MATEMATIČKI VESNIK

MATEMATIQKI VESNIK research paper

70, 4 (2018), 326–337 originalni nauqni rad
December 2018

GENERALIZED HÖLDER’S INEQUALITY IN MORREY SPACES

Ifronika, Mochammad Idris, Al Azhary Masta and Hendra Gunawan

Abstract. The aim of this paper is to present sufficient and necessary conditions for
generalized Hölder’s inequality in Morrey spaces and generalized Morrey spaces. We also
obtain similar results in weak Morrey spaces and generalized weak Morrey spaces. The
sufficient and necessary conditions for the generalized Hölder’s inequality in these spaces are
obtained through estimates for characteristic functions of balls in Rd .

1. Introduction and preliminaries

Several authors have made important observations about Hölder’s inequality in the
last three decades (see [1, 2, 7, 12]). Recently, Masta et al. [6] obtained sufficient and
necessary conditions for the generalized Hölder’s inequality in Lebesgue spaces. In
this paper, we are interested in studying the generalized Hölder’s inequality in Morrey
spaces and in generalized Morrey spaces. In particular, we shall prove sufficient and
necessary conditions for generalized Hölder’s inequality in those spaces. In addition,
we also prove similar result in weak Morrey spaces and in generalized weak Morrey
spaces.
Let us first recall the definition of Morrey spaces. For 1 ≤ p ≤ q < ∞, the Morrey
space Mpq (Rd ) is the set of all p-locally integrable functions f on Rd such that
1 1
Z  p1
kf kMpq := sup |B(a, r)| q − p |f (y)|p dy < ∞.
a∈Rd ,r>0 B(a,r)

Here, B(a, r) denotes the open ball centered at a ∈ Rd with radius r > 0, and
|B(a, r)| denotes its Lebesgue measure. One might observe that k·kMpq defines a norm
on Mpq (Rd ), and makes the space complete [9]. Also note that Mpq (Rd ) = Lp (Rd )
if q = p. Thus, Mpq (Rd ) can be viewed as a generalization of the Lebesgue space
Lp (Rd ).
The following theorem presents sufficient and necessary conditions for Hölder’s
inequality in Morrey spaces.
2010 Mathematics Subject Classification: 26D15, 46B25, 46E30
Keywords and phrases: Hölder’s inequality; generalized Hölder’s inequality; Morrey spaces;
weak Morrey spaces; generalized Morrey spaces; generalized weak Morrey spaces.

326
 Ifronika, M. Idris, A. A. Masta, H. Gunawan 327

Theorem 1.1. Let 1 ≤ p ≤ q < ∞, 1 ≤ p1 ≤ q1 < ∞, and 1 ≤ p2 ≤ q2 < ∞. Then

the following statements are equivalent:
1 1 1 1 1
(1) p1 + p2 ≤ p and q1 + q2 = 1q .

(2) kf gkMpq ≤ kf kMpq 1 kgkMpq 2 for every f ∈ Mpq11 (Rd ) and g ∈ Mpq22 (Rd ).
1 2

Let us now move to the weak Morrey spaces. For 1 ≤ p ≤ q < ∞, the weak Morrey
space wMpq (Rd ) is the set of all measurable functions f on Rd for which kf kwMpq < ∞,
1 1 1
where kf kwMpq := supa∈Rd , r,γ>0 |B(a, r)| q − p γ |{x ∈ B(a, r) : |f (x)| > γ}| p .
Note that k · kwMpq defines a quasi-norm on wMpq (Rd ). If q = p, then wMpq (Rd ) =
wLp (Rd ). Here, wMpq (Rd ) can be viewed as a generalization of the weak Lebesgue
space wLp (Rd ). The relation between wMpq (Rd ) and Mpq (Rd ) is shown in the follow-
ing lemma.
Lemma 1.2. ([5]) Let 1 ≤ p ≤ q < ∞. Then Mpq (Rd ) ⊆ wMpq (Rd ) with kf kwMpq ≤
kf kMpq for every f ∈ Mpq (Rd ).
This lemma will be useful for us to study sufficient and necessary conditions for
generalized Hölder’s inequality in weak Morrey spaces.
Next we present the definition of generalized Morrey spaces and generalized weak
Morrey spaces. For 1 ≤ p ≤ q < ∞, let Gp be the set of all functions φ : (0, ∞) →
(0, ∞) such that φ is almost decreasing (i.e. there exists C > 0 such that φ(r) ≥ C φ(s)
d
for every 0 < r < s < ∞) and r p φ(r) is almost increasing (i.e. there exists C > 0
d d
such that r p φ(r) ≤ C s p φ(s) for every 0 < r < s < ∞). Note that if φ ∈ Gp , then φ
satisfies the doubling condition, that is, there exists C > 0 such that C1 ≤ φ(r)φ(s) ≤ C
r p d
whenever 1 ≤ s ≤ 2. For φ ∈ Gp , the generalized Morrey space Mφ (R ) is defined as
the set of measurable functions f on Rd for which
 Z  p1
1 1
kf kMpφ := sup |f (x)|p dx < ∞.
a∈Rd ,r>0 φ(r) |B(a, r)| B(a,r)
d
Note that Mpφ (Rd ) = Mpq (Rd ) for φ(r) := r− q , 1 ≤ p ≤ q < ∞. Meanwhile, for
φ ∈ Gp , the generalized weak Morrey space wMpφ (Rd ) is defined to be the set of all
measurable functions f on Rd such that
1
γ |{x ∈ B(a, r) : |f (x)| > γ}| p
kf k wMp
φ
:= sup 1 < ∞.
φ(r)|B(a, r)| p
a∈Rd , r,γ>0

