5

We know that $H_n$ = $L_n + mF_n$, where $n = 0$ or $n > 0$ is simply relation between Fibonacci sequence and generalized Fibonacci-Lucas sequence. Are there any methods to prove the following identities? I am new to this site and hopefully, I will have good replies.

1) Sum of the terms:

$$H_1 + H_2 + H_3 + \ldots H_n = H_{n+2} – (m+3)$$

2) For alternate terms, we see the following:

$$H_1 + H_3 + H_5 + \ldots +H_{2n-1} = H_{2n} – 2$$ and
$$H_2 + H_4 + H_6 + \ldots + H_{2n} = H_{2n+1} – (m + 1)\;.$$

3) At the same time, if we square each terms, we see the:

$$H_1^2 + H_2^2 + H_3^2 + \ldots + H_n^2 = H_n H_{n+1} – 2 (m + 1)\;.$$

4) Finally, this interesting identity is valid:

$$H_{2n} = \sum_{k=0}^{n}\binom{n}{k}H_{n-k}$$

VividD
  • 16,196
  • Solutions to constant coefficients linear recurrence relations should be a linear combination of exponentials (with polynomial coefficients sometimes, but constant coefficients ideally), cf Binet's formula, which should allow these sorts of identities to be proven with eg geometric sum formula or the binomial theorem. – anon Aug 17 '12 at 06:54
  • @Anon! great response. Thank you. I tried Binet's formula and there is some disturbance in producing the proof. I am looking to give one proof for all questions? is it possible? I mean deducing all the results from one to other. If not solve the third and fourth one. Thanks a lot for for a quick reply. – vidyaojal Aug 17 '12 at 06:59
  • 2
    Do you know the corresponding identities for the Fibonacci sequence and for the Lucas sequence? For example, do you know the identity for $F_1+F_2+\cdots+F_n$? If you know those, you can use them to get identities for your $H_n$. – Gerry Myerson Aug 17 '12 at 07:22
  • @GerryMyerson! I got all those results by your hint. Thank U. – vidyaojal Aug 17 '12 at 14:52
  • 2
    "I got all those results by your hint." - then, please post an answer to your own question, and we can then check if your method is correct. – J. M. ain't a mathematician Aug 17 '12 at 15:58

1 Answers1

3
  1. By definition $$ \sum_{i=1}^n H_i = \sum_{i=1}^n L_i +m \sum_{i=1}^n F_i $$ where $L_{i\ge 1}=1,3,4,7,\ldots$ and $F_{i\ge 1} = 1,1,2,3,5,\ldots$. We have a well-known partial sum, see https://oeis.org/A027961 $$ \sum_{i=1}^n L_i = L_{n+2}-3 $$ and another see https://oeis.org/A000071 $$ \sum_{i=1}^n F_i = F_{n+2}-1 $$ Plugged into the first equation $$ \sum_{i=1}^n H_i = L_{n+2}-3 +m (F_{n+2}-1). $$ By definition $H_{n+2}=L_{n+2}+mF_{n+2}$, so $$ \sum_{i=1}^n H_i = H_{n+2}-mF_{n+2}- 3 +m (F_{n+2}-1) = H_{n+2}- 3 -m. $$ so the first sum of terms is proven.

  2. For the sum of the odd-index terms by definition $$ \sum_{i=0}^{n-1} H_{2i+1} = \sum_{i=0}^{n-1} L_{2i+1} +m \sum_{i=0}^{n-1} F_{2i+1} $$ where with https://oeis.org/A004146 $$ \sum_{i=0}^{n-1} L_{2i+1} = L_{2n}-2 $$ and with https://oeis.org/A001906 $$ \sum_{i=0}^{n-1} F_{2i+1} = F_{2n}. $$ Therefore by definition $$ \sum_{i=0}^{n-1} H_{2i+1} =L_{2n}-2+m F_{2n} = H_{2n}-2, $$ so the second sum of term is also proven.

For the sum of the even-index terms by definition $$ \sum_{i=1}^{n} H_{2i} = \sum_{i=1}^{n} L_{2i} +m \sum_{i=1}^{n} F_{2i} $$ where with https://oeis.org/A188378 $$ \sum_{i=0}^{n} L_{2i} = L_{2n+1}+1. $$ Removing the inital $L_0=2$ yields $$ \sum_{i=1}^{n} L_{2i} = L_{2n+1}-1. $$ With https://oeis.org/A027941 $$ \sum_{i=1}^{n} F_{2i} = F_{2n+1}-1. $$ Therefore by definition $$ \sum_{i=1}^{n} H_{2i} =L_{2n+1}-1 +m(F_{2n+1}-1) =H_{2n+1}-1-m $$ so the third sum of terms is also proven.

