Équation de Laplace sur un anneau .
En mathématiques , et plus précisément en analyse , un opérateur différentiel est un opérateur agissant sur des fonctions différentiables .
Lorsque la fonction est à une seule variable, l'opérateur différentiel est construit à partir des dérivées ordinaires.
Lorsque la fonction est à plusieurs variables, l'opérateur différentiel est construit à partir des dérivées partielles .
Un opérateur différentiel agissant sur deux fonctions
D
(
f
,
g
)
{\displaystyle D(f,g)}
est appelé opérateur bidifférentiel .
L'opération différentielle la plus commune consiste simplement à prendre la dérivée de la grandeur considérée. Les notations usuelles pour désigner la dérivée première par rapport à une variable x sont par exemple :
d
d
x
{\displaystyle {\frac {\mathrm {d} }{\mathrm {d} x}}}
ou
∂
x
{\displaystyle \partial _{x}}
, ou encore
D
{\displaystyle D}
ou
D
x
{\displaystyle D_{x}}
.
La notation en D est attribuée à Oliver Heaviside , qui l'a introduite dans son étude des équations différentielles pour noter des opérateurs différentiels de la forme :
∑
k
=
0
n
c
k
D
k
{\displaystyle \sum _{k=0}^{n}c_{k}D^{k}}
Pour des dérivées d'ordre n supérieur, ces mêmes opérateurs peuvent s'écrire :
d
n
d
x
n
{\displaystyle {\frac {\mathrm {d} ^{n}}{\mathrm {d} x^{n}}}}
,
∂
x
n
n
{\displaystyle \partial _{x^{n}}^{n}}
ou encore
D
x
n
{\displaystyle D_{x}^{n}}
La notation en "prime" s'utilise plutôt pour exprimer la valeur que prend une fonction dérivée f pour un argument x :
f
′
(
x
)
{\displaystyle f'(x)\,\!}
, ou :
[
f
(
x
)
]
′
{\displaystyle [f(x)]'\,\!}
Deux opérateurs différentiels particulièrement fréquents sont l'opérateur nabla , défini dans une base cartésienne
(
i
1
,
.
.
.
,
i
n
)
{\displaystyle (\mathbf {i} _{1},...,\mathbf {i} _{n})}
, par :
∇
=
∑
k
=
1
n
∂
∂
x
k
i
k
,
{\displaystyle \nabla =\sum _{k=1}^{n}{\partial \over \partial x_{k}}\mathbf {i} _{k},}
ainsi que l'opérateur laplacien , défini par :
Δ
=
∇
2
=
∑
k
=
1
n
∂
2
∂
x
k
2
.
{\displaystyle \Delta =\nabla ^{2}=\sum _{k=1}^{n}{\partial ^{2} \over \partial x_{k}^{2}}.}
Un autre opérateur utilisé en physique est l'opérateur Θ, dont les vecteurs propres sont les monômes homogènes, défini par[ 1]
Θ
=
z
d
d
z
{\displaystyle \Theta =z{\frac {\mathrm {d} }{\mathrm {d} z}}}
ou, dans le cas de plusieurs variables,
Θ
=
∑
k
=
1
n
x
k
∂
∂
x
k
.
{\displaystyle \Theta =\sum _{k=1}^{n}x_{k}{\frac {\partial }{\partial x_{k}}}.}
Soit
Ω
{\displaystyle \Omega }
un ouvert de
R
n
{\displaystyle \mathbb {R} ^{n}}
, et
x
{\displaystyle x}
un point de
Ω
{\displaystyle \Omega }
. On introduit les
n
{\displaystyle n}
coordonnées
x
k
(
k
=
1
,
.
.
.
,
n
)
{\displaystyle x_{k}{\mbox{ }}(k=1,...,n)}
. Supposons que l'on ait une fonction des
n
{\displaystyle n}
variables
x
k
{\displaystyle x_{k}}
.
Pour simplifier les écritures, on note usuellement la dérivée partielle première par rapport à la coordonnée
x
k
{\displaystyle x_{k}}
par le symbole :
∂
k
=
∂
∂
x
k
{\displaystyle \partial _{k}\ =\ {\frac {\partial ~~}{\partial x_{k}}}}
On est également amené à introduire l'opérateur différentiel
D
k
{\displaystyle \mathrm {D} _{k}}
du premier ordre défini par :
D
k
=
−
i
∂
k
=
−
i
∂
∂
x
k
{\displaystyle \mathrm {D} _{k}\ =\ -\ \mathrm {i} \ \partial _{k}\ =\ -\ \mathrm {i} \ {\frac {\partial ~~}{\partial x_{k}}}}
Dans cette définition,
i
{\displaystyle \mathrm {i} }
est la « racine de l'unité » complexe :
i
2
=
−
1
{\displaystyle \mathrm {i} ^{2}=-1}
. L'intérêt de définir cet opérateur
D
k
{\displaystyle \mathrm {D} _{k}}
apparaîtra plus tard, en relation avec la transformée de Fourier .
