Power functions in an o-minimal expansion of an ordered field

67 Views Asked by At

Let $\mathfrak{R}$ be a fixed o-minimal expansion of an ordered field $(R,<,+,\cdot,-,0,1)$ , A power function is a definable endomorphism of the multiplicative group $(Pos(R) , \cdot ,1)$ .

  • Claim(1) : if $f:Pos(R)\rightarrow R$ is definable, not identically $0$, and $f(xy)=f(x)f(y)$ for all $x,y\in Pos(R)$ , then $f$ is a power function. that is $f(x)>0$ for all $x\in Pos(R)$

  • Claim(2) : $(Pos(R) , \cdot ,1)$ is divisible, that is for any $x\in Pos(R)$ and for any $n\in\mathbb{N}$ , there is some $y\in Pos(R)$ such that $y^n=x$

  • Claim (3) : $x\rightarrow x^q$ where $q\in\mathbb{Q}$ is a power function. I have a problem with raising to rational powers

  • some useful facts about o-minimal expansion of an ordered field

  • fact(1) - the monotonicity theorem : Every unary definable function is ultimately differentiable (with ultimately definable derivative), ultimately either constant or strictly monotone, and has a limit at $+\infty$ in $R \cup\{-\infty, +\infty\}$

  • fact(2) : The mean value theorem and its usual consequences hold for definable functions.

  • fact(3) : The groups $(R,+)$ and $(\operatorname{Pos}(R), \cdot)$ have no proper, nontrivial definable subgroups.

  • my attempt in showing that claim(3) follows from claim(2) - assume that $(Pos(R) , \cdot ,1)$ is divisible consider the map $x\mapsto x^n$ . its kernel is a definable subgroup of $(Pos(R) , \cdot ,1)$ . by fact(3) it is injective and by divisibility it is surjective. thus we can define the nth root function, hence the rational exponents

any hints to why the three claims hold would be appreciated

1

There are 1 best solutions below

0
On

Thanks to @Atticus Stonestrom

  • Claim 1

By theorem 2.3 in the article DEFINABLE SETS IN ORDERED STRUCTURES.I by ANAND PILLAY AND CHARLES STEINHORN, $R$ is a real closed field, in particular every positive element is a square. then $f(x)=f(y\cdot y)=(f(y))^2\geq 0$.

Assume for a contradiction that $f(x)=0$ for some $x>0$. Then for any $y\in Pos(R)$ $f(y)=f(y/x\cdot x)=f(y/x)f(x)=0$ which is absurd.

  • Claim 2 follows from fact 3

consider the map $x\mapsto x^n$. restricted to $Pos(R)$, this is a definable group endomorphism of $(Pos(R),\cdot)$. certainly it has non-trivial image, and its image is a definable subgroup of $(Pos(R),\cdot)$, since $(Pos(R),\cdot)$ have no proper, non trivial definable subgroup, its image is everything.

  • Claim 3

We can use the same argument on the kernel of $x\mapsto x^n$ as in Claim 2 to establish the infectivity of that map. Or we can use the fact that every ordered group is torsion free.

Next, for every $n\in\mathbb{N}$ define $x^{\frac{1}{n}}$ to be the unique element $y$ in $Pos(R)$ such that $y^n=x$. For $q=\frac{n}{m}\in\mathbb{Q}^+$ define $x^{\frac{n}{m}}$ to be $(x^{\frac{1}{m}})^n$ and for $q \in\mathbb{Q}^-$ define $x^q$ to be $(x^{-1})^{-q}$. It remains to verify that it is well defined and have the same propreties as in the case of $\mathbb{R}$. in particular it verifies : $(x\cdot y)^q=x^q\cdot y^q$.

Finally, the map $x\mapsto x^q$ is definable:

For $q\in\mathbb{Q}^+$, the map $x\mapsto x^{\frac{n}{m}}$ where $n,m\in\mathbb{N}$ is defined by the formula $\phi_{n,m}(x,y)$ : $x^n=y^m \land 0<x \land 0<y$. And for $q<0$ $x^q=(x^{-1})^{-q}$, and since the map $x\mapsto x^{-1}$ is definable by the formula $\xi(x,y)$ ; $x\cdot y=1$. we can see that the map $x\mapsto x^q$ is definable as the composition of two definable maps.