whereΦ(s) =es−s−1, keeping in mind thatR
Ωv(t)dx= 0. Here and in what follows,LΦ(Ω) denotes the Orlicz space which corresponds to the Young functionΦ(s), andk · kLΦ(Ω)as its norm. Let's denote the Young function complementary to Φ by Ψ, so thatLΨ(Ω) denotes the Orlicz space, withk·kLΨ(Ω)is its norm. It is also known thatΨ(s) = (s+1) log(s+1)−s.
Using Hölder's inequality for Orlicz spaces [2, 78], we see that F(u(t), v(t))≥ −|Ω|
e − Z
Ω
(u(t)−1)v(t)dx+ d2
2γk∇v(t)k2L2(Ω)
≥ −|Ω|
e − kv(t)kLΦ(Ω)ku(t)−1kLΨ(Ω)+ d2
2γk∇v(t)k2L2(Ω)
≥ −|Ω|
e − K˜
4εku(t)−1kLΨ(Ω)+ d2
2γ −ε
k∇v(t)k2L2(Ω), whereε < 2γd2. Combining this with Lemma 6.3 of [26] and the fact that
Z
Ω
u(t)dx=|Ω| ∀t≥0, we get the result.
Remark 5.1. We remark that, except statement 6., all the statements of Theorem 5.14 and Lemma 5.15 are also true for blow-up solutions of (5.3) when Ω ⊂R2 has a piecewise C2 boundary.
Then there exists a time t∗ after whichF blows up, and the inequality F F00−(ν+ 1)(F0)2 ≥0
holds.
Proof. Firstly, we observe that (5.50) means that F−ν is concave with an initial positive derivative, and thus, the inequality
F−ν(t)≤F−ν(0) + (F−ν)0(0)t ⇔F−ν(t)≤F−ν(0)−νF−ν−1(0)F0(0)t, is true. From this, we get that
t≤t∗= F−ν(0)
νF−ν−1(0)F0(0) = F(0)
νF0(0). (5.51)
So,t∗ is an upper bound for the blow-up time of the functionF. Secondly, we note that
(F−ν)0 = −νF−ν−1F0
(F−ν)00 = −ν(−ν−1)F−ν−2(F0)2−νF−ν−1F00
= −νF−ν−2[F F00−(ν+ 1)(F0)2].
From the conditions in (5.50), we obtain the inequality F F00−(ν+ 1)(F0)2≥0, and the proof of the Lemma is complete.
We therefore have the following theorem regarding the system (3.1).
Theorem 5.17. Consider the problem (3.1). Assume thatv∈L∞(Ω), v0∈W1,∞(Ω), u0 ∈ L2(Ω), and set
F(t) =F = Z t
0
Z
Ω
u2(x, s)dx ds+ω, (5.52) for some constant ω >0. Then Lemma 5.16 holds true.
Proof. We start by dierentiatingF with respect tot to get F0 =
Z
Ω
u2 dx= Z t
0
d dt
Z
Ω
u2 dx
ds+ Z
Ω
u20 dx
= 2 Z t
0
Z
Ω
uutdx ds+ Z
Ω
u20 dx
= 2 Z t
0
Z
Ω
uutdx ds+C0,
(5.53)
whereC0=R
Ωu20 dx. Dierentiating in (5.53) with respect to tgives F00= 2
Z
Ω
uutdx
= 2 Z t
0
d dt
Z
Ω
uutdx
ds+ 2 Z
Ω
uutdx 0
= 2 Z t
0
d dt
Z
Ω
uutdx
ds+C1,
(5.54)
where
C1= 2 Z
Ω
uutdx 0
= 2 Z
Ω
u(d1∆u− ∇ ·(uχ∇v))dx 0
= 2
χ Z
Ω
u0∇u0∇v0 dx−d1 Z
Ω
|∇u0|2dx
= 2
χ Z
Ω
u0∇u0∇v0 dx−d1
Z
Ω
|∇u0|2dx
.
