Mathematics in Engineering, 5(2): 1–20.

DOI:10.3934/mine.2023038
Received: 03 March 2022
Revised: 24 May 2022
Accepted: 24 May 2022
http://www.aimspress.com/journal/mine Published: 10 June 2022

Research article
On parabolic Adams’s, the Chiarenza-Frasca theorems, and some other
results related to parabolic Morrey spaces†

Nicolai Krylov ∗

School of Mathematics, University of Minnesota, Minneapolis, MN 55455, USA

†
This contribution is part of the Special Issue: Nonlinear PDEs and geometric analysis
Guest Editors: Julie Clutterbuck; Jiakun Liu
Link: www.aimspress.com/mine/article/6186/special-articles

* Correspondence: Email: [email protected].

Abstract: We present several results of embedding type for parabolic Morrey and L p spaces with or
without mixed norms. Some other interpolation results for parabolic Morrey spaces are also given.
The main object of investigation is the term bi Di u and the ways to estimate it in various Morrey and L p
spaces in order to be able to treat it as a perturbation term in the parabolic equations.

Keywords: parabolic Morrey spaces; embedding theorems; Adams theorem; Chiarenza-Frasca

theorem

1. Introduction

Let Rd , d ≥ 1, be a Euclidean space of points x = (x1 , ..., xd ). In 1975 D. Adams [1] among
many other things proved that, if d ≥ 2 and we are given u ∈ C0∞ = C0∞ (Rd ) with its gradient Du =
(D1 u, ..., Dd u), Di = ∂/∂xi , satisfying
Z
|Du(y)|q dy ≤ ρd−βq , (1.1)
|x−y|<ρ

with q > 1, 1 < β ≤ d/q, and any ρ ∈ (0, ∞) and x ∈ Rd , then for all ρ ∈ (0, ∞) and x ∈ Rd we have
Z
|u(y)|r dy ≤ Nρd−(β−1)r (1.2)
|x−y|<ρ

with a constant N independent of u and r satisfying (β − 1)r = βq.

 2

This fact played a crucial role in [10] where the author investigated the solvability of elliptic
equations
ai j Di j u + bi Di u + u = f (Di j = Di D j ) (1.3)
with b < Ld,loc but rather satisfying for a sufficiently small b̂, all sufficiently small ρ and all balls B of
radius ρ Z
|b|d0 dx ≤ b̂ρd−d0
B
with certain d0 ∈ (d/2, d).
Our goal in this paper is to prepare necessary tools for developing a similar theory for parabolic
equations. In Section 2 we prove an analog of Adams’s intermediate estimate, which is the main
starting point. Section 3 contains the parabolic analog of the embedding theorem mentioned in the
beginning of the article. It also contains “local” interpolation inequalities in Morrey spaces allowing
one to deal with Morrey’s norms of expressions like bi Di u in domains when b is bounded. Section 4 is
devoted to the parabolic analog of a Chiarenza-Frasca theorem allowing to estimate the L p -norm rather
than Morrey’s norm of bi Di u. In Section 5 we treat parabolic Morrey spaces with mixed norms. The
main object of investigation is the term bi Di u and the ways to estimate it in various Morrey and L p
spaces in order to be able to treat it as a perturbation term in the parabolic analog of (1.3).
We finish the introduction with some notation and a remark. Define Bρ (x) = {y ∈ Rd : |x − y| < ρ},
Rd+1 := {z = (t, x) : t ∈ R, x ∈ Rd },

Cρ (t, x) = (s, y) ∈ ×Rd+1 : |x − y| < ρ, t ≤ s < t + ρ2 , Cρ = Cρ (0)



and let Cρ be the collection of Cρ (z), z ∈ Rd+1 , C = {Cρ , ρ > 0}. For measurable Γ ⊂ Rd+1 set |Γ| to be
its Lebesgue measure and when it makes sense set
Z Z
1
fΓ = – f dz = f dz.
Γ |Γ| Γ
Similar notation is used for f = f (x).
Remark 1.1. Formally, Adams proved (1.2) assuming that d ≥ 2. However, it is also true if d = 1.
To show this it suffices to take u depending only on one coordinate. The reader may wonder how the
restriction β ≤ d/q will become β ≤ 1/q. The point is that if d = 1 and β > 1/q, we have d − βq < 0
and condition (1.1) becomes only possible if u = 0.

2. Preliminary estimates

An important quantity characterizing L p = L p (Rd+1 ) is what we call the index which is the exponent
of ρ in the expression
d+2
kICρ kL p that is .
p
For domains Q ⊂ Rd+1 , p ∈ [1, ∞), and β ∈ (0, (d + 2)/p], introduce Morrey’s space E p,β (Q) as the
set of g such that
kgkE p,β (Q) := sup ρβ –kgIQ kL p (Cρ (t,x)) < ∞, (2.1)
ρ<∞,(t,x)∈Q

Mathematics in Engineering Volume 5, Issue 2, 1–20.

 3

where
Z 1/p
–kgkL p (Γ) = – |g| p dz .
Γ

We abbreviate E p,β = E p,β (R ). Observe that if Q = CR one can restrict ρ in (2.1) to ρ ≤ R since
d+1

β ≤ (d +2)/p. Also in that case one can allow (t, x) to be arbitrary, because, if |x| ≥ R, then BR ∩ Bρ (x) ⊂
BR ∩ Bρ (Rx/|x|). It is also useful to observe that, in case Q = CR , one gets an equivalent norm by adding
to the restrictions ρ < ∞, (t, x) ∈ CR , the requirement that the geometric center of Cρ (t, x) be in CR .
This follows from the fact that the L p (Cρ (t, x))-norm of gICR will only increase if we pull Cρ (t, x) down
the t axis to {t = 0} (if ρ2 > 2R2 ) or to the moment that the shifted Cρ (t, x) has its geometric center
inside CR .
There are many different notations for the norms in Morrey spaces. The convenience of the above
notation is well illustrated by Theorem 3.1 and Corollary 5.7.
We will often, always tacitly, use the following formulas in which u(t, x) = v(t/R2 , x/R):

–kukL p (CR ) = –kvkL p (C1 ) , kukE p,β (Q) = Rβ kvkE p,β (QR ) ,

where QR = {(t, x) : (R2 t, Rx) ∈ Q},

kDukE p,β (CR ) = Rβ−1 kDvkE p,β (C1 ) , kD2 ukE p,β (CR ) = Rβ−2 kvkE p,β (C1 ) .

For s, r > 0, α > 0, and appropriate f (t, x)’s on Rd+1 define

1
e−r /s I s>0 ,
2
pα (s, r) =
s(d+2−α)/2
Z
Pα f (t, x) = pα (s, |y|) f (t + s, x + y) dyds.
Rd+1
Observe that, if f is independent of t, then
Z
1
Pα f (t, x) = Pα f (x) = N(α) f (x + y) dy = NIα f (x),
Rd |y|d−α
where Iα is the Riesz potential. Therefore, one can get the Adams estimate found in the proof of
Proposition 3.1 of [1] from (2.3) below. In our investigation the most important values of α are 1
and 2. Set Z
β
Mβ f (t, x) = sup ρ – | f (z)| dz, 0 ≤ β ≤ d + 2,
ρ>0 Cρ (t,x)

M f = M0 f.
The following lemma is obtained by integrating by parts.
Lemma 2.1. Let β > 0 be a finite number, f (t) ≥ 0 be a function on [0, ∞) such that
Z t
−β
t f (s) ds → 0
0

as t → ∞. Then, for any S ≥ 0,

Z ∞ Z ∞ Z t 
t f (t) dt ≤ β
−β
t −β−1
f (s) ds dt.
S S S

Mathematics in Engineering Volume 5, Issue 2, 1–20.