Here k · kwMpφ is a quasi-norm on wMpφ (Rd ). Furthermore, wMpφ (Rd ) = wMpq (Rd )
d
for φ(r) := r− q . The relation between the generalized Morrey spaces and their weak
type is given in the following lemma.
Lemma 1.3. Let 1 ≤ p < ∞ and φ ∈ Gp . Then Mpφ (Rd ) ⊆ wMpφ (Rd ) with kf kwMpφ ≤
kf kMpφ for every f ∈ Mpφ (Rd ).
In Section 2 we state our main results, and in Section 3 we present the proofs.
328 Generalized Hölder’s inequality in Morrey spaces

2. Statement of the results

Our main results are presented in the following theorems. The first theorem is more
general than Theorem 1.1.
Theorem 2.1. Let m ≥ 2. If 1 ≤ p ≤ q < ∞ and 1 ≤ pi ≤ qi < ∞ for i = 1, . . . , m,
then the following statements are equivalent:
m m
1 1 1
= 1q .
P P
(1) pi ≤ p and qi
i=1 i=1
m m
Q
kfi kMpq i for every fi ∈ Mpqii (Rd ), i = 1, . . . , m.
Q
fi
(2) ≤
i
i=1 Mp
q
i=1

Theorem 2.2. Let m ≥ 2. If 1 ≤ p ≤ q < ∞ and 1 ≤ pi ≤ qi < ∞ for i = 1, . . . , m,

then the following statements are equivalent:
m m
1 1 1
= 1q .
P P
(1) pi ≤ p and qi
i=1 i=1
m m
Q
kfi kwMpq i for every fi ∈ wMpqii (Rd ), i = 1, . . . , m.
Q
fi
(2) ≤m
i
i=1 wMp
q
i=1

For generalized Morrey spaces, we have the following theorems.

m
1
≤ p1 , φ ∈ Gp , and φi ∈ Gpi for
P
Theorem 2.3. Let m ≥ 2, 1 ≤ p, pi < ∞ with pi
i=1
i = 1, . . . , m.
m
m m
Q Q Q
(1) If φi (r) ≤ φ(r) for every r > 0, then
fi
≤ kfi kMpi for every
φi
i=1 i=1 Mp
φ
i=1

fi ∈ Mpφii (Rd ), i = 1, . . . , m.

m m
kfi kMpi for every fi ∈ Mpφii (Rd ), i = 1, . . . , m, then there
Q Q
(2) If f i
≤
φi
i=1 Mp
φ
i=1
m
Q
exists C > 0 such that φi (r) ≤ C φ(r) for every r > 0.
i=1

