A SEMILINEAR PROBLEM ASSOCIATED TO THE

SPACE-TIME FRACTIONAL HEAT EQUATION IN RN

CARMEN CORTÁZAR, FERNANDO QUIRÓS, AND NOEMÍ WOLANSKI

arXiv:2405.18612v1 [math.AP] 28 May 2024

Abstract. We study the fully nonlocal semilinear equation Btα u ` p´∆qβ u “ |u|p´1 u, p ě 1, where
Btα stands for the Caputo derivative of order α P p0, 1q and p´∆qβ , β P p0, 1s, is the usual β power of
the Laplacian. We prescribe an initial datum in Lq pRN q.
We give conditions ensuring the existence and uniqueness of a solution living in Lq pRN q up to a
maximal existence time T that may be finite or infinite. If T is finite, the Lq norm of the solution
becomes unbounded as time approaches T , and u is said to blow up in Lq . Otherwise, the solution is
global in time.
For the case of nonnegative and nontrivial solutions, we give conditions on the initial datum that
ensure either blow-up or global existence. It turns out that every nonnegative nontrivial solution in Lq
blows up in finite time if 1 ă p ă pf :“ 1 ` 2β N
whereas if p ě pf there are both solutions that blow
up and global ones. The critical exponent pf , which does not depend on α, coincides with the Fujita
exponent for the case α “ 1, in which the time derivative is the standard (local) one. In contrast to
the case α “ 1, when α P p0, 1q the critical exponent p “ pf falls within the situation in which global
existence may occur. Our weakest condition for global existence and our condition for blow-up are
both related to the size of the mean value of the initial datum in large balls.

1. Introduction

1.1. Aim. We study the fully nonlocal semilinear problem

(1.1) Btα u ` p´∆qβ u “ |u|p´1 u in RN ˆ p0, T q, up¨, 0q “ u0 in RN ,
where p P r1, 8q. Here Btα , α P p0, 1q, denotes the so-called Caputo α-derivative, defined for smooth
functions by żt
α 1 upx, sq ´ upx, 0q
Bt upx, tq “ Bt ds,
Γp1 ´ αq 0 pt ´ sqα
and p´∆qβ , β P p0, 1s, is the usual β power of the Laplacian, defined for smooth functions when
β P p0, 1q by ż
vpxq ´ vpyq
p´∆qβ vpxq “ cN,β P. V. dy.
R N |x ´ y|N `2β
The positive normalization constant cN,β is chosen so that p´∆qβ “ F ´1 p| ¨ |2β Fq, where F denotes
Fourier transform. Problems like (1.1), nonlocal both in space and time, are useful to model situations
with long-range interactions and memory effects; see for instance [29].
If a solution to (1.1) is smooth enough, it is given by Duhamel’s type formula
ż żtż
(1.2) upx, tq “ u0 pyqZpx ´ y, tq dy ` |upy, sq|p´1 upy, sqY px ´ y, t ´ sq dyds,
RN 0 RN

2020 Mathematics Subject Classification. 35B44, 35B33, 35K58, 35R11, 35A01, 35A02, 35B60.
Key words and phrases. Fully nonlocal semilinear heat equation, Caputo derivative, fractional Laplacian, blow-up,
global existence, Fujita exponent.
1
2 C. CORTÁZAR, F. QUIRÓS AND N. WOLANSKI

where Z is the fundamental solution of the equation and Y “ R Bt1´α Z; see [20]. Here, R Btα stands for
the so-called Riemann-Liouville α-derivative, defined by
żt
R α 1 upx, sq
Bt upx, tq “ Bt α
ds.
Γp1 ´ αq 0 pt ´ sq

Throughout the article we consider u0 P Lq pRN q, q P r1, 8s, and look for Lq -solutions of (1.1) with
initial datum u0 and existence time T , that is, functions u P L8 q N
loc pr0, T q; L pR qq satisfying (1.2) in
Lq pRN q for almost all t P p0, T q, and such that
(1.3) ess lim }up¨, tq ´ u0 }Lq pRN q “ 0.
tÑ0`

Remark 1.1. The Caputo and the Riemann-Liouville α-derivatives are related by
Btα up¨, tq “ R Btα pup¨, tq ´ up¨, 0qq.
Thus, both derivatives coincide for functions with zero initial datum.

Our main goal is to give conditions under which problem (1.1) admits either Lq -global solutions,
that is, Lq -solutions with existence time T “ 8, or Lq -solutions with blow-up in finite time, that is,
Lq -solutions such that lim suptÕT }up¨, tq}Lq pRN q “ 8 for some T P p0, 8q.
In order to prove our results we use thoroughly the properties of Z and Y . These kernels are not
smooth in general and have a qualitative change of behavior across the critical dimensions N “ 2β and
N “ 4β. For the sake of simplicity, throughout this article we will only deal with the more involved
case N ą 4β, for which the kernels are more singular. The other cases can be treated similarly. We
recall the estimates on Z and Y required for our proofs, due to different authors, in Section 2.
Historical note. The nonlocal derivative Btα has been introduced independently by many authors; see
for instance [5, 8, 14, 16, 33]. It can be found even in a very old paper by Liouville [27]. However, we
follow the common convention of naming it after Caputo only.

1.2. Main results. The first part of the paper is devoted to basic issues such as existence, uniqueness,
prolongability and continuity. We assume always that p P r1, 8q.
We first consider, in Section 3, values of q in the good range,
N
(GR) q P pℓ, 8s with q ě p, where ℓ :“ pp ´ 1q.
2β
Notice that q “ 8 is always in the good range, while q “ 1 only belongs to it if p “ 1. If q satisfies (GR)
we are able to prove the existence of a unique maximal Lq -solution, defined up to a maximal positive
existence time T , for any initial datum in Lq pRN q. Moreover, u P Cpr0, T q; Lq pRN qq. If T is finite, the
solution blows up. When p “ 1, Lq -solutions are always global in time, that is, T “ 8.
If q P r1, pq, difficulties arise in the interpretation of (1.2), since |up¨, tq|p´1 up¨, tq need not a priori
be in Lq pRN q. On the other hand, if q P r1, ℓs, the problem is not expected to be well posed in Lq pRN q,
even if q ě p. In particular, we expect to have initial data in Lq pRN q for which no Lq -solution exists;
see Paragraph 1.4.2 below. To face these difficulties, to guarantee the existence of a solution (that
may not be unique; see Paragraph 1.4.3) outside the good range we require further integrability to the
initial data. Thus, in Section 4:
‚ If q P pℓ, pq, we consider u0 P Lq pRN q X Lp pRN q and we show that the corresponding maximal
Lp -solution is an Lq -solution with the same existence time.
 A FULLY NONLOCAL SEMILINEAR PROBLEM 3

‚ If q P r1, ℓs, we consider u0 P Lq pRN q X Lq̂ pRN q for some q̂ in the good range and show that
the corresponding maximal Lq̂ -solution is an Lq -solution with the same existence time.
In the second part of the paper we seek for conditions under which a nonnegative Lq -solution,
q P r1, 8s, is either global in time or, on the contrary, blows up in finite time when p P p1, 8q. We find
a critical exponent, pf :“ 1 ` 2β N , which, by analogy with the local case, we call the Fujita exponent,
such that every nonnegative and nontrivial Lq -solution blows up in finite time if p P p1, pf q whereas
for p P rpf , 8q there are both global and blowing up solutions. We remark that the value of the Fujita
exponent does not depend on the fractional order α.
We begin our analysis of the existence of a Fujita exponent in Section 5, where we construct global
in time Lq -solutions for all p P rpf , 8q and q P r1, 8s. We start by considering values of q in the
distinguished range,
(DR) q P pℓ, ℓps with q ě p
(we recall that we are assuming now that p ě pf , so that ℓ ě 1). Such values of q are contained in the
good range (GR). Therefore, given any u0 P Lq pRN q there is a maximal Lq -solution to (1.1). In order
to ensure that this Lq -solution is global in time we require in addition that u0 P Lℓ pRN q and one of
the two following conditions:
(1.4) }u0 }Lℓ pRN q small,
ż
α
´ Nα
(1.5) u0 ě 0, with sup R p´1 2β
α u0 pxq dx ă δ,
RąR0 |x|ăR 2β

where R0 “ R0 p}u0 }Lq pRN q q ą 0 decreases with }u0 }Lq pRN q and δ ą 0 is independent of u0 . Observe
that for every R ą 0 ż
α Nα ℓ´1
´
α u0 pxq dx ď ωN }u0 }Lℓ pRN q ,
ℓ
R p´1 2β
|x|ăR 2β
where ωN is the volume of the unit ball in RN ,
showing that the second condition is of a weaker nature
than the first one, except for p “ pf , when both conditions coincide, since in this case ℓ “ 1 and
1 N
p´1 ´ 2β “ 0.
To construct global solutions outside the distinguished range, we have to require further integrability
to the initial datum. In particular:
‚ If q P pℓ, pq, we consider u0 P Lℓ pRN q X Lp pRN q satisfying either (1.4) or (1.5), and prove that
the corresponding global Lp -solution is a global Lq -solution.
‚ If q P r1, ℓs, we consider u0 P Lq pRN q X Lq̂ pRN q for some q̂ in the good range, and satisfying
either (1.4) or (1.5), and prove that the corresponding global Lq̂ -solution is a global Lq -solution.
‚ If q P pℓp, 8q, we consider u0 P Lℓ pRN q X Lq pRN q satisfying either (1.4) or (1.5), and prove
that the corresponding global Lℓp -solution is a global Lq -solution.
Finally, in Section 6 we prove that there is a constant C depending only on N, α, β and p such that
if u0 P Lq pRN q is nonnegative and
ż
α
´ Nα
R α u0 pxq dx ą C for some R ą 0,
p´1 2β

|x|ăR 2β

then problem (1.1) cannot have a global Lq -solution with u0 as initial datum. This shows that condi-
tion (1.5) for global existence is “sharp”. As a consecuence, on the one hand there are Lq -solutions with
α N
blow-up in finite time for every p P p1, 8q and q P r1, 8s, and on the other hand, since p´1 ´ 2β ą0
4 C. CORTÁZAR, F. QUIRÓS AND N. WOLANSKI

if p P p1, pf q, every nonnegative and nontrivial Lq -solution to problem (1.1) blows up in finite time in
this range of values of p.