 4

Lemma 2.2. For any α ∈ (0, β), β ∈ (0, d + 2] there exist constants N (< ∞) such that for any f ≥ 0
and ρ ∈ (0, ∞) we have

Pα (ICρ f )(0) ≤ Nρα M f (0), Pα (ICρc f )(0) ≤ Nρα−β Mβ f (0), (2.2)

Pα f ≤ N(Mβ f )α/β (M f )1−α/β . (2.3)

In particular (by Hölder’s inequality), for any p ∈ [1, ∞], q ∈ (1, ∞], and measurable Γ

kPα f kLr (Γ) ≤ NkMβ f kα/β 1−α/β

L p (Γ) k f kLq , (2.4)

provided that
1 α 1  α1
= · + 1− .
r β p β q
Proof. We basically mimic the proof of Proposition 3.1 of [1]. Observe that (2.3) at the origin is easily
obtained from summing up the inequalities in (2.2) and minimizing with respect to ρ. At any other
point it is obtained by changing the origin. Furthermore clearly, we may assume that f is bounded with
√ √
compact support. Set Q1 = {(s, y) : |y| ≥ s}, Q2 = {(s, y) : |y| ≤ s}. Dealing with Pα ( f IQ1 ) we
√
observe that pα (s, r) ≤ Nr−(d+2−α) if r ≥ s. Therefore,
Z ∞ Z r2 Z
1 
Pα ( f IQ1 ∩Cρc )(0) ≤ N f (s, y) dσr dsdr,
ρ rd+2−α 0 |y|=r

where dσr is the element of the surface area on |y| = r. By Lemma 2.1 (α < d + 2)
Z ∞ Z r Z ρ2 Z
1  
Pα ( f IQ1 ∩Cρc )(0) ≤ N f (s, y) dσρ ds dρdr
ρ rd+3−α ρ 0 |y|=ρ

Z ∞ Z r Z r2 Z
1  
≤N f (s, y) dσρ ds dρdr
ρ rd+3−α 0 0 |y|=ρ
Z ∞
1
=N I(r) dr,
ρ rd+3−α
where Z
I(r) = f (s, y) dyds.
Cr

We use that I(r) ≤ Nrd+2−β Mβ f (0) and that α < β. Then we see that

Pα ( f IQ1 ∩Cρc )(0) ≤ Nρα−β Mβ f (0). (2.5)

Next, by using Lemma 2.1 we obtain that

Z ∞ Z
1
Pα ( f IQ2 ∩Cρc )(0) ≤ √
f (s, y) dyds
ρ2 s(d+2−α)/2 |y|≤ s

∞ ∞
√
Z Z
1 1
≤N I( s) ds = N I(r) dr.
ρ2 s(d+4−α)/2 ρ rd+3−α

Mathematics in Engineering Volume 5, Issue 2, 1–20.

 5

This along with (2.5) prove the second inequality in (2.2).

As long as the first inequality is concerned, observe that similarly to Lemma 2.1 using that α > 0
we have Z ρ Z r2  Z
1 
Pα ( f IQ1 ∩Cρ )(0) ≤ N d+2−α
f (s, y) dσ r dsdr
0 r 0 |y|=r
Z ρ Z r  Z τ2 Z
1 ∂  
=N d+2−α ∂r
f (s, y) dσ τ ds dτ dr
0 r 0 0 |y|=τ
Z ρ Z r  Z τ2  Z
1  
= J1 + N d+3−α
f (s, y) dσ τ ds dτdr
0 r 0 0 |y|=τ
Z ρ
1
≤ J1 + N d+3−α
I(r) dr,
0 r
where Z ρ  Z τ2 Z
1  1
J1 = N d+2−α f (s, y) dστ ds dτ ≤ N d+2−α I(ρ)
ρ 0 0 |y|=τ ρ
Here I(r) ≤ Nrd+2 M f (0) and α > 0, so that

Pα ( f IQ1 ∩Cρ )(0) ≤ Nρα M f (0). (2.6)

Furthermore,
Z ρ2 Z
1
Pα ( f IQ2 ∩Cρ )(0) ≤ N √
f (s, y) dyds
0 s(d+2−α)/2 |y|≤ s
ρ2 ρ
√
Z Z
1 1
≤ J2 + N I( s) ds = J2 + N I(r) dr,
0 s(d+4−α)/2 0 rd+3−α
where Z ρ2 Z
1 1
J2 = N f (τ, y) dydτ ≤ N I(ρ).
ρd+2−α 0 |y|≤ τ
√ ρd+2−α
This and (2.6) prove the first inequality in (2.2). The lemma is proved.
Remark 2.3. If d = α = 1 and f is independent of t, the inequalities (2.2) and (2.3) are useless, because
the first one in (2.2) follows by definition and the second one and (2.3) are trivial because Mβ f = ∞
(β > α = 1) unless f = 0.
If α is strictly less than the index of Lq , we have the following.
Corollary 2.4. If α ∈ (0, (d + 2)/q), q ∈ (1, ∞), then there exists a constant N such that for any f ≥ 0
we have
kPα f kLr ≤ Nk f kLq
as long as
d+2 d+2
−α= .
q r
In particular, (a classical embedding result) if 1 < q < d + 2 and u ∈ C0∞ = C0∞ (Rd+1 ), then

kDukLr ≤ Nk∂t u + ∆ukLq (∂t = ∂/∂t)

Mathematics in Engineering Volume 5, Issue 2, 1–20.

 6

as long as
d+2 d+2
−1= .
q r
Indeed, the first assertion follows from Hölder’s inequality and (2.4) with p = ∞ and β = (d + 2)/q
(> α). The second assertion follows from the first one with α = 1 (< β) and the fact that for f = ∂t u+∆u
we have Z
y
e−|y| /(4s) f (t + s, x + y) dyds,
2
Du(t, x) = c (d+2)/2
Rd+1
+
s

where c is a constant and (|y|/s1/2 )e−|y| /(4s) ≤ Ne−|y| /(8s) .

2 2

Remark 2.5. After Corollary 2.4 a natural question arises as to what power of summability b = (bi )
will be sufficient for the term bi Di u to be considered as a perturbation term in ∂t u + ∆u + bi Di u in the
framework of the Lq -theory. Observe that, in the notation of Corollary 2.4

kbi Di ukLq ≤ kbkLd+2 kDukLr ≤ NkbkLd+2 k∂t u + ∆ukLq . (2.7)

It follows that b should be of class Ld+2 and q < d + 2. Of course, if b contains just bounded part, this
part in bi Di u is taken care of by interpolation inequalities.
In the next section we will also need the following result.

Corollary 2.6. For any α ∈ (0, β), β ∈ (0, d + 2] there exists a constant N such that for any g ≥ 0,
ρ ∈ (0, ∞), and (t, x) ∈ Cρ we have

Pα (IC2ρc g)(t, x) ≤ Nρα−β Mβ g(t, x).

Indeed, since
{t + s ≥ 4ρ2 or |x + y| ≥ 2ρ} ⊂ {s ≥ ρ2 or |y| ≥ ρ}
for f = g(t + ·, x + ·) we have
Z
Pα (IC2ρc g)(t, x) ≤ ICρc (s, y)pα (s, y)g(t + s, x + y) dyds = Pα (ICρc f )(0)
Rd+1

≤ Nρα−β Mβ f (0) = Nρα−β Mβ g(t, x).

