Professional Documents
Culture Documents
ULTRAMETRIC SPACES
YOSHITO ISHIKI
arXiv:2303.17471v1 [math.MG] 30 Mar 2023
1. Introduction
For a class C of metric spaces, we say that a metric space (X, d) is
C-injective if for all pair (Y, e) and (Z, h) of metric spaces in C and
isometric embeddings φ : Y → Z and f : Y → X, there exists an iso-
metric embedding F : Z → X such that F ◦ φ = f . We denote by
F the class of all finite metric spaces. Urysohn constructed a separa-
ble complete F-injective metric space and proved its uniqueness up to
isometry, which space is today called the Urysohn universal (metric)
space. Let (U, ρ) denote the Urysohn universal metric space. Our main
subjects of this paper are non-Archimedean analogues of (U, ρ). For
more information of (U, ρ) and related topics, we refer the readers to,
for instance, [1] [10], [4], [7], [8], [14], [11] and [12].
To explain our main results and backgrounds, we prepare some no-
tations and notions. We say that a set R is a range set if it is a
subset of [0, ∞) and contains 0. A metric on a set X is said to be
an ultrametric if it satisfies the so-called strong triangle inequality
d(x, y) ≤ d(x, z) ∨ d(z, y) for all x, y, z ∈ X, where ∨ stands for the
maximal operator on R. For a range set R, an ultrametric space (X, d)
is R-valued if d(x, y) ∈ R for all x, y ∈ R. For a range set R, we denote
by N(R) the class of all finite R-valued ultrametric spaces.
For a countable range set R, we say that an ultrametric (X, d) is
the R-Urysohn universal ultrametric space if it is a separable complete
N(R)-injective R-valued ultrametric space. As is the case of (U, ρ),
2020 Mathematics Subject Classification. Primary 54E35, Secondary 51F99.
Key words and phrases. Urysohn universal ultrametric spaces.
1
2 YOSHITO ISHIKI
2. Preliminaries
κ ∈ ω0 or κ = ω0 . For a metric space (X, d), for a ∈ X, and for
r ∈ (0, ∞), we denote by B(a, r) the closed ball centered at a of radius
r.
The proofs of next three lemmas are presented in [13, Propositions
18.2, 18.4 and 18.5].
Lemma 2.1. Let (X, d) be an ultrametric space. Then the following
are true:
(1) for all x, y, z ∈ X, the inequality d(x, z) < d(z, y) implies
d(z, y) = d(x, y).
(2) For all a ∈ X, r ∈ (0, ∞), and for all q ∈ B(a, r) we have
B(a, r) = B(q, r).
(3) For all a, b ∈ X, and for r, l ∈ (0, ∞), if B(a, r) ∩ B(b, l) 6= ∅,
then we have B(a, r) ⊆ B(b, l) or B(b, r) ⊆ B(a, r).
A subset E of [0, ∞) is said to be semi-sporadic if there exists a
strictly decreasing sequence {ai }i∈Z≥0 in (0, ∞) such that limi→∞ ai = 0
and E = {0} ∪ { ai | i ∈ Z≥0 }. A subset of [0, ∞) is said to be tenuous
if it is finite or semi-sporadic (see [5]). For a range set R, we denote by
TEN(R) the set of all tenuous range subsets of R. In this paper, we
often represent a restricted metric as the same symbol to the ambient
one. We now provide the definition of the petaloid ultrametric space
that is introduced in [6] by the author.
Definition 2.1. Let R be an uncountable range set. We say that
a metric space (X, d) is R-petaloid if it is an R-valued ultrametric
space and there exists a family {Π(X, S)}S∈TEN(R) of subspaces of X
satisfying the following properties:
(P1) The set Π(X, {0}) is a singleton.
(P2) If S ∈ TEN(R) satisfies S 6= {0}, then (Π(X, S), d) is a sepa-
rable complete
S N(S)-injective S-ultrametric space.
(P3) We have S∈TEN(R) Π(X, S) = X.
(P4) If S, T ∈ TEN(R), then Π(X, S) ∩ Π(X, T ) = Π(X, S ∩ T ).
(P5) If S, T ∈ TEN(R) and x ∈ Π(X, T ), then d(x, Π(X, S)) belongs
to (T \ S) ∪ {0}.
4 YOSHITO ISHIKI
4. Hyperspaces
For a metric space (X, d), and for a subset A of X, we write d(x, A) =
inf a∈A d(x, a). For a metric space (X, d), we denote by K(X) the set
of all non-empty compact subsets of X. We also denote by HDd the
Hausdorff distance on K(X) induced from d. More precisely, we define
HD d (A, B) = max{supa∈A d(a, B), supb∈B d(b, A)}.