1.3. Precedents. In the case of the classical heat equation, α “ 1, β “ 1, Fujita proved in his classical
paper [11] that problem (1.1) admits no global L8 -solution when p P p1, 1 ` 2{N q, while it has both
global L8 -solutions and solutions which blow up in L8 in finite time if p ą 1`2{N . When p “ 1`2{N ,
it was shown later in [18] (for N “ 1, 2), [22] and [1], that all nontrivial solutions to problem (1.1)
blow up in L8 in finite time. The analysis in other Lq spaces, q P r1, 8q, was performed in [37, 38, 39];
see also [15]. These results have been generalized to deal with some more general local situations,
where the Laplacian is replaced by a different local diffusion operator and/or the reaction term up by a
nonlinearity f puq. We refer the interested reader to the surveys [6, 12, 25] and the monographs [32, 34].
In the nonlocal framework, but yet with a local time derivative, the Fujita exponent in L8 for
problem (1.1) with α “ 1 and β P p0, 1q is pf “ 1 ` 2β{N ; see [30], and also [35, 3]. Other nonlocal
diffusion operators have been considered for instance in [13].
Notice that the Fujita exponents in the cases with α “ 1 coincide with the ones for α P p0, 1q.
However, when the time derivative is nonlocal the critical exponent pf belongs to the case where there
are nonnegative and nontrivial Lq -solutions that exist globally as well as those blowing up in finite
time.
As for nonlocal time derivatives, [26] considers the Fujita phenomenon for an equation similar
to (1.1) involving the Caputo-Hadamard time derivative, instead of a Caputo derivative. On the
other hand, [40] analyzes (1.1) with α P p0, 1q when β “ 1 and considers a somewhat different meaning
of solution. Namely, the authors look at (1.2) in terms of the operator giving the solution to the
Cauchy problem for the classical heat equation, and strongly use the exponential decay in time of the
norm of such operator. Also, due to their approach, their results are only valid for continuous initial
data that tend to zero at infinity. Our methods allow us to work with any β P p0, 1s and much more
general initial data. Of course, the value that we obtain for the Fujita exponent (when β “ 1) is the
same as the one in [40]. We point out that the condition in [40] for global existence when p ě pf is
the strong (not sharp) one (1.4).
For the fully nonlocal case, α, β P p0, 1q, some (non-sharp) blow-up results were obtained in [19]
in the L8 framework. For the linear case p “ 1 there is a recent paper [7] where the authors prove
the existence and uniqueness of a fundamental solution (with initial datum a Dirac mass) to (1.1),
and obtain interesting asymptotic information for general solutions when N “ 1, α P r1{2, 1q and
β P p0, 1{2q.
Let us finally mention the paper [10], that considers blow-up issues including the Fujita phenomenon,
in L8 pRN q, for nonnegative solutions of a different fully fractional heat equation, namely pBt ´ ∆uqα “
up , p ą 0, α P p0, 1q.

1.4. Some open problems.

1.4.1. Small dimensions. In this paper we consider the case of high dimensions, N ą 4β. Other
dimensions can be treated in a similar way, since good estimates for the kernels Z and Y are available;
see for instance [20, 21]. If N ă 2β both kernels are bounded, and the proofs will be more similar to
the ones for the case α “ 1 in which the time derivative is local. In the critical cases N “ 2β and
N “ 4β logarithmic behaviors occur, and the proofs may be a bit more involved.
 A FULLY NONLOCAL SEMILINEAR PROBLEM 5

1.4.2. Instantaneous blow-up outside the good range. In the local case, α “ 1, β “ 1, if q P r1, ℓq there
are some initial data u0 P Lq pRN q for which problem (1.1) admits no local solution (for any positive
existence time) in any reasonable weak sense [4, 38, 39]; see also [24]. We expect this also to be the
case in the nonlocal setting, hence the need of extra integrability conditions on the initial datum to
ensure local existence. This interesting topic deserves being studied somewhere else.

1.4.3. Lack of uniqueness outside the good range. When q P r1, ℓq and 1 ď ℓ ď p ´ 1 (which means
that p ą pN ` 2q{pN ´ 2q), problem (1.1) does not have uniqueness of Lq -solutions in the local case
α “ 1, β “ 1 [17]; see also [2, 4, 28, 31, 36]. It would be interesting to study the possibility of such
phenomenon in our nonlocal setting.

1.5. Notations. For our analysis, it will be convenient to decompose solutions to (1.2) as
ż żtż
(1.6) upx, tq “ u0 pyqZpx ´ y, tq dy ` |upy, sq|p´1 upy, sqY px ´ y, t ´ sq dyds .
loooooooooooooomoooooooooooooon
R N
looooooooooooooooooooooooooooomooooooooooooooooooooooooooooon
0 R N

Lu0 px,tq Nu px,tq

The “linear” part of the solution, Lu0 , solves the linear problem
Btα u ` p´∆qβ u “ 0 in RN ˆ p0, 8q, up¨, 0q “ u0 in RN .
The “nonlinear part”, Nu px, tq, accounts for the effect of the reaction nonlinearity, which is responsible
of the possibility of blow up when p P p1, 8q.
N
We denote by pc :“ N ´2β a critical exponent that will appear several times along the paper.

2. Estimates for the kernels

We recall here some estimates on the kernels Z and Y in (1.2) that will be used throughout the
paper. Here and in what follows we assume always, without further mention, that N ą 4β.
Both kernels Z and Y have a selfsimilar structure,
´ Nα α
´ 2β ´p1´α` Nα q α
´ 2β
(2.1) Zpx, tq “ t 2β F pxt q, Y px, tq “ t 2β Gpxt q,
for certain radially decreasing positive profiles F and G that are singular at the origin for the dimensions
that we are considering, and smooth away from it.
For all t ą 0, }Zp¨, tq}L1 pRN q “ 1 and
# α
t´α |x|2β´N if |x| ă t 2β ,
(2.2) 0 ă Zpx, tq ď K α
tα |x|´p2β`N q if |x| ą t 2β ,
# α
t´p1`αq |x|2β´N if |x| ă t 2β ,
|Zt px, tq| ď K α
tα´1 |x|´p2β`N q if |x| ą t 2β .

The estimate for Z close to the origin is sharp. Hence, Zp¨, tq P Lr pRN q if and only if r P r1, pc q, and
in such case we have
Nα 1
(2.3) }Zp¨, tq}Lr pRN q ď Kt´ 2β p1´ r q , t ą 0,
´1´ Nα
2β
p1´ 1r q
(2.4) }Zp¨, tq ´ Zp¨, t0 q}Lr pRN q ď Kpt ´ t0 qt0 , t ě t0 ą 0.
6 C. CORTÁZAR, F. QUIRÓS AND N. WOLANSKI

Remark 2.1. Let u0 P Lq pRN q. Since Zp¨, tq is an approximation of the identity as t Ñ 0` , then
}Lu0 p¨, tq ´ u0 }Lq pRN q Ñ 0` as t Ñ 0` .
Therefore, if u satisfies (1.2), to prove (1.3) it is enough to check that
(2.5) ess lim }Nu p¨, tq}Lq pRN q “ 0.
tÑ0`

As for the kernel Y ,

# α
t´p1`αq |x|4β´N if |x| ă t 2β ,
0 ă Y px, tq ď K 2α´1 ´p2β`N q α
t |x| if |x| ą t 2β ,
# α
t´p2`αq |x|4β´N if |x| ă t 2β ,
|Yt px, tq| ď K 2α´2 ´p2β`N q α
t |x| if |x| ą t 2β .
The estimate for Y close to the origin is sharp. Thus, Y p¨, tq P Lr pRN q if and only if r P r1, N {pN ´4βqq,
and in such case we have
(2.6) }Y p¨, t ´ sq}Lr pRN q ď Kpt ´ sq´σprq , 0 ď s ă t,
´σprq´1
(2.7) }Y p¨, t ´ sq ´ Y p¨, t0 ´ sq}Lr pRN q ď Kpt ´ t0 qpt0 ´ sq , 0 ď s ă t0 ď t,
where
Nα 1
σprq :“ 1 ´ α ` p1 ´ q.
2β r
Notice that σprq ă 1 if and only if r ă pc . Hence, Y p¨, tq P L1loc pr0, 8q; Lr pRN qq if and only if r P r1, pc q.
This restriction on r will appear several times in our analysis.
For these estimates, and many more, on Z and Y , see for instance [9, 20, 21, 23].

3. Basic theory in the good range

In this section we deal with values of q satisfying (GR). In this good range we will be able to prove
the existence of a unique maximal Lq -solution for any initial datum in Lq pRN q, which is moreover a
continuous curve in Lq pRN q up to the maximal existence time. When the latter is finite, Lq -solutions
become unbounded as they approach it.

3.1. Local existence. We start by proving, via a fixed point argument, that problem (1.1) has an
Lq -solution for any u0 P Lq pRN q if T is small.
Theorem 3.1. Let p P r1, 8q and q in the good range (GR). Given u0 P Lq pRN q, there exists
t0 ą 0 such that problem (1.2) has a solution u P L8 p0, t0 ; Lq pRN qq satisfying (1.3). If u0 ě 0, u is
nonnegative. When restricted to p0, T q, T P p0, t0 s, u is an Lq -solution to (1.1) with existence time T .

Proof. Let C0 “ 2}u0 }Lq pRN q and t0 ą 0 to be chosen later. For v P L8 p0, t0 ; Lq pRN qq, let
ż żtż
T vpx, tq “ u0 pyqZpx ´ y, tq dy ` |vpy, sq|p´1 vpy, sqY px ´ y, t ´ sq dyds.
RN 0 RN
Our aim is to prove that T has a fixed point in
` ˘
K “ tv P L8 0, t0 ; Lq pRN q , ~v~ :“ sup }vp¨, tq}Lq pRN q ď C0 u
tPp0,t0 q

if t0 is small enough. Observe that K is a closed metric space with the metric dpv, v̂q :“ ~v ´ v̂~.
 A FULLY NONLOCAL SEMILINEAR PROBLEM 7

We first prove that T pKq Ă K if t0 is small enough. Let r be such that

1 p 1
(3.1) 1´ “ ´ .
r q q
1 2β
Since p ě 1 and q ą ℓ, then 0 ď 1 ´ r ă N, or equivalently, r P r1, pc q, whence estimate (2.6) holds
and σprq ă 1. Thus, for 0 ď t ď t0 ,
żt
}T vp¨, tq}Lq pRN q ď }u0 }Lq pRN q ` }|vp¨, sq|p } q }Y p¨, t ´ sq}Lr pRN q ds
0 L p pRN q
(3.2) żt ´1
C0 C0p´1 K 1´σprq ¯
ď ` C0p K pt ´ sq´σprq ds ď C0 ` t ď C0
2 0 2 1 ´ σprq 0
if t0 is small, where we have used also that }Zp¨, tq}L1 pRN q “ 1 for all t ą 0.
Let us check now that T is a strict contraction in K if t0 is small. Let v1 , v2 P K. By (2.6),
}T v1 p¨, tq ´ T v2 p¨, tq}Lq pRN q
żt
ď K }|v1 p¨, sq|p´1 v1 p¨, sq ´ |v2 p¨, sq|p´1 v2 p¨, sq} q pt ´ sq´σprq ds
0 L p pRN q

t1´σprq
ď K sup }|v1 p¨, sq|p´1 v1 p¨, sq ´ |v2 p¨, sq|p´1 v2 p¨, sq} q
sPp0,t0 q L p pRN q 1 ´ σprq
if 0 ă t ď t0 . On the other hand, if p ą 1,
}|v1 p¨, sq|p´1 v1 p¨, sq ´ |v2 p¨, sq|p´1 v2 p¨, sq} q
L p pRN q
p´1
ď p}p|v1 p¨, sq| ` |v2 p¨, sq|q |v1 p¨, sq ´ v2 p¨, sq|} q
L p pRN q