(5.55)
Note that (5.54) can be rewritten as F00=C
Z t 0
Z
Ω
u2t dx ds+
2 Z t
0
d dt
Z
Ω
uutdx
ds−C Z t
0
Z
Ω
u2t dx ds
+C1. Squaring both sides of (5.53) and using Young's inequality, we write that
(F0)2 =
2 Z t
0
Z
Ω
uutdx ds+C0
2
= 4 Z t
0
Z
Ω
uutdx ds 2
+ 4C0 Z t
0
Z
Ω
uutdx ds+C02
≤ (4 +ε) Z t
0
Z
Ω
uutdx ds 2
+CεC02,
for some constantCε. It then follows, by applying the Cauchy-Schwarz inequality, that (F0)2 ≤ (4 +ε)
Z t
0
Z
Ω
u2t dx ds· Z t
0
Z
Ω
u2dx ds+CεC02
≤ (4 +ε)F Z t
0
Z
Ω
u2t dx ds+CεC02.
Thence, F F00− C
4 +ε(F0)2 ≥F
2 Z t
0
d dt
Z
Ω
uutdx
ds−C Z t
0
Z
Ω
u2t dx ds+C1
− C
4 +εCεC02. The expression between the square brackets above can be rewritten as follows;
2 Z t
0
d dt
Z
Ω
uutdx
ds−C Z t
0
Z
Ω
ututdx ds
= 2 Z t
0
d dt
Z
Ω
u(d1∆u− ∇ ·(uχ∇v))dx
ds−C Z t
0
Z
Ω
ut(d1∆u− ∇ ·(uχ∇v))dx ds
= 2d1 Z t
0
(−k∇uk22)0ds+ 2 Z t
0
d dt
χ
Z
Ω
u∇u∇v dx
ds
−Cχ Z t
0
Z
Ω
u∇ut∇v dx ds+Cd1
2 Z t
0
(k∇uk22)0ds
= 2d1 Z t
0
(−k∇uk22)0ds+ 2χ Z t
0
Z
Ω
(ut∇u∇v+u∇ut∇v+u∇u∇vt)dx ds
−Cχ Z t
0
Z
Ω
u∇ut∇v dx ds+Cd1
2 Z t
0
(k∇uk22)0ds.
If we choose C= 2, then we get 2
Z t 0
d dt
Z
Ω
uutdx
ds−C Z t
0
Z
Ω
u2t dx ds
=d1
h
k∇u0k2L2(Ω)− k∇uk2L2(Ω)
i + 2χ
Z t 0
Z
Ω
(ut∇u∇v+u∇u∇vt)dx ds.
SinceF ≥ω, we get F F00− 2
4 +ε(F0)2 ≥F
d1
h
k∇u0k2L2(Ω)− k∇uk2L2(Ω)
i + 2χ
Z t 0
Z
Ω
(ut∇u∇v+u∇u∇vt)dx ds +2k∇v0kL∞(Ω)ku0kL2(Ω)k∇u0kL2(Ω)−2d1k∇u0k2L2(Ω)−2Cεku0k4L2(Ω)
(4 +ε)ω )
≥F d1n
k∇u0k2L2(Ω)− k∇uk2L2(Ω)
o+ 2k∇v0kL∞(Ω)ku0kL2(Ω)k∇u0kL2(Ω)
−2d1k∇u0k2L2(Ω)−2Cεku0k4L2(Ω) (4 +ε)ω
≥0,
and the proof of the Theorem is complete.
