HAL Id: hal-00974630
https://hal.archives-ouvertes.fr/hal-00974630
Preprint submitted on 7 Apr 2014
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de
Erratum to ”Multipliers and Morrey spaces”
Pierre Gilles Lemarié-Rieusset
To cite this version:
Pierre Gilles Lemarié-Rieusset. Erratum to ”Multipliers and Morrey spaces”. 2014. �hal-00974630�
Erratum to “Multipliers and Morrey spaces”.
Pierre Gilles Lemari´e–Rieusset ∗
Abstract
We correct the complex interpolation results for Morrey spaces which is false for the first interpolation functor of Calder´ on, but is exact for the Calder´ on’s second interpolation functor.
Keywords : Morrey spaces; interpolation.
2010 Mathematics Subject Classification : 42B35
In my paper “Multipliers and Morrey spaces” [3], there is a slight mistake in Theorem 3, concerning the interpolation of Morrey spaces.
Let B be the collection of all Euclidean balls B on R
d: B = B(x
B, r
B) = {x ∈ R
d/ |x − x
B| < R
B. For B ∈ B, we write |B| = R
B
dx = |B(0, 1)|r
dB. Define, for 1 < p ≤ q < +∞, the space ˙ M
p,q( R
n) as the space of locally p-integrable functions f such that
kfk
M˙p,q:= sup
B∈B
|B |
1/q−1/p( Z
B
|f(x)|
pdx)
1/p< +∞.
It is easy to check that, when 1 < p
0≤ q
0< +∞ and 1 < p
1≤ q
1< +∞, and
( 1 p , 1
q ) = (1 − θ)( 1 p
0, 1 q
0) + θ( 1 p
1, 1 q
1) for some θ ∈ (0, 1), then ˙ M
p0,q0∩ M ˙
p1,q1⊂ M ˙
p,qand
kfk
M˙p,q≤ kfk
1−θ˙Mp0,q0
kfk
θM˙p1,q1The question I studied was then whether one may find an interpolation functor F of exponent θ such that F ( ˙ M
p0,q0, M ˙
p1,q1) = ˙ M
p,q. If so, one should have the continuous embeddings
[ ˙ M
p0,q0, M ˙
p1,q1]
θ,1⊂ M ˙
p,q⊂ [ ˙ M
p0,q0, M ˙
p1,q1]
θ,∞.
∗
Laboratoire de Math´ ematiques et Mod´ elisation d’´ Evry, UMR CNRS 8071, Universit´ e
d’´ Evry; e-mail : plemarie@univ-evry.fr
The inclusion [A
0, A
1]
θ,1⊂ F (A
0, A
1) is a direct conclusion from the in- equality
kf k
F(A0,A1)≤ Ckf k
1−θA0kf k
θA1(obtained by interpolation inequalities for the operator norms of λ 7→ λf from R to A
0and from R to A
1, hence from R to F (A
0, A
1)). The inclusion F (A
0, A
1) ⊂ [A
0, A
1]
θ,∞is proven in [1] under the assumption that A
0∩ A
1is dense in A
0and in A
1. This is not the case for Morrey spaces. However, one may easily adapt the proof, as Morrey spaces are dual spaces (see for instance [6]).
If we assume that A
0= B
0′and A
1= B
′1and that B
0∩ B
1is dense in B
0and B
1; then, for b ∈ B
0∩ B
1, the linear form T
b: f 7→ hf |bi has a norm less than kbk
B0as an operator from A
0to R and less than kbk
B1as an operator from A
1to R , hence as a norm less than Ckbk
1−θB0kbk
θB1; thus, T ˜
f: b 7→ hf |bi is a continuous linear form on [B
0, B
1]
θ,1. This gives that f ∈ ([B
0, B
1]
θ,1)
′= [B
0′, B
1′]
θ,∞(since B
0∩ B
1is dense in B
0and B
1).
The theorem I proved in [3] is the following one:
Theorem 1
Let 1 < p
0≤ q
0< +∞ and 1 < p
1≤ q
1< +∞, and ( 1
p , 1
q ) = (1 − θ)( 1 p
0, 1 q
0) + θ( 1 p
1, 1 q
1)
for some θ ∈ (0, 1). Then there exists an interpolation functor F of exponent θ such that F ( ˙ M
p0,q0, M ˙
p1,q1) = ˙ M
p,qif and only if p
0/q
0= p
1/q
1.
The negative result for the case p
0/q
06= p
1/q
1was proven by a general- ization of a counterexample by Ruiz and Vega [4] which proves that, in that case, we don’t have the embedding of ˙ M
p,qinto [ ˙ M
p0,q0, M ˙
p1,q1]
θ,∞.
