Characterizing affine subspaces order-theoretically

91 Views Asked by At

Let $V$ denote a real vectorspace and $\mathrm{Con}(V)$ denote the poset of convex subsets of $V$. The goal is to identify those elements of $\mathrm{Con}(V)$ that happen to be affine subspaces of $V$ in a purely order-theoretic manner. Something like so:

Definition. Let $P$ denote a poset satisfying [blah]. Then $x \in P$ is called affine iff...

Proposition. The affine elements of $\mathrm{Con}(V)$ are precisely the affine subspaces of $V$.

What definition(s) are available to do this?

1

There are 1 best solutions below

2
On BEST ANSWER

Clearly, it suffices to define lines in a order-theoretically way, as $X$ is affine if and only if it contains (it is bigger than) all the lines through its points.

All sets are supposed convex. The order relation is "$X$ is smaller thatn $Y$ iff $X\subset Y$"

Definition: X is polygonal if $\exists$ a convex $Y\supset X$ such that $Y\setminus X$ is convex. P

(In practice $X$ is polygonal if and only if it is convex but not stricly convex)

Definition: $Y$ is long if it is polygonal and for any $Y\supset X$ such that $Y\setminus X$ is convex, we have $Y\setminus X$ is unbounded.

Definition: a line is a minimal long set, more precisely, $Y$ is a line if and only if $Y$ is long and has no long subset.

Fact: lines with this definition conincide with affine lines.

Proof: Any convex subset of a line is a segment or an half line. These are not long with respect to the present definition. On the other hand, let $Y$ be long and of dimension $>1$, and let $\pi$ be the affine plane containing $Y$. If $Y=\pi$ then it is not minimal, as it contains a line as a long subset. If $Y\neq \pi$ then it has a boundary and, by pushing that boundary inside $Y$, we find a convex subset of $Y$ wich is long.

Definition: $X$ is affine if and only if it is a single point or, for any line $L$ we have $$X\cap L\neq \emptyset \text{ in at least two points } \Rightarrow L\subset X$$

Fact: $X$ is affine iff it is an affine subspace of $V$.

So, in general, following this scheme, you need:

1) a poset $(P,<)$ such that

i) ther is a maximal element $V$

ii) for any $Y\in P$ there is an involution map on $\{X\ :\ X<Y\}$ wich reverses the order: $X\mapsto Y\setminus X$.

2) a class of convex elements of $P$.

3) a class of bounded (convex) elements of $P$.

Now, points are defined as minimal elements of $P$, and $X\cap Y\neq \emptyset$ reads: "there is a point $q$ so that $q<X$ and $q<Y$".

Next, you can give the definitions of polygonal and long as above and define lines as minimal long elements.

Finally, you have the definition of affine as an element $X$ which is either a point or so that if $X\cap L\neq\emptyset$ in two points for a line $L$, then $L<X$.