On utilise les notations sous forme de multi-indices : un multi-indice
α
{\displaystyle \alpha }
est un
n
{\displaystyle n}
-uplet d'entiers
α
=
(
α
1
,
…
,
α
n
)
;
α
k
∈
N
{\displaystyle \alpha \ =\ \left(\alpha _{1},\ \dots ,\ \alpha _{n}\right)\ ;\quad \ \alpha _{k}\,\in \,\mathbb {N} }
Sa longueur
|
α
|
{\displaystyle |\alpha |}
est définie comme la somme des
α
i
{\displaystyle \alpha _{i}}
et on définit enfin la multi-factorielle :
α
!
=
∏
k
=
1
n
(
α
k
!
)
=
α
1
!
×
…
×
α
n
!
{\displaystyle \alpha \,!\ =\ \prod _{k=1}^{n}(\,\alpha _{k}\,!\,)\ =\ \alpha _{1}\,!\ \times \ \dots \ \times \ \alpha _{n}\,!}
La dérivée partielle d'ordre
α
k
{\displaystyle \alpha _{k}}
par rapport à la coordonnée
x
k
{\displaystyle x_{k}}
correspond au symbole :
∂
k
α
k
{\displaystyle \partial _{k}^{\alpha _{k}}}
On définit alors les dérivées partielles, d'ordre global
|
α
|
{\displaystyle |\alpha |}
:
∂
α
=
∂
1
α
1
…
∂
n
α
n
{\displaystyle \partial ^{\alpha }\ =\ \partial _{1}^{\alpha _{1}}\ \dots \ \partial _{n}^{\alpha _{n}}}
Et les opérateurs différentiels
D
α
{\displaystyle \mathrm {D} ^{\alpha }}
, d'ordre global
|
α
|
{\displaystyle |\alpha |}
:
D
α
=
D
1
α
1
…
D
n
α
n
{\displaystyle \mathrm {D} ^{\alpha }\ =\ \mathrm {D} _{1}^{\alpha _{1}}\ \dots \ \mathrm {D} _{n}^{\alpha _{n}}}
Un opérateur différentiel linéaire d'ordre
m
{\displaystyle m}
est défini par :
D
=
∑
|
α
|
=
0
m
a
α
(
x
)
D
α
{\displaystyle {\mathfrak {D}}\ =\ \sum _{|\alpha |=0}^{m}\ a_{\alpha }(x)\ \mathrm {D} ^{\alpha }}
où les
a
α
(
x
)
{\displaystyle a_{\alpha }(x)}
sont des fonctions de
n
{\displaystyle n}
variables, appelées coefficients de l'opérateur
D
{\displaystyle {\mathfrak {D}}}
.
Un opérateur différentiel
D
{\displaystyle {\mathfrak {D}}}
est local au sens où, pour déterminer ses effets
D
f
(
x
)
{\displaystyle {\mathfrak {D}}\,f(x)}
sur une fonction
f
(
x
)
{\displaystyle f(x)}
suffisamment différentiable, seule la connaissance de la fonction dans le voisinage du point
x
{\displaystyle x}
est nécessaire.
On définit ici la transformée de Fourier de la fonction
f
(
x
)
{\displaystyle f(x)}
de
n
{\displaystyle n}
variables
x
k
(
k
=
1
,
.
.
.
,
n
)
{\displaystyle x_{k}{\mbox{ }}(k=1,...,n)}
par :
f
^
(
ξ
)
=
∫
R
n
d
x
e
−
i
⟨
ξ
,
x
⟩
f
(
x
)
{\displaystyle {\hat {f}}(\xi )\ =\ \int _{\mathbb {R} ^{n}}\mathrm {d} x\ \mathrm {e} ^{-\,\mathrm {i} \,\langle \,\xi \,,\,x\,\rangle }\ f(x)}
Dans cette définition :
on note
ξ
{\displaystyle \xi }
le
n
{\displaystyle n}
-uplet constitué des variables :
ξ
k
(
k
=
1
,
.
.
.
,
n
)
{\displaystyle \xi _{k}{\mbox{ }}(k=1,...,n)}
.
la mesure est :
d
x
=
∏
k
=
1
n
d
x
k
{\displaystyle \mathrm {d} x=\prod _{k=1}^{n}\mathrm {d} x_{k}}
.
le facteur
⟨
ξ
,
x
⟩
{\displaystyle \langle \xi \,,\,x\rangle }
dans l'exponentielle oscillante désigne le produit scalaire :
⟨
ξ
,
x
⟩
=
∑
k
=
1
n
x
k
ξ
k
{\displaystyle \langle \xi \,,\,x\rangle =\sum _{k=1}^{n}x_{k}\,\xi _{k}}
.
