III.4 The Spaces H
1
q
195
Hint: Use the fact that v can b e approximated by elements of D(Ω), together with
Theorem II.4.1.
The problem o f the relationship between H
1
q
(Ω) and
b
H
1
q
(Ω) is not merely
a question of mathematical completeness; rather, as pointed out for the first
time by Heywood (1976), the coincidence of the two spaces is tightly linked
with the uniqueness of flow of a viscous, incompressible fluid and, in particular,
in doma ins for which H
1
q
(Ω) 6=
b
H
1
q
(Ω) the motion of such a fluid is not
uniquely determined by the “traditional” initial and boundary data but other
extra and appropriate auxiliary conditions are needed. Referring the reader
to Chapters VI and XII for a description of these latter and related results,
in the present section we shall only consider the question of investigating for
which domains the two spaces coincide and will indicate doma ins for which
they certainly don’t. To this end, we subdivide the domains of R
n
(n ≥ 2) into
three groups:
(a) bounded domains;
(b) exterior domains;
(c) domains with a noncompact boundary.
In cases (a) and (b) one shows H
1
q
(Ω) =
b
H
1
q
(Ω) provided only Ω has a
mild degree of smoothness (for example, cone condition, or even less, would
suffice); see Theorem III.4.1 and Theorem III.4.2. On the other hand, in case
(c) one exhibits examples of domains for which the two spaces are distinct,
no matter how smooth their boundary is; see Theorem III.4.4 and Theorem
III.4.6. T herefore, the coincidence of H
1
q
(Ω) and
b
H
1
q
(Ω) does not seem related
to a high degree of regularity of Ω but rather to its shape. In this connection,
we wish to recall a remarkable general result of Masl ennikova & Bogovski
˘
i
(1983, Theorem 5), which states that, if Ω is an arbit rary strongly locally
Lipschitz domain,
3
then H
1
q
(Ω) =
b
H
1
q
(Ω), for all q ∈ [1, n/(n − 1)].
It is conjectured that, for all dom ains wi th a compact bounda ry, the two
spaces coincide for q = 2, but no proof is, to date, available in the general
case.
4
Should this coincidence fail to hold for some domain of the above type,
we would have paradoxical situations from the physical point of view. For
example, if for some bounded domain, Ω
]
(say), the two spaces did not coincide,
then the steady-state Stokes boundary-value problem formulated in Ω
]
and
corresponding to zero body force and zero (Dirichlet) data at the boundary
would admi t a non-zero and smooth solution; see Remark IV.1.2. Analogous
situation would occur if Ω
]
is an exterior domain; see Remark V.1.2
We finally notice that, as already observed, for the proof of coincidence,
it is sufficient to show
b
H
1
q
(Ω) ⊂ H
1
q
(Ω), the converse inclusion being always
satisfied.
3
This type of regularity extends t he lo cal Lipschitz one to the case of domains
with a noncompact boundary, see, e.g., Adams (1975, p. 66).
4
See, however the result of
ˇ
Sver´ak (1993) mentioned in the Notes for this Chapter.