
Number Theory Number Theory Math Forum 
 LinkBack  Thread Tools  Display Modes 
April 28th, 2014, 01:06 PM  #1 
Member Joined: Jul 2010 Posts: 83 Thanks: 2  Unique Integer Mapping with Phi and the Golden Angle
One of the most interesting properties of the golden ratio is it's supreme irrationality above all others. Of course, the golden angle is essentially nothing more than a restatement of phi so the same holds true for it as well. Now it may not seem from the outset that this property could possibly provide any means to, say, produce pseudorandom sequences of integers but in fact it can be done and the key turns out to be none other than the fibonacci sequence. Consider the following: Given some N select any greater consecutive sequence of fibonacci numbers {F1, F2} and calculate G1 = {N*F2 mod F1}. G1 is, in a sense, the fractional (noninteger) result of multiplying N with an approximation of the golden ratio; the conjecture here that G1 for any given N < F1 is unique. Similarly, given some N select any greater consecutive sequence of fibonacci numbers {F1, F2, F3} then calculate G2 = {N*F1 mod F3}. In this case, G2 rather represents the fractional (noninteger) result of multiplying N with an approximation of the golden *angle* (and likewise conjectured unique). To help visualize what's happening here, imagine a circle divided into F3 pieces. G2 then is the "position" on this circle obtained when N is multiplied by the golden angle and then truncated to fit "onto" the circle. The uniqueness of this mapping perhaps explains why nature has employed it all along in the placement of leaves and so forth; nonoverlap is kept to the most optimal level. But is it really true that this mapping is unique? Empirical evidence certainly points to such a conclusion, although a proof as yet seems elusive; in any case, the risk of it being false appears negligible at worst. Only time will tell, of course. Cheers! 
April 28th, 2014, 01:25 PM  #2  
Global Moderator Joined: Nov 2006 From: UTC 5 Posts: 16,046 Thanks: 938 Math Focus: Number theory, computational mathematics, combinatorics, FOM, symbolic logic, TCS, algorithms  Quote:
 
April 28th, 2014, 01:56 PM  #3 
Member Joined: Jul 2010 Posts: 83 Thanks: 2  
April 28th, 2014, 08:17 PM  #4 
Member Joined: Jul 2010 Posts: 83 Thanks: 2 
Also note that any other irrational is bound to have more near "collisions" with unit fractions than phi (just consider it's continued fraction expansion). So output correlations between inputs M and N which share common divisors or what have you is kept to a minimum.

April 29th, 2014, 06:49 AM  #5 
Global Moderator Joined: Nov 2006 From: UTC 5 Posts: 16,046 Thanks: 938 Math Focus: Number theory, computational mathematics, combinatorics, FOM, symbolic logic, TCS, algorithms  So you're trying to prove that F1 does not divide (nm)*F2, with 0 <= m < n < F1. The sequence of Fibonacci numbers is a gcd sequence, so gcd(F1, F2) is Fibonacci number at the gcd of their indices. Since they're consecutive the gcd of their indices is 1, so gcd(F1, F2) = F_1 = 1, and so F1 divides (nm)*F2 if and only if F1 divides nm. But this is between 1 and F11 and so F1 cannot divide it. QED.

April 29th, 2014, 09:01 AM  #6  
Member Joined: Jul 2010 Posts: 83 Thanks: 2  Quote:
 

Tags 
angle, golden, integer, mapping, phi, unique 
Search tags for this page 
Click on a term to search for related topics.

Thread Tools  
Display Modes  

Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
golden ratio  gaussrelatz  Algebra  2  February 12th, 2013 08:54 PM 
Golden Gate Bridge  baz  Algebra  1  April 30th, 2012 09:04 AM 
Trying to graph Phi (the golden ratio)  yurayurae  Algebra  2  February 26th, 2012 04:46 PM 
Golden number ratio  Voltman  Advanced Statistics  9  March 13th, 2011 08:39 PM 
Golden Ratio  Daniel Derwent  Algebra  4  June 25th, 2010 09:03 PM 