ď p}|v1 p¨, sq| ` |v2 p¨, sq|}p´1 }v p¨, sq ´ v2 p¨, sq}Lq pRN q ,
Lq pRN q 1
so that
(3.3) sup }|v1 p¨, sq|p´1 v1 p¨, sq ´ |v2 p¨, sq|p´1 v2 p¨, sq} q ď pp2C0 qp´1 }v1 p¨, sq ´ v2 p¨, sq}Lq pRN q ,
sPp0,t0 q L p pRN q

an inequality which is trivially true for p “ 1. Hence, for every p ě 1, 0 ă t ď t0 ,

1´σprq
t0 1
(3.4) }T v1 p¨, tq ´ T v2 p¨, tq}Lq pRN q ď pKp2C0 qp´1 ~v1 ´ v2 ~ ď ~v1 ´ v2 ~
1 ´ σprq 2
if t0 is small enough. Hence, T has a unique fixed point u in K, which solves (1.2). Moreover,
żt
}Nu p¨, tq}Lq pRN q ď }|up¨, sq|p } pq N }Y p¨, t ´ sq}Lr pRN q ds
0 L pR q
żt
C0p K 1´σprq
ď C0p K pt ´ sq´σprq ds ď t ,
0 1 ´ σprq
which implies (2.5). Hence, (1.3) holds; see Remark 2.1.
Now, assume u0 ě 0 and let
` ˘
K0 “ tv P L8 0, t0 ; Lq pRN q , ~v~ ď C0 and v ě 0u Ă K.
When restricted to K0 , T is still a contraction if t0 is small as above, and moreover T pK0 q Ă K0 . Since
T has a unique fixed point in K, and a unique fixed point in K0 Ă K, they are in fact the same one,
belonging to K0 . 
8 C. CORTÁZAR, F. QUIRÓS AND N. WOLANSKI

Remark 3.1. If p “ 1, we are in the good range for all q P r1, 8s, and we can choose the time step t0
in Theorem 3.1 independently of the initial datum u0 ; see (3.2) and (3.4).
Remark 3.2. Observe that p ě pc if and only if ℓ ě p. Therefore, when p ě pc , the condition q ą ℓ is
enough to guarantee that q is in the good range (GR), and hence local existence, whereas if p P r1, pc q
we just need to require that q ě p.

3.2. Continuity. We next prove that, in the good range, Lq -solutions are continuous curves in Lq in
their time interval of existence.
Proposition 3.1. Let p P r1, 8q, q in the good range (GR), and u0 P Lq pRN q. If u is an Lq -solution
to (1.2) in p0, T q, then u P Cpr0, T q; Lq pRN qq.

Proof. Let us first check that up¨, tq given by (1.2) is well defined and belongs to Lq pRN q for all
t P p0, T q. Indeed, since }Zp¨, tq}L1 pRN q “ 1 for all t ą 0, taking r as in (3.1) and using the bound (2.6),
żt
}up¨, tq}Lq pRN q ď }u0 }Lq pRN q ` }|up¨, sq|p } pq N }Y p¨, t ´ sq}Lr pRN q ds
0 L pR q
żt
t1´σprq
ď K sup }up¨, sq}pLq pRN q pt ´ sq´σprq dyds “ K sup }up¨, sq}pLq pRN q .
sPp0,tq 0 sPp0,tq 1 ´ σprq

Continuity at t “ 0 comes for free by the very definition of Lq -solution. To consider the continuity
outside the origin we make the decomposition (1.6), and prove separately the continuity of Lu0 and Nu .
Let 0 ă t1 ă t2 ă T . We have, using (2.4) with r “ 1,
}Lu0 p¨, t2 q ´ Lu0 p¨, t1 q}Lq pRN q ď }u0 }Lq pRN q }Zp¨, t2 q ´ Zp¨, t1 q}L1 pRN q ď }u0 }Lq pRN q Kpt2 ´ t1 qt´1
1 ,

from where the desired continuity for Lu0 follows immediately. As for Nu , we make the decomposition
Nu p¨, t2 q ´ Nu p¨, t1 q “ Ip¨, t2 , t1 q ` IIp¨, t2 , t1 q ` IIIp¨, t2 , t1 q, where
ż t1 ´εpt2 ´t1 q ż
` ˘
Ipx, t2 , t1 q “ |upy, sq|p´1 upy, sq Y px ´ y, t2 ´ sq ´ Y px ´ y, t1 ´ sq dyds,
0 RN
ż t1 ż
` ˘
IIpx, t2 , t1 q “ |upy, sq|p´1 upy, sq Y px ´ y, t2 ´ sq ´ Y px ´ y, t1 ´ sq dyds,
t1 ´εpt2 ´t1 q RN
ż t2 ż
IIIpx, t2 , t1 q “ |upy, sq|p´1 upy, sqY px ´ y, t2 ´ sq dyds,
t1 RN

with ε ą 0 small enough so that t1 ´ εpt2 ´ t1 q ą 0. Notice that when t2 Ñ t1 and t1 ą 0 this means
no restriction for ε.
To estimate I we take r as in (3.1) and use the bound (2.7),
ż t1 ´εpt2 ´t1 q
}Ip¨, t2 , t1 q}Lq pRN q ď }|up¨, sq|p } pq N }Y p¨, t2 ´ sq ´ Y p¨, t1 ´ sq}Lr pRN q dyds
0 L pR q
ż t1 ´εpt2 ´t1 q
ď K sup }up¨, tq}pLq pRN q pt2 ´ t1 q pt1 ´ sq´σprq´1 dyds
tPp0,t2 q 0
´σprq
ε´σprq pt2 ´ t1 q1´σprq ´ t1 pt2 ´ t1 q
“ K sup }up¨, tq}pLq pRN q .
tPp0,t2 q σprq
 A FULLY NONLOCAL SEMILINEAR PROBLEM 9

As for II, using the bound (2.6),

ż t1
` ˘
}IIp¨, t2 , t1 q}Lq pRN q ď }|up¨, sq|p } q }Y p¨, t2 ´ sq}Lr pRN q ` }Y p¨, t1 ´ sq}Lr pRN q dyds
t1 ´εpt2 ´t1 q L p pRN q
ż t1
` ˘
ď K sup }up¨, tq}pLq pRN q pt2 ´ sq´σprq ` pt1 ´ sq´σprq dyds
tPp0,t2 q t1 ´εpt2 ´t1 q
` ˘
p1 ` εq1´σprq ´ 1 ` ε1´σprq
“ K sup }up¨, tq}pLq pRN q pt2 ´ t1 q 1´σprq
.
tPp0,t2 q 1 ´ σprq

In order to estimate III we use again the bound (2.6),

ż t2 ż
}IIIp¨, t2 , t1 q}Lq pRN q ď }|up¨, sq|p } q }Y p¨, t2 ´ sqLr pRN q dyds
t1 RN L p pRN q
ż t2
ď K sup }up¨, tq}pLq pRN q pt2 ´ sq´σprq dyds
tPp0,t2 q t1

pt2 ´ t1 q1´σprq
“ K sup }up¨, tq}pLq pRN q . 
tPp0,t2 q 1 ´ σprq

Remark 3.3. If sup }up¨, tq}Lq pRN q ă 8, we may repeat the arguments in the proof of Proposition 3.1
tPp0,T q
to check first that
ż żTż
up¨, T q :“ u0 pyqZpx ´ y, T q dy ` |upy, sq|p´1 upy, sqY px ´ y, T ´ sq dyds
RN 0 RN

is well defined and belongs to Lq pRN q, and then that lim up¨, tq “ up¨, T q in Lq pRN q.
tÕT

3.3. Extension. We now show that if the Lq -norm of an Lq -solution is bounded up to its existence
time T , then we can construct an Lq -solution with a larger existence time.
Theorem 3.2. Let p P r1, 8q and q in the good range (GR). Let u be an Lq -solution to (1.1) with
initial datum u0 P Lq pRN q and existence time T . If suptPp0,T q }up¨, tq}Lq pRN q ă 8, then there exists
some τ ą 0 such that (1.1) has an Lq -solution ū with initial datum u0 P Lq pRN q and existence time
T̄ “ T ` τ . Moreover, suptPp0,T̄ q }ūp¨, tq}Lq pRN q ă 8. If u ě 0, then ū ě 0.

Proof. The idea is to construct a solution ū that coincides with u for t P p0, T ´ τ q for some τ P p0, T {2q
small to be further specified later. Thus, for t P pT ´ τ, T ` τ q we should have
ż ż T ´τ ż
ūpx, tq “ u0 pyqZpx ´ y, tq dy ` |upy, sq|p´1 upy, sqY px ´ y, t ´ sq dyds
RN 0 RN
żt ż
` |ūpy, sq|p´1 ūpy, sqY px ´ y, t ´ sq dyds.
T ´τ RN

With this in mind, we consider the set

` ˘
K̄ :“ tv P L8 T ´ τ, T ` τ ; Lq pRN q , ~v~ :“ sup }vp¨, tq}Lq pRN q ď C1 u,
tPpT ´τ,T `τ q
10 C. CORTÁZAR, F. QUIRÓS AND N. WOLANSKI

where C1 :“ 2 sup }up¨, tq}Lq pRN q , and define in it an operator T̄ by

tPp0,T q
ż ż T ´τ ż
T̄ vpx, tq “ u0 pyqZpx ´ y, tq dy ` |upy, sq|p´1 upy, sqY px ´ y, t ´ sq dyds
R N 0 R N
(3.5) żt ż
` |ūpy, sq|p´1 ūpy, sqY px ´ y, t ´ sq dyds.
T ´τ RN

We will prove that T̄ has a (unique) fixed point ũ in K̄ if τ ą 0 is small enough. Then,
#
upx, tq, t P p0, T ´ τ q,
ūpx, tq “
ũpx, tq, t P pT ´ τ, T ` τ q,

will be an Lq -solution to (1.1) with existence time T̄ “ T ` τ .