3. A parabolic analog of the Adams Theorem 3.1 of [1]

Theorem 3.1. For any α ∈ (0, β), β ∈ (0, (d + 2)/q], q ∈ (1, ∞), and r such that

r(β − α) = qβ,

there is a constant N such that for any f ≥ 0 we have

kPα f kEr,β−α ≤ Nk f kEq,β . (3.1)

Mathematics in Engineering Volume 5, Issue 2, 1–20.

 7

Proof. It suffices to prove that for any ρ > 0

Z 1/r
β−α
ρ – |Pα f |r dz ≤ Nk f kEq,β ,
Cρ

that is
Z 1/r
β−α−(d+2)/r
ρ |Pα f |r dz ≤ Nk f kEq,β , (3.2)
Cρ

Observe that by Hölder’s inequality Mβ f ≤ Nk f kEq,β and by definition

Z 1/q
IC2ρ f q dz ≤ Nρ(d+2)/q−β k f kEq,β .
Rd+1

It follows from Lemma 2.2 with p = ∞ that

Z 1/r
|Pα (IC2ρ f )|r dz ≤ Nρ((d+2)/q−β)(1−α/β) k f kEq,β
Cρ

= Nρ(d+2)/r−β+α k f kEq,β .
Furthermore, by Corollary 2.6
Z 1/r
|Pα (IC2ρc f )|r dz ≤ Nρ(d+2)/r sup Pα (IC2ρc f )
Cρ Cρ

≤ Nρ(d+2)/r+α−β Eq,β f.
By combining these estimates we come to (3.2) and the theorem is proved.
Remark 3.2. We did not explicitly used that β ≤ (d + 2)/q and formally the proof is valid for any
β ∈ (0, ∞) if in Definition 2.1 we allow any β > 0. However, if β > (d + 2)/q and f , 0, the right-hand
side of (3.1) is infinite. Therefore, to make Theorem 3.1 nontrivial one requires β ≤ (d + 2)/q.
Remark 3.3. There is a simple relation of Pα1 Pα2 to Pα1 +α2 , which, in light of Theorem 3.1, implies that,
if β > α2 ≥ α1 > 0, q1 , q2 ∈ (1, ∞), q1 (β − α1 ) = q2 (β − α2 ) ≤ d + 2, then kPα2 f kEq2 ,β−α2 ≤ NkPα1 f kEq1 ,β−α1 .
We leave details of the proof to the reader and we do not use this fact in what follows.
The following, obtained similarly to Corollary 2.4, was communicated to the author by Hongjie
Dong.
Corollary 3.4. If 1 < q < d + 2, β ∈ (1, (d + 2)/q], and u ∈ C0∞ , then

kDukEr,β−1 ≤ Nk∂t u + ∆ukEq,β

as long as
1 1 1
r(β − 1) = qβ, that is = − . (3.3)
r q βq
Remark 3.5. For β = (d + 2)/q Corollary 3.4 yields the second part of Corollary 2.4 once more. This
is because Eq,(d+2)/q = Lq .

Mathematics in Engineering Volume 5, Issue 2, 1–20.

 8

Remark 3.6. In the framework of the Morrey spaces Corollary 3.4 opens up the possibility to treat
the terms like bi Di u as perturbation terms in operators like ∂t u + ∆u + bi Di u even with rather low
summability properties of b = (bi ). To show this, observe that for f, g ≥ 0 in the notation of
Corollary 3.4
ρβ –k f gICρ kLq ≤ ρ –k f ICρ kLβq · ρβ−1 –kgICρ kLr .
It follows that
kbi Di ukEq,β ≤ kbkEβq,1 kDukEr,β−1 ≤ NkbkEβq,1 k∂t u + ∆ukEq,β . (3.4)
For β = (d + 2)/q estimate (3.4) coincide with (2.7), but for β < (d + 2)/q in the framework of Morrey
spaces we allow b to be summable to the power βq < d + 2 in contrast with Remark 2.5. However, we
need kbkEβq,1 < ∞ and, if we ask ourselves what r should be in order for b ∈ Lr to have kbkEβq,1 < ∞, the
answer is r = d + 2 at least. Still we gain the possibility to have higher singularities of b than functions
from Ld+2 . Elliptic versions of (3.4) for usual or generalized Morrey spaces are found in many papers,
see, for instance, [5] and the references therein.
Next we move to deriving “local” versions of the above results. A statement somewhat weaker than
Corollary 3.4 can be obtained from the following general result by taking (S , T ) to be large enough and
then sending S → −∞, T → ∞.

Theorem 3.7. Let 1 < q < d + 2, β ∈ (1, (d + 2)/q] and let (3.3) hold. Then there is a constant N such
that for any u ∈ C0∞ , −∞ < S < T < ∞, and QS ,T = (S , T ) × Rd

kDukEr,β−1 (QS ,T ) ≤ Nk|∂t u| + |∆u| kEq,β (QS ,T ) + N(T − S )−1 kukEq,β (QS ,T ) . (3.5)

Proof. Shifting and changing the scales in Rd+1 allow us to assume that S = −1 = −T . In that case
consider the mapping Φ : [−3/2, 3/2] → [−1, 1], Φ(t) = t 2/(|t| ∨ 1) − 1 that preserves [−1, 1], is


Lipschitz continuous and has Lipschitz continuous inverse if restricted to [−3/2, 3/2] \ (−1, 1). Then,
obviously, for w(t, x) = v(Φ(t), x) we have

kwIQ−3/2,3/2 kEq,β ≤ NkvkEq,β (Q−1,1 ) , (3.6)

where N = N(q).
Now take (t, x) ∈ Q−1,1 , ρ ∈ (0, ∞), and take ζ ∈ C0∞ (R) such that ζ = 1 on (−1, 1), ζ = 0 outside
(−3/2, 3/2), and |ζ| + |ζ 0 | ≤ 4.
Although the function ζw, where w(s, y) = u(Φ(s), y), is not as smooth as required in Corollary 3.4
the argument leading to it applies to ζ(s)w(s, y) (we have a general Remark 5.14 to that effect) and
since r(β − 1) = qβ we have

ρβ−1 –kDuIQ−1,1 kLr (Cρ (t,x)) ≤ Nρβ−1 –kD(ζw)kLr (Cρ (t,x))

≤ NkIQ−3/2,3/2 (|∂t (ζw)| + |ζ∆w|)kEq,β .

It only remains to note that the last expression is less than the right-hand side of (3.5) in light of (3.6).
The theorem is proved.
To prove an interpolation theorem in CR we need two lemmas.

Mathematics in Engineering Volume 5, Issue 2, 1–20.

