Topology and its Applications 339 (2023) 108585

Contents lists available at ScienceDirect

Topology and its Applications

journal homepage: www.elsevier.com/locate/topol

A result concerning the Lipschitz realcompactification of the

product of two metric spaces ✩
M. Isabel Garrido a,∗ , Ana S. Meroño b
a
Instituto de Matemática Interdisciplinar (IMI) and Departamento de Álgebra, Geometría y Topología,
Universidad Complutense de Madrid, 28040-Madrid, Spain
b
Departamento de Matemática Aplicada a la Ingeniería Industrial, Universidad Politécnica de Madrid,
28006-Madrid, Spain

a r t i c l e i n f o a b s t r a c t

Article history: For a metric space (X, d), we consider the so-called Lipschitz realcompactification
Received 22 December 2021 of X, denoted by H(Lipd (X)). In this note we give a result concerning the equality
Received in revised form 1 February
2022
Accepted 2 May 2023 H(Lipd+ρ (X × Y )) = H(Lipd (X)) × H(Lipρ (Y ))
Available online 16 May 2023

Dedicated to Professor José M.R. for the product of the two metric spaces (X, d) and (Y, ρ). More precisely, we prove
Sanjurjo on the occasion of his 70th that such equality holds if and only if H(Lipd (X)) = X  or H(Lipρ (Y )) = Y , where
birthday  and Y denote the completion of X and Y respectively, or equivalently, if and
X
MSC: only if the Lipschitz realcompactification of one of the factors X or Y is as simple
primary 54D30, 54D60 as possible. We also point out that our result is, in fact, a true generalization of a
secondary 54D35, 54E40, 54C20 known theorem by Woods about the Samuel compactification of the product of two
metric spaces.
Keywords: © 2023 The Authors. Published by Elsevier B.V. This is an open access article
Metric spaces under the CC BY-NC-ND license (http://
Real-valued Lipschitz functions creativecommons.org/licenses/by-nc-nd/4.0/).
Lipschitz realcompactification
Samuel compactification

1. Introduction

For a metric space (X, d) we will denote by sd X its Samuel compactification. It is known that sd X is the
smallest compactification of X (with the usual order in the family of the compactifications) with the property
that each function in Ud∗ (X) (the set of all the bounded real-valued uniformly continuous on X) can be
continuously extended to it. We refer to the nice work by Woods [6] where this space is thoroughly studied.
Precisely, in that paper, it is proved the following theorem concerning the Samuel compactification of the

✩
Partially supported by Ministerio de Economía y Competitividad - PGC2018-097286-B-I00 (Spain).
* Corresponding author.
E-mail addresses: maigarri@mat.ucm.es (M.I. Garrido), anasoledad.merono@upm.es (A.S. Meroño).

https://doi.org/10.1016/j.topol.2023.108585
0166-8641/© 2023 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
2 M.I. Garrido, A.S. Meroño / Topology and its Applications 339 (2023) 108585

product of two metric spaces. Recall that, for the metric spaces (X, d) and (Y, ρ), we consider on the product
space X × Y the metric (d + ρ)((x1 , y1 ), (x2 , y2 )) = d(x1 , x2 ) + ρ(y1 , y2 ), for every (x1 , y1 ), (x2 , y2 ) ∈ X × Y .
As we can read in [6], the metric d + ρ is the “correct” metric in a categorical sense.

Theorem. [Woods] For two metric spaces (X, d) and (Y, ρ), the following are equivalent:

(a) sd+ρ (X × Y ) = sd X × sρ Y .
(b) At least one of (X, d) and (Y, ρ) is totally bounded.