To prove that T̄ pK̄q Ă K̄ if τ is small enough, we rewrite T̄ v as T¯ v “ I ` II ` III ` IV+V, where
ż żTż
Ipxq “ u0 pyqZpx ´ y, T q dy ` |upy, sq|p´1 upy, sqY px ´ y, T ´ sq dyds,
RN 0 RN
ż
` ˘
IIpx, tq “ u0 pyq Zpx ´ y, tq ´ Zpx ´ y, T q dy,
RN
ż T ´τ ż
` ˘
IIIpx, tq “ |upy, sq|p´1 upy, sq Y px ´ y, t ´ sq ´ Y px ´ y, T ´ sq dyds,
0 RN
żT ż
IVpx, tq “ |upy, sq|p´1 upy, sqY px ´ y, T ´ sq dyds,
T ´τ RN
żt ż
Vpx, tq “ |vpy, sq|p´1 vpy, sqY px ´ y, t ´ sq dyds.
T ´τ RN

On the one hand }I}Lq pRN q “ }up¨, T q}Lq pRN q ď C1 {2. On the other hand, by Young’s inequality
and (2.4) with r “ 1, if τ P p0, T {p8Kqq we have
#
KC1 pt ´ T qT ´1 , t P pT, T ` τ q,
}IIp¨, tq}Lq pRN q ď }u0 }Lq pRN q }Zp¨, tq ´ Zp¨, T q}L1 pRN q ď
2 pT ´ tqt´1 , t P pT ´ τ, T q,
C1
ď KC1 T ´1 τ ď for all t P pT ´ τ, T ` τ q.
8

In order to estimate III, IV and V, we take r as in (3.1), so that, since we are in the good range,
r P r1, pc q and σprq P p0, 1q. Young’s inequality plus (2.7), in the estimate for III, or (2.6), in the
estimates for IV and V, yield
ż T ´τ
τ 1´σprq C1
}IIIp¨, tq}Lq pRN q ď C1p Kτ pT ´ sq´σprq´1 ds ď C1p K ď ,
0 σprq 8
żT
τ 1´σprq C1
}IVp¨, tq}Lq pRN q ď C1p K pT ´ sq´σprq ds “ C1p K ď ,
T ´τ 1 ´ σprq 8
żt
pt ´ T ` τ q1´σprq p2τ q1´σprq C1
}Vp¨, tq}Lq pRN q ď C1p K pt ´ sq´σprq ds “ C1p K ď C1p K ď
T ´τ 1 ´ σprq 1 ´ σprq 8
 A FULLY NONLOCAL SEMILINEAR PROBLEM 11

if τ is small enough, namely, if

!´ C 1´p σprq ¯ 1
1 ´ C11´p p1 ´ σprqq ¯ 1´σprq
1 )
1 1´σprq
τ ď min , .
8K 2 8K
Now that we know that T̄ pK̄q Ă K̄, we will prove that T̄ is a contraction in K̄ for the distance
associated to the norm ~v~ if τ is small. Let v1 , v2 P K̄. For t P pT ´ τ, T ` τ q,
}T̄ v1 p¨, tq ´ T̄ v2 p¨tq}Lq pRN q
żt
ď }|v1 p¨, sq|p´1 v1 p¨, sq ´ |v2 p¨, sq|p´1 v2 p¨, sq} q }Y p¨, t ´ sq}Lr pRN q ds.
T ´τ L p pRN q
´ ¯ 1
1 p2C1 q1´p p1´σprqq 1´σprq
Hence, by (3.3) with C0 replaced by C1 , if τ ď 2 2pK , then
żt
}T̄ v1 p¨, tq ´ T̄ v2 p¨, tq}Lq pRN q ď pp2C1 q p´1
~v1 ´ v2 ~K pt ´ sq´σprq ds
T ´τ
pt ´ T ` τ q1´σprq
“ pp2C1 qp´1 K ~v1 ´ v2 ~
1 ´ σprq
p2τ q1´σprq 1
ď pp2C1 qp´1 K ~v1 ´ v2 ~ ă ~v1 ´ v2 ~.
1 ´ σprq 2
Let now u0 ě 0 and
K̄0 :“ tv P L8 pT ´ τ, T ` τ ; Lq pRN qq , ~v~ ď C1 and v ě 0u.
Let us see that if u ě 0 and τ is chosen as above, then T̄ : K̄0 Ñ K̄0 and it is a contraction. To this
aim we just have to see that v ě 0 implies that T v ě 0. This follows immediately from the definition
of T̄ ; see (3.5). 
Remark 3.4. If p “ 1 the increment τ of the existence time in Theorem 3.2 can be chosen indepen-
dently of sup }up¨, tq}Lq pRN q .
tPp0,T q

Remark 3.5. By construction, ūp¨, tq “ up¨, tq for t P p0, T ´ τ q. In the next paragraph we will prove
that in fact the equality holds up to time T .

3.4. Uniqueness. In the good range any two Lq -solutions with the same initial datum coincide in
their common existence time. In particular, there is at most one Lq -solution with a given existence
time.
Theorem 3.3. Let p P r1, 8q and q in the good range (GR). Let u1 and u2 be two Lq -solutions to (1.1)
with initial datum u0 P Lq pRN q and corresponding existence times T1 and T2 . Then u1 p¨, tq “ u2 p¨, tq
for all t P p0, mintT1 , T2 uq.

Proof. Let t̄ P p0, mintT1 , T2 uq. We define K and T as in the proof of Theorem 3.1, but now with
C0 “ maxt2}u0 }Lq pRN q , sup }u1 p¨, tq}Lq pRN q , sup }u2 p¨, tq}Lq pRN q u
tPp0,t̄q tPp0,t̄q

Repeating that proof, we see that if t0 ą 0 is small enough then T : K Ñ K is a contraction. Therefore,
T has a unique fixed point in K. As u1 and u2 belong to K if t0 ď t̄ and are fixed points of T , then
u1 p¨, tq “ u2 p¨, tq if 0 ă t ď t0 .
12 C. CORTÁZAR, F. QUIRÓS AND N. WOLANSKI

Let Tp :“ suptt̂ P p0, mintT1 , T2 uq : u1 p¨, tq “ u2 p¨, tq for 0 ă t ď t̂u. Assume for contradiction that
T ă mintT1 , T2 u. Let us call upx, tq “ u1 px, tq “ u2 px, tq for 0 ă t ă Tp and
p

C1 “ maxt2 sup }up¨, tq}Lq pRN qq , sup }u1 p¨, tq}Lq pRN q , sup }u2 p¨, tq}Lq pRN q u.
tPp0,Tpq tPp0,Tpq tPp0,Tpq

Let K̄ and T̄ as in Theorem 3.2 with T replaced by Tp. As in that theorem, we see that T̄ : K̄ Ñ K̄ is
a contraction if τ is small enough, so that T̄ has a unique fixed point in K̄. Besides, if in addition τ
is such that Tp ` τ ď mintT1 , T2 u, then, u1 and u2 belong to K̄ and they are both fixed points
of T̄ . Hence, u1 p¨, tq “ u2 p¨, tq for 0 ă t ď Tp ` τ . But this contradicts the definition of Tp. Hence,
Tp “ mintT1 , T2 u. 

3.5. Maximal solutions. If q is in the good range (GR), we define the maximal existence time of
Lq -solutions to (1.1) with initial datum u0 P Lq pRN q by
Tm pu0 q :“ suptT ą 0 : D an Lq -solution to (1.1) with initial datum u0 and existence time T u.
In this range, an Lq -solution with existence time T belongs to Cpr0, T q; Lq pRN qq, so that it makes sense
to consider its value (in Lq pRN q) at a time t P r0, T q. Given t P r0, Tm q we define um p¨, tq “ up¨, tq,
where u is any Lq -solution to (1.1) with initial datum u0 and existence time T ą t. Thanks to
Theorem 3.3, um is well defined, and it is obviously an Lq -solution to (1.1) with existence time Tm .
Hence, by Proposition 3.1 um P Cpr0, Tm q; Lq pRN q. This function um is known as the maximal Lq -
solution to (1.1) with initial datum u0 .
As corollaries to Theorem 3.2, we have the following results for maximal Lq -solutions in the good
range.
Corollary 3.1. Let p “ 1, q P r1, 8s and u0 P Lq pRN q. The unique maximal Lq -solution to prob-
lem (1.1) with initial datum u0 is global in time.
Corollary 3.2. Let p P p1, 8q, q in the good range (GR), and u0 P Lq pRN q. Let u be the maximal
Lq -solution to (1.1) with initial datum u0 P Lq and T its maximal existence time. If T ă 8, then
lim sup }up¨, tq}Lq pRN q “ `8.
tÕT

3.6. More general nonlinearities. Arguments similar to the ones that we have used in the previous
paragraphs allow to deal with some nonlinearities that are not powers. A case that only requires
obvious changes in the proofs is that of globally Lipschitz nonlinearities f such that f p0q “ 0.
Theorem 3.4. Let f : R Ñ R be a Lipschitz continuous function such that f p0q “ 0. Let u0 P Lq pRN q
for some q ě 1. Then, there exists a unique global Lq -solution to
(3.6) Btα u ` p´∆qβ u “ f puq in RN ˆ p0, 8q, up¨, 0q “ u0 in RN ,
which belongs moreover to Cpr0, 8q; L8 pRN qq.

If f is yet globally Lipschitz, but f p0q ‰ 0. we still have a global existence result if the initial datum
is bounded.
Theorem 3.5. Let f : R Ñ R be a Lipschitz continuous function and let u0 P L8 pRN q. Then, there
exists a unique global L8 -solution to (3.6), which belongs moreover to Cpr0, 8q; L8 pRN qq.
 A FULLY NONLOCAL SEMILINEAR PROBLEM 13

Proof. The first step is to prove existence for small times. We proceed as in Theorem 3.1, now with
C0 “ maxt2}u0 }L8 pRN q ; |f p0q|u, and find that T : K Ñ K is a contraction if t0 is small, how small
not depending on the value of C0 . Indeed, let L be a Lipschitz constant for f . Then, if v P K, and
` α ˘1{α
0 ă t ď t0 ď 2p1`Lq ,

C0 ` ˘ tα ´1 tα ¯
}T vp¨, tq}L8 pRN q ď ` |f p0q| ` LC0 ď C0 ` p1 ` Lq 0 ď C0 .
2 α 2 α
` α ˘1{α
On the other hand, if v1 , v2 P K and 0 ă t ď t0 ă 2L ,
tα0
}T v1 p¨, tq ´ T v2 p¨, tq}L8 pRN q ď L sup }v1 p¨, sq ´ v2 p¨, sq}L8 pRN q
α sPp0,t0 q
1
ă sup }v1 p¨, sq ´ v2 p¨, sq}L8 pRN q .
2 sPp0,t0 q

In a similar way we can prove a result as the(one in Theorem 3.2. As τ can be taken independently
of C1 “ max 2 suptPp0,T q }up¨, tq}L8 pRN q ; |f p0q| , by iterating the result we find that the solution is
global.
Uniqueness follows easily as in Theorem 3.3, and continuity as in Proposition 3.1. 

When f is only locally Lipschitz, we cannot ascertain global existence, but we can yet show the
existence of a unique maximal L8 -solution to (3.6) for any u0 P L8 pRN q. We omit the proof, since it
is an easy adaptation of previous arguments.
Theorem 3.6. Let f : R Ñ R be locally Lipschitz continuous and u0 P L8 pRN q. Then, there exists a
unique maximal L8 -solution to (3.6). Moreover, if the maximal existence time T is finite, then
lim sup }up¨, tq}L8 pRN q “ `8.
tÕT

4. Basic theory outside the good range

There are two reasons why q may fail to be in the good range: either p P r1, pc q and q P r1, pq, or
p P rpc , 8q and q P r1, ℓs (see Remark 3.2). In order to guarantee the existence of an Lq -solution in
these cases we require the initial data to satisfy an extra integrability condition.
If p P r1, pc q, q P r1, pq, and u0 P Lp pRN q, we know from the previous section that there exists a
maximal Lp -solution to (1.1). If in addition u0 P Lq pRN q, then this Lp -solution is also an Lq -solution,
as we show next.
Theorem 4.1. Let p P r1, pc q, q P r1, pq, and u0 P Lp pRd q X Lq pRd q. If u is an Lp -solution to (1.1), it
is also an Lq -solution. Moreover, u P Cpr0, T q; Lq pRN qq.

