(r/s) = r/(r+s). In terms of these two functions, for a given fraction r/s, the algorithm used in the problem generates two descendants: fL(r/s) and 1/fR(r/s).

However, by definition,

(2) 1/fR(r/s) = fL(s/r).

Recollect an observation we made in the discussion on the Stern-Brocot tree: fL transforms a row into the left half of the next row, fR transforms a row into the right half of the next row. From this remark and (2), it follows that the algorithm fills up the left side of the Stern-Brocot tree, a row at a time. The order of the fractions is, of course, all screwed up. For example, the 4th row will appear as 1/5, 4/5, 3/7, 4/7, 2/7, 5/7, 3/8, 5/8.

An upshot of having two solutions to the problem is that we can interpret Solution 1 as it applies to the Stern-Brocot tree. Unwittingly, we got a second proof that all fractions on the Stern-Brocot tree emerge in lowest terms. We also got a second proof of the fact that the Stern-Brocot tree contains all positive fractions.

Copyright © 1996-2008 Alexander Bogomolny

29436906Page copy protected against web site content infringement by Copyscape


Search:
Keywords:


Latest on CTK Exchange
try this puzzle ?/?? + ?/?? + ?/? ...
Posted by albert1950
0 messages
07:05 AM, Jul-10-08

Monty Hall Problem
Posted by linkdon
72 messages
06:07 PM, Jul-24-08

Missing information
Posted by roboknight
2 messages
07:32 AM, Jun-22-08

Can You See The Patterns..?
Posted by wustvn
0 messages
10:08 AM, Jul-23-08

Mistake on the page (an aside, Be ...
Posted by Max
4 messages
10:28 AM, Feb-28-08

Central Limit Theorem proof problem
Posted by Manuel
1 messages
01:54 PM, Jul-22-08

You can drill a square hole
Posted by Giorgis
1 messages
10:15 PM, Jul-12-08