I'm just unable to comprehend what would $$f(x) = (\log(n))!$$ be.
Most of what I saw online talks about $$f(x) = (\log(n!)).$$
Any idea into the intuition for this would be helpful!
I'm just unable to comprehend what would $$f(x) = (\log(n))!$$ be.
Most of what I saw online talks about $$f(x) = (\log(n!)).$$
Any idea into the intuition for this would be helpful!
The closest thing to the factorial for non-integer numbers is the Gamma function $\Gamma(x)$ which, if $x$ is a positive integer, is equal to $(x-1)!$, and if $x$ is positive and not an integer, lies between the values of the factorial. You can then attach a meaning to $\log(n)!$ of $\Gamma(\log(n)+1)$. But the natural logarithm $\log(n)$ is not an integer for integer $n$ for any except the "trivial" case $\log(1)=0$.
I'm not how much there is to say here. At any rate in order to make sense of the function, we should interpret $y!$ as $\Gamma(y + 1)$, where $\Gamma$ denotes the gamma function, and hence our function $x \mapsto (\log x)!$ as $x \mapsto \Gamma(\log x + 1)$, with which it coincides on the nonnegative integers and which is defined everywhere except at the nonpositive integers. (This extension is the best one possible in a sense made precise by the Bohr-Mollerup Theorem.)
We can immediately read off properties of $x \mapsto (\log x)!$ from those of $\Gamma$: For example, $x \mapsto (\log x)!$ has simple poles precisely at $e^{-1}, e^{-2}, e^{-3}, \ldots$ and hence domain $(0, \infty) \setminus \{e^{-1}, e^{-2}, e^{-3}, \ldots\}$, and its range is $\Bbb R \setminus \{0\}$. The function's (unique) minimum occurs at $\beta = e^{\alpha - 1} = 1.58666\ldots$, where $\alpha = 1.46163\ldots$ is the location of the unique minimum of $\Gamma$ on $(0, \infty)$, and $x \mapsto (\log x)!$ is strictly monotone on $[\beta, \infty)$.
Stirling's approximation gives the (peculiar-looking) asymptotic formula as $n \to \infty$: $$(\log x)! = \sqrt{2 \pi} \frac{(\log x)^{\log x + \frac{1}{2}}}{x} \left(1 + O((\log x)^{-1})\right) .$$ Using an additional term of the expansion of $\Gamma(y + 1)$ at $\infty$ gives a better approximation for large $x$: $$(\log x)! \sim \sqrt{2 \pi} \frac{\log x^{\log x - \frac{1}{2}} (12 \log x + 1)}{12 x} .$$ In fact, already at $x = 2$ the relative error of the latter formula is smaller than $1$ part in $1\,300$.
The red curve is the graph of $(\log x)!$, the blue curve is the graph of the leading term of the above asymptotic formula, and the green curve is the graph of the second, better approximation.
As some of the commenters have mentioned, it's more common to consider $\log (x!) = \log \Gamma(x + 1)$; indeed $\log \Gamma$ is simply called the log-gamma function.