3
$\begingroup$

I have asked this question here (*), but there are no answer.

Let $n \in \mathbb N^*$, $\{a_0,\ldots,a_n\} \subset \left] 0,+\infty\right]$. We suppose $Eq : \sum\limits_{k=0}^n a_k f^k(x)=0$ have no linear solution.

Determinate the solution of $Eq$ for $f \in C^\infty (\mathbb R)$, where $f^2(x)=f \circ f (x)$ and $f^0(x)=x$.

Reference: https://artofproblemsolving.com/community/c6h3164353_functional_equation

$\endgroup$
6
  • $\begingroup$ Are the $a_k$ operators or constants? In general this type of equation would be extremely hard to work with. An idea that comes to mind is to replace the functions with Carlemann matrices so this becomes a Carlemann Matrix polynomial equation, but it’s not clear to me if that helps with solving at all. $\endgroup$ Commented Oct 1, 2023 at 17:30
  • $\begingroup$ $\mathbb R_+^*=]0,+\infty[$ $\endgroup$
    – Dattier
    Commented Oct 1, 2023 at 17:34
  • 2
    $\begingroup$ A linear function $f(x)=\lambda x$ is a solution iff $\sum_{k=0}^n a_k\lambda^k=0$, so the assumption may be rephrased: “the polynomial $P(x):=\sum_{k=0}^n a_kx^k$ has no real roots”, right? $\endgroup$ Commented Oct 7, 2023 at 18:53
  • 2
    $\begingroup$ When you write =]0,+\infty[ or \subset]0,+\infty[ you don't have proper horizontal spacing between $=$ or $\subset$ and the bracket. $$ \begin{align} \text{right: } & \subset\left]0,+\infty\right[ \\ {} \\ \text{wrong: } & \subset]0,+\infty[ \end{align}$$ $\endgroup$ Commented Oct 8, 2023 at 16:46
  • 2
    $\begingroup$ Do you have a simple example of such an equation? What is the motivating example? $\endgroup$
    – Somos
    Commented Feb 5 at 23:29

2 Answers 2

2
$\begingroup$

There are no solutions, even if we only assume that $f$ has a derivative at every point.

Note that if $f(x)>x$ for all $x$, then $f(f(x))>f(x)>x$ etc, thus $\sum a_k f^k(x)\geqslant (\sum a_k)x$ that is positive for $x>0$. Analogously, if $f(x)<x$ for all $x$, then $f(f(x))<f(x)<x$ etc, and $\sum a_k f^k(x)\leqslant (\sum a_k)x$ that is negative for $x<0$. Thus, by Intermediate Value Theorem, there exists $x_0$ such that $f(x_0)=x_0$. If $p=f'(x_0)$, taking the derivative of our equation at $x_0$ we get $\sum a_kp^k=0$, thus, $f(x)=px$ is a solution of Eq.

$\endgroup$
1
$\begingroup$

As it was pointed in the comments, the case of a linear function $px$ should be excluded. Let's investigate the given sum on a linear function $f(x)=px+q$ with $q\ne0$.

We have $$\sum_{k=0}^n a_k (px+q)^{\circ k} = g_n x + qh_n,$$ where $g_n$ and $h_n$ are polynomials in $p,a_0,\dots,a_n$. It can be shown by induction on $n$ that we also have the following identity: $$g_n + (1-p) h_n = \sum_{k=0}^n a_k.$$ It implies that both $h_n$ and $g_n$ cannot be zero since the right-hand side is strictly positive. In other words, $\sum_{k=0}^n a_k (px+q)^{\circ k} \ne 0$ for any $p$ and $q\ne0$.

$\endgroup$

You must log in to answer this question.

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