Proof. We make the decomposition (1.6). On the one hand,

}Lu0 p¨, tq}Lq pRN q ď }u0 }Lq pRN q }Zp¨, tq}L1 pRN q ď }u0 }Lq pRN q .

Hence, to check that u is an Lq -solution it is enough to show that Nu P L8 p0, τ ; Lq pRN qq for all
τ P p0, T q, and (2.5); see Remark 2.1.
14 C. CORTÁZAR, F. QUIRÓS AND N. WOLANSKI

Let τ P p0, T q. By hypotheses, u P L8 p0, τ ; Lp pRN qq. Since p P r1, pc q, then q P r1, pc q, so that
σpqq ă 1. Therefore,
żt
}Nu p¨, tq}Lq pRN q ď }|up¨, sq|p }L1 pRN q }Y p¨, t ´ sq}Lq pRN q ds
0
żt
t1´σpqq
ď K sup }up¨, sq}pLp pRN q pt ´ sq´σpqq ds ď K sup }up¨, sq}pLp pRN q
sPp0,τ q 0 sPp0,τ q 1 ´ σpqq
if t P p0, τ q, which implies both Nu P L8 p0, τ ; Lq pRN qq, and (2.5).
The proof of continuity is similar to that of Proposition 3.1. We omit the details. 

If p P rpc , 8q, q P r1, ℓs, and u0 P Lq̂ pRN q for some q̂ in the distinguished range (DR), we know from
the results in Section 3 that there exists a maximal Lq̂ -solution to (1.1). If in addition u0 P Lq pRN q,
we will show that this Lq̂ -solution is also an Lq -solution.
Remark 4.1. If p P rpc , 8q, q P r1, ℓs, and u0 P Lq pRN q X Lq̃ pRN q for some q̃ in the good range (GR),
then u0 P Lq̂ pRN q for some q̂ in the distinguished range.
Theorem 4.2. Let p P rpc , 8q, q P r1, ℓs, q̂ in the distinguished range (DR), and u0 P Lq pRd qXLq̂ pRd q.
If u is an Lq̂ -solution to (1.1), it is also an Lq -solution. Moreover, u P Cpr0, T q; Lq pRN qq.

Proof. As in the proof of Theorem 4.1, to see that u is an Lq -solution it is enough to check that
Nu P L8 p0, τ ; Lq pRN qq for all τ P p0, T q, and (2.5).
1 p
We first prove the result for q “ ℓ. To this aim we take r such that 1 ´ r “ q̂ ´ 1ℓ . Since q̂ P pℓ, ℓps,
1 p 1 p´1 2β
0 ď1´ “ ´ ă “ ,
r q̂ ℓ ℓ N
or equivalently, r P r1, pc q. Hence, estimate (2.6) holds and σprq ă 1. Then, since u P L8 p0, τ ; Lq̂ pRN qq
for all τ P p0, T q by hypothesis,
żt
}Nu p¨, tq}Lℓ pRN q ď }|up¨, sq|p } q̂ }Y p¨, t ´ sq}Lr pRN q ds
0 L p pRN q
żt
t1´σprq
ď K sup }up¨, sq}pLq̂ pRN q pt ´ sq´σprq ds ď K sup }up¨, sq}pLq̂ pRN q
sPp0,τ q 0 sPp0,τ q 1 ´ σprq
if t P p0, τ q, which implies both Nu P L8 p0, τ ; Lℓ pRN qq, and (2.5). The proof of the continuity in Lℓ of
an Lq̂ -solution is similar to that of Proposition 3.1.
Let now q P r pℓ , ℓq. Since u0 P Lq̂ pRd q X Lq pRd q, then u0 P Lq̂ pRd q X Lℓ pRd q. Hence, as we have just
proved, the Lq̂ -solution is an Lℓ -solution. To prove that it is also an Lq -solution we take r such that
and 1 ´ 1r “ pℓ ´ 1q . In the range of q that we are considering r P r1, pc q, and therefore
t1´σprq
}Nu p¨, tq}Lq pRN q ď K sup }up¨, sq}pLℓ pRN q
sPp0,τ q 1 ´ σprq
if t P p0, τ q, hence the result. The proof of the continuity in Lq of an Lℓ -solution for q in this range
follows easily.
To cover the whole range down to q “ 1 we proceed by induction. Assume that we have already
ℓ
proved the result for q P r pℓj , pj´1 q for some j P N. We will prove that then it is also valid for
 A FULLY NONLOCAL SEMILINEAR PROBLEM 15

ℓ
q P r pj`1 , pℓj q. Notice that, by hypothesis, p ě pc ą 1. Hence, ℓ
pj
Ñ 0 as j Ñ 8, which means that we
will reach q “ 1 in a finite number of steps.
ℓ
Let q P r pj`1 , pℓj q. Then qp P r pℓj , pj´1
ℓ
q. Since u0 P Lq̂ pRd q X Lq pRd q, then u0 P Lq̂ pRd q X Lqp pRd q,
and, by the induction hypothesis, the Lq̂ -solution is an Lqp -solution. Therefore, using (2.6),
żt
}Nu p¨, tq}Lq pRN q ď }|up¨, sq|p }Lq pRN q }Y p¨, t ´ sq}L1 pRN q ds
0
żt
tα
ď K sup }up¨, sq}Lqp pRN q pt ´ sq´1`α ds ď K sup }up¨, sq}pLqp pRN q
p
sPp0,τ q 0 sPp0,τ q α

if t P p0, τ q, hence the result. The proof of the continuity in Lq of an Lqp -solution is similar to that of
Proposition 3.1. 

5. Global existence above the Fujita exponent

If p P rpf , 8q, which implies that ℓ P r1, 8q, we will construct global Lq -solutions, for certain initial
data, for all q P r1, 8s.

5.1. The distinguished range. We start by considering values of q in the distinguished range (DR).

5.1.1. The strategy. If q satisfies (DR), it is in the good range (GR). Hence, given u0 P Lq pRN q, there
is a (unique) maximal Lq -solution with initial datum u0 , with a maximal existence time T P p0, `8s.
Moreover, if T ă `8, then lim suptÕT }up¨, tq}Lq pRN q “ 8. Hence, to prove that an Lq -solution is
global in time, it is enough to prove that }up¨, tq}Lq pRN q is bounded by a constant that does not depend
on t up to time T . This will guaranteed if we are able to obtain such a bound for the function
Nα´1 1¯ α Nα
(5.1) f : p0, T q Ñ r0, 8q, f ptq :“ sup }sb up¨, sq}Lq pRN q , b :“ ´ “ ´ ,
sPp0,tq 2β ℓ q p ´ 1 2βq

since b ą 0 when q ą ℓ. We will show that there is some value η ą 0 such that if
(5.2) tb }Lu0 p¨, tq}Lq pRN q ď η for every t ą 0,
then the function f will never reach the value 2η. The goal will be then to obtain conditions on the
initial data ensuring that (5.2) holds.

5.1.2. The basic lemma. Following the above strategy, we first prove that (5.2) implies the desired
bound for f .
Lemma 5.1. Let p P rpf , 8q and q satisfying (DR). There is a value η ą 0 depending only on p, q, α, β
and N such that if u is a maximal Lq -solution to (1.1) with initial datum u0 P Lq pRN q satisfying (5.2),
and f is as in (5.1), then f ptq ď 2η for all t P p0, T q. As a corollary, u is global in time.

Proof. We borrow ideas from the paper [39], which deals with the local case α, β “ 1. But we have to
face the extra difficulty stemming from the singularity of the kernels Z and Y at x “ 0 that appears
when α P p0, 1q.
16 C. CORTÁZAR, F. QUIRÓS AND N. WOLANSKI

Let 1 ´ 1r “ pq ´ 1q “ 2β ℓ 1 2β
N q . Since p ě pf ą 1 and q ą ℓ, then 1 ´ r P p0, N q, so that r P p1, pc q.
Hence, we may use the bound (2.6) and the hypothesis (5.2) to obtain
żt
α´1´ Nα p1´ 1r q
tb }up¨, tq}Lq pRN q ď tb }Lu0 p¨, tq}Lq pRN q ` Ktb }up¨, sq}pLq pRN q pt ´ sq 2β ds
0
żt ´ s ¯´a
ď η ` Kt b´a
}sb up¨, sq}pLq pRN q s´bp 1 ´ ds,
0 t
with
Nα ℓ
a :“ 1 ´ α ` pp ´ 1q “ 1 ´ α ` α .
2βq q
Notice that a P p0, 1q, since α P p0, 1q and q ą ℓ. After a change of variables we arrive at
ż1
(5.3) f ptq ď η ` Kf ptq p
τ ´bp p1 ´ τ q´a dτ,
0
where we have used that b ´ a ` 1 ´ bp “ 0, an identity that can be easily checked. Remember that
b ą 0, because q ą ℓ. Moreover, since q ď ℓp,
N α ´ p p ¯ N αpp ´ 1q
bp “ ´ ď “ α ă 1.
2β ℓ q 2βℓ
Therefore, the integral on the second term in the right-hand side of (5.3) is bounded, and we have
f ptq ď η ` K0 f ptqp
with K0 a universal constant.
We now notice that f ptq Ñ 0 as t Ñ 0` , since }up¨, tq}Lq pRN q is bounded in r0, τ s for any τ P p0, T q.
On the other hand, f is nondecreasing. Hence, if K0 2p η p´1 ă 1 and tb }Lu0 p¨, tq}Lq pRN q ď η for every
t ą 0, the function f ptq can never reach the value 2η. Indeed, if t1 is such that f pt1 q “ 2η, then
1 ` ˘
f pt1 q ď η ` K0 p2ηqp´1 f pt1 q “ f pt1 q 1 ` K0 2p η p´1 ă f pt1 q
2
a contradiction, and the theorem is proved. 
Remark 5.1. The global Lq -solutions provided by Lemma 5.1 satisfy }up¨, tq}Lq pRN q ď 2ηt´b . Hence
they decay to 0 in Lq pRN q as t Ñ 8.
5.1.3. A strong condition for global existence. Our first condition ensuring (5.2), and hence global
existence, is of the same type as the one obtained in [39] for the local case α “ 1, β “ 1.
Theorem 5.1. Let p P rpf , 8q and q satisfying (DR). There exists a value η0 ą 0 depending only on
p, q, α, β and N such that if u0 P Lℓ pRN q X Lq pRN q and
(5.4) }u0 }Lℓ pRN q ď η0 ,
then the maximal Lq -solution to (1.1) with initial datum u0 is global in time.
2β
Proof. Let r such that 1 ´ 1r “ 1ℓ ´ 1q . The hypotheses on q imply that 0 ă 1 ´ 1r ă N, or equivalently,
r P p1, pc q. Hence, we may use the bound (2.3) to obtain
` ˘
b b b ´ Nα 1´ r1
t }Lu0 p¨, tq}Lq pRN q ď t }u0 }Lℓ pRN q }Zp¨, tq}Lr pRN q ď Kt }u0 }Lℓ pRN q t 2β
` ˘
b ´ Nα 1´ 1r
“ Kt }u0 }Lℓ pRN q t 2β ď Kη0 .
The result now follows from Lemma 5.1, taking η0 “ η{K. 
 A FULLY NONLOCAL SEMILINEAR PROBLEM 17

5.1.4. A weaker (“sharp”) condition for global existence. Our second condition on the initial datum
ensuring (5.2), and hence global existence, is of a different, weaker, nature. We will obtain in Section 6
a condition for blow-up in the same spirit, which shows that condition (5.2) is “sharp”. We may assume
without loss of generality that p ą pf , since when p “ pf both conditions for global existence are in
fact the same.
Theorem 5.2. Let p P ppf , 8q, q satisfying (DR), and u0 P Lℓ pRN q X Lq pRN q. Let η ą 0 as in
Lemma 5.1 and R0 “ pη{}u0 }Lq pRN q q1{b . There is a constant δ ą 0 depending only on η such that if
u0 ě 0 and
ż
α
´ Nα
(5.5) sup R α u0 pyq dy ă δ,
p´1 2β
RąR0 |y|ăR 2β

then the maximal Lq -solution to (1.1) exists globally in time. Moreover, it exists globally in Lℓ pRN q.

