1
$\begingroup$

Macpherson in a survey of homogeneous structures, states that there are many $\aleph_0$-categorical structures which are not homogeneous. Here homogeneity is the ultrahomogeneity that is defined as every isomorphism between two finite substructures of a structure $M$ can be extended to an automorphism of $M$. $\omega$-homogeneity means that any finite partial elementary mapping can be extended so that its domain includes any given element.

I am confused on this because it is well known that a $\aleph_0$-categorical structure is both atomic and countably saturated, and both atomic and countably saturated structures are $\omega$-homogeneous. This actually means that a $\aleph_0$-categorical structure is ultrahomogeneous. Where is wrong here?

$\endgroup$
1
  • 3
    $\begingroup$ By the way, $\aleph_0$-homogeneity and $\omega$-homogeneity mean the exact same thing. So, in addition to Emil's answer, you can refer to the discussion on your other question about why $\omega$-homogeneity is not the same as ultrahomogeneity. $\endgroup$ Commented Jul 19, 2020 at 12:32

1 Answer 1

9
$\begingroup$

You are confusing several notions of homogeneity. Saturated structures, and therefore also $\aleph_0$-categorical structures, are homogeneous, but not necessarily ultrahomogeneous. This means that every finite partial elementary mapping extends to an automorphism.

$\omega$-homogeneity is in fact an even weaker property: it says that any finite partial elementary mapping can be extended so that its domain includes any given element. However, this is equivalent to the property above for countable structures.

Ultrahomogeneity of $\omega$-saturated structures implies quantifier elimination, hence it is not implied by any standard model-theoretic properties that are invariant by expansion of the language with definable predicates.


In more detail, let me try to deconfuse Macpherson’s terminology by reviewing the relevant properties (using more standard terminology that does not drop the ultra- prefixes) and their connections. In what follows, $M$ is a structure, and $\kappa$ is an infinite cardinal.

  • $M$ is $\kappa$-homogeneous if for every partial elementary map $f\colon M\rightharpoonup M$ such that $|f|<\kappa$, and for every $a\in M$, there exists a partial elementary map $g\supseteq f$ such that $a\in\operatorname{dom}(g)$.

  • $M$ is strongly $\kappa$-homogeneous if every partial elementary map $f\colon M\rightharpoonup M$ such that $|f|<\kappa$ extends to an automorphism of $M$.

  • If $\kappa=|M|$, and $M$ is $\kappa$-homogeneous, it is in fact strongly $\kappa$-homogeneous. Such structures are simply called homogeneous.

  • $M$ is $\kappa$-ultrahomogeneous if for every partial isomorphism $f\colon M\rightharpoonup M$ such that $|f|<\kappa$, and for every $a\in M$, there exists a partial isomorphism $g\supseteq f$ such that $a\in\operatorname{dom}(g)$.

  • $M$ is strongly $\kappa$-ultrahomogeneous if every partial isomorphism $f$ such that $|f|<\kappa$ extends to an automorphism of $M$.

  • $M$ is ultrahomogeneous if it is $\kappa$-ultrahomogeneous (or equivalently, strongly $\kappa$-ultrahomogeneous) for $\kappa=|M|$.

The basic properties are:

  • If $M$ is $\kappa$-saturated, it is $\kappa$-homogeneous.

  • If $M$ is atomic, it is $\omega$-homogeneous.

  • The following are equivalent:

    1. $M$ is $\kappa$-ultrahomogeneous;
    2. $M$ is $\kappa$-homogeneous, and every partial isomorphism $M\rightharpoonup M$ is elementary.
  • Likewise for strong $\kappa$-ultrahomogeneity.

  • If $M$ is in a finite relational language, or if it is $\omega$-saturated, the following are equivalent:

    1. Every partial isomorphism $M\rightharpoonup M$ is elementary.
    2. $M$ has quantifier elimination.
  • Consequently, if $M$ is in a finite relational language, or if it is $\omega$-saturated, the following are equivalent:

    1. $M$ is $\kappa$-ultrahomogeneous.
    2. $M$ is $\kappa$-homogeneous, and $M$ has quantifier elimination.
$\endgroup$
5
  • $\begingroup$ What is the exact difference between strong ultrahomogeneous and ultrahomogeneous? The first part of each definition is the same, ie. "if every partial isomorphism f such that". So the difference lies in the second part of each difinition. $\endgroup$ Commented Mar 22, 2023 at 22:06
  • 1
    $\begingroup$ No, that is important. For example, by back and forth, this definition of κ-ultrahomogeneous means there is an automorphism. So it is the same as strongly κ -homogeneous. If I am not wrong, strongly κ-homogeneous should be stronger than κ -homogeneous as the name suggests. $\endgroup$ Commented Mar 23, 2023 at 7:35
  • $\begingroup$ This definition of $\kappa$-ultrahomogeneous means there is an automorphism: nonsense. Of course it is strictly stronger. For example, let $M$ be the lexicographic product $(\omega_1+1)\times\mathbb Q$ as a linearly ordered set. Then $M$ is $\omega$-ultrahomogeneous (every dense linear order without endpoints is), but it is not strongly $\omega$-ultrahomogeneous, as the partial isomorphism $(0,0)\mapsto(\omega_1,0)$ does not extend to an automorphism of $M$ (there are $\aleph_1$ elements above the former element, but only $\aleph_0$ above the latter). $\endgroup$ Commented Mar 23, 2023 at 8:32
  • $\begingroup$ These two definitions are the same for countable models in which automorphism can follow for sure by back and forth. $\endgroup$ Commented Mar 23, 2023 at 16:10
  • $\begingroup$ It is better noted that for countable structure, both definitions are equivalent $\endgroup$ Commented Mar 23, 2023 at 21:32

You must log in to answer this question.

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