Theorem 2.4. Let m ≥ 2 and 1 ≤ p, pi < ∞ for i = 1, . . . , m. If φ ∈ Gp and φi ∈ Gpi

m
Q 
such that φi (r) = φ(r) for every r > 0 and there exists  > 0 such that r pi φi (r)
i=1
are almost decreasing for i = 1, . . . , m, then the following statements are equivalent:
m
1
≤ p1 .
P
(1) pi
i=1

m m
kfi kMpi for every fi ∈ Mpφii (Rd ), i = 1, . . . , m.
Q Q
(2) f i
≤
φi
i=1 Mp
φ
i=1
 Ifronika, M. Idris, A. A. Masta, H. Gunawan 329

m m
Q Q
fi
Remark 2.5. In [10,11], Sugano states that ≤ kfi kMpi holds for every
φi
i=1 Mp
φ
i=1
m m
fi ∈ Mpφii (Rd ), i = 1, . . . , m, provided that 1 1
P Q
pi = p and φi (r) = φ(r). Theorems
i=1 i=1
2.3 and 2.4 may be viewed as counterparts of Sugano’s results.
Finally, for generalized weak Morrey spaces, the following theorems hold.
m
1
≤ p1 , φ ∈ Gp , and φi ∈ Gpi for
P
Theorem 2.6. Let m ≥ 2, 1 ≤ p, pi < ∞ with pi
i=1
i = 1, . . . , m.
m
m m
Q Q Q
(1) If φi (r) ≤ φ(r) for every r > 0, then
fi
≤m kfi kwMpi for every
φi
i=1 i=1 wMp
φ
i=1

fi ∈ wMpφii (Rd ), i = 1, . . . , m.

m m
kfi kwMpi for every fi ∈ wMpφii (Rd ), i = 1, . . . , m, then
Q Q
(2) If fi
≤m
φi
i=1 wMp
φ
i=1
m
Q
there exists C > 0 such that φi (r) ≤ C φ(r) for every r > 0.
i=1

Theorem 2.7. Let m ≥ 2 and 1 ≤ p, pi < ∞ for i = 1, . . . , m. If φ ∈ Gp and φi ∈ Gpi

m
Q 
such that φi (r) = φ(r) for every r > 0 and there exists  > 0 such that r pi φi (r)
i=1
are almost decreasing for i = 1, . . . , m, then the following statements are equivalent:
m
1
≤ p1 .
P
(1) pi
i=1
m m
kfi kwMpi for every fi ∈ wMpφii (Rd ), i = 1, . . . , m.
Q Q
(2)
fi
≤m
φi
i=1 wMp
φ
i=1

3. Proofs of the theorems

Here, the letter C denotes a constant that may change from line to line. To prove our
results, we shall use Lemma 1.2, Lemma 1.3, and the following lemma.
Lemma 3.1. ([3–5]) Let 1 ≤ p < ∞ and φ ∈ Gp . Then there exists C > 0 (depending
on φ) such that
1 C
≤ kχB(a0 ,R) kwMpφ ≤ kχB(a0 ,R) kMpφ ≤
φ(R) φ(R)
for every a0 ∈ Rd and R > 0. In particular, we have
d d
R q ≤ kχB(a0 ,R) kwMpq ≤ kχB(a0 ,R) kMpq ≤ C R q
for every a0 ∈ Rd and R > 0.
330 Generalized Hölder’s inequality in Morrey spaces

Proof. This fact is proved in [3–5]; we rewrite it here for convenience. Let B0 :=
B(a0 , R) ⊆ Rd where a0 ∈ Rd and R > 0. If r ≤ R, then φ(r) ≥ C φ(R), so that
! p1  1
|B(a, r) ∩ B0 | p
Z
1 1 p C C
|χB0 (x)| dx ≤ ≤
φ(r) |B(a, r)| φ(R) |B(a, r)| φ(R)
B(a,r)
d d
d
for every a ∈ R . If r ≥ R, then r p φ(r) ≥ C R p φ(R), so that
Z ! p1
1 1 C 1 C
|χB0 (x)|p dx ≤ d |B(a, r) ∩ B0 |
p ≤
φ(r) |B(a, r)| φ(R)R p φ(R)
B(a,r)
d C
for every a ∈ R . Hence we conclude that kχB0 kMpφ ≤ φ(R) .
Next, by Lemma 1.3, we have kχB0 kwMpφ ≤ kχB0 kMpφ . Finally, by using the
definition of k · kwMpφ , we have
 1  1
γ |{x ∈ B0 : |χB0 (x)| > γ}| p γ |B0 | p γ
kχB0 kwMpφ ≥ = = ,
φ(R) |B0 | φ(R) |B0 | φ(R)
1 
for every γ ∈ (0, 1). Therefore kχB0 kwMp ≥ φ(R)
φ
, and the lemma is proved.