 9

Lemma 3.8. Let 0 < R1 < 1 < R2 < ∞, 1 ≤ q < ∞, β ∈ (0, (d + 2)/q]. Define Γ1 = B̄1 \ BR1 ,
Γ2 = B̄R2 \ B1 and let Φ : Γ2 → Γ1 be a smooth one-to-one mapping with |DΦ|, |DΦ−1 | ≤ K, where K is
a constant. Let v(t, x) ≥ 0 be zero outside G2 := (0, 1) × Γ2 and set u(t, x) = v(t, Φ−1 (x))IΓ1 (x). Then

kvkEq,β ((0,1)×BR2 ) ≤ N(d, q, β, K)kukEq,β (C1 ) . (3.7)

Proof. Take (t, x) ∈ (0, 1) × BR2 and ρ > 0. Then

 1 Z 1/q
β
ρ d+2 I(0,1)×BR2 vq dyds
ρ Cρ (t,x)

 1 Z 1/q
β
≤ Nρ d+2 IC1 uq dyds =: I,
ρ Ψ(Cρ (t,x)∩G2 )

where Ψ(s, y) = (s, Φ(y)). Observe that, if Cρ (t, x) ∩ G2 , ∅, then |y1 − y2 | ≤ 2ρ for any y1 , y2 ∈
Cρ (t, x) ∩ G2 . It follows that Φ(Cρ (t, x) ∩ G2 ) ⊂ B, where B is a ball of radius 2Kρ with center in B1 ,
and Z
β
 1 1/q
I ≤ N(2Kρ) IC uq
dyds ≤ NkukEq,β (C1 ) .
(2Kρ)d+2 (t,t+(2Kρ)2 )×B 1
This proves the lemma.
The following lemma about the interpolation inequality (3.9) is quite natural and obviously useful,
but its elliptic counterpart was proved only rather late in [10]. One of its goals is to be able to treat
bi Di u, when b is bounded, as a perturbation term.
Lemma 3.9. Let p ∈ (1, ∞), 0 < β ≤ (d+2)/p. Then there is a constant N such that, for any R ∈ (0, ∞),
ρ ≤ 2R, C ∈ Cρ with its geometric center in CR , ε ∈ (0, 1], and u ∈ C0∞ , we have

ρβ –kICR DukL p (C) ≤ NεR sup sβ –kICR (|∂t u| + |D2 u|)kL p (C(s))
ρ≤s≤2R

+ Nε−1 R−1 sup sβ –kICR (u − c)kL p (C(s)) , (3.8)

ρ≤s≤2R

where c is any constant and C(s) ∈ C s with the geometric center the same as C. In particular,

kDukE p,β (CR ) ≤ NεRk|∂t u| + |D2 u| kE p,β (CR ) + Nε−1 R−1 kukE p,β (CR ) . (3.9)

Proof. Changing scales shows that we may assume that R = 1. Obviously we may also assume that
c = 0. Then denote v = Du, w = |∂t u| + |D2 u|, G s = C(s) ∩ C1 ,

U = sup sβ –kukL p (G s ) , W = sup sβ –k(|∂t u| + |D2 u|)kL p (G s ) ,

ρ≤s≤2 ρ≤s≤2

By Poincaré’s inequality (see, for instance, Lemma 5.9), for ρ ≤ s ≤ 2,

–kv − vG s kL p (G s ) ≤ N(d, p)s –kwkL p (G s ) ≤ N s1−β W.

Also by interpolation inequalities, there exists a constant N = N(d, p) such that, for ε ∈ (0, 1] and
ε≤ s≤2,
–kv − vG s kL p (G s ) ≤ 2 –kvkL p (G s ) ≤ N –kwk1/2 kuk1/2
L p (G s ) – L p (G s )

Mathematics in Engineering Volume 5, Issue 2, 1–20.

 10

+ N s−1 –kukL p (G s ) ≤ N –kwk1/2 L p (G s ) + Nε

kuk1/2
L p (G s ) –
−1
–kukL p (G s ) , (3.10)
which for 2 ≥ s ≥ ε ∨ ρ yields
sβ –kv − vG s kL p (G s ) ≤ NW 1/2 U 1/2 + Nε−1 U.
Hence, for any ε ∈ (0, 1] and ρ ≤ s ≤ 2
sβ –kv − vG s kL p (G s ) ≤ N1 εW + N2 ε−1 U,
where N1 = N1 (d, p), N2 = N2 (d, p).
Following Campanato, one can transform this result to estimate vG s going along ρ, 2ρ,... and, since
β ∈ (0, (d + 2)/p], by Campanato’s results (cf. for instance, Proposition 5.4 in [8]) one gets that
ρβ –kvkL p (Gρ ) ≤ N3 (N1 εW + N2 ε−1 U) + N3 –kvkL p (G2 ) ,
where N3 = N3 (d, p, β). We estimate the last term as in (3.10) and come to what implies (3.8). The
lemma is proved.
The following is a local version of Corollary 3.4. It allows us to draw the same conclusions as in
Remark 3.6 in bounded domains.
Theorem 3.10. Let 1 < q < d + 2, β ∈ (1, (d + 2)/q] and let r(β − 1) = qβ. Then there is a constant N
such that for any R ∈ (0, ∞], u ∈ C0∞ ,
kDukEr,β−1 (CR ) ≤ Nk|∂t u| + |D2 u| kEq,β (CR ) + NR−2 kukEq,β (CR ) . (3.11)
Proof. The case of R = ∞ is obtained by passing to the limit. In case R < ∞, as usual, we may assume
that R = 1. In that case, mimicking the Hestenes formula, for 1 ≤ |x| ≤ 6/5 define
v(t, x) = 6u(t, x(2/|x| − 1)) − 8u(t, x(3/|x| − 2)) + 3u(t, x(4/|x| − 3))
=: 6v1 − 8v2 + 3v3
and for |x| ≤ 1 set v(t, x) = u(t, x). One can easily check that v ∈ C 1,2 ([0, 1] × B6/5 ). In light of
Lemmas 3.8 and 3.9, for instance,
kD2 vkEq,β ((0,1)×B6/5 ) ≤ kD2 ukEq,β (C1 ) + NkIB6/5 \B1 D2 v1 kEq,β ((0,1)×B6/5 ) + ...
+NkIB6/5 \B1 D2 v3 kEq,β ((0,1)×B6/5 ) ≤ NkD2 ukEq,β (C1 ) + NkDukEq,β (C1 )
≤ NkD2 ukEq,β (C1 ) + NkukEq,β (C1 ) . (3.12)
Now take (t, x) ∈ C1 , ρ ∈ (0, ∞), and take ζ ∈ C0∞ (Rd ) such that ζ = 1 on B1 , ζ = 0 outside B6/5 , and
|ζ| + |Dζ| + |D2 ζ| ≤ N = N(d).
By using Theorem 3.7 we get
ρβ−1 –kDuIC1 kLr (Cρ (t,x)) ≤ Nρβ−1 –kIQ0,1 D(ζv)kLr (Cρ (t,x))
≤ NkD(ζv)kEr,β−1 (Q0,1 ) ≤ Nk |∂t (ζv)| + |∆(ζv)| kEq,β (Q0,1 ) + NkζvkEq,β (Q0,1 )
≤ Nk |∂t (ζv)| + |∆(ζv)| kEq,β ((0,1)×B6/5 ) + NkvkEq,β ((0,1)×B6/5 )
It only remains to note that the last expression is less than the right-hand side of (3.11) as is well
seen from (3.12). The theorem is proved.

Mathematics in Engineering Volume 5, Issue 2, 1–20.

