5
$\begingroup$

According to Wikipedia, it is possible to invert Binet's formula for Fibonacci numbers:

$$F_n = \frac{\varphi^n-\psi^n}{\varphi-\psi} = \frac{\varphi^n-\psi^n}{\sqrt 5}$$ where $\varphi = \frac{1 + \sqrt{5}}{2} \approx 1.61803\,39887\ldots$ and $\psi = \frac{1 - \sqrt{5}}{2} = 1 - \varphi = - {1 \over \varphi} \approx -0.61803\,39887\ldots$, or more specifically the truncating variant of that:

$$F_n = \left\lfloor\frac{\varphi^n}{\sqrt 5} + \frac{1}{2}\right\rfloor$$

to find the index $n(F)$ of the largest Fibonacci number that is not greater than a real number $F > 1$:

$$n(F) = \left\lfloor \log_\varphi (F\sqrt5 + 1/2)\right\rfloor$$

Since no justification is given, besides the floor function being monotonic, I would like to prove this.

So assuming the truncating formula holds: $$F_n = \left\lfloor\frac{\varphi^n}{\sqrt 5} + \frac{1}{2}\right\rfloor$$ $$F_n = \frac{\varphi^n}{\sqrt 5} + \frac{1}{2} + E, \text{where $0 \le E < 1$}$$ $$\varphi^n = \sqrt5(F_n - \frac{1}{2} - E), \text{where $0 \le E < 1$}$$ $$n \log\varphi = \log(\sqrt5(F_n - \frac{1}{2} - E)), \text{where $0 \le E < 1$}$$ $$n = \log_\varphi(\sqrt5(F_n - \frac{1}{2} - E)), \text{where $0 \le E < 1$}$$

Since n is a whole number we can take the floor:

$$n = \lfloor n \rfloor = \left\lfloor \log_\varphi(\sqrt5(F_n - \frac{1}{2} - E)) \right\rfloor, \text{where $0 \le E < 1$}$$ $$n = \left\lfloor \log_\varphi(\sqrt5 F_n \left(1 - \frac{1 + 2E}{2F_n}\right) \right\rfloor, \text{where $0 \le E < 1$}$$ $$n = \left\lfloor \log_\varphi(\sqrt5 F_n) + \log_\varphi \left(1 - \frac{1 + 2E}{2F_n}\right) \right\rfloor, \text{where $0 \le E < 1$}$$

And this does not seem to be the right path...

$\endgroup$

1 Answer 1

6
$\begingroup$

We have $$F_n =\frac{\phi^n-(-\frac{1}{\phi})^n}{\sqrt{5}}$$ what can be inversed writing $$\phi^{2n}-F_n\sqrt{5}\,\phi^n-(-1)^n = 0$$ which is a quadratic in $\phi^n$. So $$\phi^n=\frac{F_n\sqrt{5}+\sqrt{5F_n^2\pm 4}}{2}\implies n_\pm \log(\phi)=\log\Bigg[\frac{F_n\sqrt{5}+\sqrt{5F_n^2\pm 4}}{2} \Bigg]$$ One of the two solution is an integer $$\left( \begin{array}{cccc} n & F_n & n_- & n_+ \\ 3 & 2 & 3.000000000 & 3.209573980 \\ 4 & 3 & 3.907487979 & 4.000000000 \\ 5 & 5 & 5.000000000 & 5.033256487 \\ 6 & 8 & 5.987011534 & 6.000000000 \\ 7 & 13 & 7.000000000 & 7.004918572 \\ 8 & 21 & 7.998115113 & 8.000000000 \\ 9 & 34 & 9.000000000 & 9.000719061 \\ 10 & 55 & 9.999725212 & 10.00000000 \end{array} \right)$$

If this is not the case, then the number is not Fibonacci. Trying for $12345678987654321$, the result is $n_\pm=78.669724$; so, it is not a Fibonacci number (but we know the closest one). $$F_{78}=8944394323791464 \quad < \quad 12345678987654321$$ $$F_{79}=14472334024676221 \quad > \quad 12345678987654321$$

Edit

Consider what is the difference between the two roots : consider $F_n$ to be large and make a series expansion $$n_+-n_-\sim\frac{2}{5 F_n^2 \log (\phi )} <\frac{1}{ F_n^2 }$$

$\endgroup$
2
  • $\begingroup$ Why can we ignore the other root the abc formula gives? $\endgroup$ Commented Dec 24, 2021 at 15:11
  • $\begingroup$ If you neglect the +-4, than you can simplify the RHS to log(F*sqrt5). Rounding will recover n- or n+ (whichever is integer). Rounding is the same as adding 1/2 and taking the floor. This should get you to the formula in the question, but I am not convinced it has the property described there... $\endgroup$ Commented Jan 3, 2022 at 11:52

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.