Attraction-Repulsion KS Equations in Scale of Banach Spaces
6.1 Introduction
In this chapter, we study the well-posedness and asymptotic global dynamics of the attraction- repulsion Keller-Segel system of equations admitting the following abstract formulation:
Ut+ApU =P(u)U,
U(0) =U0 ∈Eαq ×Erβ ×Erβ,0≤α−β <1, q, r≥1,
(6.1)
modelling aggregation of microglia in Alzheimer's disease, whereU = (u, v, w)>have entries holding meanings as in (4.2). For notational convenience, we will set v =ψ2, w= ψ3. In (6.1), we have that
M3×3(Lp(Ω,R3))3 Ap = diag[−∆,−∆ +λ2,−∆ +λ3]
:D(Ap)⊂Lp(Ω,R3)→Lp(Ω,R3) (6.2) with domain
D(Ap) :=
ϕ∈H2,p(Ω,R3) :∂~nϕ= 0 on Γ , (6.3) 144
considered for real valued vector functions dened on an open bounded subset Ω ⊂ RN possessing a smooth boundaryΓ =∂Ω,~n denotes the unit outward pointing normal vector to Γ,
P(u)U =
−
3
X
j=2
Div(u(−1)jχj∇ψj), a2u, a3u
>
, (6.4)
and the biophysical constants are as dened in (4.5), withdi = 1, i= 1,2,3.
More precisely, the evolutionary equation (6.1) reads the following chemotaxis system of equations
ut = ∆u−P3
j=2Div(u(−1)jχj∇ψj), vt = ∆v−λ2v2+a2u,
wt = ∆w−λ3w+a3u, inΩ×I,˙ 0 = ∂~nu=∂~nψj on Γ×I˙,
u(0) = u0, ψj(0) =ψ0 inΩ,
(6.5)
whereI˙= (0, T),I = [0, T), and in simplication we have written
−
3
X
j=2
Div(u(−1)jχj∇ψj) =−∇ ·(u(χ2∇v−χ3∇w)) =:P(u)ψ. (6.6) Recall that (6.6) can be viewed in the sense of distributions as the weak form
PΩ(u)ψ:=hP(u)ψ, ϕip0,p=
3
X
j=2
Z
Ω
u(−1)jχj∇ψj∇ϕ (6.7) in adequate function spaces.
It is clear to see that the system of equations (6.5) has L1−spatial integrable solutions in taking the Lp− dual product with as test function ϕ= (1,1,1)> in distributions sense.
More concretely, d dt
Z
Ω
u= 0⇒uΩ(t) = Z
Ω
u0(x), ∀t∈I,˙ d
dt Z
Ω
ψ=−λ Z
Ω
ψ+a Z
Ω
u⇒ψΩ(t) =e−λtψ0+a|Ω|¯u0
λ (1−e−λt), ∀t∈I,˙
whereuΩ =R
Ωu=|Ω|u.Thus, if T =∞ we obtain M=
(
A∈R3:A=
|Ω|¯u0,a2|Ω|u¯0
λ2 ,a3|Ω|u¯0 λ3
>)
, (6.8)
as the time limit set of L1−spatially integrable solutions.
On the other hand, the stationary equations to (6.5), using La-Salle- Hale-Henry [30]
invariance principle, can be deduced associated to the system of equations (6.5) in following the works of [36] by means of the Lyapunov function
J(u, ψ) = Z
Ω
ulnu−κ Z
Ω
uψ+ κ ar
Z
Ω
(|∇ψ|r+λ|ψ|r), (6.9) where ψ = (ψ2, ψ3), λ = (λ2, λ3), a = (a2, a3), κ = sgnP3
j=2(−1)jχj > 0 to which holding onto, if the system of equations is globally well-posedness in time, then it implies studying of the non-local elliptic problem
∆ψ−λψ+µeκψ= 0 inΩ,
∂ψ
∂~n = 0 onΓ =∂Ω, (6.10)
where
µ=a
R
Ωu R
ΩeP3j=2(−1)jχjψ
=a R
Ωu0 R
ΩeP3j=2(−1)jχjψ ,
using the implied conclusion in (6.8). Henceforth, in alternatives we can distinguish the following possible situations:
eP3j=2(−1)jχj
1 if χ2 χ3,
≥ 1 ifχ2≥χ3, 1 ifχ2χ3
(6.11)
corresponding respectively to that:
Repulsion coecient dominating strongly the attraction coecient.