 11

Remark 3.11. By considering functions depending only on x we naturally obtain “elliptic” analogs of
our results. For instance, for G ⊂ Rd by defining

kgkE p,β (G) = sup ρβ –kgIG kL p (Bρ (x)) ,

ρ<∞,x∈G

we get from (3.11) for u ∈ C0∞ (Rd ) that

kDukEr,β−1 (BR ) ≤ Nk|D2 u kEq,β (BR ) + NR−2 kukEq,β (BR ) , (3.13)

whenever 1 < q < d, β ∈ (1, d/q] and r(β − 1) = qβ. Actually, formally, one gets (3.13) even for
β ≤ (d + 2)/q, but for β > d/q, both sides of (3.13) are infinite unless u = 0.
After that arguing as in (3.4) we see that for 1 < q < d, β ∈ (1, d/q]

kbi Di ukEq,β (B1 ) ≤ NkbkEβq,1 (B1 ) k∆ukEq,β (B1 ) + NkukEq,β (B1 ) . (3.14)

From the point of view of the theory of elliptic equations the most desirable version of (3.14) would
be
kbi Di ukEq,β (B1 ) ≤ εk∆ukEq,β (B1 ) + N(ε)kukEq,β (B1 ) (3.15)
for any ε > 0 with N(ε) independent of u. This fact is, actually, claimed in Theorem 5.4 of [5]. We
will show that (3.15) cannot hold if ε is small enough.
Let h(t) be a smooth nondecreasing function on R such that h(t) = 0 for t ≤ 0, h(t) = t for t ≥ 1 and
for δ > 0 set uδ (x) = h(ln(δ/|x|)). Let 1 < q < d/2, β = 2, b(x) = 1/|x|.
Then
kukEq,β (B1 ) ≤ N(d)kuδ kLd (B1 ) → 0
as δ ↓ 0. At the same time

xi 0 1  xi x j  1 xi x j
Di uδ = − h, Di j uδ = 2 − δ i j h0 + 2 2 h00 .
|x|2 |x|2 |x|2 |x| |x|

It is seen that |D2 uδ | ≤ N(d)/|x|2 and, since q < d/2, the Eq,β (B1 )-norm of D2 uδ is bounded as δ ↓ 0.
Also, for |x| ≤ δ/e, we have b|Duδ | = 1/|x|2 , so that for r ≤ δ/e
Z 1/q
– bq |Duδ |q dx = N(d, p)r−2 .
|x|≤r

It follows that the Eq,β (B1 )-norm of b|Duδ | is bounded away from zero as δ ↓ 0 and this shows that
(3.15) cannot hold for all δ > 0 if ε is small enough.

4. A parabolic version of Chiarenza–Frasca result [4]

In Remark 3.6 we have shown how to estimate a Morrey norm of |b| |Du| in terms of a Morrey
norm of b. Here, following [4], we show how to estimate an L p -norm of the same quantity through the
L p -norms of ∂t u and D2 u.

Mathematics in Engineering Volume 5, Issue 2, 1–20.

 12

Theorem 4.1. Let d + 2 ≥ q > p > 1, b ∈ Eq,1 . Then for any f ≥ 0 we have
Z
I := |b| p (P1 f ) p dz ≤ NkbkEp q,1 k f kL p , (4.1)
Rd+1

where N depends only on d, p, q. In particular (see the proof of Corollary 2.4), for any u ∈ C0∞
Z
|b| p |Du| p dz ≤ NkbkEp q,1 K, (4.2)
Rd+1

where K = kD2 u, ∂t ukLp p and N depends only on d, p, q.

Observe that we already know this result if q = d + 2 from Remarks 2.5 or 3.6.
In the proof we are going to use “parabolic” versions of some results from Real Analysis associated
with balls and cubes. These versions are obtained by easy adaptation of the corresponding arguments
by replacing balls with parabolic cylinders and cubes with parabolic boxes. To make the adaptation
more natural we introduce the “symmetric” maximal parabolic function operator by
Z
M̂ f (t, x) = sup – | f | dz,
C∈C, C
C3(t,x)

where (recall that) C is the set of Cr (z), r > 0, z ∈ Rd+1 . To prove the theorem we need the following.
Lemma 4.2. a) For r ∈ (0, ∞) define Dr = {|t| ≤ r2 , |x| ≤ r}. Then
rd+2
M̂IDr (t, x) ≤ ID2r + NIDc2r ≤ N 2 M̂IDr (t, x), (4.3)
|t|(d+2)/2 ∨ |x|d+2
where N = N(d).
b) For any nonnegative g(t, x), q ∈ [1, ∞), β ∈ (0, d + 2], α > 0, α > 1 − qβ/(d + 2), and r ∈ (0, ∞)
Z

α
gq M̂IDr dz ≤ N(d, q, α, β)rd+2−qβ kgkqEq,β . (4.4)
Rd

Proof. Assertion a) is proved by elementary means. To prove b), we use a) and split Dc2r into two parts
Dc2r ∩ {|x|2 ≥ |t|} and Dc2r ∩ {|x|2 < |t|} and, taking into account obvious symmetries, we see that it suffices
to show that Z ∞Z
gq (t, x)
I1 := α(d+2)/2
dxdt ≤ Nr(d+2)(1−α)−qβ kgkqEq,β ,
4r 2 B √t t
Z Z |x|2 q
g (t, x)
I2 := α(d+2)
dtdx ≤ Nr(d+2)(1−α)−qβ kgkqEq,β .
|x|≥2r 0 |x|
By observing that
t
t(d+2)/2−qβ/2
Z Z
1 
gq (s, x) dx ds ≤ α(d+2)/2 kgkqEq,β → 0
tα(d+2)/2 4r2 B √s t

as t → ∞, we have Z ∞ Z t Z
1 d  
I1 = gq (s, x) dx ds dt
4r2 tα(d+2)/2 dt 4r2 B √s

Mathematics in Engineering Volume 5, Issue 2, 1–20.

 13
Z ∞ Z t Z
1 
=N gq (s, x) dxds dt
4r2 tα(d+2)/2+1 4r2 B √s
Z ∞ (d+2)/2−qβ/2
t
≤ NkgkqEq,β dt = Nr(d+2)(1−α)−qβ kgkqEq,β .
4r2 tα(d+2)/2+1
Also as is easy to see
Z ∞ Z ρ2 Z
1 
I2 = N g p (t, x) dσρ dtdρ
2r ρα(d+2) 0 |x|=ρ

∂ ρ 
Z ∞ Z 2 Z
1 p
 
≤N g (t, x) dx dt dρ
2r ρ
α(d+2) ∂ρ
4r2 |x|≤ρ

ρ
Z ∞ d+2−qβ
q
≤ NkgkEq,β dρ = Nr(d+2)(1−α)−qβ kgkqEq,β .
2r ρ
α(d+2)+1

This proves the lemma.

Proof of Theorem 4.1. We follow some arguments in [4] and may assume that b ≥ 0. First set
r0 = (p + q)/2 and assume that there is a constant N0 such that M̂(|b|r0 ) ≤ N0 |b|r0 , that is, |b|r0 is in
the class A1 of Muckenhoupt. Observe that by Hölder’s inequality kbkEr0 ,1 ≤ kbkEq,1 . It is convenient to
prove the following version of (4.1) (notice r0 in place of q)

I ≤ NkbkEp r ,1 k f kL p , (4.5)
0

Then assume that b ≥ 0, set u = P1 f , and write

Z Z
I= p p−1

P1 f dz = P∗1 b p u p−1 f dz ≤ k f kL p P∗1 b p u p−1 ,



b u L p0
(4.6)
Rd+2 Rd+2

where p0 = p/(p − 1) and P∗1 is the conjugate operator for P1 , namely, for any g ≥ 0,

(P∗1 g)(s, x) = P1 (g(−·, −·) (−s, −x).