Taking into account that sd X and sd˜X  are equivalent compactifications of X (see [6]), where (X, d)
˜
denotes the completion of (X, d), and that condition (b) can be translated as “the completion of at least
one of these metric spaces must be compact”, then we can add in the above result the following equivalent
condition:

 or sρ Y = Y .
(c) sd X = X

And this means that the equality between the two compactifications of X × Y appearing in (a) only
happens when the Samuel compactification of one of the factors X or Y is as simple as possible.
We are mainly interested here in real-valued Lipschitz functions on metric spaces. At this respect, and
using the classical Katětov theorem asserting that every bounded real-valued uniformly continuous function
can be uniformly approximated by Lipschitz functions (see e.g. [2]), we have that sd X is also the smaller
compactification of X where every bounded real-valued Lipschitz function can be continuously extended.
Therefore, the Samuel compactification, also known as the uniform compactification, could be called the
Lipschitz compactification of a metric space.
Then, it is natural to ask if there exists an analogous theorem to the above when we consider unbounded
Lipschitz functions, that is, for the so-called Lipschitz realcompactification of a metric space X. We refer to
[1] and [3] where this realcompactification was introduced and studied. Recall that such space is denoted
by H(Lipd (X)), being Lipd (X) the family of all the real-valued Lipschitz functions on X. The letter H
cames from the word “homomorphism”, since in fact H(Lipd (X)) is the space of all the real unital vector
lattice homomorphisms on Lipd (X), endowed with the topology inherited as a subspace of the product
space RLipd (X) .
In this paper, we are interested in the equality

H(Lipd+ρ (X × Y )) = H(Lipd (X)) × H(Lipρ (Y )).

At this respect, we will see that an analogous result to the above for the Lipschitz realcompactifications can
be stated (Theorem 9). Moreover, we will prove that in fact our theorem is a true generalization of the already
mentioned Woods’ Theorem (Corollary 14). But, before obtaining these results, we will devote sections 2
and 3 to give some preliminary results about realcompactifications in the context of the general Tychonoff
spaces, and to summarize some relevant facts in relation with this particular Lipschitz realcompactification,
all of them mainly taken from [1] and [3].

2. Preliminaries on realcompactifications

Most of the results contained in this section can be seen in [1] (see also [3]) and we include them here in
order to follow more easily the rest of the paper. Thus, for a Tychonoff topological space X, we consider L
to be a unital vector lattice of real-valued continuous functions on X, i.e., a vector lattice containing the
 M.I. Garrido, A.S. Meroño / Topology and its Applications 339 (2023) 108585 3

constant function 1, and hence all the constant functions. We define H(L) as the set of all the real unital
vector lattice homomorphisms on L. That is, ϕ : L → R belongs to H(L) whenever it satisfies:

(i) ϕ(λf + μg) = λϕ(f ) + μϕ(g), for all f, g ∈ L and all λ, μ ∈ R.

(ii) ϕ(|f |) = |ϕ(f )|, for all f ∈ L.
(iii) ϕ(1) = 1.

Note that in this case ϕ is positive, i.e., ϕ(f ) ≥ 0 when f ≥ 0, and also ϕ preserves the supremum and the
infimum of functions, i.e., ϕ(f ∨ g) = ϕ(f ) ∨ ϕ(g) and ϕ(f ∧ g) = ϕ(f ) ∧ ϕ(g).
We regard H(L) as a subspace of the product space RL , where the real line R is endowed with the usual
topology. It is easy to check that H(L) is closed in RL , and therefore it is a realcompact space. In the same
way, we can consider L∗ the unital vector sublattice formed by the bounded functions in L. Now the space
H(L∗ ) is in fact compact, and it is easy to see that H(L) can be considered as a topological subspace of
H(L∗ ). Hence, we can write H(L) ⊂ H(L∗ ).
Moreover, when the family L separates points and closed sets of X, i.e., when for every closed subset F
of X and x ∈ X \ F there exists some f ∈ L such that f (x) ∈ / f (F ), then we can embed the topological
space X (in a dense way) in H(L) and also in H(L∗ ). And this means, in particular, that H(L) is a
realcompactification of X and H(L∗ ) is a compactification of X. And thus, we have

