280 7 Interpolation and Monotonicity
In applying Step 3 to the interval (k
i−1
,k
i
), C = C(k
i−1
,k
i
,d(Ω)) can be
replaced by C(a, b, d(Ω)) since the k
j
s can be expressed as explicit functions
of a and b.
Theorem 7.3.5 If 0 ≤ a<c<band >0, there is a constant C =
C(a, b, c, , d(Ω)) such that
[u]
c
≤ C[u]
a
+ [u]
b
(7.7)
and
|u|
c
≤ C|u|
a
+ |u|
b
. (7.8)
Proof: Letting c = λa +(1−λ)b, 0 <λ<1, the result follows from Inequal-
ity (7.3) when [u]
b
= 0. It therefore can be assumed that [u]
b
= 0. It suffices
to prove the result when [u]
b
=1forthen
(
u
[u]
b
)
c
≤ C
(
u
[u]
b
)
a
+
implies that [u]
c
≤ C[u]
a
+ [u]
b
.Inthecase|u|
b
=1,[u]
c
≤ C[u]
λ
a
by
Theorem 7.3.4, where C = C(a, b, d(Ω)). It is easily seen that there is a
linear function Kx + /C such that x
λ
≤ Kx + /C for x ≥ 0whereK
depends upon , c,andC;infact,
K = λ
C(1 −λ)
(λ−1)/λ
where λ is determined from c.Thus,
[u]
c
≤ C[u]
λ
a
≤ C(K[u]
a
+ /C)=CK[u]
a
+ ,
where CK may depend upon c through λ. The second assertion follows in
exactly the same way.
7.4 Interpolation of Weighted Norms
In this section, Ω can be any bounded open subset of R
n
.Convexityisnot
required. The section applies to any weight function φ>0 satisfying the
inequality |φ(x) − φ(y)|≤M|x − y|
α
,x,y ∈ Ω,forsomefixedα ∈ (0, 1]. As
the results will be applied only to the weight function φ(x)=d(x)=d(x, ∂Ω),
the results of this section will be stated only for d. In the proof of the following
lemma, D
j
u will denote a generic symbol for a jth order partial derivative of
u. The lemma is a special case of a more general result that can be found in
[25] allowing arbitrary j, k ≥ 0 . As the consideration of many cases is required
in the proof, only a few cases will be dealt with to illustrate techniques.