2
$\begingroup$

I don't know the method to solve the following inequation

$$C_1 (1+ \sqrt{a})^{n-1} + C_2(1-\sqrt{a})^{n-1} \geq k$$

where $C_1$, $C_2$, and $a$ are constants. I would like to express a lower bound for $n$ as a function of $k$.

  • 0
    Perhaps you meant to say "express *a lower bound for* $n$ as a function of $k$"?2017-01-16
  • 0
    @JoseArnaldoBebitaDris Yes ! :-) I correct my post. Thank you2017-01-16
  • 0
    Are $C_1,C_2$ and $a$ in some known ranges?2017-01-18
  • 1
    @dxiv I am interested in a general method, but you can consider that $a=2$, and $C_1$ and $C_2$ strictly positive.2017-01-18

1 Answers 1

2

This is not simple. The addition doesn't play nicely with the powers so you will need to find a numeric solution. The best approach will depend upon the constants you have. We set $m=n-1$ to save typing.

The easiest case is when $k$ is rather large, especially where $C_2$ is not much greater than $C_1$. In that case the first term will dominate. You can ignore the second to get a starting value, then iterate. Set $$C_1(1+\sqrt a)^{m_0}=k \\m_0=\frac {\log k - \log C_1}{\log (1+\sqrt a)}\\m_{i+1}=\frac {\log (k-C_2(1-\sqrt a)^{m_i}) - \log C_1}{\log (1+\sqrt a)}$$ and iterate to convergence

Another easy case is when $a$ is small and $m$ is not too large. You can use the binomial expansion then and write $$(1+\sqrt a)^m \approx 1+m\sqrt a\\C_1 (1+ \sqrt{a})^{m} + C_2(1-\sqrt{a})^{m}\approx C_1+C_2 +(C_1-C_2)m\sqrt a$$ and you have a linear equation to solve for $m$. Again, you can use that as a starting value for iteration if it is not accurate enough. The test for accuracy is $m \sqrt a \ll 1$

If $C_2$ is rather large compared with $C_1$ you can have two ranges of $m$. An example is in the plot below where I took $C_1=1, C_2=100, \sqrt a=0.1$ You can see that for $k=40$ you have two regions of solution. The minimum value for the higher region will be found easily by the first approach. The root around $10$ will be harder to find. enter image description here

  • 0
    Thank you very much for your answer. What do you mean by "The root around 10 will be harder to find" ?2017-01-18
  • 0
    I don't see a nice way of getting a starting value for root finding. You can take the derivative to find the minimum and check $m=0$ to see if there is a root in there.2017-01-18
  • 0
    Sorry, I don't understand. What do you call root?2017-01-20
  • 0
    I was assuming $k=40$. In that case there will be a case of equality with $m$ somewhere close to $10$ and another close to $40$. Those are roots of the equality.2017-01-20
  • 0
    Thank you for your detailed answer.2017-01-25