X ⊂ H(L) ⊂ H(L∗ ),

where the first inclusion must be understood as x  (f (x))f ∈L , whereas the second one is just (zf )f ∈L 
(zf )f ∈L∗
On the other hand, it is clear that every function in L (respectively in L∗ ) admits a unique continuous
extension to H(L) (resp. to H(L∗ )). In fact, H(L) (resp. H(L∗ )) is characterized (up to equivalence) as
the smallest realcompactification (resp. compactification) of X with this property. Note that we are here
considering the usual order in the set of all the realcompactifications and compactifications of X. Namely,
we say that α1 X ≤ α2 X whenever there is a continuous mapping h : α2 X → α1 X leaving X pointwise
fixed. And we say that α1 X and α2 X are equivalent, we write α1 X = α2 X, whenever α1 X ≤ α2 X and
α2 X ≤ α1 X, and this implies the existence of a homeomorphism between α1 X and α2 X leaving X pointwise
fixed.
Note that if we consider L = C(X), the set of all the real-valued continuous functions on X, then
H(C(X)) = υX is the Hewit-Nachbin realcompactification of X and H(C ∗ (X)) = βX is now its Stone-
Čech compactification.
When (X, d) is a metric space, two important unital vector lattices of real-valued functions can be also
considered. Namely, the set Lipd (X) of all the real-valued Lipschitz functions, and the set Ud (X) of all the
real-valued uniformly continuous functions defined on X. At this point we can say that H(Ud∗ (X)) is the
Samuel compactification sd X of X since, as we said in the Introduction, this compactification is characterized
as the smallest compactification where all the real and bounded uniformly continuous functions on X can
be continuously extended ([6]). Now, using the fact that L and its uniform closure L define equivalent
realcompactifications (see [1]), together with the above mentioned Katětov theorem, we can derive that

H(Lip∗d (X)) = H(Ud∗ (X)) = sd X.

Nevertheless, we have that H(Ud (X)) and H(Lipd (X)), which we will call respectively the Samuel
realcompactification and the Lipschitz realcompactification of the metric space (X, d), could be differ-
ent realcompactifications. Indeed, this is the case for an infinite space X with the 0 − 1 metric, since
4 M.I. Garrido, A.S. Meroño / Topology and its Applications 339 (2023) 108585

H(Ud (X)) = H(C(X)) = υX and H(Lipd (X)) = H(Lip∗ (X)) = sd X = βX. We refer to [3] where both
realcompactifications are studied in detail.

3. The Lipschitz realcompactification

In this section, we present some results (without proofs) taken from [3] that will be useful for our purposes.
First of all, note that H(Lipd (X)) is the smallest realcompactification of the metric space (X, d) where every
function f ∈ Lipd (X) can be continuously extended, and also that it is contained in H(Lip∗d (X)) = sd X,
the Samuel compactification of X.
The following result gives us a description of H(Lipd (X)) as a subspace of sd X.

Proposition 1. Let (X, d) be a metric space and x0 ∈ X. Then


H(Lipd (X)) = clsd X Bd [x0 , n]
n∈N

where Bd [x0 , n] denotes the closed ball in X around x0 and radius n, and clsd X denotes the closure in sd X.

Now, from the above, we can easily obtain a characterization of Lipschitz realcompactness, where a space
X is said to be Lipschitz realcompact whenever X = H(Lipd (X)).

Proposition 2. A metric space (X, d) is Lipschitz realcompact, if and only if it satisfies the Heine-Borel
property, i.e., every closed and bounded subset in X is compact.

The next Corollary follows at once, but we consider it has some interest in order to see the difference
between Lipschitz realcompactness and just realcompactness in the context of metric spaces.

Corollary 3. A Banach space is Lipschitz realcompact, if and only if it has finite dimension.

In [6] we can see that for a metric subspace (B, dB ) of (X, d) we have that clsd X B = sdB B. The following
result says that the same happens in the Lipschitz realcompactification whenever B is in addition bounded.

Proposition 4. Let (X, d) a metric space and (B, dB ) a bounded metric subspace. Then, clH(Lipd (X)) B =
clsd X B = sdB B.

As a consequence of all the above, we can derive a similar result to the already noticed for the Samuel
 are equivalent compactifications of X ([6]).
compactification. Namely, sd X and sd˜X

Proposition 5. Let (X, d) be a metric space and (X,  its completion. Then H(Lipd (X)) and H(Lip  (X))
 d) 
d
are equivalent realcompactifications of X.

We finish this section giving some examples where we use some of the above results.

