Lemma 2.2.
[16, Lemma 2] Let be a measure space,
some Banach space of -measurable functions on , a -compact quasi-metric space with positive Radon measure , and a function on . Assume that
(a) the convergence of a sequence in norm in yields the convergence of some
subsequence to the same function for -a. e. ;
(b) for -a. e. ;
(c) the map is Bochner integrable with respect to .
Then for -a. e. one has
|
|
|
( stands for the Bochner integral in ).
In the following stands for a domain in endowed with the Euclidean metric and Lebesgue measure , and denotes the group of bijective isometries of . Every such isometry is smooth by the Myers–Steenrod theorem [14]. Equipped with the compact-open topology becomes a locally compact topological group [10, p. 46, Theorem 4.7]. We call a family measurable if the map is measurable as a map between the measure space and the topological space . We shall wright
|
|
|
where for all .
Recall that the Sobolev space () is the Banach space of functions
with weak derivatives () equipped with the norm
|
|
|
(see, e. g., [1] for details).
According to [6] a domain is called a -extension domain if for every Sobolev
function there is a Sobolev function such that
almost everywhere in and that
|
|
|
where the norm on the left-hand side is the Sobolev norm in and that on
the right-hand side is the Sobolev norm in . The whole space and every its bounded subdomain with
smooth boundary is an extension domain, but an extension domain can be quite
complicated. See [6, Example 8.24 (f)].
Theorem 2.5.
Let
be a -compact quasi-metric space with Radon
measure . Let be a -extension domain in , and a family be measurable.
(i) The Hausdorff-type operator is bounded in the Sobolev space if (), and in this case
for some positive constant .
(ii) Let in addition for some constant and for all one has (). Then is bounded in , , and for a function in () we have for all
|
|
|
|
(2.1) |
|
|
|
|
Consequently,
|
|
|
|
Proof.
(i) Let .
It is known (see, e.g., [6, p. 40])
that equals the Hajłasz-Sobolev space in a sense that
a function belongs to if and only if there exists
so that
|
|
|
(2.2) |
holds for a. e. (recall that stands for the Euclidean distance in ).
In this case the following norm
|
|
|
where infimum is taken over all satisfying the defining inequality (2.2) is comparable with the usual norm in .
We shall verify the conditions of Lemma 2.2 where , .
(a) Since , this follows from the well known F. Riesz theorem.
(b) Note that every preserve the Lebesgue measure on by Lemma 2.3. For and we have for a. e.
|
|
|
|
|
|
|
|
where . Thus, for all and (b) holds.
(c) We shall use the criterium of Bochner integfrability (see, e. g., [8]). The space is separable (see, e. g., [1, Theorem 3.6]). Therefore to verify that the -valued function is strongly -measurable it suffices to prove that the -valued function is weakly -measurable.
To do this, note that every linear bounded functional on has the form
|
|
|
for some functions () see, e. g., [1, Theorem 3.19]. Thus, as a family is measurable and all operators are continuous on , the function
|
|
|
is measurable, too. Indeed, for each the evaluation map , is continuous with respect to the compact-open topology in , and thus the map is measurable as a composition of measurable mappings.
Next, for each and every we have , , and
|
|
|
where infimum is taken over all satisfying the condition
|
|
|
for a. e. . Since every such function has the form where satisfies the condition (2.2), and , we have . This implies (c) due to the criterium of Bochner integfrability, because and for all .
Thus, by Lemma 2.2,
|
|
|
(the Bochner integral for ).
Therefore
|
|
|
|
|
|
|
|
|
|
Since the norm is comparable with the usual norm in , the assertion (i) follows.
(ii) Let . We shall use the following special case of the definition of the generalized partial derivative of a locally integrable function on . Let . If
|
|
|
where is the space of test functions (indefinitely differential functions with compact support in ) then we say that (in the
sense of distributions).
Now let and . Then by the Fubuni theorem
|
|
|
|
(2.3) |
|
|
|
|
The application of the Fubuni theorem is justified by the following estimate
|
|
|
|
|
|
|
|
|
Since
|
|
|
|
|
|
|
(2.4) |
|
|
|
|
we deduce from (2.3) (again by the Fubuni theorem) that for all
|
|
|
|
|
|
which proves (2.1).
To finish the proof first note that
|
|
|
|
|
|
|
|
Similarly, formula (2.1) implies that for all
|
|
|
|
|
|
|
|
Therefore
|
|
|
|
|
|
|
|
which completes the proof.
Example 1.
Let be the group of real orthogonal matrices, (, is a column vector), , and the normalized Haar measure of the (compact) group . In this case, the Hausdorff operator turns into the following integral transform (the averaging operator over the action of )
|
|
|
This operator is bounded in () where is the ball in (possibly
with infinite radius so that ) by Theorem 2.5(i) and
|
|
|
since in this case where . Moreover, this operator is bounded in and by Theorem 2.5(ii), because for all for .
A similar result is valid for an averaging operator
|
|
|
where stands for a measurable set with .
Corollary 2.6.
Let , for all , and a family of isometries of be measurable.
(i) If then the operator
|
|
|
(2.5) |
is bounded in (), and for some positive constant . Moreover, this operator is bounded in , and in this case .
(ii) If the range of the map is bounded in the condition is necessary for the boundedness of in
.