What function grows slower than $\ln(x)$ as $x \rightarrow\infty$? How am I supposed to find it besides just trying finding limits of all known functions?
-
1$\begingroup$ I think $x\mapsto 1$ "grows" slower than $\ln x$ as $x$ approaches to infinity. $\endgroup$– peterwhyAug 7, 2015 at 21:07
-
4$\begingroup$ You probably ought to add that the function is unbounded and strictly increasing. $\endgroup$– marty cohenAug 7, 2015 at 21:35
7 Answers
For $x>1$, $x \mapsto \ln(\ln(x))$.
-
6$\begingroup$ To expand on this ever-so-slightly, one can find a function that grows slower than the $n^{\text{th}}$ iteration of the logarithm by just applying another logarithm. $\endgroup$– ClaytonAug 7, 2015 at 21:01
-
$\begingroup$ Given any sequence of functions $g_k\to \infty$, you can always find an $f\to \infty$ such that $f/g_k \to 0$ for every $k.$ See math.stackexchange.com/questions/1370380/… $\endgroup$– zhw.Aug 7, 2015 at 21:16
$$f(x)=\sqrt{\ln x}$$
should do, for $x\ge 1$.
-
$\begingroup$ In my opinion this is a much simpler answer than the accepted one. Based on my policy "simpler is better" you have my +1. $\endgroup$– Paramanand Singh ♦Aug 8, 2015 at 7:52
If we want to really go to the extreme, we can look at some smooth continuation of the Inverse Ackermann function.
Also, we can look at the inverse of a transexponential function. A function, $f(x)$ is transexponential if there exists some there exists points $p_1, p_2,..., p_n,...$ such that $\forall x > p_1$ $f(x) > e^x$, $\forall x > p_2$, $f(x) < e^{e^x}$, $\forall x > p_3$ $f(x) > e^{e^{e^x}}...$ The inverse of this $f(x)$ will grow very slowly.
You have to define what you mean by "grows slower than", but I'll assume you want to find a function $f(x)$ such that $\frac{f(x)}{\ln x}\rightarrow 0$ as $x\rightarrow\infty$ where $f(x)\rightarrow\infty$ as well.
L'Hopital's rule is applicable here so that $\displaystyle \lim_{x\rightarrow\infty}\frac{f(x)}{\ln x}=\lim_{x\rightarrow\infty} xf'(x)$. So a function that diverges but whose derivative goes to zero faster than $\frac{1}{x}$ works.
Generally, let $f'(x)=\frac{1}{x}g(x)$ so that $g(x)\rightarrow0$ as $x\rightarrow\infty$ and $\sum \frac{1}{n}g(n)$ diverges, then $f(x)=\int_a^x \frac{1}{t}g(t) dt$ works for a suitably chosen $a$.
For example, $f(x)=\int_2^x \frac{1}{t}\frac{1}{(\ln t)^p} dt$ for $0<p\leq1$ works.
$$f(x)=\underbrace{\ln(\ln(\ln(\dots(}_{\lfloor x\rfloor\text{ times}}x)\dots)))$$ Slightly overkill. (Of course, we could than do $f_2(x)=f(f(\dots(x)\dots))$, and then $f_3(x)=f_2(\dots(x)\dots)$, and then… and then $f_\omega(x)=f_{\lfloor x\rfloor}(x)$, though that last one might not tend to infinity…)
$\ln(x)$ is monotonic over the relevant interval. So, all one has to do to find a a function that maps $x$ to any function with a derivative less than $1$ at infinity.
In other words,
$$\ln(f(x))$$
Where ${{df} \over {dx}} \lt 1$ as $x$ approaches infinity.
Here are some examples for functions that fit the requirements.
$$f(x)=\sqrt x$$ $$f(x)=e^{-x}$$
You should note that,
$$\ln(f(x)) \gt \ln(x)$$
May hold in the limit. What matters is the rate of growth, not the amount of growth.
-
$\begingroup$ I would say that $\ln \sqrt x$ grows at the same rate as $\ln x$ in terms of asymptotics since their ratio tends to $1/2$. $\endgroup$– jdodsAug 7, 2015 at 21:18
-
1
-
-
$\begingroup$ @jdods I was looking at the derivative, which does get arbitrarily close to being equal... $\endgroup$ Aug 7, 2015 at 21:22
You could compare the derivatives of two functions, and prove that the derivative of ln(x) is always larger than the derivative of your other function
$f(x) = ln(x)$
$f'(x) = \frac{1}{x}$
$g(x) = arctan(x)$
$g'(x) = \frac{1}{x^2 + 1}$
For all $x$
$x^2 + 1 > x \Rightarrow \frac{1}{x^2 + 1}< \frac{1}{x}$
What this means is that the slope of $f(x)$ is always greater than the slope of $g(x)$ for any given $x$. Which means $g(x)$ is growing slower than $f(x)$.
-
5$\begingroup$ But arctan is a bounded function - it never exceeds $\pi/2$. This question is asking about a sequence of unbounded functions. $\endgroup$ Aug 7, 2015 at 21:33
-
$\begingroup$ @martycohen Oh, my bad. I didn't realize it specifically asking about an unbounded function since it wasn't written in the OP. $\endgroup$– Shane TAug 7, 2015 at 21:40