Examples 6. We are going to consider different metrics on the space of natural numbers N:

(1) If du denotes the usual metric, then H(Lipdu (N)) = N, since (N, du ) is Heine-Borel.
(2) If d0-1 is the 0 − 1 metric, then H(Lipd0-1 (N)) = H(Lip∗d0-1 (N)) = sd0-1 N = βN.
(3) If d is given by d(n, m) = |1/n − 1/m|, then the completion is N  = {1/n : n ∈ N} ∪ {0} endowed with
 
the usual metric. And, H(Lipd (N)) = H(Lipd(N)) = N = sd N.
 M.I. Garrido, A.S. Meroño / Topology and its Applications 339 (2023) 108585 5

(4) Let ϕ : N → N × N be a bijection. If we consider the product space (N × N, d0-1 + du ), we can define d
on N as d(n, m) = (d0-1 + du )(ϕ(n), ϕ(m)). Clearly, (N, d) and (N × N, d0-1 + du ) are isometric spaces.
Then, in this case, (N, d) is complete, not Heine-Borel, and H(Lipd (N)) cannot be compact since there
exist unbounded Lipschitz functions. In fact, as we will see as consequence of Theorem 9, we have that
H(Lipd (N)) = H(Lipd0-1 +du (N × N)) = βN × N (see Example 11).

4. The main result

As we have already announced, we are interested in determining the equivalence between the realcom-
pactifications H(Lipd+ρ (X × Y )) and H(Lipd (X)) × H(Lipρ (Y )), of the metric space (X × Y, d + ρ). At
this respect, we have a first result saying that one of the inequalities is always true.

Proposition 7. Let (X, d) and (Y, ρ) be metric spaces. Then,

H(Lipd+ρ (X × Y )) ≥ H(Lipd (X)) × H(Lipρ (Y )).

Proof. In order to see this inequality we need to prove that the inclusion map,

i : X × Y → H(Lipd (X)) × H(Lipρ (Y ))

can be continuously extended to H(Lipd+ρ (X × Y )). For that, it is sufficient to extend the two coordinate
functions,

π1 ◦ i : X × Y → H(Lipd (X))
π2 ◦ i : X × Y → H(Lipρ (Y ))

where π1 and π2 denote, respectively, the first and the second projection map.
Next, in order to extend π1 ◦ i : X × Y → H(Lipd (X)) ⊂ RLipd (X) it is again sufficient to extend
every coordinate, namely, for every f ∈ Lipd (X), the corresponding function πf ◦ (π1 ◦ i), that maps
(x, y)  f (x). Since, clearly, this function belongs to Lipd+ρ (X × Y ) then it can be continuously extended
to H(Lipd+ρ (X × Y )). The same is true for π2 ◦ i, and that completes the proof. 

Proposition 8. Let (X, d) and (Y, ρ) be metric spaces. If (X, d) or (Y, ρ) satisfies that every bounded subset
is totally bounded then

H(Lipd (X)) × H(Lipρ (Y )) ≥ H(Lipd+ρ (X × Y )).

Proof. Firstly note that, according to Proposition 5, the above inequality can be written as,

 × H(Lipρ̃ (Y )) ≥ H(Lip ˜ (X

H(Lipd˜(X))  × Y )).
d+ρ̃

Then, we can suppose without loss of generality that both spaces are complete. Now, if we assume that the
complete metric space (X, d) satisfies that every bounded set is totally bounded, then every bounded and
closed set is compact, i.e., it must have the Heine-Borel property. Hence, from Proposition 2, we have that
H(Lipd (X)) = X.
Summarizing, we have to prove that