3.1 The proof of Theorem 2.1

Pm Pm Pm
Proof. (1)⇒(2) Let i=1 p1i ≤ p1 and i=1 q1i = 1q hold. Put p1∗ := i=1 p1i . Clearly
p∗ ≥ p. Now take an arbitrary B := B(a, R) ⊆ Rd and fi ∈ Mpqii (Rd ), where
i = 1, . . . , m. By the generalized Hölder’s inequality in Lebesgue spaces [2], we have
m
Z Y ! p1 Z Ym
! p1∗
1 1 1 1 ∗
−
|B| q − p |fi (x)|p dx ≤ |B| q p∗ |fi (x)|p dx
B i=1 B i=1
m Z  p1
1 i
− p1
Y
≤ |B| qi i |f (x)|pi dx .
B
i=1
Q m m
Q
fi p ≤
Taking the supremum over B, we obtain kfi kMpq i .

i
i=1 Mq i=1
m m
Q
kfi kMpq i for every fi ∈ Mpqii (Rd ), i = 1, . . . , m.
Q
(2)⇒(1) Suppose that
fi
≤
i
i=1 Mp
q
i=1
Take an arbitrary R > 0 and choose fi := χB(0,R) for i = 1, . . . , m. It follows from
Qm Qm m
Q
the hypothesis that kχB(0,R) kMpq = fi
≤ i=1 kfi k pi =
Mqi kχB(0,R) kMpq i .
p i
i=1 Mq i=1
d
Pm d
q−
Hence,
Pby Lemma 3.1, we have R i=1 qi
≤ C. Since R > 0 is arbitrary, we conclude
m
that i=1 q1i = 1q .
Next, choose 0 <  < min{ dp dpm
q1 , . . . , qm }. Clearly  < d. For arbitrary K ∈ N, we
1

PK
define g,K (x) := χ{0≤|x|<1} (x)+ j=1 χ{j≤|x|≤j+j − } (x). (If one prefers, one may re-
duce to the case of d = 1 and then consider the tensor product G,K (x1 , x2 , . . . , xn ) =
 Ifronika, M. Idris, A. A. Masta, H. Gunawan 331

g,K (x1 )g,K (x2 ) · · · g,K (xn ) for the general case,
Qm working with cubes instead of balls.)
Qm p
We define fi := g,K , i = 1, . . . , m. Note that i=1 fi = g,K and so  i=1 fi  = g,K .
Hence, we obtain

m
Z m
! p1
Y 1 1 Y p
|B(a, r)| q − p
 
fi = sup  fi (x) dx

p a∈Rd ,r>0 B(a,r)
i=1 Mq i=1

Z ! p1
1 1
≥|B(0, K + K − )| q − p g,K (x) dx
B(0,K+K − )
  p1
K Z
d d X
=C(K + K − ) q−p |B(0, 1)| + dx
j=1 j≤|x|≤j+j −

  p1
K
d d X
=C(K + K − ) q − p |B(0, 1)| + (|B(0, j + j − )| − |B(0, j)|)
j=1
  p1   p1
K K
d d X d d X
≥C(K + K − ) q − p (j + j − )d − j d  ≥ C(K + K − ) q − p  j d−−1  .
 

j=1 j=1

The last inequality follows from the Mean Value Theorem applied to the function
t 7→ td on the interval [j, j + j − ] for each j = 1, . . . , K. Observe now that for K ≥ 2,
we have
K Z K
X
d−−1
 1 
j ∼ td−−1 dt = C(K d− − 1) = C 1 − d− K d−
j=1 1 K

≥ C K d− ≥ C(K + K − )d− .

We note that this inequality also holds for K = 1 provided that C ≤ 1. With this
observation, we obtain
m
Y d d d  d 
fi ≥ C(K + K − ) q − p (K + K − ) p − p = C(K + K − ) q − p .

p
i=1 Mq
Meanwhile, for each i = 1, . . . , m, we claim that
1 1
Z  p1 1 1
Z  p1
qi − pi pi qi − pi
fi (x)pi dx
i i
sup |B(a, r)| fi (x) dx . |B(0, L)|
a∈Rd ,r>0 B(a,r) B(0,L)
for some integer L with 2 ≤ L ≤ K + 1. Observe that fi = g,K is symmetrical about
0 and has most mass around 0, and so for each a ∈ Rd and r > 0, we have
1
Z  p1 Z  p1
− 1 1
− 1
fi (x)pi dx fi (x)pi dx
i i
|B(a, r)| qi pi ≤ |B(0, r)| qi pi .
B(a,r) B(0,r)
Now, as a function of r only, the value of the last expression on the right-hand side
gets larger and larger as r grows from 0 to 2 but decreases for r > K + K − . This
verifies our claim about the supremum.
332 Generalized Hölder’s inequality in Morrey spaces

