I'm interested in $\theta(N):=\int_0^1 (1-x)^{N-1} e^{xN} dx$. I'd like to show that $\theta(N)\sim c/\sqrt{N}$ as $N\to\infty$ and determine $c$. Any ideas?
-
5$\begingroup$ This is homework. Presumably an application of the principle of stationary phase. $\endgroup$– Denis SerreDec 19, 2010 at 20:50
-
$\begingroup$ no, research. I'll check out the stationary phase approximation page on wikipedia... Thanks. $\endgroup$– B ArmbrusterDec 19, 2010 at 21:07
-
$\begingroup$ The function doesn't oscillate though. So I'm not sure stationary phase approximation is appropriate. Mathematica gives $e^N N^{-N} (\Gamma(N)-\Gamma(N,N))$. However, now I'm stuck with finding the asymptotics of $\Gamma(N,N)$ which doesn't seem easier. $\endgroup$– B ArmbrusterDec 19, 2010 at 21:12
-
3$\begingroup$ To the OP, a hint: re-write $(1-x) = \exp \ln (1-x)$ and apply Laplace's method. $\endgroup$– Willie WongDec 19, 2010 at 21:16
3 Answers
Denote by $I$ your integral. Then, $I = e^N \int_0^1 {x^{N - 1} e^{ - xN} \,{\rm d}x} = \frac{{\Gamma (N)e^N }}{{N^N }}\int_0^N {\frac{{x^{N - 1} e^{ - x} }}{{\Gamma (N)}}\,{\rm d}x}.$ Now, if $X_1,\ldots,X_N$ are independent and identically distributed exponential(1) random variables, then their sum $X_1 + \cdots + X_N$ has gamma density $x^{N-1}e^{-x}/\Gamma(N)$, $x>0$. Thus, the last integral above is equal to ${\rm P}(X_1 + \cdots + X_N \leq N)$, or equivalently to ${\rm P}(X_1 + \cdots + X_N - N \leq 0)$. By the central limit theorem, ${\rm P}(X_1 + \cdots + X_N - N \leq 0) \to 1/2$ as $N \to \infty$ (since the $X_i$ have expectation equal to $1$). So, using $N^N \sim N!e^N /\sqrt {2\pi N} $, we get $ I \sim \sqrt {\frac{\pi }{2}} \frac{1}{{\sqrt N }}.$
-
$\begingroup$ I need to check this, but this seems elegant. Thanks. $\endgroup$ Dec 19, 2010 at 21:37
Or just type Series[Integrate[(1 - x)^N Exp[x N], {x, 0, 1}], {N, Infinity, 1}] into Mathematica, and preserve those valuable neurons.
For an accurate calculation.
Using the gamma functions $$\theta(n)=\int_0^1 (1-x)^{n-1} e^{xn} dx=e^n n^{-n} (\Gamma (n)-\Gamma (n,n))$$ Rewrite it as $$e^n\, n^{-n}\,\Gamma (n) \left(1-\frac{\Gamma (n,n)}{\Gamma (n)}\right)$$ Using 8.11.12 for the asymptotics of $\Gamma (n,n)$ leads to
$$\theta(n)=\sqrt{\frac{\pi }{2n}}\,\,\Bigg(1+\frac 13 \sqrt{\frac{2}{\pi n} } +\frac{1}{12 n}+\frac 4{135n}\sqrt{\frac{2}{\pi n}} +\frac{1}{288 n^2}+O\left(\frac{1}{n^{5/2}}\right) \Bigg) $$ which is quite accurate (relative error of $0.010$% if $n=4$ and $0.001$% if $n>9$)