4
$\begingroup$

I encountered this problem that I could not figure out, the first form of the Euler Maclaurin summation is:

$$ \sum_{a \leq k \leq b}f(k) = \int_a^b f(x)dx + \frac{f(a)+f(b)}{2} + \sum_{1 \leq i \leq m}\frac{B_{(2i)}}{(2i)!}f^{(2i-1)}(x)|^b_a $$

while its second form is:

$$ \sum_1^N f(k) = \int_1^Nf(x)dx + \frac{f(N)}{2} + \gamma + \sum_{1\leq k \leq m} \frac{B_{2i}}{(2i)!}f^{(2i-1)}(N)+R_m $$

where for both forms $B_i$ represents the $i$th Bernoulli number and $f^{(i)}$ represents the $i$th derivative.

Using (either of ) these formulas, am supposed so show that given the Gamma function $\Gamma(n)=(n-1)!$, we have the following relationship over generalized binomial coefficients:

$$ \binom{n+\alpha}{n} \equiv \frac{\Gamma(n+\alpha+1)}{\Gamma(n+1)\Gamma(\alpha+1)} \sim \frac{n^\alpha}{\Gamma(\alpha+1)} $$

If I use the definition $\Gamma(n)=(n-1)!$, the first equivalence $\equiv$ relation could be derived easily ... But I could not figure out how the Euler Maclaurin equation comes to play ... I am having a hunch it has something to do with proving the $\sim$ relation... Any help ?

$\endgroup$
1
  • $\begingroup$ It is not so clear what you are given and what you should prove. If e.g. $\alpha$ or $n$ is not an integer, the first equivalence is actually a definition and the second is an equivalence for fixed $\alpha$ as $n$ goes to infinity. Is this a homework exercise of some sort? $\endgroup$ Commented Aug 6, 2020 at 8:21

1 Answer 1

3
+100
$\begingroup$

I think you can get this from the Bernoulli integral for the $\Gamma$ function: $$ \Gamma(z) = \int_0^\infty x^{z-1} e^{-x} dx. $$

If we use the first Euler Maclaurin summation and swap the integral and summation, we get: $$ \int_a^b f(x) dx = \sum_{a\leq k\leq b} f(k) - \frac{f(a) + f(b)}{2} - \sum_{1\leq i\leq m} \frac{B_{2i}}{(2i)!} f^{(2i-1)}(N) - R(m). $$

We then truncate the last expansion term summation leaving the simple (0-th order "expansion"): $$ \int_a^b f(x) dx \approx \sum_{a\leq k\leq b} f(k) - \frac{f(a) + f(b)}{2}. $$

For our $\Gamma$ function integral, we let $f(x)=x^{z-1}e^{-x}$ to get: $$ \Gamma(z) = \int_0^\infty x^{z-1} e^{-x} dx \approx \sum_{0\leq k\leq \infty} k^{z-1} e^{-k} - \lim_{a\downarrow 0}\frac{a^{z-1}}{2e^a} - \lim_{b\to\infty} \frac{b^{z-1}}{2e^b} = \sum_{0\leq k\leq \infty} k^{z-1} e^{-k}. $$

Note that for the above to hold, we require $z>0$ so that $f(x)$ is bounded above or $z$ such that $f$ is dominated by a function $g$ with a finite integral $\int_0^\infty g(x) dx$.

Now rewrite $\binom{n+\alpha}{n}$: $$ \binom{n+\alpha}{n} = \frac{\Gamma(n+\alpha+1)}{\Gamma(n+1)\Gamma(\alpha+1)} \approx \frac{\sum_{0\leq k\leq \infty} k^{n+\alpha} e^{-k}}{\Gamma(\alpha+1)\sum_{0\leq k\leq \infty} k^{n} e^{-k}} = \frac{k^\alpha}{\Gamma(\alpha+1)} \frac{\sum_{0\leq k\leq \infty} k^{n} e^{-k}}{\sum_{0\leq k\leq \infty} k^{n} e^{-k}} = \frac{k^\alpha}{\Gamma(\alpha+1)}. $$

$\endgroup$
1
  • $\begingroup$ Small note: we need the Dominated Convergence Theorem to swap summation and integration, but that goes through which is why the swap may be made. $\endgroup$ Commented May 4, 2021 at 15:40

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.