La formule de transformation inverse s'écrit alors :
f
(
x
)
=
∫
R
n
d
ξ
~
e
+
i
⟨
ξ
,
x
⟩
f
^
(
ξ
)
{\displaystyle f(x)\ =\ \int _{\mathbb {R} ^{n}}\mathrm {d} {\tilde {\xi }}\ \mathrm {e} ^{+\,\mathrm {i} \,\langle \,\xi \,,\,x\,\rangle }\ {\hat {f}}(\xi )}
où la mesure est :
d
ξ
~
=
d
ξ
(
2
π
)
n
{\displaystyle \mathrm {d} {\tilde {\xi }}\ =\ {\frac {\mathrm {d} \xi }{(2\pi )^{n}}}\quad }
avec
d
ξ
=
∏
k
=
1
n
d
ξ
k
{\displaystyle \mathrm {d} \xi =\prod _{k=1}^{n}\mathrm {d} \xi _{k}}
.
On applique l'opérateur différentiel
D
k
=
−
i
∂
k
{\displaystyle \mathrm {D} _{k}=-\,\mathrm {i} \,\partial _{k}}
à la représentation de Fourier de la fonction
f
{\displaystyle f}
. En supposant qu'on puisse intervertir la dérivation et l'intégration, on obtient :
D
k
f
(
x
)
=
∫
R
n
d
ξ
~
(
−
i
∂
k
e
+
i
⟨
ξ
,
x
⟩
)
f
^
(
ξ
)
=
∫
R
n
d
ξ
~
e
+
i
⟨
ξ
,
x
⟩
ξ
k
f
^
(
ξ
)
{\displaystyle \mathrm {D} _{k}\,f(x)\ =\ \int _{\mathbb {R} ^{n}}\mathrm {d} {\tilde {\xi }}\ \left(\ -\ \mathrm {i} \ \partial _{k}\ \mathrm {e} ^{+\,\mathrm {i} \,\langle \,\xi \,,\,x\,\rangle }\ \right)\ {\hat {f}}(\xi )\ =\ \int _{\mathbb {R} ^{n}}\mathrm {d} {\tilde {\xi }}\ \mathrm {e} ^{+\,\mathrm {i} \,\langle \,\xi \,,\,x\,\rangle }\ \xi _{k}\ {\hat {f}}(\xi )}
qu'on peut écrire :
(
D
k
f
^
)
(
ξ
)
=
ξ
k
f
^
(
ξ
)
{\displaystyle ({\widehat {\mathrm {D} _{k}\,f}})(\xi )=\xi _{k}\ {\hat {f}}(\xi )}
. On en déduit que :
(
D
α
f
^
)
(
ξ
)
=
ξ
α
f
^
(
ξ
)
{\displaystyle ({\widehat {\mathrm {D} ^{\alpha }\,f}})(\xi )\ =\ \xi ^{\alpha }\ {\hat {f}}(\xi )}
où :
ξ
α
=
ξ
1
α
1
×
…
×
ξ
n
α
n
{\displaystyle \xi ^{\alpha }=\xi _{1}^{\alpha _{1}}\ \times \ \dots \ \times \ \xi _{n}^{\alpha _{n}}}
. L'opérateur différentiel
D
{\displaystyle {\mathfrak {D}}}
d'ordre
m
{\displaystyle m}
vérifie donc la relation :
(
D
f
)
(
x
)
=
∑
|
α
|
=
0
m
a
α
(
x
)
∫
R
n
d
ξ
~
e
+
i
⟨
ξ
,
x
⟩
ξ
α
f
^
(
ξ
)
{\displaystyle ({\mathfrak {D}}\,f)(x)\ =\ \sum _{|\alpha |=0}^{m}\ a_{\alpha }(x)\ \int _{\mathbb {R} ^{n}}\mathrm {d} {\tilde {\xi }}\ \mathrm {e} ^{+\,\mathrm {i} \,\langle \,\xi \,,\,x\,\rangle }\ \xi ^{\alpha }\ {\hat {f}}(\xi )}
On peut intervertir la somme et l'intégrale pour écrire :
(
D
f
)
(
x
)
=
∫
R
n
d
ξ
~
e
+
i
⟨
ξ
,
x
⟩
(
∑
|
α
|
=
0
m
a
α
(
x
)
ξ
α
)
f
^
(
ξ
)
{\displaystyle ({\mathfrak {D}}\,f)(x)\ =\ \int _{\mathbb {R} ^{n}}\mathrm {d} {\tilde {\xi }}\ \mathrm {e} ^{+\,\mathrm {i} \,\langle \,\xi \,,\,x\,\rangle }\ \left(\ \sum _{|\alpha |=0}^{m}\ a_{\alpha }(x)\ \xi ^{\alpha }\ \right)\ {\hat {f}}(\xi )}
On appelle symbole de l'opérateur différentiel
D
{\displaystyle {\mathfrak {D}}}
d'ordre
m
{\displaystyle m}
la fonction
σ
(
x
,