Since 1
qi − 1
pi ≤ 0 for i = 1, . . . , m and j + j − ≤ 2j for j = 1, . . . , K, we have
! p1
i
Z
1
− p1
kfi kMpq i = sup |B(a, r)| qi i fi (x)pi dx
i
a∈Rd ,r>0 B(a,r)
! p1 L
! p1
i i
Z
1
− p1 d
− pd
X
pi
(j + j − )d − j d
 
≤C |B(0, L)| qi i fi (x) dx ≤CL qi i |B(0, 1)| +
B(0,L) j=1

L
! p1
i
d
− pd d
− pd d
− p d
− p
X
≤C L qi i |B(0, 1)| + j d−−1 ≤ C L qi i L pi i = C L qi i .
j=1
d 
−
Moreover, since L ≤ K + 1 ≤ 2(K + K − ), we obtain kfi kMpq i ≤ C(K + K − ) qi pi
i
for i = 1, . . . , m.
m Qm m
d d
P Q
Knowing that qi = q and
fi ≤ kfi kMpq i , we conclude from the
p i
i=1 i=1 M i=1

Pqm 
− − p + i=1 pi
two
Pm inequalities above that (K + K ) ≤ C for every K ∈ N. Therefore
1 1 
i=1 pi ≤ p , as desired.
Remark 3.2. For m = 2, we obtain the proof of Theorem 1.1.

3.2 The proof of Theorem 2.2

Proof. If (1) holds, then by similar arguments as in [8] we can prove that (1) implies
(2). It thus remains to prove that (2) implies (1). To do so, take an arbitrary R > 0
and let fi := χB(0,R) for i = 1, . . . , m. By the hypothesis, we then have

Y m Ym m
Y
kχB(0,R) kwMpq = fi ≤m kfi kwMpq i = m kχB(0,R) kwMpq i .

i i
p i=1 i=1 i=1
wMq
d
Pm d
q−
Pm 1
Hence R i=1 qi
≤ C. Since this holds for every R > 0, it follows that i=1 qi = 1q .
Next, let 0 <  < min{ dp dpm
q1 , . . . , qm } and define g,K (x) := χ{0≤|x|<1} (x) +
1

PK
j=1 χ{j≤|x|≤j+j − } (x), for arbitrary K ∈ N. For i = 1, . . . , m, let fi := g,K .
We observe
m that
 1
Y 1 1 
− q1 − p 1  p
fi ≥ |B(0, K + K )|  x ∈ B(a, r) : |fi (x)| > 2 


i=1
p
2
wMq
d d d  d 
≥ C(K + K − ) q − p (K + K − ) p − p = C(K + K − ) q − p .
Meanwhile, by Lemma 1.2 and the Morrey-norm estimate for fi , we obtain kfi kwMpq i ≤
m i
d
− − p Pm 1 1
Q
kfi kMpq i ≤ C(K +K ) i i for i = 1, . . . , m. Since i=1 qi = q and fi
q ≤
i p i=1 wMq
m 
Pm 
− − p + i=1
Q
m kfi k p
wMqii , we have (K + K ≤ C. ) pi

i=1 Pm 1
Since it holds for every K ∈ N, we must have i=1 pi ≤ p1 . 
 Ifronika, M. Idris, A. A. Masta, H. Gunawan 333

