No label defined (Q5552797)

Put \(F_0=0\), \(F_1 = 1\), \(F_{n+1}=F_{n}+F_{n-1}\), \(n \ge 1\). Let \(R(N)\) denote the number of representations \[ N = F_{k_1}+F_{k_2}+\ldots+F_{k_r}, \tag{*} \] where \(k_1>k_2>\cdots>k_r \ge 2\). This is equivalent to \[ \prod_{n=2}^{\infty}\left(1+y^{F_n}\right)=\sum_{N=0}^{\infty} R(N) y^N. \] It is known that the representation (*) is unique provided \(k_j - k_{j+1} \ge 2\) \((j=1, \ldots, r-1)\), \(k_r \ge 2\). Such a representation is called the \textit{canonical} representation of \(N\). If we put \[ \prod_{n=1}^{\infty}\left(1+x^{F_n} y^{F_{n+1}}\right) = \sum_{m,n=0}^{\infty} A(m, n) x^m y^n \] it is shown that \(R(N) = A(e(N), N)\), where \(e(N) = F_{k_1-1}+F_{k_2-1}+\ldots+F_{k_r-1}\) and the \(k_s\) are those occurring in (*); the value of \(e(N)\) is independent of the particular representation employed. The first main result of the paper is the following reduction formula: \[ R(N) = R\left(e^{k_r-1}(N_1)\right)+\left([\tfrac12 k_r]-1\right) R\left(e^{k_r-2}(N_1)\right), \tag{**} \] where \(N_1 = F_{k_1} + cdots+F_{k_{r-1}}\) and the \(k_s\) are determined by the canonical representations. By means of (**) explicit formulas are determined for \(r=1,2,3\) but the general case is complicated. However if \(N= F_{2k_1}+\cdots+ F_{2k_r}\) \((k_1>\cdots>k_r \ge 1)\) and \(j_s = k_s - k_{s+1}\) \((s=1, \ldots, r-1)\), \(j_r = k_r\), \(f_r=f\left(j_1, \ldots, j_s\right)=R(N)\), \(S_r = 1+f_1+\cdots+f_r\), then \[ S_0=1,\ S_1= j_1+1,\ S_r = (j_r+1) S_{r-1} - S_{r-2}\quad (r \ge 2). \] In particular, if \(j_1=\cdots=j_r=j\), then \[ S_r = \sum_{2t \le r} (-1)^t \binom{r-t}{t} (j+1)^{r-2t}. \] Finally it is proved that if \(N = F_{k_1} + \cdots+F_{k_r}\) is the canonical representation of \(N\), then \(e(N+1)=e(N)\) if and only if \(k_r=2\); also if \(k_r>2\), then \(e(N)=[N/\alpha]\), the nearest integer to \(N/\alpha\), while if \(k_r=2\) then \(e(N)=[N/\alpha]+1 \).