We can also generalize the notion of a differential equation.
Definition:
The differential equation
L
u
=
f
{\displaystyle {\mathcal {L}}u=f}
is a differential equation in the sense
of a distribution (i.e., in the weak sense) if
f
{\displaystyle f}
and
u
{\displaystyle u}
are distributions
and all the derivatives are interpreted in the weak sense.
Suppose
L
{\displaystyle {\mathcal {L}}}
is the generalized differential operator
L
=
a
n
(
x
)
d
n
d
x
n
+
a
n
−
1
(
x
)
d
(
n
−
1
)
d
x
(
n
−
1
)
+
⋯
+
a
0
(
x
)
{\displaystyle {\mathcal {L}}=a_{n}(x){\cfrac {d^{n}}{dx^{n}}}+a_{n-1}(x){\cfrac {d^{(n-1)}}{dx^{(n-1)}}}+\dots +a_{0}(x)}
where
a
i
(
x
)
{\displaystyle a_{i}(x)}
is infinitely differentiable.
We seek a
u
{\displaystyle u}
such that
⟨
L
u
,
ϕ
⟩
=
⟨
f
,
ϕ
⟩
∀
ϕ
∈
D
{\displaystyle \left\langle {\mathcal {L}}u,\phi \right\rangle =\left\langle f,\phi \right\rangle \qquad \forall \phi \in D}
which is taken to mean that
⟨
u
,
L
∗
ϕ
⟩
=
⟨
f
,
ϕ
⟩
∀
ϕ
∈
D
.
{\displaystyle \left\langle u,{\mathcal {L}}^{*}\phi \right\rangle =\left\langle f,\phi \right\rangle \qquad \forall \phi \in D~.}
Note that
⟨
a
k
(
x
)
d
k
u
d
x
k
,
ϕ
⟩
=
⟨
d
k
u
d
x
k
,
a
k
(
x
)
ϕ
⟩
=
(
−
1
)
k
⟨
u
,
d
d
x
k
[
a
k
(
x
)
ϕ
]
⟩
{\displaystyle \left\langle a_{k}(x){\cfrac {d^{k}u}{dx^{k}}},\phi \right\rangle =\left\langle {\cfrac {d^{k}u}{dx^{k}}},a_{k}(x)~\phi \right\rangle =(-1)^{k}\left\langle u,{\cfrac {d}{dx^{k}}}[a_{k}(x)~\phi ]\right\rangle }
Therefore,
L
∗
ϕ
=
(
−
1
)
n
d
n
(
a
n
ϕ
)
d
x
n
+
(
−
1
)
n
−
1
d
(
n
−
1
)
(
a
n
−
1
ϕ
)
d
x
(
n
−
1
)
+
⋯
+
a
0
ϕ
.
{\displaystyle {\mathcal {L}}^{*}\phi =(-1)^{n}~{\cfrac {d^{n}(a_{n}\phi )}{dx^{n}}}+(-1)^{n-1}~{\cfrac {d^{(n-1)}(a_{n-1}\phi )}{dx^{(n-1)}}}+\dots +a_{0}~\phi ~.}
Here
L
∗
{\displaystyle {\mathcal {L}}^{*}}
is the formal adjoint of
L
{\displaystyle {\mathcal {L}}}
. We can check that
(
L
∗
)
∗
=
L
{\displaystyle ({\mathcal {L}}^{*})^{*}={\mathcal {L}}}
. If
L
=
L
∗
{\displaystyle {\mathcal {L}}={\mathcal {L}}^{*}}
we say that
L
{\displaystyle {\mathcal {L}}}
is
formally self adjoint .
For example, if
n
=
2
{\displaystyle n=2}
then
L
=
a
2
(
x
)
d
2
d
x
2
+
a
1
(
x
)
d
d
x
+
a
0
(
x
)
{\displaystyle {\mathcal {L}}=a_{2}(x){\cfrac {d^{2}}{dx^{2}}}+a_{1}(x){\cfrac {d}{dx}}+a_{0}(x)}
Then
L
∗
ϕ
=
d
2
(
a
2
ϕ
)
d
x
2
−
d
(
a
1
ϕ
)
d
x
+
a
0
ϕ
{\displaystyle {\mathcal {L}}^{*}\phi ={\cfrac {d^{2}(a_{2}\phi )}{dx^{2}}}-{\cfrac {d(a_{1}\phi )}{dx}}+a_{0}~\phi }
or,
L
∗
ϕ
=
a
2
ϕ
″
+
(
2
a
2
′
−
a
1
)
ϕ
′
+
(
a
2
″
−
a
1
′
+
a
0
)
ϕ
.
