Home

[This page is auto-generated; please prefer to read the article’s PDF Form.]



[Prev]   [Up]   [Next]   

The Three Gap-Lengths

We will be using a few facts from the theory of continued fractions. Say, pk∕qk represent the convergents from the continued fraction of a real number α > 0. Then, it is known that for all k 0:

     ||      ||
     |α − pk| < ---1--
     |    qk|   qkqk+1
                --1-
⇔   |qkα−  pk| < qk+1
For all k 0, since qk+1 1, so:
|qkα− pk| < 1

It is also known that, for all k 0:

               |      |
      α − pk-= ||α − pk||(− 1)k
          qk   |    qk|
⇔   q α − p  = |q α− p |(− 1 )k
     k     k     k    k

Due to above relations, and the fact that any integer can be added or subtracted inside {}, {qkα} can also be written as:

{qkα} = {qkα − pk}
      = {1 + qkα − pk}
      = {1 + |qkα − pk|(− 1)k}

      = qkα − pk                                 (for even k)
        {1 − (pk − qkα )} = 1 − (pk − qkα)         (for odd k)  (3)

Now we compute the gap-lengths for ϕ. As defined in section “Three Distance Theorem” of [1], we will use u1 and uN to refer to the two “corner points”.

Consider the case when t is even. Using theorem 4 in [1], the u1 and uN for ϕ will be:

 u1 = qt = Ft+1

uN  = qt− 1 + rqt = Ft + (1)Ft+1 = Ft+2

We noted earlier, due to relation (2), that Ft+2 is the largest Fibonacci number not exceeding N. Now we see that the two corner points correspond to the two largest Fibonacci numbers not exceeding N.

Now we use theorem 2 in [1] to find the three gap-lengths, which are specified as L1, L2 and L1 + L2 in that theorem.

L1 = {u1α }
   = {qtα}
   = qtα − pt                (due to (3) with t even)

   = Ft+1ϕ − Ft+2

L  = 1 − {u α }
 2         N
   = 1 − {qt+1α }             (since Ft+2 is also qt+1)
   = 1 − (1− (pt+1 − qt+1α)) (due to (3) with t + 1 odd )

   = pt+1 − qt+1α
   = Ft+3 − Ft+2ϕ
L1 + L2 = Ft+1ϕ − Ft+2 + Ft+3 − Ft+2ϕ = Ft+1 − Ftϕ

This was for even t. Similarly, it can be proved that for t as odd, the expressions of the three lengths will be:

     L1 = − (Ft+3 − Ft+2ϕ)
     L2 = − (Ft+1ϕ − Ft+2)

L1 + L2 = − (Ft+1 − Ftϕ)

Corollary 1. The two smallest gap-lengths (L1 and L2, in this order or reverse) for α = ϕ are:

|Ft+2ϕ−  Ft+3 | and |Ft+1ϕ− Ft+2 |

where, Ft+2 (t 0) is the largest Fibonacci number not exceeding N. The gap-length L1 + L2 can also be written as:

|Ftϕ− Ft+1 |.

Note how u1 and uN, and hence the three gap-lengths, change only when t changes. Due to the relation (2) which defines t, that occurs whenever N attains any Fibonacci number.

Note that, gaps of length L1 + L2 need not exist for all N. From corollary 5 in [1], gap-length L1 + L2 will not exist if s = qt 1. For case of ϕ, that can be rewritten as:

                s = qt − 1
⇔   N − qt−1 − rqt = qt − 1  (due to relation (4) in [1 ])

⇔   N − Ft − Ft+1 = Ft+1 − 1
⇔              N  = Ft+3 − 1

Corollary 2. For α = ϕ, the gap-length of L1+L2 does not exist whenever N + 1 is some Fibonacci number.

So, in Figure 1 for N = 12, which is F7 1, we should expect only two distinct gap-lengths.

[Prev]   [Up]   [Next]