(4.7)

Next, take γ > 0, such that (1 + γ)p ≤ r0 , 1 + γp0 ≤ r0 , and p ≥ 1 + γ. Note that


 1/p  0
(p−1)/p
P∗1 b p u p−1 = P∗1 b1+γ b p−1−γ u p−1 ≤ P∗1 b(1+γ)p ) P∗1 b p−γp u p .

It follows that
Z 0
h 1/(p−1) i (p−1)/p
P∗1 p p−1

b u L p0
≤ b p−γp u p P1 P∗1 b(1+γ)p ) dz .
Rd
Now in light of (4.6) we see that, to prove (4.5) in our particular case, it only remains to show that
1/(p−1) i 0
≤ Nbγp kbkEp r ,1 .
0
h
P1 P∗1 b(1+γ)p ) (4.8)
0

For α = 1 and β = (1 + γ)p (> α) it follows from (2.3) and (4.7) that


1−1/(p+γp)
P∗1 b(1+γ)p ≤ NkbkEβ,1 M̂ b(1+γ)p .

Mathematics in Engineering Volume 5, Issue 2, 1–20.

 14

where the last factor by assumption (and Hölder’s inequality) is dominated by Nb(1+γ)p−1 and kbkEβ,1 ≤
kbkEr0 ,1 . After that to obtain (4.8) it suffices to use again (2.3) with α = 1 and β = 1 + γp0 to see that
0

1−1/(1+γp0 )
≤ NkbkEr0 ,1 bγp .
0
0
P1 b1+γp ≤ NkbkE1+γp0 ,1 M b1+γp

We now get rid of the assumption that M̂(|b|r0 ) ≤ N0 |b|r0 as in [4].

For r1 = (r0 + q)/2 we have |b|r0 ≤ ( M̂(|b|r1 ))r0 /r1 := b̃r0 and since r0 /r1 < 1, b̃r0 is an A1 -weight with
N0 = N0 (r0 /r1 ) (see, for instance, [7] p. 158). Therefore, (4.5) holds with b̃ in place of b and it only
remains to show that
kb̃kEr0 ,1 ≤ NkbkEq,1 ,
that is, for any t, x, ρ, Z
b̃r0 dz ≤ Nρd+2−r0 kbkrE0q,1 . (4.9)
Cρ (t,x)

Of course, we may assume that t = 0, x = 0. Then by Hölder’s inequality we see that the left-hand side
of (4.9) is less than
Z r0 /q
Nρ (d+2)(q−r0 )/q
( M̂(|b|r1 ))q/r1 ICρ dz ,
Rd+1

where the integral by a Fefferman-Stein Lemma 1, p. 111 of [6] and the fact that q/r1 > 1 is dominated
by Z
N |b|q M̂ICρ dz ≤ Nρd+2−q kbkqEq,1 ,
Rd+1

where we used Lemma 4.2 b) for α = β = 1. Hence,

Z
b̃r0 dz ≤ Nρ(d+2)(q−r0 )/q+(d+2−q)r0 /q kbkrE0q,1 ,
Cr

which is (4.9).
An alternative way to get the result is to follow the proof of Theorem 3 of [3]. We have
Z Z
r1 q/r1
( M̂(|b| )) ICρ dz ≤ ( M̂(|b|r0 ))q/r0 ( M̂ICρ )α dz =: J,
Rd+1 Rd+1

where α ∈ (0, 1). An easy exercise leads to the well-known result that ( M̂ICρ )α is an A1 -weight, and,
hence, an Aq/r0 -weight. By the Muckenhoupt theorem
Z
J≤N |b|q ( M̂ICρ )α dz
Rd+1

and it only remains to use Lemma 4.2 b) again with β = 1 and any appropriate α. The theorem is
proved.
Remark 4.3. In the above proof we tacitly assumed that I < ∞. One √ can easily avoid it by taking f
with compact support, replacing |b| with |b| ∧ bn , where n bn = 1 ∧ ( |t| + |x|) , observe that brn0 ∈ A1 ,
−1 −1

and while checking that the new I is finite use Hölder’s inequality and Corollary 2.4.

Mathematics in Engineering Volume 5, Issue 2, 1–20.

 15

5. Mixed-norm estimates

For q1 , q2 ∈ [1, ∞] and measurable f and Γ ⊂ Rd+1 introduce

Z Z q2 /q1 1/q2
k f kLq1 ,q2 = | f (t, x)|q1 dx dt ,
R Rd

–k f kLq1 ,q2 (Γ) = kIΓ k−1

Lq ,q k f IΓ kLq1 ,q2 .
1 2

Here the index of Lq1 ,q2 which is the exponent of ρ in the expression
d 2
kICρ kLq1 ,q2 is + .
q1 q2
If in addition 0 < β ≤ d/q1 + 2/q2 , set
k f kEq1 ,q2 ,β (Q) = sup ρβ –kIQ f kLq1 ,q2 (Cρ (t,x)) .
ρ<∞,(t,x)∈Q

We also introduce the spaces Lq1 ,q2 (Q) and Eq1 ,q2 ,β (Q) as the spaces of functions whose respective norms
are finite. We abbreviate Lq1 ,q2 = Lq1 ,q2 (Rd+1 ), Eq1 ,q2 ,β = Eq1 ,q2 ,β (Rd+1 ).
The following is certainly well known.
Lemma 5.1. Let f be a nonnegative function on Rd+1 , p, q ∈ (1, ∞). Then for any w x (x), wt (t) which
are A p Muckenhoupt weights on Rd and R, respectively, we have
Z Z
p
| M̂ f | w x wt dxdt ≤ N | f | p w x wt dxdt, (5.1)
Rd+1 Rd+1

where N depends only on d, p, and the A p -constants of w x , wt . Furthermore,

Z ∞Z q/p Z ∞Z q/p
p
| M̂ f | dx dt ≤ N | f | p dx dt, (5.2)
−∞ Rd −∞ Rd

where N depends only on d, p, q.

Proof. Estimate (5.1) follows by application of the Muckenhoupt theorem to w x wt , which is an A p -
weight on Rd+1 . Then observe that in the particular case that w x ≡ 1, (5.1) means that
Z ∞ h Z 1/p i p Z ∞ h Z 1/p i p
p
| M̂ f | dx wt dt ≤ N | f | p dx wt dt
−∞ Rd −∞ Rd

for any A p -weight wt , which implies (5.2) by the Rubio de Francia extrapolation theorem. The lemma
is proved.
This lemma, (2.3), and Hölder’s inequality immediately yield the following.
Lemma 5.2. For any α ∈ (0, β), β ∈ (0, d + 2], p ∈ [1, ∞], q1 , q2 ∈ (1, ∞], there exists a constant N
such that for any f ≥ 0 and measurable Γ we have
kPα f kLr1 ,r2 (Γ) ≤ NkMβ f kα/β
L p (Γ) k f kLq ,q ,
1−α/β
(5.3)
1 2

provided that
1 α 1  α 1
= · + 1− , i = 1, 2.
ri β p β qi

Mathematics in Engineering Volume 5, Issue 2, 1–20.