{\displaystyle {\mathcal {L}}^{*}\phi =a_{2}~\phi ''+(2a'_{2}-a_{1})\phi '+(a''_{2}-a'_{1}+a_{0})\phi ~.}
Therefore, for
L
∗
{\displaystyle {\mathcal {L}}^{*}}
to be self adjoint,
L
∗
ϕ
=
L
ϕ
=
a
2
ϕ
″
+
a
1
ϕ
′
+
a
0
ϕ
{\displaystyle {\mathcal {L}}^{*}\phi ={\mathcal {L}}\phi =a_{2}\phi ''+a_{1}\phi '+a_{0}\phi }
Hence
a
2
′
=
a
1
⟹
a
2
″
=
a
1
′
.
{\displaystyle a'_{2}=a_{1}\implies a''_{2}=a_{1}'~.}
In such a case,
L
{\displaystyle {\mathcal {L}}}
is called a Sturm-Liouville operator .
To solve the differential equation
x
d
u
d
x
=
0
{\displaystyle x{\cfrac {du}{dx}}=0}
we seek a distribution
u
{\displaystyle u}
which satisfies
⟨
x
u
′
,
ϕ
⟩
=
−
⟨
u
,
(
x
ϕ
)
′
⟩
=
0
{\displaystyle \left\langle xu',\phi \right\rangle =-\left\langle u,(x\phi )'\right\rangle =0}
Define
ψ
:=
(
x
ϕ
)
′
{\displaystyle \psi :=(x\phi )'}
. Then
ψ
{\displaystyle \psi }
must be a test function. We can
show that
ψ
{\displaystyle \psi }
is a test function if and only if
(2)
∫
∞
∞
ψ
(
x
)
d
x
=
0
and
∫
0
∞
ψ
(
x
)
d
x
=
0
{\displaystyle {\text{(2)}}\qquad \int _{\infty }^{\infty }\psi (x){\text{d}}x=0\qquad {\text{and}}\qquad \int _{0}^{\infty }\psi (x){\text{d}}x=0}
Now let us pick two test functions
ϕ
0
{\displaystyle \phi _{0}}
and
ϕ
1
{\displaystyle \phi _{1}}
satisfying
∫
−
∞
∞
ϕ
0
(
x
)
d
x
=
0
and
∫
0
∞
ϕ
0
(
x
)
d
x
=
1
{\displaystyle \int _{-\infty }^{\infty }\phi _{0}(x){\text{d}}x=0\qquad {\text{and}}\qquad \int _{0}^{\infty }\phi _{0}(x){\text{d}}x=1}
and
∫
−
∞
∞
ϕ
1
(
x
)
d
x
=
1
and
∫
0
∞
ϕ
1
(
x
)
d
x
=
0
{\displaystyle \int _{-\infty }^{\infty }\phi _{1}(x){\text{d}}x=1\qquad {\text{and}}\qquad \int _{0}^{\infty }\phi _{1}(x){\text{d}}x=0}
Then we can write any arbitrary test function
ϕ
(
x
)
{\displaystyle \phi (x)}
as a linear
combination of
ϕ
0
{\displaystyle \phi _{0}}
and
ϕ
1
{\displaystyle \phi _{1}}
plus a terms which has the form of
ψ
{\displaystyle \psi }
:
ϕ
(
x
)
=
ϕ
0
(
x
)
∫
0
∞
ϕ
(
s
)
d
s
+
ϕ
1
(
x
)
∫
−
∞
∞
ϕ
(
s
)
d
s
+
ψ
(
x
)
{\displaystyle \phi (x)=\phi _{0}(x)\int _{0}^{\infty }\phi (s)~ds+\phi _{1}(x)\int _{-\infty }^{\infty }\phi (s)~ds+\psi (x)}
which serves to define
ψ
(
x
)
{\displaystyle \psi (x)}
. Note that
ψ
{\displaystyle \psi }
satisfies
equation (2).
Since
⟨
u
,
ψ
⟩
=
0
{\displaystyle \left\langle u,\psi \right\rangle =0}
, the action of
u
{\displaystyle u}
on
ϕ
{\displaystyle \phi }
is given by
⟨
u
,
ϕ
⟩
=
⟨
u
,
ϕ
0
⟩
∫
∞
∞
H
(
x
)
ϕ
(
s
)
d
s
+
⟨
u
,
ϕ
1
⟩
∫
∞
∞
ϕ
(
s
)
d
s
{\displaystyle \left\langle u,\phi \right\rangle =\left\langle u,\phi _{0}\right\rangle \int _{\infty }^{\infty }H(x)\phi (s)ds+\left\langle u,\phi _{1}\right\rangle \int _{\infty }^{\infty }\phi (s)ds}
Therefore the solution is
u
=
C
1
H
(
x
)
+
C
2
{\displaystyle u=C_{1}~H(x)+C_{2}}
where
C
1
:=
⟨
u
,
ϕ
0
⟩
{\displaystyle C_{1}:=\left\langle u,\phi _{0}\right\rangle }
and
C
2
:=
⟨
u
,
ϕ
1
⟩
{\displaystyle C_{2}:=\left\langle u,\phi _{1}\right\rangle }
.
Template:Lectures