  1. For the sum of the squares by binomial expansion $$ \sum_{i=1}^n H_i^2 = \sum_{i=1}^n (L_i+mF_i)^2 = \sum_{i=1}^n L_i^2 +m^2 \sum_{i=1}^n F_i^2 +2m\sum_{i=1}^n L_iF_i. $$ Here by https://oeis.org/A005970 $$ \sum_{i=1}^n L_i^2 = L_n L_{n+1}-2. $$ By https://oeis.org/A001654 $$ \sum_{i=1}^n F_i^2 = F_n F_{n+1}. $$ and again by https://oeis.org/A027941 $$ \sum_{i=1}^n F_iL_i = F_{2n+1}-1. $$ Inserted above $$ \sum_{i=1}^n H_i^2 = L_n L_{n+1}-2 + m^2 F_n F_{n+1} +2m (F_{2n+1}-1). $$ Furthermore by definition $$ H_nH_{n+1} = (L_n+mF_n)(L_{n+1}+mF_{n+1}) = L_nL_{n+1} +mF_nL_{n+1} +mL_nF_{n+1}+m^2F_nF_{n+1}. $$ Subtracting the previous two equations yields $$ \sum_{i=1}^n H_i^2 - H_nH_{n+1} = L_n L_{n+1}-2 + m^2 F_n F_{n+1} +2m (F_{2n+1}-1) -[L_nL_{n+1} +mF_nL_{n+1} +mL_nF_{n+1}+m^2F_nF_{n+1}] $$ $$ = -2-2m +2m F_{2n+1} -mF_nL_{n+1} -mL_nF_{n+1} $$ By Bergot's formula in https://oeis.org/A001519 we know that the last 3 terms cancel, which establishes validity of 3).

The generating function of $H$ is $$ \sum_{i=0}^\infty H_{i} x^i = \sum_{i=0}^\infty L_{i} x^i +m \sum_{i=0}^\infty F_{i} x^i, $$ and the individual g.f. of the Lucas and Fibonacci sequenecs are very well known, see https://oeis.org/A000032 and https://oeis.org/A000045 $$ h(x)\equiv \sum_{i=0}^\infty H_{i} x^i = \frac{2-x}{1-x-x^2} +m\frac{x}{1-x-x^2}.\quad (1) $$

The generating function of the even-indexed $H$ is $$ \sum_{i=0}^\infty H_{2i} x^i = \sum_{i=0}^\infty L_{2i} x^i + m \sum_{i=0}^\infty F_{2i} x^i $$ where the individual g.f. are known by https://A005248 and https://oeis.org/A001906: $$ \sum_{i=0}^\infty L_{2i} x^i = \frac{2-3x}{1-3x+x^2}; $$ $$ \sum_{i=0}^\infty F_{2i} x^i = \frac{x}{1-3x+x^2}. $$ Therefore $$ h^{(2)}(x)\equiv \sum_{i=0}^\infty H_{2i} x^i = \frac{2-3x+mx}{1-3x+x^2}.\quad (2) $$ Furthermore the sequene of the inverse binomial transform of $H$ is $$ \sum_{k=0}^n\binom{n}{k}H_{n-k} = \sum_{k=0}^n\binom{n}{n-k}H_{n-k} = \sum_{k=0}^n\binom{n}{k}H_k.\quad (3) $$ It's a well-known property that the binomial transform replaces the argument of the generating function as follows see e.g. https://dx.doi.org/10.1016/0024-3795(94)00245-9 $$ b_n=\sum_{k=0}^n \binom{n}{k}a_k \leftrightarrow b(x) = \frac{1}{1-x}a(\frac{x}{1-x}). $$ If we apply this to (1) and (3) we have $$ \frac{1}{1-x}h(\frac{x}{1-x}) = \frac{2-3x+mx}{1-3x+x^2} $$ which matches exactly $h^{(2)}(x)$. Since the two generating functions match, the individual terms must also match, and 4) is also proven.

R. J. Mathar
  • 4,450