X × H(Lipρ (Y )) ≥ H(Lipd+ρ (X × Y )),

6 M.I. Garrido, A.S. Meroño / Topology and its Applications 339 (2023) 108585

or, equivalently, that the inclusion map

i : X × Y → H(Lipd+ρ (X × Y )) ⊂ RLipd+ρ (X×Y )

can be continuously extended to X × H(Lipρ (Y )). To this end, we are going to extend every coordinate
function πf ◦ i, for every f ∈ Lipd+ρ (X × Y ).
Indeed, let f ∈ Lipd+ρ (X × Y ) and, for x ∈ X fixed, define fx (y) = f (x, y), for every y ∈ Y . Then
fx ∈ Lipρ (Y ), and hence fx can be extended to a continuous function fx∗ on H(Lipρ (Y )). Next, we can
define (πf ◦i)∗ (x, ξ) = fx∗ (ξ), for every (x, ξ) ∈ X ×H(Lipρ (Y )). In order to check that (πf ◦i)∗ is continuous
on X × H(Lipρ (Y )), it is enough to prove that (πf ◦ i)∗ is continuous on (X × Y ) ∪ {(x0 , ξ0 )}, for every
x0 ∈ X and ξ0 ∈ H(Lipρ (Y ))\Y (see 6H in [4]).
Let x0 ∈ X, ξ0 ∈ H(Lipρ (Y ))\Y , and ε > 0. Since fx∗0 is continuous on H(Lipρ (Y )), there exists an open
set V ξ0 ⊂ H(Lipρ (Y )) such that ξ0 ∈ V ξ0 and

fx∗0 (V ξ0 ) ⊂ Bdu (fx∗0 (ξ0 ), ε/2)

where du denotes the usual metric in R.

On the other hand, as f is in particular uniformly continuous in X × Y , there exists δ > 0 such that
|f (x, y) − f (u, v)| < ε/2 whenever (x, y), (u, v) ∈ X × Y and (d + ρ)((x, y), (u, v)) < δ.
Finally, for every (x, y) = (x0 , ξ0 ) belonging to (Bd (x0 , δ) × V ξ0 ) ∩ ((X × Y ) ∪ {(x0 , ξ0 )}) we have

|(πf ◦ i)∗ (x, y) − (πf ◦ i)∗ (x0 , ξ0 )| = |f (x, y) − fx∗0 (ξ0 )| ≤

≤ |f (x, y) − f (x0 , y)| + |f (x0 , y) − fx∗0 (ξ0 )| < ε/2 + ε/2 = ε

and (πf ◦ i)∗ is continuous in (x0 , ξ0 ), as we claimed. 

Note that the proof given in the last Proposition in order to extend the corresponding inclusion map is
standard in this kind of results. For instance, it can be seen in the classical paper by Glicksberg [5] devoted
to the Stone-Čech compactifications of products, as well as in the already mentioned Woods’ paper ([6]) for
the Samuel compactifications.
We are now ready to establish our main result.

Theorem 9. For two metric spaces (X, d) and (Y, ρ), the following are equivalent:

(a) H(Lipd+ρ (X × Y )) = H(Lipd (X)) × H(Lipρ (Y )).

(b) At least one of (X, d) and (Y, ρ) satisfies that every bounded subset is totally bounded.
(c) H(Lipd (X)) = X  or H(Lipρ (Y )) = Y .

Proof. First of all note that the equivalence between (b) and (c) follows from Propositions 2 and 5. Moreover,
that (b) implies (a) is clear from Propositions 7 and 8.
Then, in order to prove (a) implies (b), suppose on the contrary that there exist bounded sets B of
(X, d) and B  of (Y, ρ) which are not totally bounded. Then there exist infinite countable uniformly discrete
subsets D ⊂ B and D ⊂ B  . In particular, D × D is an infinite countable uniformly discrete subset of the
metric subspace B × B  .
According to Proposition 4 we have,

clH(Lipd (X)) D = clsd X D = sdD D = βN

clH(Lipρ (Y )) D = clsρ Y D = sρD D = βN
 M.I. Garrido, A.S. Meroño / Topology and its Applications 339 (2023) 108585 7

clH(Lipd+ρ (X×Y )) (D × D ) = clsd+ρ (X×Y ) (D × D ) = β(N × N)

But then, using condition (a), we have that

clH(Lipd+ρ (X×Y )) (D × D ) = clH(Lipd (X)×H(Lipρ (Y )) (D × D ) =

= clH(Lipd (X)) D × clH(Lipρ Y )) D = βN × βN

So, we get that βN × βN = β(N × N), which contradicts the classical Glicksberg’s Theorem, saying that
for infinite Tychonoff spaces X and Y , β(X × Y ) = βX × βY if and only if X × Y is pseudocompact (see
[5]). 

