2

Is there a continuous function $f$ defined on real number satisfying:

$$f(f(x))=-x^3+x.$$

I'm shame to say it's my homework and I've spend several hours on it. Also, I've tried to construct a function, but all failed.

Any hint will be appreciated, at least I want to konw whether $f$ exist.

Tianlalu
  • 5,387
Oolong Milktea
  • 218
  • 3
  • 17
  • 3
    You have spent several hours on it, so please include in your post all results you derived during these several hours, so that others attempting the question do not take several hours to derive things you already know from your work. – Sarvesh Ravichandran Iyer Nov 22 '18 at 02:41
  • If we restrict to an interval, $\frac{-1}{\sqrt 3} < x <\frac{1}{\sqrt 3}, $ we can construct a $C^\infty$ function $f$ that does this. My guess is that we cannot extend the function to the whole line, critical points and decreasing functions cause trouble – Will Jagy Nov 22 '18 at 02:47
  • $f(-1)$, $f(0)$ and $f(1)$ will all be zeroes of $f(x)$ – Seth Nov 22 '18 at 03:35
  • I'm sorry for being away.I've thought f must be a bijection,so unbounded,moreover f(f) is monotonic when x is big enough.If f is monotonic when x is big enough,then f(f) must be monotonically increasing.But it's just a assumption.My homework always has an elegant solution,so I think it won't take too long.Sorry for my poor English and ugly typing style. – Oolong Milktea Nov 22 '18 at 04:45
  • @Fume There is no way $f$ can be a bijection. If it is one-to-one then, since $f(f(0))=f(f(1))$ we get $0=1$. – Kavi Rama Murthy Nov 22 '18 at 05:48
  • Typo,I mean surjection. – Oolong Milktea Nov 22 '18 at 06:10

2 Answers2

2

The Range of $f$

If the range of $f$ is a proper subset of $\mathbb{R}$, then the range of $f \circ f$ is a proper subset of a proper subset of $\mathbb{R}$, which is itself a proper subset of $\mathbb{R}$.
The range of $-x^3 + x$ is $\mathbb{R}$, so it follows that the range of $f$ is $\mathbb{R}$.

Suppose $f$ is not monotonic.

It has a range of $\mathbb{R}$. and is continuous. Therefore there have to be (at least) three disjoint intervals values $A,B,C$ such that $\forall a\in A|\exists b\in B,c\in C|f(a)=f(b)=f(c)$. Assume we're talking about a "maximal" $A,B,C$, in the sense that there's not a similar triplet of which these three intervals are subsets.

$\{x|\exists a\in A | f(a)=x\}$ is an interval; call it $O$.

If $O$ isn't disjoint with $(A\cup B\cup C)$,

Define $O' = A\cap O \neq\emptyset$. (or $B$ or $C$; this part's symmetric)
Consider $a_0\in O'$. Because $a_0\in O$, we know there are $a_{-1}^a,b_{-1}^a,c_{-1}^a|f(a_{-1}^a)=f(b_{-1}^a)=f(c_{-1}^a)=a_0$.
Because $a_0\in A$, we know there exist $b_0,c_0|f(a_0)=f(b_0)=f(c_0)$. There must be $b_{-1}^b|f(b_{-1}^b)=b_0$ and $c_{-1}^c|f(c_{-1}^c)=c_0$.
Because $b_0\neq c_0\neq a_0$, and $a_{-1}^a\neq b_{-1}^a\neq c_{-1}^a$, we now have five distinct numbers $x\in\{a_{-1}^a,b_{-1}^a,c_{-1}^a,b_{-1}^b,c_{-1}^c\}$ such that $f(f(x)) = f(a_0)$
But if $f(f(x))=-x^3+x$, then there could be at most three such points. A contradiction.

If $O$ is disjoint with $(A\cup B\cup C)$,

then $\not\exists x\in(A\cup B\cup C)|f^{-1}(x)\in(A\cup B\cup C)$. Then we have
$\forall a\in A|\exists b\in B,c\in C|f(f(a))=f(f((b))=ff(((c))$
and we have intervals $A_{-1},B_{-1},C_{-1}$ where $\forall a_{-1}\in A_{-1}|\exists b_{-1}\in B_{-1},c_{-1}\in C_{-1}|f(f(a_{-1}))=f(f((b))=f(f((c))\in O$
But $(A\cup B\cup C)$ is distinct from $(A_{-1}\cup B_{-1}\cup C_{-1})$, and $-x^3+x$ has only one such triplet of overlapping intervals, another contradiction. (It's important here that $A,B,C$ were "maximal", as mentioned.)

If $f$ is monotonic?

If $f$ is monotonic, then $f\circ f$ is also monotonic, but $-x^3+x$ isn't, so that's not an option either.

We conclude that no such function exists.

Your teacher probably won't give you many points for the above proof; it needs a lot of cleaning up. But I haven't found any holes in it yet.

1

One of my classmate show me the answer below.I think it's probably true.
First it's easy to see while $x\to+\infty$ ,$f(f(x))\to-\infty$,while $x\to-\infty$ ,$f(f(x))\to+\infty$.

If $\limsup_{x\to+\infty}f(x)=+\infty$,then $\limsup_{x\to+\infty}f(f(x))=+\infty$,a contradiction.
(Notice $\limsup_{x\to+\infty}f(x)$ always exists.)

The same for $\limsup_{x\to-\infty}f(x)=-\infty$ And for $\limsup_{x\to+\infty}f(x)=-\infty$,$\limsup_{x\to+\infty}f(f(x))=-\infty$ can't be true. So $A=\limsup_{x\to+\infty}f(x)$ is bounded.

The same for $B=\liminf_{x\to+\infty}f(x)$,then $\liminf_{x\to+\infty}f(x)$ is bounded.

Choose any $\epsilon>0$, there is a $M$,while $x>M$,$B-\epsilon<f(x)<A+\epsilon$.

So $f(f(x))\to+\infty$ can't be true since $f$ is continuous on $\mathbb{R}$,bounded on the interval.

Sorry for my poor English and ugly typing.

Oolong Milktea
  • 218
  • 3
  • 17