Proof. Since }Zp¨, tq}L1 pRN “ 1, then }Lu0 p¨, tq}Lq pRN q ď }u0 }Lq pRN q . As b ą 0 when q ą ℓ, then
tb }Lu0 p¨, tq}Lq pRN q ď R0b }u0 }Lq pRN q “ η for t P p0, R0 s.
It remains then to show that (5.2) holds for t ě R0 .
Let Λ ą 0 to be chosen later. We make the decomposition Lu0 “ L1 ` L2 , where
ż ż
L1 px, tq “ α u0 pyqZpx ´ y, tq dy, L 2 px, tq “ α u0 pyqZpx ´ y, tq dy.
|y|ąΛt 2β |y|ăΛt 2β
1 1
Let r be such that 1 ´ r “ ℓ ´ 1q . The hypotheses on q imply that
1 2β 1 2β
0ď1´ “ ´ ă ,
r N pp ´ 1q q N
or equivalently, r P r1, pc q. Hence, we may use the bound (2.3) to obtain, for all t ě R0 ą 0,
Nα 1 η
}L1 p¨, tq}Lq pRN q ď }u0 } ℓ α }Zp¨, tq} r
L pR N q ď K}u0 } α t´ 2β p1´ r q ď t´b
L p|y|ąΛt q
2β 2β
Lℓ p|y|ąΛR0 q 2
if Λ ě 1 is large enough, since Nα
2β p1 ´ 1r q “ b.
In order to estimate L2 we make the decomposition L2 “ L21 ` L22 , where
ż
L21 px, tq “ α
|y| ă Λt 2β
u0 pyqZpx ´ y, tq dy,
α
|x ´ y| ă λt 2β
ż
L22 px, tq “ α
|y| ă Λt 2β
u0 pyqZpx ´ y, tq dy,
α
|x ´ y| ą λt 2β

for some λ P p0, 1q small to be further specified later. We take r as above and use the bound (2.2) to
obtain
1 ´ Nα p1´ 1r q η
}L21 p¨, tq}Lq pRN q ď }u0 } ℓ α }Zp¨, tq} α ď }u0 }Lℓ pRN q K̃λ2β´N p1´ r q t 2β ď t´b
L p|y|ăΛt 2β q Lr p|z|ăλt 2β q 4
1
if λ is small enough, how small depending on }u0 }Lℓ pRN q , because 2β ´ N p1 ´ r q ą 0.
From now on we fix λ. We have
ż
}L22 p¨, tq}Lq pRN q ď α u0 pyq dy }Zp¨, tq} α .
|y|ăΛt 2β Lq p|z|ěλt 2β q
18 C. CORTÁZAR, F. QUIRÓS AND N. WOLANSKI

The bound (2.2) for the kernel Z yields

` ˘ ` ˘
p ´2β´N 1´ q1 ´ Nα 1´ q1
}Zp¨, tq} α ă Kλ t 2β
Lq p|z|ěλt 2β q

for all t ą 0 for some constant K p depending only on λ, α, β and N . Therefore, for all t ě R0 ,
` ˘ ż
α
b p ´2β´N 1´ q1 p´1 ´ Nα
t }L22 p¨, tq}Lq pRN q ď Kλ t 2β
α u0 pyq dy
|y|ăΛt 2β
` ˘ 2β ż
1 2β α Nα
p ´2β´N 1´ q Λ p´1 ´N pΛ α tq p´1 ´ 2β
“ Kλ α u0 pyq dy
|y|ăΛt 2β
` ˘ 2β ż
α Nα
p ´2β´N 1´ q1 ´N ´
ď Kλ Λ p´1 sup R p´1 2β α u0 pyq dy
2β |y|ăR 2β
RąΛ α R0
ż
α
´ Nα
ď sup R p´1 2β
α u0 pyq dy
RąR0 |y|ăR 2β
` ˘ 2β
p ´2β´N 1´ q1
if we choose Λ ě 1 so large that Kλ Λ p´1 ´N ď 1, as 2β
p´1 ´ N ă 0. Hence,
η
tb }L22 p¨, tq}Lq pRN q ď
4
if (5.5) holds with δ “ η4 .
Global existence in Lq now follows from Lemma 5.1, and then global existence in Lℓ comes as a
corollary, thanks to Theorem 4.2. 

5.2. Global existence outside the distinguished range. We will now use the result for q satis-
fying (DR) to prove the existence of global Lq -solutions, whenever p ě pf , outside the distinguished
range. Let us recall that p ě pf if and only if ℓ ě 1.

5.2.1. Outside the good range. If 1 ď ℓ ă q ă p, then p P rpf , pc q and q P pℓ, ℓps. The existence of a
global Lq -solution is therefore an immediate corollary of Theorem 4.1 together with either Theorem 5.1
or Theorem 5.2.
Corollary 5.1. Let p P rpf , pc q, q P pℓ, pq, and u0 P Lℓ pRN q X Lp pRN q. If either (5.4) or (5.5) hold,
then the global Lp -solution to problem (1.1) with initial datum u0 is a global in time Lq -solution, which
moreover belongs to Cpr0, 8q; Lq pRN qq.

If q P r1, ℓs, the existence of a global Lq -solution, if p P rpf , 8q, is an immediate corollary of
Theorem 4.2 together with either Theorem 5.1 or Theorem 5.2.
Corollary 5.2. Let p P rpf , 8q, q P r1, ℓs, q̂ satisfying (DR), and u0 P Lq pRd q X Lq̂ pRd q. If either (5.4)
or (5.5) hold, then the global Lq̂ -solution to problem (1.1) with initial datum u0 is a global in time Lq -
solution, which moreover belongs to Cpr0, 8q; Lq pRN qq.

5.2.2. Outside the distinguished range, but yet in the good range. We yet have to show the existence of
global solutions for q P pℓp, 8s, with p P rpf , 8q. This will be a corollary of the following result.
Theorem 5.3. Let p P rpf , 8q, q P pℓp, 8s, and u0 P Lℓ pRN q X Lq pRN q. Let u be an Lℓp -solution
to (1.1) with initial datum u0 and existence time T . Then u is also an Lq -solution with initial datum
u0 and existence time T . Moreover, u P Cpr0, T q; Lq pRN qq.
 A FULLY NONLOCAL SEMILINEAR PROBLEM 19

Proof. We start by introducing some auxiliary functions and numbers. For any j P N Y t0u and
p P r1, 8q we define fj ppq “ 1 ´ pj pp ´ 1q. For any fixed j, the function fj is strictly decreasing in
p, and has a unique root, that we denote by ? pj , so that fj ppq ą 0 for p P r1, pj q and fj ppq ă 0 for
p ą pj . For instance, p0 “ 2, p1 “ p1 ` 5q{2. On the other hand, for any fixed p P p1, 8q the
function j Ñ fj ppq is strictly decreasing in j and goes to ´8 as j Ñ 8. Therefore, pj Œ 1 as j Ñ 8,
so that pj ă pf if j is large enough. While pj ą pf , which occurs at least for j “ 0, 1, we define
qj : rpf , pj q Ñ R` by qj ppq “ ℓ{fj ppq. An easy analysis of derivatives shows that pfj ppq P p0, 1q if
p P rpf , pj q, so that ℓp ă qj ppq in that interval.
1 1
Let r0 such that 1 ´ r0 “ ℓ ´ 1q . Since q ą ℓp ě ℓpf ą ℓ, then 1 ´ 1
r0 ą 0. Besides,
1 1 1 2β p´1
´ “1´ ă “
ℓ q r0 N ℓ
1 f0 ppq 1
if and only if q ą ℓ “ q0 ppq . This holds, when q ą ℓp, if and only if

q P pℓp, 8s if p ą p0 , q P pℓp, 8q if p “ p0 , q P pℓp, q0 ppqq if p P rpf , p0 q.

Under these restrictions on q, if 0 ă t ă τ ă T , then, using (2.6),
żt
τ 1´σpr0 q
}Nu p¨, tq}Lq pRN q ď }|up¨, sq|p }Lℓ pRN q }Y p¨, t ´ sq}Lr0 pRN q ds ď K sup }up¨, sq}pLℓp pRN q ,
0 sPp0,τ q 1 ´ σpr0 q
whence the result.
To cover the case p “ p0 “ 2, q “ 8, we take r̃0 such that 1 ´ r̃10 “ 2ℓ 1
“ Nβ P p0, 2β
N q. Using that u
is an L 2ℓp 0 -solution (see the previous step) and also estimate (2.6), for 0 ă t ă τ ă T we have
τ 1´σpr̃0 q
}Nu p¨, tq}L8 pRN q ď K sup }up¨, sq}pL02ℓp0 pRN q .
sPp0,τ q 1 ´ σpr̃0 q

We next consider the case p P rpf , p0 q, q “ q0 ppq. Let ε P p0, q0ℓpppq ´ 1q, so that q0 ppq
1`ε P pℓp, q0 ppqq.
q0 ppq
Since u0 P Lℓ pRN q X Lq0 ppq pRN q, then u0 P L 1`ε pRN q, and by a previous step we already know that
q0 ppq
1 pp1`εq 1
u is an L 1`ε -solution. Let r̄0 such that 1 ´ r̄0 “ q0 ppq ´ q0 ppq . Assuming in addition that
´ p ´ 1 ` q ppq ˘¯
0
ε P 0, ´1 ,
p ℓ
then 1 ´ 1
r̄0 P p0, 2β
N q. Therefore, using once more (2.6), for 0 ă t ă τ ă T we deduce that