3.3 The proof of Theorem 2.3

m
Q
Proof. (1) Suppose that φi (r) ≤ φ(r)
for every r > 0. Take an arbitrary B :=
i=1 Pm
pi
B(a, R) ⊆ Rd and fi ∈ i = 1, . . . , m. Putting p1∗ := i=1 p1i , it
Mφi (Rd ), where
follows from the generalized Hölder’s inequality in Lebesgue spaces that
  p1   p1∗
Z Ym Z Ym
1  1 1 1 ∗
|fi (x)|p dx ≤  |fi (x)|p dx
φ(R) |B| i=1 φ(R) |B| i=1
B B
 p1 
Z m i
1  1 pi
Y
≤ |fi (x)| dx  .
φ (R) |B|
i=1 i B
m m
Q Q
We can now take the supremum over B to obtain
fi
≤ kfi kMpi .
φi
i=1 Mp
φ
i=1

m m
kfi kMpi for every fi ∈ Mpφii (Rd ), i = 1, . . . , m.
Q Q
(2) Suppose that fi
≤
φi
i=1 Mp
φ
i=1
Take an arbitrary R > 0 and define fi := χB(0,R) for i = 1, . . . , m. Then there exists
m
1
Q
C > 0 (independent of R) such that φ(R) ≤ kχB(0,R) kMpφ ≤ kχB(0,R) kMpi ≤
φi
i=1
m m
C 
Q Q
φi (R) . Thus φi (R) ≤ C φ(R), as desired.
i=1 i=1

3.4 The proof of Theorem 2.4

Pm
Proof. (1)⇒(2) Suppose that i=1 p1i ≤ p1 . As before, one may easily observe that
m m
kfi kMpi for every fi ∈ Mpφii (Rd ), i = 1, . . . , m.
Q Q
fi ≤
φi
i=1 Mp
φ
i=1

m m
kfi kMpi for every fi ∈ Mpφii (Rd ), where i = 1, . . . , m.
Q Q
(2)⇒(1) Let fi
≤
φi
i=1 Mp
φ
i=1
PK
For arbitrary K ∈ N, we define g,K (x) := χ{0≤|x|<1} (x) + j=1 χ{j≤|x|≤j+j − } (x),
where  > 0 satisfies the hypothesis (which forces us to have  < d). For i = 1, . . . , m,
let fi := g,K . It is easy to check that

m
Z ! p1
Y 1 1
fi ≥ g,K (x) dx


i=1
p φ (K + K − ) |B(0, K + K − )|
Mφ B(0,K+K − )
  p1
K
C X C
≥ d
 j d−−1  ≥  .
(K + K − ) φ(K + K − )
p
j=1
(K + K − ) p φ(K + K − )
Meanwhile, for i = 1, . . . , m, by using similar arguments as in the proof of Theorem
334 Generalized Hölder’s inequality in Morrey spaces

2.1 one may observe that for 2 ≤ L ≤ K + 1,

! p1
i
Z
C 1 C
kfi kMpi ≤ g,K (x) dx ≤  .
φi φi (L) |B(0, L)| (K + K ) φi (K + K − )
− pi
B(0,L)

Qm m
Q m
Q
Since f i
≤ kf i k M
pi and φi (r) = φ(r) for every r > 0, then for arbi-
φi
i=1 Mp
φ
i=1 i=1

Pm 
− − p + i=1
Pm 1
trary K ∈ N it holds (K + K ) pi
≤ C. Hence, i=1 pi ≤ p1 . 

3.5 The proof of Theorem 2.6

m
φi (r) ≤ φ(r) for every r > 0. Let fi ∈ wMpφii (Rd ), where
Q
Proof. (1) Suppose that
i=1
i = 1, . . . , m. For an arbitrary B := B(a, R) ⊆ Rd and γ > 0, let
" ( m
)# p1
1 
p
Y  fi (x)  
A(B, γ) := γ x ∈ B : > γ  .
  
φp (R)|B|  kf k

pi
i=1 i wM 
φi

1
Pm 1
Putting p∗ := i=1 pi , we observe that
  p1∗
( )
m
 1 
p∗ 
Y  fi (x)  
A(B, γ) ≤  m γ x ∈ B :   > γ 
kfi kwMpi

p∗ 
Q 
|B| φi (R)

i=1 φi
i=1
" (
m
)# p1∗
1 p∗  Y  f i (x)  
= γ0  x ∈ B :   > γ0  ,
|B| 
i=1 i
φ (R)kfi kwMpi 
φi
γ
where γ0 := m
Q . Furthermore, by using Young’s inequality for products, we have
φi (R)
i=1
 # 1
o p∗
" m
1 p∗ n Y  fi (x) 
A(B, γ) ≤ γ  x∈B : > γ0 
  
|B| 0  i=1
φ i (R)kf i kwM
pi 
φi
" ( m
)# p1∗
1 p∗  X p∗  fi (x)  p∗i 
≤ γ0  x ∈ B :  p > γ0 
|B| 
i=1 i
p φi (R)kfi kwMpi 
φi
"m ( )# p1∗
X 1 p∗  p∗  fi (x)  p∗i γ 0

