a) Show that the function
G
(
x
)
=
1
/
2
e
−
|
x
|
{\displaystyle G(x)=1/2e^{-|x|}}
is a solution in the distribution sense of the equation
−
G
″
+
G
=
δ
(
x
)
;
−
∞
<
x
<
∞
{\displaystyle -G''+G=\delta (x);\quad -\infty <x<\infty }
.
b) Use part (a) to write a solution of
−
u
″
+
u
=
f
(
x
)
;
−
∞
<
x
<
∞
{\displaystyle -u''+u=f(x);\quad -\infty <x<\infty }
We want to show
∫
R
G
(
−
φ
″
+
φ
)
d
x
=
φ
(
0
)
{\displaystyle \int _{\mathbb {R} }G(-\varphi ''+\varphi )\,dx=\varphi (0)}
for every test function
φ
∈
C
c
∞
(
R
)
{\displaystyle \varphi \in C_{c}^{\infty }(\mathbb {R} )}
.
One can compute
G
(
x
)
=
G
″
(
x
)
{\displaystyle G(x)=G''(x)}
and
G
′
(
x
)
=
1
/
2
(
s
g
n
(
−
x
)
)
e
−
|
x
|
{\displaystyle G'(x)=1/2(sgn(-x))e^{-|x|}}
. Therefore, away from 0, we have
−
G
″
+
G
=
0
{\displaystyle -G''+G=0}
, that is,
−
G
″
+
G
=
0
{\displaystyle -G''+G=0}
a.e. and
∫
−
G
″
+
G
=
0
{\displaystyle \int -G''+G=0}
.
We now compute by an integration by parts:
∫
−
∞
0
(
−
G
″
+
G
)
φ
=
∫
−
∞
0
G
′
φ
′
+
G
φ
d
x
−
φ
G
|
−
∞
0
=
∫
−
∞
0
G
′
φ
′
+
G
φ
d
x
−
φ
(
0
)
lim
x
→
0
−
G
(
x
)
=
∫
−
∞
0
G
(
−
φ
″
+
φ
)
d
x
−
1
/
2
φ
(
0
)
+
φ
′
G
|
−
∞
0
=
∫
−
∞
0
G
(
−
φ
″
+
φ
)
d
x
−
1
/
2
φ
(
0
)
+
1
/
2
φ
′
(
0
)
.
{\displaystyle {\begin{aligned}\int _{-\infty }^{0}(-G''+G)\varphi =&\int _{-\infty }^{0}G'\varphi '+G\varphi \,dx-\left.\varphi G\right|_{-\infty }^{0}\\=&\int _{-\infty }^{0}G'\varphi '+G\varphi \,dx-\varphi (0)\lim _{x\to 0^{-}}G(x)\\=&\int _{-\infty }^{0}G(-\varphi ''+\varphi )\,dx-1/2\varphi (0)+\left.\varphi 'G\right|_{-\infty }^{0}\\=&\int _{-\infty }^{0}G(-\varphi ''+\varphi )\,dx-1/2\varphi (0)+1/2\varphi '(0).\\\end{aligned}}}
A similar calculation gives
∫
0
∞
(
−
G
″
+
G
)
φ
=
∫
0
∞
G
(
−
φ
″
+
φ
)
d
x
−
1
/
2
φ
(
0
)
−
1
/
2
φ
′
(
0
)
.
{\displaystyle \int _{0}^{\infty }(-G''+G)\varphi =\int _{0}^{\infty }G(-\varphi ''+\varphi )\,dx-1/2\varphi (0)-1/2\varphi '(0).}
So we have shown that for all
φ
∈
C
c
∞
(
R
)
{\displaystyle \varphi \in C_{c}^{\infty }(\mathbb {R} )}
0
=
∫
φ
(
−
G
″
+
G
)
=
∫
G
(
−
φ
″
+
φ
)
d
x
−
φ
(
0
)
{\displaystyle 0=\int \varphi (-G''+G)=\int G(-\varphi ''+\varphi )\,dx-\varphi (0)}
which gives the desired result.
We guess
u
=
G
∗
f
{\displaystyle u=G\ast f}
. Then by part (a),
−
u
″
(
x
)
+
u
(
x
)
=
∫
R
[
−
G
″
(
x
−
y
)
+
G
(
x
−
y
)
]
f
(
y
)
d
y
=
∫
R
δ
(
x
−
y
)
f
(
y
)
d
y
=
f
(
x
)
{\displaystyle -u''(x)+u(x)=\int _{\mathbb {R} }[-G''(x-y)+G(x-y)]f(y)\,dy=\int _{\mathbb {R} }\delta (x-y)f(y)\,dy=f(x)}
.
Multiply the PDE by
u
{\displaystyle u}
and integrate:
λ
∫
B
u
2
=
−
∫
B
u
Δ
u
=
∫
B
|
D
u
|
2
−
∫
∂
B
u
∂
ν
u
=
∫
B
|
D
u
|
2
+
∫
∂
B
u
2
≥
0
{\displaystyle \lambda \int _{B}u^{2}=-\int _{B}u\Delta u=\int _{B}|Du|^{2}-\int _{\partial B}u\partial _{\nu }u=\int _{B}|Du|^{2}+\int _{\partial B}u^{2}\geq 0}
.
Of course we know that
λ
=
0
{\displaystyle \lambda =0}
is an eigenvalue of
−
Δ
{\displaystyle -\Delta }
corresponding to a constant eigenfunction. But a constant function has
∂
v
u
≡
0
{\displaystyle \partial _{v}u\equiv 0}
which implies
u
≡
0
{\displaystyle u\equiv 0}
by the boundary condition. Hence
λ
=
0
{\displaystyle \lambda =0}
is no longer an eigenvalue. This forces
λ
>
0
{\displaystyle \lambda >0}
.
To see orthogonality of the eigenfunctions, let
φ
n
,
φ
m
{\displaystyle \varphi _{n},\varphi _{m}}
be two eigenfunctions corresponding to distinct eigenvalues
λ
n
,
λ
m
{\displaystyle \lambda _{n},\lambda _{m}}
, respectively. Then by an integration of parts,
∫
B
−
φ
n
Δ
φ
m
+
φ
m
Δ
φ
n
=
∫
B
D
φ
n
D
φ
m
−
D
φ
n
D
φ
m
d
x
+
∫
∂
B
−
φ
n
∂
ν
φ
m
+
∂
ν
φ
n
φ
m
d
S
=
0
{\displaystyle \int _{B}-\varphi _{n}\Delta \varphi _{m}+\varphi _{m}\Delta \varphi _{n}=\int _{B}D\varphi _{n}D\varphi _{m}-D\varphi _{n}D\varphi _{m}\,dx+\int _{\partial B}-\varphi _{n}\partial _{\nu }\varphi _{m}+\partial _{\nu }\varphi _{n}\varphi _{m}\,dS=0}
So by the PDE,
0
=
∫
B
−
φ
n
Δ
φ
m
+
φ
m
Δ
φ
n
=
∫
B
(
λ
n
−
λ
m
)
φ
n
φ
m
{\displaystyle 0=\int _{B}-\varphi _{n}\Delta \varphi _{m}+\varphi _{m}\Delta \varphi _{n}=\int _{B}(\lambda _{n}-\lambda _{m})\varphi _{n}\varphi _{m}}
.
Since
λ
n
−
λ
m
≠
0
{\displaystyle \lambda _{n}-\lambda _{m}\neq 0}
this implies that
{
φ
n
}
{\displaystyle \{\varphi _{n}\}}
are pairwise orthogonal in
L
2
(
B
)
{\displaystyle L^{2}(B)}
.