τ 1´σpr̄0 q
}Nu p¨, tq}Lq0 ppq pRN q ď K sup }up¨, sq}p q0 ppq .
sPp0,τ q L 1`ε pRN q 1 ´ σpr̄0 q

We now proceed by iteration, until pj ď pf , something that will happen after a finite number of
steps, starting from the already studied case j “ 0. So we assume that pj´1 ą pf , and we have yet to
cover the case p P rpf , pj´1 q, q ą qj´1 ppq.
1 p
Let rj such that 1 ´ rj “ qj´1 ppq ´ 1q . Since q ą qj´1 ppq, then 1 ´ 1
r0 ą 0. Besides,

p 1 1 2β p´1
´ “1´ ă “
qj´1 ppq q rj N ℓ
20 C. CORTÁZAR, F. QUIRÓS AND N. WOLANSKI

1 p p´1 pfj´1 ppq`1´p fj ppq 1

if and only if q ą qj´1 ppq ´ ℓ “ ℓ “ ℓ “ qj ppq . This holds, when q ą ℓp, if and only if
q P pℓp, 8s if p ą pj , q P pℓp, 8q if p “ pj , q P pℓp, qj ppqq if p P rpf , pj q,
and these cases are covered reasoning as above.
To deal with the case p “ pj , q “ 8, we take r̃j such that 1 ´ 1
r̃j “ 1
2ℓ “ β
N P p0, 2β
N q. Using that u
is an L2ℓpj -solution, something that we have just proved, we conclude, reasoning as in the case j “ 0,
that u is an L8 -solution.
To complete the argument we have to consider the case p P rpf , pj q, q “ qj ppq. We take
´ !` q ppq ˘ p ´ 1 ` qj ppq ˘)¯
j
ε P 0, min ´1 , ´1 ,
ℓp p ℓ
1 pp1`εq 1
and r̄j such that 1 ´ r̄j “ qj ppq ´ qj ppq and reason as in the case j “ 0 to conclude that j is an
Lqj ppq -solution.
We omit the proof of continuity, since it is similar to that of Proposition 3.1. 
Corollary 5.3. Let p P rpf , 8q, q P pℓp, 8s, and u0 P Lℓ pRN q X Lq pRN q. If either (5.4) or (5.5) hold,
then the global Lℓp -solution to problem (1.1) with initial datum u0 is a global in time Lq -solution, which
moreover belongs to Cpr0, 8q; Lq pRN qq.

6. Blow-up results

We devote this section to the proof of a blow-up result which shows, in particular, that:
‚ for any p P p1, 8q and q P r1, 8s there are Lq -solutions that blow up in finite time;
‚ all nontrivial nonnegative Lq -solutions of (1.1) blow up if p P p1, pf q, which combined with the
results of the previous section shows that pf is the Fujita exponent for this problem.
‚ condition (5.5) for global existence is “sharp”.
Theorem 6.1. Let p P p1, 8q and q P r1, 8s. There is a constant C “ CpN, α, β, pq such that, if u is
a nontrivial nonnegative Lq -solution to (1.1) with initial datum in Lq pRN q satisfying
ż
α
´ Nα
(6.1) R p´1 2β α u0 pyq dy ą C for some R ą 0,
|y|ăR 2β

then lim sup }up¨, tq}Lq pRN q “ 8 for some T P p0, 8q.
tÕT

Before proceeding to the proof of the theorem, let us give two important corollaries.
Corollary 6.1. Let p P p1, 8q and q P r1, 8s. There are Lq -solutions to (1.1) that blow up in finite
time.

Proof. Let M ą 0 and u0 “ M χB1 p0q . Since u0 P L1 pRN q X L8 pRN q, there are Lq -solutions to (1.1)
with this initial datum for all q P r1, 8s; see theorems 3.1, 4.1, and 4.2. For this u0 , any fixed value
R ą 1, and C as in Theorem 6.1,
ż
α
´ Nα α
´ Nα
α u0 pyq dy “ M ωN R
R p´1 2β p´1 2β ą C

|y|ăR 2β

if M is large enough. Hence, any Lq -solution with this initial datum blows up in finite time. 
 A FULLY NONLOCAL SEMILINEAR PROBLEM 21

Corollary 6.2. Let p P p1, pf q and q P r1, 8s. Let u be a nontrivial nonnegative Lq -solution to (1.1).
There exists T P p0, 8q such that lim sup }up¨, tq}Lq pRN q “ 8.
tÕT

ş
Proof. Let R̄ ą 0 such that α u0 pyq dy “ B ą 0 and C ą 0 as in Theorem 6.1. Since p ă pf ,
|y|ăR̄ 2β
α Nα
then α
p´1 ´ Nα
2β ą 0, and (6.1) holds for any R ą R̄ such that R p´1 ´ 2β B ą C. 

From now on we always assume, without further mention, that u is an nontrivial nonnegative Lq -
solution to (1.1) with an initial datum u0 P Lq pRN q for some q P r1, 8s.
The proof of Theorem 6.1, which is organized in three lemmas, goes as follows. Let G be the profile
of the kernel Y ; see (2.1). For any b ą 0 and T ą 0 we define
żT ż
JpT ; bq “ upx, tqYb px, T ´ tq dxdt, where
0 RN
pα
Yb px, T ´ tq “ χ α pT ´ tq p´1 ´1 Ḡpbq and Ḡpbq :“ min Gpξq.
t|x|ăpT ´tq 2β u t|ξ|ăbu

We first prove that J is finite as long as }up¨, tq}Lq pRN q remains bounded.

Proposition 6.1. Let p ą 1 and b ą 0. If }up¨, tq}Lq pRN q ď C for t P r0, T q, then JpT ; bq is finite.

Proof. Let r be such that 1{r ` 1{q “ 1. Notice that r P r1, 8s. Then
żT
JpT ; bq ď }up¨, tq}q }Yb p¨, T ´ tq}r dt.
0
Nα pα
Since }Yb p¨, T ´ tq}Lr pRN q “ ḠpbqCpN, rqpT ´ tq 2βr pT ´ tq p´1 ´1 , then JpT ; bq ă 8. 

The next step is to obtain an estimate from above for JpT ; bq, whenever this quantity is finite,
independent of the initial data.
α
Proposition 6.2. If b ě 2 2β `1 , there exists C1 “ C1 pN, α, β, p, bq ą 0 such that, if JpT ; bq ă 8, then
Nα
(6.2) JpT ; bq ă C1 T α` 2β .

Proof. Since u is a solution of the integral equation (1.2),

żT ż żtż
JpT ; bq ą up py, sqY px ´ y, t ´ sq dyds Yb px, T ´ tq dxdt
0 RN 0 RN
żTż żTż
p
“ u py, sq Y px ´ y, t ´ sqYb px, T ´ tq dxdt dyds.
0 RN s RN
2β α α
If s ă T , t ą s ` p 2b q α pT ´ sq, |x| ă pT ´ tq 2β and |y| ă pT ´ sq 2β , then
α α
|x ´ y| ă 2pT ´ sq 2β ă bpt ´ sq 2β .
22 C. CORTÁZAR, F. QUIRÓS AND N. WOLANSKI

α
Nα `1
Hence, since α ´ 1 ´ 2β ă 0, if b ě 2 2β we get, using the selfsimilar structure (2.1) of Y ,
żTż żT ż
pα
p ´1
JpT ; bq ą Ḡpbq α u py, sq 2β α Y px ´ y, t ´ sqpT ´ tq
p´1 dxdt dyds
0 |y|ăpT ´sq 2β s`p 2b q α pT ´sq |x|ăpT ´tq 2β
żT ż żT ż
pα
α´1´ Nα ´1
ě Ḡpbq2 α up py, sq 2β α pt ´ sq 2β pT ´ tq p´1 dxdt dyds
0 |y|ăpT ´sq 2β s`p 2b q α pT ´sq |x|ăpT ´tq 2β
żTż żT
pα
2 p α´1´ Nα ´1` Nα
ě CpN qḠpbq α u py, sqpT ´ sq 2β
2β
pT ´ tq p´1 2β dt dyds.
0 |y|ăpT ´sq 2β s`p 2b q α pT ´sq

We conclude that
żTż
pα
JpT ; bq ą CpN, α, β, p, bq up py, sqŶb py, T ´ sqpT ´ sq p´1 dyds, where
0 RN
Ŷb px, T ´ tq “ χ α pT ´ tqα´1 Ḡpbq
t|x|ăpT ´tq 2β u

for some constant CpN, α, β, pq ą 0. Using Jensen’s inequality,

żTż
pα
JpT ; bq ą CpN, α, β, p, bq up py, sqŶb py, T ´ sqpT ´ sq p´1 dyds
0 RN
´ż T ż Ŷb py, T ´ sq α
¯p
ě CpN, α, β, p, bq upy, sq pT ´ sq p´1 dyds }Ŷb }L1 pRN ˆp0,T qq
0 RN }Ŷb }L1 pRN ˆp0,T qq
´ż T ż ¯p 1
“ CpN, α, β, p, bq upy, sqYb py, T ´ sq dyds .
0 RN }Ŷb }p´1
A direct computation shows that
żTż żT
α´1` Nα
}Ŷb }L1 pRN ˆp0,T qq “ Ḡpbq α pT ´ tqα´1 dxdt “ CpN qḠpbq pT ´ tq 2β dt
0 |x|ăpT ´tq 2β 0
α` Nα
“ CpN, α, β, pqḠpbqT 2β .
´pα` Nα qpp´1q
Hence, JpT ; bq ą C̄pN, α, β, p, bqJpT ; bqp T 2β , whence (6.2). 

Finally, we obtain an estimate from below for JpT ; bq that shows that this quantity is big for certain
initial data.
Proposition 6.3. There exists C2 “ C2 pN, α, β, p, bq ą 0 such that
ż
α` Nα α
´ Nα
JpT ; bq ě C2 ApT qT 2β , where ApT q :“ T p´1 2β
α u0 pyq dy.
|y|ăpT {4q 2β

α α α α
Proof. If |x| ă t 2β and |y| ă t 2β , then |x ´ y| ă 2t 2β . Hence, if |x| ă t 2β ,
ż ż ż
´ Nα
upx, tq ą u0 pyqZpx ´ y, tq dy ě α u 0 pyqZpx ´ y, tq dy ě CpN, α, βqt 2β
α u0 pyq dy,
RN |y|ăt 2β |y|ăt 2β
`
´ Nα ´α ˘ ´ Nα α
since Zpx ´ y, tq “ t G px ´ yqt 2β ě t 2β Ḡp2q if |x ´ y| ă 2t 2β . Therefore,
2β

ż T {2 ż ż
pα
´ Nα ´1
JpT ; bq ě CpN, α, βqḠpbq α u0 pyq dy α α t 2β pT ´ tq p´1 dxdt.
T {4 |y|ăt 2β |x|ămintt 2β ,pT ´tq 2β u
 A FULLY NONLOCAL SEMILINEAR PROBLEM 23

α α α α
But mintt 2β , pT ´ tq 2β u “ t 2β ě pT {4q 2β if t P rT {4, T {2s. Thus,
ż ż T {2
pα
´1
JpT ; bq ě C̄pN, α, βqḠpbq α u0 pyq dy pT ´ tq p´1 dt
|y|ăpT {4q 2β T {4
ż
pα
ě C2 pN, α, β, p, bq α u0 pyq dy T
p´1

|y|ăpT {4q 2β
´ α Nα ż ¯
α` Nα Nα
“ C2 pN, α, β, p, bq T 2β T p´1 ´ 2β α u 0 pyq dy “ C2 pN, α, β, p, bqApT qT α` 2β . 
|y|ăpT {4q 2β

We have now all the ingredients to prove Theorem 6.1.

