2
$\begingroup$

Assume that $f$ is a function from $(0,1]$ to $[0, \infty]$ that is strictly decreasing and satisfies $f(1)=0$, $f(0^{+})= \infty$ and $f(u^{2})=2 f(u)$. I do not know if it helps, but let us assume that $f$ is convex. There exists one family of solutions given by the set $\lbrace x \to -k \ln x, k>0 \rbrace$. Is there another family of solutions that does not lie in the above set?

$\endgroup$
2
$\begingroup$

Let $g(x):=f\big(\exp(-x)\big)$ for $x\in\mathbb{R}_{>0}$. First, I shall prove that every solution $f:(0,1)\to\mathbb{R}$ to the functional equation $f(u^2)=2\,f(u)$ for all $u\in(0,1)$ takes the form $$f(u)=-\ln(u)\,h\left(\frac{\ln\big(-\ln(u)\big)}{\ln(2)}\right)\tag{*}$$ for every $u\in(0,1)$, where $h:\mathbb{R}\to\mathbb{R}$ is a periodic function of period $1$.

We have $$g(2x)=f\big(\exp(-2x)\big)=f\Big(\big(\exp(-x)\big)^2\Big)=2\,f\big(\exp(-x)\big)=2\,g(x)$$ for all $x>0$.

Let $h(t):=\dfrac{g\big(2^t\big)}{2^t}$ for all $t\in\mathbb{R}$. Then, $$h(t+1)=\dfrac{g\big(2^{t+1}\big)}{2^{t+1}}=\frac{g\big(2\cdot 2^t\big)}{2\cdot 2^t}=\frac{2\cdot g(2^t)}{2\cdot 2^t}=\frac{g(2^t)}{2^t}=h(t)\,.$$ Therefore, $h$ is a periodic function with period $1$. Ergo, $$f(u)=g\big(-\ln(u)\big)=-\ln(u)\,h\left(\frac{\ln\big(-\ln(u)\big)}{\ln(2)}\right)$$ for every $u\in(0,1)$.

With the additional assumption that $f$ is continuous and nonnegative, $h$ is also continuous and nonnegative. From the limit conditions on $f$, we have $$\lim\limits_{x\to 0^+}\,g(x)=0\text{ and }\lim\limits_{x\to\infty}\,g(x)=\infty\,.$$
This shows that $h$ is bounded below above $0$, namely, $h(t)\geq l$ for every $t\in\mathbb{R}$, where $l>0$ is fixed. These are all required properties of $h$. That is, all solutions $f$ with the continuity, nonegativity, and limit conditions must be of the form (*), where $h:\mathbb{R}\to\mathbb{R}_{>0}$ is a continuous periodic function with period $1$.

In particular, if $h\equiv k$ for some constant $k>0$, then we get the family of solutions that the OP proposed. While it has not been taken into account that $f$ is strictly decreasing, it is an easy exercise to show that there are nonconstant periodic functions $h$ that makes $g$ a strictly increasing function, which in turns makes $f$ a strictly decreasing function. I have not thought about the convexity condition, but it is highly probable that only convex solutions take the form proposed by the OP.

If you only need $f$ to be strictly decreasing, then $h(t):=1+\epsilon\,\sin(2\pi t)$ for all $t\in\mathbb{R}$ works for any real number $\epsilon$ such that $|\epsilon|\leq \dfrac{\ln(2)}{\sqrt{4\pi^2+\big(\ln(2)\big)^2}}\approx 0.10965$. This gives another family with parameter $\epsilon$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.