Remark 10. Note that according to Proposition 5, condition H(Lipd (X)) = X  tells us that H(Lipd (X)) is
as simple as possible. Indeed, H(Lipd (X)) = H(Lipd(X))  = X. And, in particular, this happens if and only

if X is Heine-Borel. Thus, we can add a new equivalent item in the above result, namely: “The completion
of at least one of the metric spaces X and Y must be Heine-Borel”. On the other hand, in [3] it is shown that
condition H(Lipd (X)) = X  is also equivalent to say that H(Lipd (X)) is metrizable. And hence, another
item that can be added in Theorem 9 would be: “At least one of H(Lipd (X)) and H(Lipρ (Y )) is metrizable”.

Example 11. In Examples 6, item (4), we assured that H(Lipd0-1 +du (N × N)) = βN × N. Now we can
confirm that, according to Theorem 9 (or directly from the above Remark 10). Indeed, since (N, du ) is
Heine-Borel we have that

H(Lipd0-1 +du (N × N)) = H(Lipd0-1 (N)) × H(Lipdu (N)) = βN × N.

In the setting of Banach spaces and from Corollary 3, we can derive the following.

Corollary 12. For Banach spaces X and Y , the equality H(Lip(X × Y )) = H(Lip(X)) × H(Lip(Y )) holds
if and only if one of them has finite dimension.

Example 13. Consider the euclidean norm · in Rn . Then, for any metric space (X, d) we have
H(Lipd+· (X × Rn )) = H(Lipd (X)) × Rn .

Finally, we are going to see that the Woods Theorem given in the Introduction can be now obtained as
Corollary of our main Theorem 9.

Corollary 14. (Woods’ Theorem) For two metric spaces (X, d) and (Y, ρ), the following are equivalent:

(a) sd+ρ (X × Y ) = sd X × sρ Y .
(b) At least one of (X, d) and (Y, ρ) is totally bounded.

Proof. Since uniformly equivalent metrics define equivalent Samuel compactifications for any metric space,
then we can suppose that the metrics d and ρ are bounded. Taking into account that the corresponding
Lipschitz functions on the bounded metric spaces X, Y and X × Y must be bounded, then the Lispchitz
realcompactifications of them must be their Samuel compactifications. Now applying Theorem 9 we get
immediately the above equivalence. Indeed, sd+ρ (X×Y ) = sd X×sρ Y ⇔ H(Lipd+ρ (X×Y )) = H(Lipd (X))×
H(Lipρ (Y )) ⇔ at least one of (X, d) and (Y, ρ) satisfies that every bounded subset is totally bounded ⇔
at least one of (X, d) and (Y, ρ) is totally bounded. 
8 M.I. Garrido, A.S. Meroño / Topology and its Applications 339 (2023) 108585

Question 15. A natural question is what happens for the Samuel realcompactifications of products of metric
spaces. Namely, what about the equality

H(Ud+ρ (X × Y )) = H(Ud (X)) × H(Uρ (Y )) ?

We are currently studying this problem but in the general setting of the uniform spaces, not necessarily
metrizable. So far, we can reveal that in that context we will need to include some additional hypotheses
related to “measurable cardinals”.

References

[1] M.I. Garrido, J.A. Jaramillo, Homomorphism on function lattices, Monatshefte Math. 141 (2004) 127–146.
[2] M.I. Garrido, J.A. Jaramillo, Lipschitz-type functions on metric spaces, J. Math. Anal. Appl. 340 (2008) 282–290.
[3] M.I. Garrido, A.S. Meroño, The Samuel realcompactification of a metric space, J. Math. Anal. Appl. 456 (2017) 1013–1039.
[4] L. Gillman, J.M. Jerison, Rings of Continuous Functions, Graduate Texts in Mathematics, vol. 43, Springer-Verlag, New
York-Heidelberg, 1976.
[5] I. Glicksberg, Stone-Čech compactifications of products, Trans. Am. Math. Soc. 90 (1959) 369–382.
[6] R.G. Woods, The minimum uniform compactification of a metric space, Fundam. Math. 147 (1995) 39–59.