 16

Similarly to Corollary 2.4 we have

Corollary 5.3. Let q1 , q2 ∈ (1, ∞],
d 2
β := + > 0,
q1 q2
α ∈ (0, β). Then for any f ≥ 0 we have

kPα f kLr1 ,r2 ≤ Nk f kLq1 ,q2

as long as qi β = ri (β − α), i = 1, 2.
In particular, (almost follows from Theorem 10.2 of [2]) if β > 1, and u ∈ C0∞ , then

kDukLr1 ,r2 ≤ Nk∂t u + ∆ukLq1 ,q2 (5.4)

as long as qi β = ri (β − 1), i = 1, 2.
Corollary 5.4. Under the assumptions of Corollary 5.3, if β > 1, there is a constant N such that, for
any b = (bi ) ∈ Lβq1 ,βq2 and u ∈ C0∞ ,

kbi Di ukLq1 ,q2 ≤ NkbkLβq1 ,βq2 k∂t u + ∆ukLq1 ,q2 . (5.5)

Indeed, by Hölder’s inequality

kbi Di ukLq1 ,q2 ≤ kbkLβq1 ,βq2 kDukLr1 ,r2 .

Remark 5.5. It is instructive to compare this result with Remark 2.5. Now we can treat b ∈ L s1 ,s2 with
si ∈ (1, ∞] satisfying d/s1 + 2/s2 = 1.
Since Eq1 ,q2 ,β = Lq1 ,q2 if β = d/q1 + 2/q2 , the following is a generalization of Corollary 5.3.
Theorem 5.6. Let q1 , q2 ∈ (1, ∞],
d 2
+ ≥ β > 0,
q1 q2
α ∈ (0, β). Then there is a constant N such that for any f ≥ 0 we have

kPα f kEr1 ,r2 ,β−α ≤ Nk f kEq1 ,q2 ,β , (5.6)

where ri (β − α) = qi β, i = 1, 2.
Proof. It suffices to prove that for any ρ > 0

 Z ρ2  Z r2 /r1 1/r2

β−α
ρ – – |Pα f |r1 dy ds ≤ Nk f kEq1 ,q2 ,β ,
0 Bρ

that is
Z ρ2 Z r2 /r1 1/r2
β−α−(d/r1 +2/r2 )
ρ |Pα f |r1 dy ds ≤ Nk f kEq1 ,q2 ,β . (5.7)
0 Bρ

Mathematics in Engineering Volume 5, Issue 2, 1–20.

 17

Observe that by Hölder’s inequality Mβ f ≤ Nk f kEq1 ,q2 ,β and by definition

 Z 4ρ2  Z q2 /q1 1/q2

d/q1 +2/q2
kIC2ρ f kLq1 ,q2 = Nρ – – | f |q1 dy ds
0 B2ρ

≤ Nρd/q1 +2/q2 −β k f kEq1 ,q2 ,β = Nρ(d/r1 +2/r2 )β/(β−α)−β k f kEq1 ,q2 ,β .

It follows from Lemma 5.2 with p = ∞ that (5.7) holds with IC2ρ f in place of f on the left.
Furthermore, by Corollary 2.6 we have |Pα (IC2ρc f )| ≤ Nρα−β Mβ f in Cρ . Therefore,

 Z ρ2  Z r2 /r1 1/r2

β−α
ρ – – |Pα (IC2ρc f )|r1 dy ds ≤ N sup Mβ f ≤ Nk f kEq1 ,q2 ,β .
0 Bρ

By combining these results we come to (3.2) and the theorem is proved.

Corollary 5.7. Under the assumptions of Theorem 5.6, if β > 1, for any u ∈ C0∞ , we have

kDukEr1 ,r2 ,β−1 ≤ Nk∂t u + ∆ukEq1 ,q2 ,β ,

where ri (β − 1) = qi β, i = 1, 2. This coincides with (5.4) if β is equal to the index of Lq1 ,q2 .
Remark 5.8. Corollary 5.7 opens up the possibility to treat the terms like bi Di u as perturbation terms in
operators like ∂t u + ∆u + bi Di u with even lower summability properties of b = (bi ) than in Remark 5.5.
To show this observe that for q1 , q2 , β as in Theorem 5.6 with β > 1 and si = βqi ∈ (1, ∞], i = 1, 2, we
have
ρβ –kICρ bi Di ukLq1 ,q2 ≤ ρ –kbICρ kLs1 ,s2 · ρβ−1 –kICρ DukLr1 ,r2
implying that

kbi Di ukEq1 ,q2 ,β ≤ kbkE s1 ,s2 ,1 kDukEr1 ,r2 ,β−1 ≤ NkbkE s1 ,s2 ,1 k∂t u + ∆ukEq1 ,q2 ,β , (5.8)

where d/s1 + 2/s2 ≥ 1.

However, note that we also need
ρ –kbICρ (t,x) kLs1 ,s2
to be bounded as a function of ρ, t, x. If we ask ourselves what τ > 0 should be to guarantee this
boundedness if b ∈ Lτs1 ,τs2 , if d/s1 + 2/s2 > 1, the slightly disappointing answer is that τ = d/s1 + 2/s2 ,
so that d/(τs1 ) + 2/(τs2 ) = 1. Still functions in E s1 ,s2 ,1 may have higher singularities than those in
Lτs1 ,τs2 .
Another advantage of (5.8) in comparison with (5.5) is seen when b depends only on t or |b(t, x)| ≤
b̂(t). In that case (5.8) becomes

kbi Di ukEq1 ,q2 ,β ≤ Nkb̂kEβq2 ,1/2 (R) k∂t u + ∆ukEq1 ,q2 ,β ,

and if βq2 = 2, then

kb̂kEβq2 ,1/2 (R) = kb̂kL2 (R) .
Thus for any q1 ∈ (1, ∞] and q2 ∈ (1, 2)

kbi Di ukEq1 ,q2 ,2/q2 ≤ Nkb̂kL2 (R) k∂t u + ∆ukEq1 ,q2 ,2/q2 .

Mathematics in Engineering Volume 5, Issue 2, 1–20.

 18

In case q1 ∈ (1, d), q2 ∈ (1, ∞], 1 < β ≤ d/q1 , and kb(·, t)kEβq1 ,1 (Rd ) ≤ b̂ < ∞ for any t, we also have

kbi Di ukEq1 ,q2 ,β ≤ N b̂k∂t u + ∆ukEq1 ,q2 ,β .

An application of the last inequality in case u, b are independent of t, β = d/q1 , q1 ∈ (1, d), and q2 = ∞,
yields the well-known estimate
kbi Di ukLq1 (Rd ) ≤ NkbkLd (Rd ) k∆ukLq1 (Rd ) .
To extend the embedding and interpolation results to Morrey spaces with mixed norms we need the
following result very useful also in other circumstances.
Lemma 5.9 (Poincaré’s inequality). Let 1 ≤ r1 , r2 < ∞, u ∈ C0∞ , ρ ∈ (0, ∞). Then
–kDu − (Du)Cρ krL2r ,r (Cρ ) ≤ N(d, r1 , r2 )ρr2 –k |∂t u| + |D2 u|krL2r ,r (Cρ ) . (5.9)
1 2 1 2

Proof. We follow the usual way (see, for instance, Lemma 4.2.2 of [9]). First, due to self-similar
transformations, we may take ρ = 1. In that case, for a ζ ∈ C0∞ (B1 ) with unit integral, introduce
Z
v(t) = ζ(y)Du(t, y) dy.
B1

