I was trying to prove that $(N+\sqrt{N^2-1})^k$, where k is a positive integer, differs from the integer nearest to it by less than $(2N-\frac{1}{2})^{-k}$. Note: N is an integer greater than 1.
So, I tried to look for the answer of the question, which I have taken it from an exam paper.
It said that: We let T=$(N+\sqrt{N^2-1})^K +(N+\sqrt{N^2-1})^{-k}$= $(N+\sqrt{N^2-1})^k + (N-\sqrt{N^2-1})^k=2(N^k + kC2 N^{K-2}(N^2-1)+...$ which is clear it is an integer.
We know that: $(N-\frac{1}{2})^2 =N^2-N+\frac{1}{4}=\frac{5}{4}-N$ This is <0 when N > 1, since N is an integer.
So,
$N-\frac{1}{2}<\sqrt{N^2-1}$.
$2N-\frac{1}{2} < N+\sqrt{N^2-1}$
$(2N-\frac{1}{2})^{-k}>(N+\sqrt{N^2-1}^{-k}$
Let $|T-(N+\sqrt{N^2-1})^{-k}$| we would be able to prove the question.
However, my question is that, we need to have the nearest integer, so, the integer that the answer used is
$(N+\sqrt{N^2-1})^k + (N + \sqrt{N^2-1})^{-k}$.
Why is this the nearest integer?
Thank you so much for your reply.
Let $N$ be a fixed integer.
Using the expression you give,
$$\underbrace{(N+\sqrt{N^2-1})^k}_{A_k} + \underbrace{(N-\sqrt{N^2-1})^k}_{B_k>0}=\underbrace{2(N^k + \binom{k}{2} N^{k-2}(N^2-1)+...}_{C_k, \ \text{an integer}},\tag{1}$$
$C_k$ will be the closest integer to $A_k$ if their difference $B_k=C_k-A_k$ is proven to be $< 1/2$, whatever the values of $N>1$ and $k>0$ ;
Indeed :
$$B_k=N(1-\sqrt{1-1/N^2})^k < N(1-(1-1/N^2))^k=1/N^{2k-1} \leq 1/2^{2k-1}<1/2$$
due to inequality
$$\sqrt{1-x} < 1-x$$ valid for any $x$ such that $0 < x < 1$.
Remarks :
a) Let us show a numerical example in the case $N=2$ , with columns $k, A_k, B_k, C_k$ in that order :
$$\begin{array}{|r|r|r|r|} \hline 1&3.732050807 & 0.267949192 & 4\\ 2&13.928203230 & 0.071796769 & 14\\ 3&51.980762113 & 0.019237886 & 52\\ 4&193.994845223 & 0.005154776 & 194\\ 5&723.998618781 & 0.001381218 & 724\\ 6&2701.999629903 & 0.000370096 & 2702\\ \hline \end{array}$$
It is to be noted that numbers $C_k$ obey a second order recurrence relationship : $$C_{k+1}=4 C_k-C_{k-1}\tag{2}$$
(see https://oeis.org/A003500) and more generally :
$$C_{k+1}=2N C_k-C_{k-1}$$
(it comes from the fact that $A_k$ and $B_k$ can be considered as the two roots of quadratic characteristic equation associated with (2).
b) [in connection with (a)] : (1) can be considered for example as the analog of Binet formula for Fibonacci sequence defined by second order recurrence :
$$F_{k+1}=F_k+F_{k-1}$$
http://www.maths.surrey.ac.uk/hosted-sites/R.Knott/Fibonacci/fibFormula.html#section1)