≤ γ0  x ∈ B :  >
p
.

|B| pi φi (R)kfi kwMpi m 


i=1 φi
 Ifronika, M. Idris, A. A. Masta, H. Gunawan 335

  p∗i
pp∗
p∗  fi (x) p γ0 γ0 pi
Since pi  φi (R)kfi kwMpi  > m is equivalent to |fi (x)| > mp∗
i φi (R)kfi kwMpi =: γi ,
φi
φi
we obtain
  1∗
m 
γi (mp∗ )p
∗
/pi pi |{x ∈ B : |f (x)| > γ }| p
i i 
X
A(B, γ) ≤  p∗ /pi
i=1 φi (R)pi kfi kwMpi |B|
φi
  p1∗ #1 # 1∗
 p∗ p∗ p∗
m  ∗  p∗ pi
"m "m
∗ p
X p γ i |{x ∈ B : |f i (x)| > γ }|
i 
X X p
= m pi |B|kf kpi p
≤m ≤m = m,
i=1
p i φ i (R) i wM i
i=1
pi p
i=1 i
φi
∗
because 1 ≤ p ≤ pi for i = 1, . . . , m. We
mthen
take the supremum of A(B, γ) over
d
Q
B := B(a, R) ⊆ R and γ > 0 to obtain fi

.
i=1 wMpφ

m m
kfi kwMpi for every fi ∈ wMpφii (Rd ), i = 1, . . . , m. Take
Q Q
(2) Let f i
≤m
φi
i=1 wMp
φ
i=1
an arbitrary R > 0 and define fi := χB(0,R) for i = 1, . . . , m. By the hypothesis,
m
Q
we have kχB(0,R) kwMpφ ≤ m kχB(0,R) kwMpi . It thus follows from Lemma 3.1 that
φi
i=1
m

Q
there exists C > 0 (independent of R) such that φi (R) ≤ C φ(R).
i=1

3.6 The proof of Theorem 2.7

Pm
Proof. (1)⇒(2) Suppose that i=1 p1i ≤ p1 . As before, for every fi ∈ wMpφii (Rd ),
m m
Q Q
i = 1, . . . , m, we obtain f i
≤ m kfi kwMpi .
φi
i=1 wMp
φ
i=1

m m
kfi kwMpi for every fi ∈ wMpφii (Rd ),
Q Q
(2)⇒(1) Suppose that fi
≤ m
φi
i=1 wMp
φ
i=1
PK
where i = 1, . . . , m. Define g,K (x) := χ{0≤|x|<1} (x) + j=1 χ{j≤|x|≤j+j − } (x), for
arbitrary K ∈ N (where  > 0 satisfies the hypothesis), and for i = 1, . . . , m put
fi := g,K . By using the same arguments as in the proof of Theorem 2.2, we have
m
Y C
fi ≥ .


(K + K − ) p φ(K + K − )
i=1 p
wMφ

Next, using Lemma 1.3 and the generalized Morrey-norm estimate for fi , we have
C
kfi kwMpi ≤ kfi kMpi ≤  ,
φi φi
(K + K ) pi φi (K + K − )
−
m m m
Q Q Q
for i = 1, . . . , m. Since
fi
≤ m kfi kwMpi and φi (r) = φ(r) for ev-
p φi
i=1 wMφ i=1 i=1

Pm 
− − p + i=1
ery r > 0, it follows that (K + K ) pi
≤ C. We therefore conclude that
m 1 1 
P
i=1 pi ≤ p .
336 Generalized Hölder’s inequality in Morrey spaces

4. Concluding remarks

We have shown sufficient and necessary conditions for generalized Hölder’s inequality
in several spaces, namely Morrey spaces and their weak type versions. From Theorems
2.1 and 2.2, we see that both generalized Hölder’s inequality in Morrey
Pm spaces and in
weak Morrey spaces are equivalent to the same condition, namely i=1 p1i ≤ p1 and
Pm 1 1
i=1 qi = q . Accordingly, we have the following corollary.