Then by the usual Poincaré inequality

Z Z Z
r1
|Du(t, x) − v(t)| dx =
r1
[Du(t, x) − Du(t, y)]ζ(y) dy dx
B1 B1 B1
Z Z Z
≤N |Du(t, x) − Du(t, y)| dxdy ≤ N r1
|D2 u(t, x)|r1 dx. (5.10)
B1 B1 B1
Next, observe that for any constant vector v the left-hand side of (5.9) is less than a constant times
(recall that ρ = 1) Z 1Z r2 /r1
|Du(t, x) − v|r1 dx dt
0 B1
Z 1Z r2 /r1 Z 1
≤N |Du(t, x) − v(t)| dx r1
dt + N |v(t) − v|r2 dt.
0 B1 0
By (5.10) the first term on the right is less than the right-hand side of (5.9). To estimate the second
term, take Z 1
v= v(t) dt.
0
Then by Poincaré’s inequality
Z 1 Z 1 Z Z 1 Z
r2 r2
r2
|v(t) − v| dt ≤ N ζ∂t Du dx dt = N (Dζ)∂t u dx dt
0 0 B1 0 B1

and to finish the proof it only remains to use Hölder’s inequality. The lemma is proved.
The usual Poincaré inequality was used in the proof of Lemma 3.9. Also observe that mixed-norms
estimates like (3.10) are available in [2] (see Theorem 9.5 there). Therefore, by using Lemma 5.9 and
following very closely the proofs of Lemmas 3.8, 3.9, and Theorems 3.10 we arrive at the following
results about interpolation and embedding for Morrey spaces with mixed norms.

Mathematics in Engineering Volume 5, Issue 2, 1–20.

 19

Lemma 5.10. Let q1 , q2 ∈ (1, ∞), 0 < β ≤ d/q1 + 2/q2 . Then there is a constant N such that, for any
R ∈ (0, ∞), ε ∈ (0, 1], and u ∈ C0∞ ,

kDukEq1 ,q2 ,β (CR ) ≤ NεRk|∂t u| + |D2 u| kEq1 ,q2 ,β (CR ) + Nε−1 R−1 kukEq1 ,q2 ,β (CR ) . (5.11)

Theorem 5.11. Let q1 , q2 ∈ (1, ∞), 1 < β ≤ d/q1 + 2/q2 and let ri (β − 1) = qi β, i = 1, 2. Then there is
a constant N such that for any R ∈ (0, ∞], u ∈ C0∞ we have

kDukEr1 ,r2 ,β−1 (CR ) ≤ Nk|∂t u| + |D2 u| kEq1 ,q2 ,β (CR ) + NR−2 kukEq1 ,q2 ,β (CR ) . (5.12)

Remark 5.12. By taking u depending only on x we recover from Lemma 5.10 and Theorem 5.11 their
“elliptic” counterpart stated as Lemmas 4.4 and 4.7 in [10], respectively.
Remark 5.13. Theorem 5.11 is the most general results of the paper containing as particular cases
our previous results on embeddings. Thus, Corollary 5.7 (in an obvious rougher form) follows from
Theorem 5.11 when R = ∞ and contains embedding results for Lebesgue spaces with mixed norms as
β = d/q1 + 2/q2 and for Lq -spaces as q = q1 = q2 .
Remark 5.14. We stated our results only for u ∈ C0∞ just for convenience. Let us show why, for
instance, Theorem 5.11 is valid as long as ∂t u, Du, D2 u ∈ Eq1 ,q2 ,β (CR ). For that, it suffices to prove that
for any R0 < R, ρ > 0, (t, x) ∈ CR0 the quantity

I := ρβ –kICR0 DukLr1 ,r2 (Cρ (t,x))

is less than the right-hand side of (5.12) with (R0 )−2 in place of R−2 . For ε > 0 define u(ε) = (ICR u) ∗ ζε ,
where ζε (x) = ε−d−1 ζ(t/ε, x/ε), nonnegative ζ ∈ C0∞ has integral one and ζ(t, x) = 0 for t ≥ 0. Also
introduce I ε by replacing u in the definition of I with u(ε) . Of course, I ε → I as ε ↓ 0 and by
Theorem 5.11

I ε ≤ Nk|∂t u(ε) | + |D2 u(ε) | kEq1 ,q2 ,β (CR0 ) + N(R0 )−2 ku(ε) kEq1 ,q2 ,β (CR0 ) =: J ε .

Observe that if ε is small enough and (s, y) ∈ CR0 , then ∂t u(ε) (s, y) = (ICR ∂t u) ∗ ζε (s, y). Similar
formulas are valid for D2 u(ε) and by Minkowski’s inequality (the norm of a sum is less then the sum of
norms) we have Z
ε

J ≤ ζ(s, y) NkICR (|∂t u| + |D2 u|)(· − ε(s, y) kEq1 ,q2 ,β (CR0 )
Rd+1

+N(R0 )−2 kICR u(· − ε(s, y)kEq1 ,q2 ,β (CR0 ) dyds
Z 
= ζ(s, y) NkICR (|∂t u| + |D2 u|) kEq1 ,q2 ,β (CR0 −ε(s,y))
Rd+1

+N(R0 )−2 kICR ukEq1 ,q2 ,β (CR0 −ε(s,y)) dyds.
Since in the last integral CR0 − ε(s, y) ⊂ CR if ε is small enough, it follows that for small ε

J ε ≤ NkICR (|∂t u| + |D2 u|) kEq1 ,q2 ,β (CR ) + N(R0 )−2 kICR ukEq1 ,q2 ,β (CR )

which yields the desired result.

Mathematics in Engineering Volume 5, Issue 2, 1–20.

 20

Acknowledgments

The author brings his gratitude to the referees for their useful comments.

Conflict of interest

The author declares no conflict of interest.

References

1. D. Adams, A note on Riesz potentials, Duke Math. J., 42 (1975), 765–778.

http://doi.org/10.1215/s0012-7094-75-04265-9
2. O. V. Besov, V. P. Il’in, S. M. Nikol’skii, Integral representations of functions and imbedding
theorems, Vol. 1, Washington, D.C.: Halsted Press [John Wiley & Sons], 1978.
3. F. Chiarenza, M. Frasca, Morrey spaces and Hardy-Littlewood maximal function, Rend. Mat. Appl.,
7 (1987), 273–279.
4. F. Chiarenza, M. Frasca, A remark on a paper by C. Fefferman, Proc. Amer. Math. Soc., 108 (1990),
407–409. http://doi.org/10.1090/S0002-9939-1990-1027825-X
5. G. Di Fazio, D. I. Hakim, Y. Sawano, Elliptic equations with discontinuous coefficients in
generalized Morrey spaces, Eur. J. Math., 3 (2017), 729–762. http://doi.org/10.1007/s40879-017-
0168-y
6. C. Fefferman, E. Stein, Some maximal inequalities, Amer. J. Math., 93 (1971), 107–115.
http://doi.org/10.2307/2373450
7. J. Garcia-Guevra, J. L. Rubio de Francia, Weighted norm inequalities and related topics, Elsevier,
1985.
8. M. Giaquinta, L. Martianazzi, An introduction to the regularity theory for elliptic
systems, Harmonic maps and minimal graphs, Edizioni della Normale Pisa, 2012.
http://doi.org/10.1007/978-88-7642-443-4
9. N. V. Krylov, Lectures on elliptic and parabolic equations in Sobolev spaces, Providence, RI:
Amer. Math. Soc., 2008. http://doi.org/10.1090/gsm/096/04
10. N. V. Krylov, Linear and fully nonlinear elliptic equations with Morrey drift, arXiv:2108.06840.

c 2023 the Author(s), licensee AIMS Press. This

is an open access article distributed under the
terms of the Creative Commons Attribution License
(http://creativecommons.org/licenses/by/4.0)

Mathematics in Engineering Volume 5, Issue 2, 1–20.