Let (X, d) be an ultrametric space, and r ∈ (0, ∞). For A ∈ K(X)
we denote by hAic(r) the set { B(a, r) | a ∈ A }. Remark that it is can
happen that B(a, r) ∈ hAic(r) contains some elements of X \ A. By
(3) in Lemma 2.1 and the compactness of A, the set hAic(r) is finite.
The next theorem is an analogue of [9, Theorem 5.1] and [5, Corollary
2.30].
Proposition 4.1. Let (X, d) be an ultrametric space. Then for all
subsets A and B of X, the value HD d (A, B) is equal to the minimum
r ∈ R such that hAic(r) = hBic(r) .
Proof. Let u be the infimum of r ∈ R such that hAic(r) = hBic(r) . Put
S = d(A2 )∪d(B 2 ). Then S is tenuous (see [5, Corollary 2.28]). If u 6∈ S,
then we can take a, b ∈ S such that u ∈ (a, b) and (a, b) ∩S = ∅. In this
setting, we have B(x, r) = B(x, a) for all x ∈ A∪B and for all r ∈ (u, b).
Thus, we have hAic(a) = hBic(a) , which is a contradiction. Therefore
u ∈ S, and hence u ∈ R. We also notice that u is the minimum. By
the definition of the Hausdorff distance, we have HD d (A, B) ≤ u. To
obtain the opposite inequality, take arbitrary l with HD d (A, B) < l.
Then we see that supa∈A d(a, B) ≤ l and supb∈B d(b, A) ≤ l. Then
hAic(l) = hBic(l) . Hence u ≤ HDd (A, B). This completes the proof.
URYSOHN 7
We now show that the property (P5) is true for (K(X), HDd ). Take
S, T ∈ TEN(R) with S 6= T and B ∈ Π(K(X), T ). Let L stand
for the distance between B and Π(K(X), S). We shall show that L
belongs to {0} ∪ (T \ S). We may assume that B 6∈ Π(K(X), S). Then
B ⊆ Π(X, T ) and B 6⊆ Π(X, S). We put H = { d(y, Π(X, S)) | y ∈ B }.
Due to (P5) for (X, d), we see that H ⊆ {0} ∪ (T \ S). Put h = max H
and take z ∈ B such that d(z, Π(X, S)) = h. The existence of h is
guaranteed by the fact that H is a subset of the tenuous set T . Note
that h ∈ T \ S. We also take a maximal finite h-net Q of B with z ∈ Q.
For all p ∈ Q, take v(p) ∈ Π(X, S) with d(p, v(p)) = d(p, Π(X, S)) (see
Lemma 2.2). Put V = { v(p) | p ∈ Q }. Then HD d (B, V ) ≤ h.
For the sake of contradiction, suppose that there exists F ∈ Π(X, S)
with HD d (B, F ) < h. Put l = HD d (B, F ). Then Proposition 4.1
implies that hBic(l) = hF ic(l) . Hence there exists f ∈ F such that
d(z, f ) ≤ l < h (see (2) in Lemma (2)). This is a contradiction to the
fact that h = inf x∈Π(X,S) d(z, x). Thus, for all F ∈ Π(X, S), we have
h ≤ HD d (B, F ) and HD d (B, V ) = h. This means that L = h and
h ∈ T \ S. This finishes the proof.
5. Existence theorem
This section is devotes to the proof of the following theorem:
Theorem 5.1. Let R be a range set. Let (X, d) be a complete R-valued
N(R, ω0 )-injective ultrametric space. Then (X, d) contains a subspace
F isometric to the R-Urysohn universal ultrametric space. Moreover,
for every compact subset K of X satisfying that d(K 2 ) ⊆ R, we can
take F so that K ⊆ F .
In what follows, the symbols R, (X, d) and K are the same objects
as the statement in Theorem 5.1.
Definition 5.1. We say that a family {Ω[a, r]}a∈X,r∈R\{0} of subsets
of X is an R-seed of X if the following conditions are satisfied for all
a ∈ X, and r ∈ R \ {0}:
(1) We have a ∈ Ω[a, r];
(2) the set Ω[a, r] is r-equidistant;
(3) we have Card(Ω[a, r]) = ℵ0 .
Remark that since (X, d) is (R, ω0 )-haloed, there exists an R-seed
(see Theorem 3.2).
Definition 5.2. Fix an R-seed {Ω[a, r]}a∈X,r∈R\{0} and a point P ∈ X.
For S ∈ TEN(R), we say that a point x ∈ X is an S-heir of P if there
m−1
exists sequences {vi }m i=0 in X and {ri }i=0 in S such taht
(1) v0 = P ;
(2) vm = x;
(3) vi+1 ∈ Ω[vi , ri ] for all i ∈ {0, . . . , m − 1};
URYSOHN 9
(Yoshito Ishiki)
12 YOSHITO ISHIKI