Attraction coecient dominating mildly the repulsion coecient. (6.12) Attraction coecient dominating signicantly the repulsion coecient.
We point-out here that, as technical basis for our analysis, we use abstract dynamical systems theory for evolutionary equations [30, 68, 66], where the approach dictates that, to solve the equations (6.5) and to understand their qualitative properties one has to seeks for solutions satisfying the integral equations
F(u, u0)(t) :=e∆tu0+ Z t
0
e∆(t−s)P(u)ψ(s)ds,
ψ(t, ψ0) :=e(∆−λ)tψ0+a Z t
0
e(∆−λ)(t−s)u(s)ds, ψ=ψj, j = 2,3,
(6.13)
and vice-versa. Note that if (6.13) is to be properly dened, then the non homogeneous terms of the equations (6.5) need to be such that they are mapped into the spaces of the initial data.
This chapter is organized as follows. In Section 6.2, we give some preliminaries on the function spaces, and Eqα −Epβ heat kernel estimates of the semigroup associated to the operator (6.2), which might not have been covered in Chapter 1.
Section 6.3, is devoted to the well-posedness of the system of equations (6.5) inLσ( ˙I;Lp(Ω)) takingα, β = 0, i.e. u0 ∈Lq(Ω), v0, w0 ∈Lr(Ω). In order, to gain control over the coupled
term in (6.6) of the cell density equation in (6.5) we introduce the Banach space Zq0 :=
n
∇z∈Lp(Ω);−Div(a(x)∇z)∈Lq0(Ω), a(x)∈LΘ(Ω)⊃H1,q(Ω) xed, 1
q = 1 Θ+1
p
⊂H−1,q0(Ω),
(6.14)
endowed with the norm
ka(x)∇zkZ
q0 =ka(x)∇zkq+kDiv(a(x)∇z)kq0 ∼=kDiv(a(x)∇z)kq0 <∞.
Then we prove the following lemma.
Lemma 6.1. Assume in (6.6) that uχ∇ψ∈ Zq0, q0 ≥ N−22N . Then, PΩ(u)ψ∈ H−1,q0(Ω) is well dened, and
kPΩkL(H1,q(Ω),H−1,q0(Ω)) = sup
k∇ϕkq0≤1
|hP(u)ψ, ϕiq,q0|
kuχ∇ψkq ≤2(N eπ)−1<1. (6.15) Moreover,
p≥q ≥ p
2 ⇐⇒ N ≥q≥ N
2, (6.16)
is a valid Sobolev spaces embedding relation, with q0 ≥p ifN <4 and p > q0 ≥q if N ≥4. Important to take note of is that, (6.16) imply studying the cells density equation up to the critical space H1,N(Ω), and the reduction of the system of equations in the large time asymptotic dynamics to the non-local elliptic problem (6.10), to which the Moser-Trudinger inequality imply well-posedness only if
κ≤N ω
1 N−1
N−1, where ωN−1 = 2πN2
Γ(N/2), (6.17)
denote the measure of the unit sphere in RN, N ≥ 2. In the context of Lemma 6.1, we
obtain that the system of equations (6.5) admits a unique solution of at least class
Xq,rp (I) := C(I;Lq(Ω))∩ L∞N
2
1 q−1
p
(I;Lp(Ω))∩Lσ(I;Lp(Ω))×
×[C(I;Lr(Ω))∩C(I;H1,p(Ω))∩C1(I;Lp(Ω))]2
= V ×Z×Z. (6.18)
More precisely, we have the following theorem.
Theorem 6.2. Consider the system of equations (6.5) with u0 ∈ Lq(Ω),ψ0 ∈ Lr(Ω), and assume that Lemma 6.1 holds. Let u∈Lσ( ˙I;LΘ(Ω)), forr, p≥Θ, be such that
1 σ + N
2Θ ≤min
1 +N 2r,1
2+ N 2p
. (6.19)
Then,
(i) ψ∈C(I;Lr(Ω))∩C( ˙I;H1,p(Ω)). (ii) If (i) holds, then F(u, u0)∈ L∞N
2Θ
( ˙I;Lp(Ω)) satises that the mapping Lq(Ω)×Lσ( ˙I;LΘ(Ω))3(u0, u)→ F(u, u0)∈ L∞N
2Θ
( ˙I;Lp(Ω)) (6.20) is linear and continuous. Furthermore, F(u, u0) is locally Hölder continuous with values inLp(Ω).
