The generating function $\sum_{n\geq 0} D(n) x^n = \frac{1}{\sqrt{1-6x+x^2}}$ is the gf of the Delannoy number.
See last paragraph in https://en.wikipedia.org/wiki/Delannoy_number
In this link, they mentioned that the coefficient of this gf is $D(n) = \sum_{k=0}^{n} {n \choose k} { n+k \choose k}$.
My question is how can this coefficient be calculated/extracted from its gf.And how they conclude that it behaves asymptotically as
${\displaystyle D(n)={\frac {c\,\alpha ^{n}}{\sqrt {n}}}\,(1+O(n^{-1}))}$
where $\alpha =3+2{\sqrt {2}}\approx 5.828$
and
$c=(4\pi (3{\sqrt {2}}-4))^{-1/2}\approx 0.5727$.
How to calculate/extract $D(n)$ from its gf and how to analyze it asymptotically then? Is that possible to do that with a direct calculations?
Edit I tried to find the asymptotic of the codfficidnt in the above gf this way:
The gf has two poles: $r_1=3+\sqrt{8}$, $r_2=3-\sqrt{8}$. So the smallest/dominant pole is $r_2=3-\sqrt{8}$. We can write the gf as: $A(x)=1/ \sqrt{r_2-x} * 1/ \sqrt{r_1-x}$ Then A(x) behaves as:
$1/ \sqrt {r_1-r_2} * 1/ \sqrt{r_2-x}= 1/ \sqrt {2 * 8^{1/4}} * \sum_{n\geq 0} {-1/2 \choose n} (-1)^n (r_1)^{n+1}x^n$ Here I used the expansion's formula of $(1+x)^a$ for $a=-1/2$ and $x=-x/r_2$. So the coefficient behaves asymptotically as: $1/ \sqrt 2 * 1/ 8^{1/4} * (3+\sqrt 8)^{n+1/2} {-1/2 \choose n} (-1)^n$.
We can simplify it more, by using the asymptotic of ${-1/2 \choose n}$.
Is that approach fine?
For the asymptotics we use the Wilf text, Theorem 5.3.1 (page 179) (Darboux) as suggested in the comments. We seek the asymptotics of
$$[z^n] \frac{1}{\sqrt{1-6z+z^2}} = [z^n] \frac{1}{\sqrt{(z-(3+2\sqrt{2}))(z-(3-2\sqrt{2}))}} \\ = \frac{1}{(3-2\sqrt{2})^n} (3-2\sqrt{2})^n [z^n] \frac{1}{\sqrt{(z-(3+2\sqrt{2}))(z-(3-2\sqrt{2}))}} \\ = (3+2\sqrt{2})^n [z^n] \frac{1}{\sqrt{((3-2\sqrt{2})z-(3+2\sqrt{2})) ((3-2\sqrt{2})z-(3-2\sqrt{2}))}} \\ = (3+2\sqrt{2})^n [z^n] \frac{1}{\sqrt{((17-12\sqrt{2})z-1) (z-1)}} \\ = (3+2\sqrt{2})^n [z^n] \frac{1}{\sqrt{(1-(17-12\sqrt{2})z) (1-z)}}.$$
Now here we have one as the dominant singularity on the circle of convergence and the theorem applies, taking the parameter $\beta = -1/2.$ Expanding the term containing the subdominant singularity around one we get for the first (constant) term
$$\frac{1}{\sqrt{1-(17-12\sqrt{2})\times 1}} = \frac{1}{\sqrt{12\sqrt{2}-16}} = \frac{1}{2\sqrt{3\sqrt{2}-4}}.$$
This gives the asymptotic
$$\bbox[5px,border:2px solid #00A000]{ \frac{1}{2\sqrt{3\sqrt{2}-4}} (3+2\sqrt{2})^n {n-1/2\choose n}.}$$
The Wilf text gives $O(n^{-3/2})$ for the error in this approximation, in sync with Wikipedia.
Wilf quotes on the same page an asymptotic for the remaining binomial coefficient namely
$${n-\alpha-1\choose n} \sim \frac{n^{-\alpha-1}}{\Gamma(-\alpha)}$$
with $\alpha$ not a nonnegative integer. In the present case we have $\alpha = -1/2$ so we obtain
$$\frac{1}{\sqrt{n}} \frac{1}{\Gamma(1/2)} = \frac{1}{\sqrt{n}} \frac{1}{\sqrt{\pi}}.$$
This gives the form from the Wikipedia entry which is
$$\bbox[5px,border:2px solid #00A000]{ \frac{1}{2\sqrt{\pi(3\sqrt{2}-4)}} \frac{1}{\sqrt{n}} (3+2\sqrt{2})^n.}$$