Lower bound for combined Gamma functions

88 Views Asked by At

I am asking for your help with an exercise from my general physics I class. The exercise is about the Gamma function and reads in this way.

Let $b\in\mathbb{N}, b\ge 2$ and $a\in (0, 1)$. Prove that there exists an lower bound which does not depend on $a$ for $$\frac{\Gamma((a+b)/2)}{\Gamma((b-a)/2)}\cdot \left(\frac{\Gamma(b/2)}{\Gamma(b)}\right)^{a/b},$$ i.e. that there exists a positive constant $c$, non depending on $a$, such that $$\frac{\Gamma((a+b)/2)}{\Gamma((b-a)/2)}\cdot \left(\frac{\Gamma(b/2)}{\Gamma(b)}\right)^{a/b}\ge c.$$

Here $\Gamma$ is defined, for $Re(z)>0$, as $$\Gamma(z) = \int_0^\infty t^{z-1} e^{-t} dt$$ (Note that here $a, b, a-b, a+b$ are all real and nonnegative numbers).

The only idea I have so far is to split $\int_0^\infty \dots = \int_0^1 \dots +\int_1^\infty \dots$. This led me to a huge number of integrals which I do not know how to manage and, honestly, I do not think is the right way to prove the result. Anyone could please help in proving this result?

Finally, just out of my curiosity, I would like to understand if one can get an upper bound, too (but this second thing is not important and this thing will not affect the acceptance of the answer).

1

There are 1 best solutions below

5
On BEST ANSWER

Lower bound. By Gautschi’s Inequality $(5.6.4)$, \begin{align*} \frac{{\Gamma \big( {\frac{b}{2} + \frac{a}{2}} \big)}}{{\Gamma \big( {\frac{b}{2} - \frac{a}{2}} \big)}} & = \frac{{b - a}}{2}\frac{{\Gamma \big( {\frac{{b - a}}{2} + a} \big)}}{{\Gamma \big( {\frac{{b - a}}{2} + 1} \big)}} \ge \frac{{b - a}}{2}\left( {\frac{{b - a}}{2} + 1} \right)^{a - 1} \\ & \ge \frac{b}{2}\left( {\frac{b}{2} + 1} \right)^{a - 1} \ge \frac{1}{2}\left( {\frac{b}{2} + 1} \right)^a \end{align*} for any $b\ge 2$ and $0<a<1$. We also have $$ \left( {\frac{{\Gamma (b/2)}}{{\Gamma (b)}}} \right)^{1/b} = \left( {\sqrt \pi \,2^{1 - b} \frac{1}{{\Gamma ((b + 1)/2)}}} \right)^{1/b} \ge \frac{1}{2}\left( {\frac{1}{{\Gamma (b)}}} \right)^{1/b} \ge \frac{1}{2}\frac{1}{b} $$ if $b\ge 2$. Thus, \begin{align*} \frac{{\Gamma \big( {\frac{b}{2} + \frac{a}{2}} \big)}}{{\Gamma \big( {\frac{b}{2} - \frac{a}{2}} \big)}}\left( {\frac{{\Gamma (b/2)}}{{\Gamma (b)}}} \right)^{a/b} & \ge \frac{1}{2}\left( {\left( {\frac{b}{2} + 1} \right)\left( {\frac{{\Gamma (b/2)}}{{\Gamma (b)}}} \right)^{1/b} } \right)^a \\ & \ge \frac{1}{2}\left( {\left( {\frac{b}{2} + 1} \right)\frac{1}{2}\frac{1}{b}} \right)^a \ge \frac{1}{2}\left( {\frac{1}{4}} \right)^a \ge \frac{1}{8}. \end{align*} Upper bound. The quantity is unbounded. Assume that $0<a<1$. By the asymptotics $(5.11.12)$, \begin{align*} \frac{{\Gamma \big( {\frac{b}{2} + \frac{a}{2}} \big)}}{{\Gamma \big( {\frac{b}{2} - \frac{a}{2}} \big)}} \sim \left( {\frac{b}{2}} \right)^a \end{align*} as $b\to+\infty$. By Stirling's formula $$ \left( {\frac{{\Gamma (b/2)}}{{\Gamma (b)}}} \right)^{1/b} \sim \sqrt {\frac{{\rm e}}{2}} \frac{1}{{b^{1/2} }} $$ as $b\to+\infty$. Consequently, $$ \frac{{\Gamma \big( {\frac{b}{2} + \frac{a}{2}} \big)}}{{\Gamma \big( {\frac{b}{2} - \frac{a}{2}} \big)}}\left( {\frac{{\Gamma (b/2)}}{{\Gamma (b)}}} \right)^{a/b} \sim \left( {\frac{{\rm e}}{8}} \right)^{a/2} b^{a/2} $$ as $b\to+\infty$.