(iii) The system of equations (6.5) admits a unique solution in the class (6.18). That is U ∈Xq,rp (I).
A priori uniform boundedness inΩ×I˙of the cells density solution is proven in Subsection 6.3.1, yielding, as a result, that the complete system solution is a global classical solution.
Independent to this conclusion, we obtain the following proposition.
Proposition 6.3. Consider the system of equations (6.5) in the context of Theorem 6.2.
Then,
lim sup
t%+∞
ku(t)kp = 0, lim sup
t%+∞
k∇ψ(t)kp= 0, (6.21) and the system of equations dene an extended or perturbed analytic semigroup in Lp− spaces. Moreover, in the global asymptotic dynamics ,it holds that
lim sup
t%∞
k(u(t), v(t), w(t))>kp =A∗ ∈ M ∪ {0}, (6.22) where the limit setM, as dened in (6.8), corresponds to theL1−spatial integrable solutions of the system equations in distributions sense.
In Section 6.4, we prove similar results to those of Section 6.3, but in a much more general function space setting, which includes one used in Chapter 4 of the scale of Hilbert spaces. More precisely, we give a treatment of the equations in Bessel potential spaces Eqα, α∈R,1< q <∞. To this end, we rst establish the following nesting relation between the spaces;
Epα7−→Eqα 7−→Epβ 7−→Eqβ 7−→Eβr, (6.23) and
Eqα0 ←−−→ Epα 7−→Eqα7−→Epβ0 7−→Eqβ0
←−−→ Epβ. (6.24)
Then, prove the following counterpart to Lemma 6.1.
Lemma 6.4. Assume (6.23)-(6.24) hold, and let u ∈ Eqα, ψ ∈ Epβ, β ≥ 12, 0 ≤ α−β <
1, p≥q. Then, if
1 2 +N
2p ≤α+β, and 1 + N
2p ≤2α+β, (6.25)
hold, then the product uχ∇ψ ∈ Eqα, and the weak form PΩ(u) ∈ Eq00 ⊆ Ep−β0 , for q0 ≥
2N
N−4α, p≥ 2αN, are well dened, and kPΩkL(Eα
q,E−β
q0 ):= sup
kϕkα,q0≤1
|hP(u)ψ, ϕi
Eqα,E−βq0 | χku∇ψkα,q ≤
2 N eπ
α+β2−12
. (6.26)
In particular,PΩ∈ Llip(Eqα, Eqβ0) is satised.
To conclude, we give similar results to those in Theorem 6.2-Proposition 6.3 in the following theorem.
Theorem 6.5. Assume in the system of equations (6.1)-(6.4) that Lemma 6.4 holds. Then, (i) The system of equations admits a uniqueC1−strong solution. Furthermore, there exists
a constant
ω = min
1−
2 N eπ
α+β−12 3
X
j=2
χj+aj q
,
1− 2
N eπ
α+β2−12 3
X
j=2
χj+aj r
>0 (6.27) such that the coupled system dierential operator in (6.5) denes a perturbed analytic semigroup in Zα+β =Eαq ×Erβ ×Erβ spaces, and
lim sup
t%∞
k(u(t), v(t), w(t))>kα+β =A∗∈ M ∪ {0} (6.28) where the limit set M is dened as in (6.8) corresponds to L1− spatial integrable solutions of the system (6.5) of equations in distributions sense.
(ii) Assume that the rst condition in (6.25) is veried strictly. Then, the solution semi- group is a classical solution.
To make the proof of the results in above theorems accessible to the reader we give, in the next section, some preliminaries.