with W
1
· is a complex Gaussian white noise on R
n
. 2 If m
1
κ
1
m
2
κ
2
and d
1
d
2
, then the finite-dimensional distributions of the rescaled random field [ǫ
m2κ2 α
L
m
2
ǫ
−
1 α
]
−
1 2
e
−d
1 t
ǫ
n
w
t ǫ
, x
ǫ
1 α
; w
· − Q t
ǫ ; d
1
, d
2
C
1
C
2
o , t
0, x ∈ R
n
, converge weakly, as
ǫ → 0, to the finite-dimensional distributions of the random field
T
2 m
2
t, x :=
−p
11
p
12
X
2 m
2
t, x −p
21
p
12
X
2 m
2
t, x
, t 0, x ∈ R
n
, where
X
2 m
2
t, x := C
2 m
2
p m
2
Kn, κ
2
m2 2
Z
′
R
n ×m2
e
i x,z
1
+...+z
m2
−µt|z
1
+...+z
m2
|
α
|z
1
| · · · |z
m
2
|
n −κ2
2
m
2
Y
l=1
W
2
dz
l
, 28
and W
2
· is a complex Gaussian white noise on R
n
. 3 If m
1
= m
2
:= m, κ
1
= κ
2
:= κ, and d
1
d
2
, then the finite-dimensional distributions of the rescaled random field
[ǫ
m κ
α
L
m
ǫ
−
1 α
]
−
1 2
e
−d
1 t
ǫ
n
w
t ǫ
, x
ǫ
1 α
; w
· − Q t
ǫ ; d
1
, d
2
C
1
C
2
o , t
0, x ∈ R
n
, converge weakly, as
ǫ → 0, to the finite-dimensional distributions of the random field
T
3 m
t, x := T
1 m
t, x + T
2 m
t, x, 29
where T
1 m
t, x and T
2 m
t, x, are defined in the case 1 and the case 2 with m
1
= m
2
= m and κ
1
= κ
2
= κ.
4.2 The subdiffusive case
We extend the above results to the subdiffusive case, meant for which the time-fractional derivative is
∂
β
∂ t
β
, β ∈ 0, 1, see Section 1 in the system 2, that is,
∂
β
∂ t
β
u
v
= −µI − ∆
γ 2
−∆
α 2
u
v
+ B
u v
,
µ, α, γ 0, 30
The time-fractional derivative
∂
β
∂ t
β
, for any β 0, can be seen, for example, the book of Djrbashian
[8]. For 0 β 1, it is d
β
f d t
β
t = 1
Γ1 − β Z
t
f
′
τ t − τ
β
d τ,
31 where f t is causal i.e., f vanishes for t
0. 971
The solution of 30 can be obtained via the fractional both in time and in space procedure see, for example, [20, 21]; under the Condition DM, we express the solution as the convolution of the
fractional both in time and in space Green function and the initial data, as follows.
wt, x; w · =
Z
R
n
P
G
β
t, x − y; d
1
G
β
t, x − y; d
2
P
−1
u
y v
y
d y 32
with the fractional Green function G
β
t, x; d
j
is defined via the transformation E
β
−µ|λ|
α
1 + |λ|
2
γ 2
t
β
+ d
j
t
β
= Z
R
n
e
i x,λ
G
β
t, x; d
j
d x, j ∈ {1, 2}, where E
β
· is the Mittag-Leffler function defined by see, for example, [2] or [8, Chapter 1] E
β
z =
∞
X
p=0
z
p
Γβ p + 1 , z
∈ C. 33
For the properties of the Mittag-Leffler function, we refer the classic book by Erdélyi et.al. [10] pp. 206-212, in particular p. 206 7 and p. 210 21.
The following results are time-fractional versions of those in the above; however, the sub-diffusivity brings some new feature. For the large-scalings of homogenization of the system, we need to take
an additional scaling on the matrix B in the system 30 in order to compromise the effect of this sub-diffusivity upon the interaction between u and v; while the sub-diffusivity has no influence on
the small-scalings and thus it is the same as that in Section 3. To make the situation clear, in the following we denote the vector solution by
wt, x; w ·, B. We only state some partial assertions
in the below.
Proposition 4. Let {wt, x; w
·, B, t 0, x ∈ R
n
} be the solution-vector of the initial value prob- lem 30 and 3, satisfying Conditions DM, SGRID, and LD. Denote again the long-range-dependence
indices κ
j
and the Hermite ranks m
j
, j ∈ {1, 2}. Assume that m
1
κ
1
m
2
κ
2
. Large-scalings: under m
1
κ
1
, m
2
κ
2
min{2α, n} the finite-dimensional distributions of the rescaled random field
T
1 ǫ
t, x := ǫ
m1κ1 α
L
m
1
ǫ
−
β α
−
1 2
n
w ǫ
−1
t, ǫ
−
β α
x; w
·, ǫ
β
B − Cǫ
−1
t; ǫ
β
B o
, t
0, x ∈ R
n
, converge weakly, as ǫ → 0, to the finite-dimensional distributions of the random field
T
1
t, x =
p
11
p
22
T
1
t, x; d
1
− p
12
p
21
T
1
t, x; d
2
p
21
p
22
T
1
t, x; d
1
− p
21
p
22
T
1
t, x; d
2
, t
0, x ∈ R
n
, where
C ǫ
−1
t; ǫ
β
B = Ct; B = P
E
β
d
1
t
β
E
β
d
2
t
β
P
−1
C
1
C
2
,
and for j ∈ {1, 2} T
1
t, x; d
j
is expressed by the following multiple-Wiener integral: C
1 m
1
p m
1
Kn, κ
1
m1 2
Z
′
R
n ×m1
e
i x,λ
1
+···+λ
m1
E
β
−µ|λ
1
+ · · · + λ
m
1
|
α
t
β
+ d
j
t
β
|λ
1
| · · · |λ
m
1
|
n −κ1
2
m
1
Y
l=1
W
1
dλ
l
. 34
972
Small-scalings: under m
1
κ
1
, m
2
κ
2
min{2α + γ, n} the finite-dimensional distributions of the rescaled random field
M
1 ǫ
:= [ ǫ
m
1
κ
1
χ
L
m
1
ǫ
−χ
]
−
1 2
n
w ǫt, ǫ
β α+γ
x; w
ǫ
−
β α+γ
−χ
·, B −
C
1
C
2
o , t
0, x ∈ R
n
, converge weakly, as
ǫ → 0, to the finite-dimensional distributions of the random field
M
1
t, x =
M
1
t, x
, t 0, x ∈ R
n
, where
M
1
= C
1 m
1
p m
1
Kn, κ
1
m1 2
Z
′
R
n ×m1
e
i x,λ
1
+···+λ
m1
E
β
−µ|λ
1
+ · · · + λ
m
1
|
α+γ
t
β
|λ
1
| · · · |λ
m
1
|
n −κ1
2
m
1
Y
l=1
W
1
dλ
l
.
4.3 Two remarks for future study