In mathematics, the fractional Laplacian is an operator, which generalizes the notion of Laplacian spatial derivatives to fractional powers. This operator is often used to generalise certain types of Partial differential equation, two examples are [1] and [2] which both take known PDEs containing the Laplacian and replacing it with the fractional version.
In literature the definition of the fractional Laplacian often varies, but most of the time those definitions are equivalent. The following is a short overview proven by Kwaśnicki, M in.[3]
Let
p\in[1,infty)
l{X}:=Lp(Rn)
l{X}:=
n) | |
C | |
0(R |
l{X}:=Cbu(Rn)
n) | |
C | |
0(R |
f:Rn\toR
\forall\varepsilon>0,\existsK\subsetRn
|f(x)|<\epsilon
x\notinK
Cbu(Rn)
f:Rn\toR
\forall\epsilon>0,\exists\delta>0
|f(x)-f(y)|<\epsilon
x,y\inRn
|x-y|<\delta
\existsM>0
|f(x)|\leqM
x\inRn
Additionally, let
s\in(0,1)
If we further restrict to
p\in[1,2]
(-\Delta)sf:=
-1 | |
l{F} | |
\xi |
(|\xi|2sl{F}(f))
This definition uses the Fourier transform for
f\inLp(Rn)
p\in[1,infty)
The Laplacian can also be viewed as a singular integral operator which is defined as the following limit taken in
l{X}
(-\Delta)sf(x)=
| ||||||||||
\pid/2|\Gamma(-s)| |
\lim | |
r\to0+ |
\int\limits | { | |
Rd\setminusBr(x) |
f(x)-f(y) | |
|x-y|d+2s |
dy}
Using the fractional heat-semigroup which is the family of operators
\{Pt\}t
-(-\Delta)sf(x)=
\lim | |
t\to0+ |
Ptf-f | |
t |
It is to note that the generator is not the fractional Laplacian
(-\Delta)s
-(-\Delta)s
Pt:l{X}\tol{X}
Ptf:=pt*f
where
*
pt:=
-1 | |
l{F} | |
\xi |
-t|\xi|2s | |
(e |
)
For all Schwartz functions
\varphi
\int | |
Rd |
(-\Delta)sf(y)\varphi(y)dy=
\int | |
Rd |
f(x)(-\Delta)s\varphi(x)dx
where
(-\Delta)s\varphi
The fractional Laplacian can be expressed using Bochner's integral as
(-\Delta)sf=
1 | ||||
|
infty | |
\int | |
0 |
\left(etf-f\right)t-1dt
where the integral is understood in the Bochner sense for
l{X}
Alternatively, it can be defined via Balakrishnan's formula:
(-\Delta)sf=
| ||||||
\pi |
infty | |
\int | |
0 |
(-\Delta)\left(sI-\Delta\right)-1fss/2ds
with the integral interpreted as a Bochner integral for
l{X}
Another approach by Dynkin defines the fractional Laplacian as
(-\Delta)sf=
\lim | |
r\to0+ |
| ||||||||||
|
\int | |
Rd\setminus\overline{B |
(x,r)}
f(x+z)-f(x) | |
|z|d\left(|z|2-r2\right)s/2 |
dz
with the limit taken in
l{X}
In
l{X}=L2
\langle(-\Delta)sf,\varphi\rangle=l{E}(f,\varphi)
where
l{E}(f,g)=
| ||||||||||
|
\int | |
Rd |
\int | |
Rd |
(f(y)-f(x))(\overline{g(y) | |
- |
\overline{g(x)})}{|x-y|d
When
s<d
l{X}=Lp
p\in[1,
d | |
s |
)
| |||||||||
|
\int | |
Rd |
(-\Delta)sf(x+z) | |
|z|d |
dz=f(x)
The fractional Laplacian can also be defined through harmonic extensions. Specifically, there exists a function
u(x,y)
\begin{cases} \Deltaxu(x,y)+\alpha2
2/\alpha | |
c | |
\alpha |
y2
2 | |
\partial | |
y |
u(x,y)=0&fory>0,\\ u(x,0)=f(x),\\ \partialyu(x,0)=-(-\Delta)sf(x), \end{cases}
where
c\alpha=2-\alpha
| ||||||
|
u( ⋅ ,y)
l{X}
y\in[0,infty)
\|u( ⋅ ,y)\|l{X
y\geq0