Corollary 4.1. For m ≥ 2, the following statements are equivalent:

Qm m
kfi kMpq i for every fi ∈ Mpqii (Rd ), where i = 1, . . . , m.
Q
(1) fi
≤
i
i=1 Mp
q
i=1
m m
Q
kfi kwMpq i for every fi ∈ wMpqii (Rd ), where i = 1, . . . , m.
Q
fi
(2) ≤m
i
i=1 wMp
q
i=1

Similarly, from Theorems 2.4 and 2.7, we have the following corollary about
Hölder’s inequality in generalized Morrey spaces and in generalized weak Morrey
spaces.

Corollary 4.2. Let m ≥ 2 and 1 ≤ p, pi < ∞ for i = 1, . . . , m. If φ ∈ Gp and

m
Q
φi ∈ Gpi such that φi (r) = φ(r) for every r > 0 and there exists  > 0 such that
i=1

r pi φi (r) are almost decreasing for i = 1, . . . , m, then the following statements are
equivalent:
m m
kfi kMpi for every fi ∈ Mpφii (Rd ), where i = 1, . . . , m.
Q Q
(1) fi p ≤

φi
i=1 Mφ i=1

m m
kfi kwMpi for every fi ∈ wMpφii (Rd ), where i = 1, . . . , m.
Q Q
(2)
fi
≤m
φi
i=1 wMp
φ
i=1

Acknowledgement. The research is supported by ITB Research & Innovation

Program 2017.
The authors thank the referee for his/her useful remarks on the earlier version of
this paper.

References
[1] F. Avram, L. Brown, A generalized Hölder’s inequality and a generalized Szego theorem, Proc.
Amer. Math. Soc., 107(3) (1989), 687–695.
[2] W. S. Cheung, Generalizations of Hölder’s inequality, Int. J. Math. Math. Sci., 26(1) (2001),
7–10.
[3] Eridani, H. Gunawan, E. Nakai, Y. Sawano, Characterizations for the generalized fractional
integral operators on Morrey spaces, Math. Ineq. Appl., 17 (2014), 761–767.
[4] Eridani, H. Gunawan, M. I. Utoyo, A characterization for fractional integral operators on
generalized Morrey spaces, Anal. Theory Appl., 28(3) (2012), 263–267.
 Ifronika, M. Idris, A. A. Masta, H. Gunawan 337

[5] H. Gunawan, D. I. Hakim, K. Limanta, A. A. Masta, Inclusion properties of generalized Morrey

spaces, Math. Nachr., 290 (2017), 332–340.
[6] A. A. Masta, H. Gunawan, W. Setya-Budhi, An inclusion property of Orlicz-Morrey spaces,
J. Phys.: Conf. Ser., 893 (2017), 1–8.
[7] J. Matkowski, The converse of the Hölder’s inequality and its generalizations, Stud. Math.
109(2) (1994), 171–182.
[8] I. Rukmana, Multiplication Operators on Morrey Spaces and Weak Morrey Spaces (in Indone-
sian), Master Thesis, Institut Teknologi Bandung, Bandung, 2016.
[9] Y. Sawano, S. Sugano, H. Tanaka, Generalized fractional integral operators and fractional
maximal operators in the framework of Morrey spaces, Trans. Amer. Math. Soc., 363(12)
(2011), 6481–6503.
[10] S. Sugano, Some inequalities for generalized fractional integral operators on generalized Mor-
rey spaces, Math. Ineq. Appl., 14(4) (2011), 849–865.
[11] S. Sugano, Some inequalities for generalized fractional integral operators on generalized Mor-
rey spaces and their remarks, Sci. Math. Japon., 76(3) (2013), 471–495.
[12] V. Vasyunin, The sharp constant in the reverse Hölder’s inequality for Muckenhoupt weights,
St. Petersburg Math. J., 15(1) (2004), 49–79.

(received 03.11.2017; in revised form 21.03.2018; available online 10.07.2018)

Analysis and Geometry Group, Faculty of Mathematics and Natural Sciences, Bandung
Institute of Technology, Bandung 40132, Indonesia
E-mail: ifronika@math.itb.ac.id
Department of Mathematics, Universitas Lambung Mangkurat, Banjarbaru 70714, Indonesia
E-mail: idemath@gmail.com
Department of Mathematics Education, Universitas Pendidikan Indonesia, Bandung 40154,
Indonesia
E-mail: alazhari.masta@upi.edu
Analysis and Geometry Group, Faculty of Mathematics and Natural Sciences, Bandung
Institute of Technology, Bandung 40132, Indonesia
E-mail: hgunawan@math.itb.ac.id