3
$\begingroup$

I have noticed experimentally that the following question has a positive answer.

Is it true that for all even and convex functions $f$, $g$:

$$\int_0^1 f(\sin(1/x)) \times g(\cos(1/x)) dx \leq \int_0^1 f(\sin(1/x)) dx \times \int_0^1 g(\cos(1/x))dx? $$

$\endgroup$
0

1 Answer 1

7
$\begingroup$

$\newcommand\abs[1]{\lvert#1\rvert}$It already suffices that $f$ and $g$ be even and nondecreasing on $[0,1]$ (which of course is the case if $f$ and $g$ are even and convex). Indeed, then the identity $\abs\sin^2+\abs\cos^2=1$ implies that for all real $u$, $v$ we have $$(\abs{\sin u}-\abs{\sin v})(\abs{\cos u}-\abs{\cos v})\le0\tag{1}\label{1}$$ and hence $$\begin{aligned}h(u,v)&:=[f(\sin u)-f(\sin v)][g(\cos u)-g(\cos v)] \\ &=[f(\abs{\sin u})-f(\abs{\sin v})][g(\abs{\cos u})-g(\abs{\cos v})]\le0, \end{aligned}$$ so that the difference between the left-hand side of your inequality and its right-hand side is $$\frac12\,\int_0^1\int_0^1 dx\,dy\,h\Big(\frac1x,\frac1y\Big)\le0. $$


One may note that the above reasoning holds if in the inequality in question one replaces all instances of $1/x$ by $k(x)$, where $k$ is any Borel-measurable function from $(0,1)$ to $\mathbb R$. Also, one can replace $\sin$ and $\cos$ by any functions $S$ and $C$ from $\mathbb R$ to $\mathbb R$ that are Borel-measurable and (say) bounded and "negatively dependent" in the sense that \eqref{1} holds with $S$ and $C$ in place of $\sin$ and $\cos$: $$(\abs{S(u)}-\abs{S(v)})(\abs{C(u)}-\abs{C(v)})\le0 \tag{2}\label{2}$$ for all real $u$, $v$. In particular, inequality \eqref{2} will hold if $|S|$ is any increasing function and $|C|$ is any decreasing one (or vice versa).

$\endgroup$
4
  • $\begingroup$ TeX note: || for the absolute value is usually harmless, but sometimes TeX needs explicit pointers to the difference between opening and closing bars, and other uses of bars; consider $|\sin|$ |\sin| compared to $\lvert\sin\rvert$ \lvert\sin\rvert. I have edited accordingly. \\ Doubtless such pairs $S$ and $C$ exist abstractly, but are there any obvious interesting pairs other than modifications of the sine and cosine functions? $\endgroup$
    – LSpice
    Commented Mar 4, 2022 at 21:04
  • 1
    $\begingroup$ @LSpice : The latter displayed inequality in the answer will hold if e.g. $|S|$ is any increasing function and $|C|$ is any decreasing one (or vice versa). I do not know if this is interesting, though. $\endgroup$ Commented Mar 4, 2022 at 21:12
  • 1
    $\begingroup$ @LSpice : Thank you for your edits. Indeed, $\lvert\sin\rvert$ looks much better than $|\sin|$. $\endgroup$ Commented Mar 4, 2022 at 21:35
  • 1
    $\begingroup$ Why the downvote? Care to explain? $\endgroup$ Commented Apr 13, 2022 at 16:53

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .