Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Source. V. E. Hoggatt, Jr. and Marjorie Bicknell, A reciprocal series of Fibonacci numbers with subscripts 2nk2^nk, Fibonacci Quart. 14 (1976), no. 5, 453--455. The paper numbers no theorems; the closed form is the display after "Finally," on p. 455, and its derivation runs from p. 453 to p. 455. Bibliographic details are on the source card.

Statement

Here FmF_m and LmL_m are the Fibonacci and Lucas numbers, which the paper uses without restating their definitions. For a fixed index kk (the paper states no range; its derivation and its check at k=1k=1 treat kk as a positive integer),

∑n=0∞1F2nk={2Lk−F2k5+5Fk22F2k,k odd;2−Fk5+Lk2Fk,k even.\sum_{n=0}^{\infty}\frac{1}{F_{2^nk}}= \begin{cases} \dfrac{2L_k-F_{2k}\sqrt5+5F_k^2}{2F_{2k}}, & k\ \text{odd};\\[2ex] \dfrac{2-F_k\sqrt5+L_k}{2F_k}, & k\ \text{even}. \end{cases}

Just before this display (p. 455) the paper writes the two cases as 1/Fk−5/2+5Fk/(2Lk)1/F_k-\sqrt5/2+5F_k/(2L_k) for kk odd and 1/Fk−5/2+Lk/(2Fk)1/F_k-\sqrt5/2+L_k/(2F_k) for kk even; the displayed forms follow from these through F2k=FkLkF_{2k}=F_kL_k.

The case k=1k=1 (p. 454). The paper evaluates the limit at k=1k=1 as 1+5(5−1)/2=(7−5)/21+\sqrt5(\sqrt5-1)/2=(7-\sqrt5)/2, the value of ∑n≥01/F2n\sum_{n\ge0}1/F_{2^n} found by Good and posed by Millin, which the paper cites as its references [1] and [2].

Odd and even cases (p. 455). For k=2s+1k=2s+1 odd, the paper writes BB for the sum over the indices (2s+1)2n(2s+1)2^n and CC for the sum over the indices 2(2s+1)2n2(2s+1)2^n, writes each in the three-term form of its case from the limit computation (1/Fk1/F_k, a quotient of Fibonacci and Lucas numbers, and −5/2-\sqrt5/2), and notes that B=C+1/F2s+1B=C+1/F_{2s+1}. (Since 2(2s+1)2n=(2s+1)2n+12(2s+1)2^n=(2s+1)2^{n+1}, the second series is the first with its n=0n=0 term removed; this remark is the page's, not the paper's.)

Proof sketch (pp. 453--455)

  • The identity F2k=FkLkF_{2k}=F_kL_k, applied to the first few partial sums and rewritten with Lm+p+Lm−p=LmLpL_{m+p}+L_{m-p}=L_mL_p for even pp, together with the Lucas identity L2nk(L(2n−2)k+⋯+L2k+1)=L(2n+1−2)k+⋯+L2kL_{2^nk}\bigl(L_{(2^n-2)k}+\cdots+L_{2k}+1\bigr)=L_{(2^{n+1}-2)k}+\cdots+L_{2k}, gives the finite sum (1) on p. 453: the partial sum up to nn equals (F2nk/Fk+L(2n−2)k+⋯+L2k+1)/F2nk\bigl(F_{2^nk}/F_k+L_{(2^n-2)k}+\cdots+L_{2k}+1\bigr)/F_{2^nk}.
  • A summation formula for ∑Faj−b\sum F_{aj-b} quoted from K. Siler (the paper's reference [3]), applied with a=2ka=2k and b=±1b=\pm1 and added termwise, sums the Lucas numbers L2kjL_{2kj} in closed form (p. 454).
  • Writing the partial sum to NN through these closed forms and letting N→∞N\to\infty with α=(1+5)/2\alpha=(1+\sqrt5)/2 and β=(1−5)/2\beta=(1-\sqrt5)/2 gives the limit 1/Fk+(5−5β2k)/(L2k−2)1/F_k+(\sqrt5-\sqrt5\beta^{2k})/(L_{2k}-2), which the paper reduces to 1/Fk−5/2+5F2k/(2(L2k−2))1/F_k-\sqrt5/2+5F_{2k}/(2(L_{2k}-2)) (p. 454).
  • The identity (2) Lk2=L2k+2(−1)kL_k^2=L_{2k}+2(-1)^k turns L2k−2L_{2k}-2 into Lk2L_k^2 for kk odd and into 5Fk25F_k^2 for kk even, which yields the two cases (p. 455).

This sketch follows the paper's structure; the algebra was not re-derived here.

Read depth. Claims checked: the statement was read clause by clause on p. 455 of the printed article, and the k=1k=1 evaluation on p. 454; the derivation was read for structure only.

Dependencies

Siler's summation formula for ∑k=1nFak−b\sum_{k=1}^{n}F_{ak-b}, quoted from the paper's reference [3] (K. Siler, Fibonacci Quart. 1 (1963), 67--69) and not proved in the paper, and the standard identities F2k=FkLkF_{2k}=F_kL_k, Lm+p+Lm−p=LmLpL_{m+p}+L_{m-p}=L_mL_p (pp even) and Lk2=L2k+2(−1)kL_k^2=L_{2k}+2(-1)^k.

Bears on

  • Problem 267: the index sequences nj=2jkn_j=2^jk have ratio nj+1/nj=2n_{j+1}/n_j=2, and the theorem evaluates ∑j1/Fnj\sum_j1/F_{n_j} for them in closed form. In both cases the coefficient of 5\sqrt5 is −1/2-1/2 and the other terms are rational. The paper evaluates the sums and does not discuss their irrationality; it says nothing about index sequences of any other form.