The proof for the positive result (on the case p
0/q
0= p
1/q
1) was inex-
act. I claimed that in that case we have the complex interpolation ˙ M
p,q=
[ ˙ M
p0,q0, M ˙
p1,q1]
θ. But this is false as pointed to me by Sickel (who has
recently characterized the intermediate space ˙ M
p,q⊂ [ ˙ M
p0,q0, M ˙
p1,q1]
θin
a joint work with Yang and Yuan [5]). Indeed, it is easy to see that,
when p
0/q
0= p
1/q
1= p
0/q
0< 1 and p
06= p
1, M ˙
p0,q0∩ M ˙
p1,q1is not
dense in ˙ M
p,q, while it is always true that A
0∩ A
1is dense in [A
0, A
1]
θ(see [1]). Sickel’s counterexample is very clear : if r = min(p
0, p
1) and
s = max(q
0, q
1), we have ˙ M
p,q⊂ M ˙
r,qand ˙ M
p0,q0∩ M ˙
p1,q1⊂ M ˙
r,s; thus
the applications f 7→ ρ
d(1/q−1/r)1
B(0,ρ)f are equicontinuous from ˙ M
p,qto L
r;
for f ∈ M ˙
p0,q0∩ M ˙
p1,q1, we have lim
ρ→0ρ
d(1/q−1/r)k1
B(0,ρ)fk
r= 0, while for
f
0= |x|
−d/q∈ M ˙
p,q, we have lim
ρ→0ρ
d(1/q−1/r)k1
B(0,ρ)f
0k
r> 0; thus f
0does
not belong to the closure of ˙ M
p0,q0∩ M ˙
p1,q1However, a slight modification of the proof of [3] gives the following the- orem :
Theorem 2
Let 1 < p
0≤ q
0< +∞ and 1 < p
1≤ q
1< +∞ , and ( 1
p , 1
q ) = (1 − θ)( 1 p
0, 1 q
0) + θ( 1 p
1, 1 q
1) If p
0/q
0= p
1/q
1, then
M ˙
p,q= [ ˙ M
p0,q0, M ˙
p1,q1]
θ.
Let us recall that Calder´on [2] defined two complex interpolation functors : [A
0, A
1]
θand [A
0, A
1]
θ. We have [A
0, A
1]
θ= [A
0, A
1]
θ(with equality when at least one of the two spaces A
0and A
1is reflexive).
Proof :
Let us recall the definition of [A
0, A
1]
θand [A
0, A
1]
θ.
Let Ω be the open complex strip Ω = {z ∈ C / 0 < ℜz < 1}. F is the space of functions F defined on the closed complex strip ¯ Ω such that :
1. F is continuous and bounded from ¯ Ω to A
0+ A
12. F is analytic from Ω to A
0+ A
13. t 7→ F (it) is continuous from R to A
0, and lim
|t|→+∞kF (it)k
A0= 0 4. t 7→ F (1+ it) is continuous from R to A
1, and lim
|t|→+∞kF (1+ it)k
A0=
0 Then
f ∈ [A
0, A
1]
θ⇔ ∃F ∈ F , f = F (θ) and
kfk
[A0,A1]θ= inf
f=F(θ)
max(sup
t∈R
kF (it)k
A0, sup
t∈R
kF (1 + it)k
A1).
On the other hand, G is the space of functions G defined on the closed complex strip ¯ Ω such that :
1.
1+|z|1G is continuous and bounded from ¯ Ω to A
0+ A
12. G is analytic from Ω to A
0+ A
13. t 7→ G(it) − G(0) is Lipschitz from R to A
04. t 7→ G(1 + it) − G(1) is Lipschitz from R to A
1Then
f ∈ [A
0, A
1]
θ⇔ ∃G ∈ G, f = G
′(θ) and
kf k
[A0,A1]θ= inf
f=G′(θ)
max( sup
t1,t2∈R
k G(it
2) − G(it
1) t
2− t
1k
A0, sup
t1,t2∈R
k G(1 + it
2) − G(1 + it
1) t
2− t
1k
A1).
Let us remark that, for continuous functions, (strong) analyticity is equiv- alent to weak analyticiy or even *-weak analyticity when A
0and A
1are dual spaces of B
0and B
1with B
0∩ B
1dense in B
0and B
1. Indeed, analyticity is equivalent to the fact that, whenever the closed ball ¯ B(z
0, r) is contained in Ω and |w − z
0| < r, then F (w) =
2iπ1R
|z−z0|=r
F (z)
z−wdz. As F is continuous, we have, for b ∈ B
0∩ B
1,
hb| 1 2iπ
Z
|z−z0|=r
F (z) dz
z − w i
B0∩B1,A0+A1= 1 2iπ
Z
|z−z0|=r
hb|F (z)i
B0∩B1,A0+A1dz z − w The equvalence remains true for *-weakly continuous functions.
However, of course, there is no equivalence between (strong) continuity and *-weak continuity. In the original proof of [3], one made two remaks :
1. Let 1 < p
0≤ q
0< +∞ and 1 < p
1≤ q
1< +∞, and ( 1
p , 1
q ) = (1 − θ)( 1 p
0, 1
q
0) + θ( 1 p
1, 1
q
1).
If F is an interpolation functor of exponent θ that satisfies F (L
p0, L
p1) = L
p, then F ( ˙ M
p0,q0, M ˙
p1,q1) ⊂ M ˙
p,q. Thus, we have the embeddings of [ ˙ M
p0,q0, M ˙
p1,q1]
θ,p, [ ˙ M
p0,q0, M ˙
p1,q1]
θand [ ˙ M
p0,q0, M ˙
p1,q1]
θinto ˙ M
p,q. 2. When moreover p
0/q
0= p
1/q
1= p/q we may define for f ∈ M ˙
p,qthe
function F (z) =
|ff||f |
(1−z)p0p+zp1p. This is a bounded *-weakly continu- ous function of z = x+iy (for 0 ≤ x ≤ 1) with values in ˙ M
p0,q0+ ˙ M
p1,q1, holomorphic on the strip 0 < x < 1, with sup
∈RkF (iy)k
M˙p0,q0< +∞, sup
∈RkF (1 + iy )k
M˙p1,q1< +∞, and F (θ) = f .
If F was strongly continuous, we would find that f = F (θ) would belong to [ ˙ M
p0,q0, M ˙
p1,q1]
θ. But F is only *-weakly continuous. We may define G(z) = R
z0