α
Proof of Theorem 6.1. We take b ě 2 2β `1 . If }up¨, tq}Lq pRN q is bounded in r0, T q, then JpT ; bq is well
α` Nα
defined and finite; see Proposition 6.1. Hence, JpT ; bq ă C1 T 2β , with C1 as in Proposition 6.2. If
ApT q ě C1 {C2 , we get a contradiction; see Proposition 6.3. This inequality for ApT q will hold true for
Nα α
T “ 4R ą 0 if (6.1) holds for some R ą 0 and C “ 4 2β ´ p´1 C1 {C2 . 
Remark 6.1. The proof of Theorem 6.1 gives an upper bound for the existence time when (6.1) holds.
Indeed, T ď 4R.

Acknowledgments

This research was supported by the European Union’s Horizon 2020 research and innovation pro-
gramme under the Marie Sklodowska-Curie grant agreement No. 777822, and by grant CEX2019-
000904-S, funded by MCIN/AEI/10.13039/501100011033.
C. Cortázar was also supported by FONDECYT grant 1190102 (Chile).
F. Quirós was also supported by grants PID2020-116949GB-I00 and RED2022-134784-T, both of
them funded by MCIN/AEI/10.13039/501100011033, and by the Madrid Government (Comunidad
de Madrid – Spain) under the multiannual Agreement with UAM in the line for the Excellence of
the University Research Staff in the context of the V PRICIT (Regional Programme of Research and
Technological Innovation).
N. Wolanski was also supported by ANPCyT PICT2016-1022.

Data availability statement

The manuscript has no associated data.

References
[1] Aronson, D. G.; Weinberger, H. F. Multidimensional nonlinear diffusion arising in population genetics. Adv. Math.
30 (1978), 33–76.
[2] Baras, P. Non-unicité des solutions d’une equation d’évolution non-linéaire. (French) [Nonuniqueness of the solutions
of a nonlinear evolution equation]. Ann. Fac. Sci. Toulouse Math. (5) 5 (1983), no. 3–4, 287–302.
[3] Birkner, M.; López-Mimbela, J. A.; Wakolbinger, A. Blow-up of semilinear PDE’s at the critical dimension. A prob-
abilistic approach. Proc. Amer. Math. Soc. 130 (2002), no. 8, 2431–2442.
[4] Brezis, H.; Cazenave, Th. A nonlinear heat equation with singular initial data. J. Anal. Math. 68 (1996), 277–304.
[5] Caputo, M. Linear models of dissipation whose Q is almost frequency independent–II. Geophys. J. R. Astr. Soc. 13
(1967), 529–539.
24 C. CORTÁZAR, F. QUIRÓS AND N. WOLANSKI

[6] Deng, K.; Levine, H. A. The role of critical exponents in blow-up theorems: the sequel. J. Math. Anal. Appl. 243
(2000), 85–126.
[7] Dipierro, S.; Pellacci, B.; Valdinoci, E.; Verzini, G. Time-fractional equations with reaction terms: fundamental
solutions and asymptotics. Discrete Contin. Dyn. Syst. 41 (2021), no. 1, 257–275.
[8] Dzherbashyan, M. M.; Nersesian, A. B. Fractional derivatives and the Cauchy problem for differential equations of
fractional order. (Russian) Izv. Akad. Nauk Arm. SSR, Mat. 3 (1968), 3–29.
[9] Eidelman, S. D.; Kochubei, A. N. Cauchy problem for fractional diffusion equations. J. Differential Equations 199
(2004), no. 2, 211–255.
[10] Ferreira, R.; de Pablo, A. Blow-up for a fully fractional heat equation. Discrete Contin. Dyn. Syst. 44 (2024), no. 2,
569–584.
[11] Fujita, H. On the blowing up of solutions of the Cauchy problem for ut “ ∆u ` uα`1 . J. Fac. Sci. Univ. Tokyo Sect. I
13 (1966), 109–124.
[12] Galaktionov, V. A.; Levine, H. A. A general approach to critical Fujita exponents in nonlinear parabolic problems.
Nonlinear Anal. 34 (1998), 1005–1027.
[13] García-Melián, J.; Quirós, F. Fujita exponents for evolution problems with nonlocal diffusion. J. Evol. Equ. 10 (2010),
no. 1, 147–161.
[14] Gerasimov, A. N. A generalization of linear laws of deformation and its application to problems of internal friction.
(Russian) Akad. Nauk SSSR. Prikl. Mat. Meh. 12 (1948), 251–260.
[15] Giga, Y. Solutions for semilinear parabolic equations in Lp and regularity of weak solutions of the Navier-Stokes
system. J. Differential Equations 62 (1986), no. 2, 186–212.
[16] Gross, B. On creep and relaxation. J. Appl. Phys. 18 (1947), 212–221.
[17] Haraux, A.; Weissler, F. B. Nonuniqueness for a semilinear initial value problem. Indiana Univ. Math. J. 31 (1982),
no. 2, 167–189.
[18] Hayakawa, K. On nonexistence of global solutions of some semilinear parabolic differential equations. Proc. Japan
Acad. 49 (1973), 503–505.
[19] Kirane, M.; Laskri, Y.; Tatar, N.-e. Critical exponents of Fujita type for certain evolution equations and systems
with spatio-temporal fractional derivatives. J. Math. Anal. Appl. 312 (2005), no. 2, 488–501.
[20] Kemppainen, J.; Siljander, J.; Zacher, R. Representation of solutions and large-time behavior for fully nonlocal
diffusion equations. J. Differential Equations 263 (2017), no. 1, 149–201.
[21] Kim, K.-H.; Lim, S. Asymptotic behaviors of fundamental solution and its derivatives to fractional diffusion-wave
equations. J. Korean Math. Soc. 53 (2016), no. 4, 929–967.
[22] Kobayashi, K.; Sirao, T.; Tanaka, H. On the growing up problem for semilinear heat equations. J. Math. Soc. Japan
29 (1977), no. 3, 407–424.
[23] Kochubei, A. N. Diffusion of fractional order. (Russian) Differentsial’nye Uravneniya 26 (1990), no. 4, 660–670,
733–734; translation in Differential Equations 26 (1990), no. 4, 485–492.
[24] Laister, R.; Robinson, J. C.; Sierżęga, M.; Vidal-López, A. A complete characterisation of local existence for semi-
linear heat equations in Lebesgue spaces. Ann. Inst. H. Poincaré C Anal. Non Linéaire 33 (2016), no. 6, 1519–1538.
[25] Levine, H. A. The role of critical exponents in blow-up theorems. SIAM Rev. 32 (1990), 262–288.
[26] Li, Ch.; Li, Zh. The blow-up and global existence of solution to Caputo-Hadamard fractional partial differential
equation with fractional Laplacian. J. Nonlinear Sci. 31 (2021), no. 5, Paper No. 80, 35 pp.
[27] Liouville, J. Memoire sur quelques questions de géometrie et de méecanique, et sur un nouveau gentre pour resoudre
ces questions. J. Ecole Polytech. 13 (1832), 1–69.
[28] Matos, J.; Terraneo, E. Nonuniqueness for a critical nonlinear heat equation with any initial data. Nonlinear Anal.
55 (2003), no. 7–8, 927–936.
[29] Metzler, R.; Klafter, J. The random walk’s guide to anomalous diffusion: a fractional dynamics approach. Phys.
Rep. 339 (2000), no. 1, 77 pp.
[30] Nagasawa, M.; Sirao, T. Probabilistic treatment of the blowing up of solutions for a nonlinear integral equation.
Trans. Amer. Math. Soc. 139 (1969), 301–310.
[31] Ni, W.-M.; Sacks, P. Singular behavior in nonlinear parabolic equations. Trans. Amer. Math. Soc. 287 (1985), no. 2,
657–671.
[32] Quittner, P.; Souplet, Ph. “Superlinear parabolic problems. Blow-up, global existence and steady states.”
Second edition. Birkhäuser Advanced Texts: Basler Lehrbücher. [Birkhäuser Advanced Texts: Basel Textbooks]
Birkhäuser/Springer, Cham, 2019. ISBN: 978-3-030-18220-5; 978-3-030-18222-9.
[33] Rabotnov, Yu. N. “Polzuchest Elementov Konstruktsii”. (Russian) Nauka, Moscow (1966); English translation:
“Creep Problems in Structural Members”. North-Holland, Amsterdam (1969).
 A FULLY NONLOCAL SEMILINEAR PROBLEM 25

[34] Samarskii, A. A.; Galaktionov, V. A.; Kurdyumov, S. P.; Mikhailov, A. P. “Blow-up in quasilinear parabolic equa-
tions”. Translated from the 1987 Russian original. De Gruyter Expositions in Mathematics, 19. Walter de Gruyter &
Co., Berlin, 1995. ISBN: 3-11-012754-7.
[35] Sugitani, S. On nonexistence of global solutions for some nonlinear integral equations. Osaka Math. J. 12 (1975),
45–51.
[36] Terraneo, E. Non-uniqueness for a critical non-linear heat equation. Comm. Partial Differential Equations 27 (2002),
no. 1–2, 185–218.
[37] Weissler, F. B. Semilinear evolution equations in Banach spaces. J. Functional Analysis 32 (1979), no. 3, 277–296.
[38] Weissler, F. B. Local existence and nonexistence for semilinear parabolic equations in Lp . Indiana Univ. Math. J. 29
(1980), no. 1, 79–102.
[39] Weissler, F. B. Existence and nonexistence of global solutions for a semilinear heat equation. Israel J. Math. 38
(1981), no. 1–2, 29–40.
[40] Zhang, Q.-G.; Sun, H.-R. The blow-up and global existence of solutions of Cauchy problems for a time fractional
diffusion equation. Topol. Methods Nonlinear Anal. 46 (2015), no. 1, 69–92.

Carmen Cortázar
Departamento de Matemática, Pontificia Universidad Católica de Chile
Santiago, Chile.
Email address: [email protected]

Fernando Quirós
Departamento de Matemáticas, Universidad Autónoma de Madrid,
28049-Madrid, Spain,
and Instituto de Ciencias Matemáticas ICMAT (CSIC-UAM-UCM-UC3M),
28049-Madrid, Spain.
Email address: [email protected]

Noemí Wolanski
IMAS-UBA-CONICET,
Ciudad Universitaria, Pab. I,
(1428) Buenos Aires, Argentina.
Email address: [email protected]