ξ
)
{\displaystyle \sigma (x,\xi )}
des
2
n
{\displaystyle 2n}
variables
(
x
,
ξ
)
{\displaystyle (x,\xi )}
polynomiale en
ξ
{\displaystyle \xi }
de degré
m
{\displaystyle m}
:
σ
(
x
,
ξ
)
=
∑
|
α
|
=
0
m
a
α
(
x
)
ξ
α
{\displaystyle \sigma (x,\xi )=\sum _{|\alpha |=0}^{m}\ a_{\alpha }(x)\ \xi ^{\alpha }}
de telle sorte que :
(
D
f
)
(
x
)
=
∫
R
n
d
ξ
~
e
+
i
⟨
ξ
,
x
⟩
σ
(
x
,
ξ
)
f
^
(
ξ
)
{\displaystyle ({\mathfrak {D}}\,f)(x)\ =\ \int _{\mathbb {R} ^{n}}\mathrm {d} {\tilde {\xi }}\ \mathrm {e} ^{+\,\mathrm {i} \,\langle \,\xi \,,\,x\,\rangle }\ \sigma (x,\xi )\ {\hat {f}}(\xi )}
On constate que cette formule pourrait en fait permettre de définir l'opérateur
D
{\displaystyle {\mathfrak {D}}}
à partir de son symbole
σ
{\displaystyle \sigma }
. Cette idée sera mise à profit dans la théorie des opérateurs pseudo-différentiels .
Attention : pour un opérateur différentiel dont les coefficients
a
α
(
x
)
{\displaystyle a_{\alpha }(x)}
ne sont pas constants, le symbole
σ
(
x
,
ξ
)
{\displaystyle \sigma (x,\xi )}
dépend des coordonnées d'espace
x
{\displaystyle x}
, et l'expression
σ
(
x
,
ξ
)
f
^
(
ξ
)
{\displaystyle \sigma (x,\xi )\,{\hat {f}}(\xi )}
n'est pas la transformée de Fourier de
(
D
f
)
(
x
)
{\displaystyle ({\mathfrak {D}}\,f)(x)}
, c’est-à-dire que :
(
D
f
^
)
(
ξ
)
≠
σ
(
x
,
ξ
)
f
^
(
ξ
)
{\displaystyle ({\widehat {{\mathfrak {D}}\,f}})(\xi )\ \neq \ \sigma (x,\xi )\ {\hat {f}}(\xi )}
La formule correcte de la transformée de Fourier est calculée dans le paragraphe « Cas général ».
On appelle symbole principal de l'opérateur différentiel
D
{\displaystyle {\mathfrak {D}}}
d'ordre
m
{\displaystyle m}
la fonction :
σ
m
(
x
,
ξ
)
=
∑
|
α
|
=
m
a
α
(
x
)
ξ
α
{\displaystyle \sigma _{m}(x,\xi )=\sum _{|\alpha |=m}\ a_{\alpha }(x)\ \xi ^{\alpha }}
L'opérateur différentiel
D
{\displaystyle {\mathfrak {D}}}
est dit elliptique au point
x
∈
Ω
{\displaystyle x\ \in \ \Omega }
si et seulement si :
∀
ξ
∈
R
n
∖
{
0
}
,
σ
m
(
x
,
ξ
)
≠
0
{\displaystyle \forall \ \xi \ \in \ \mathbb {R} ^{n}\backslash \{0\}\ ,\quad \sigma _{m}(x,\xi )\ \neq \ 0}
D
{\displaystyle {\mathfrak {D}}}
est dit elliptique dans
Ω
{\displaystyle \Omega }
s'il est elliptique pour tout point
x
∈
Ω
{\displaystyle x\ \in \ \Omega }
.
L'opérateur différentiel
D
{\displaystyle {\mathfrak {D}}}
est dit hyperbolique dans la direction
η
{\displaystyle \eta }
au point
x
∈
Ω
{\displaystyle x\ \in \ \Omega }
si et seulement si :
σ
m
(
x
,
η
)
≠
0
{\displaystyle \sigma _{m}(x,\eta )\neq 0}
et si, pour tout
ξ
{\displaystyle \xi }
non colinéaire à
η
{\displaystyle \eta }
, les racines
λ
i
{\displaystyle \lambda _{i}}
de l'équation :
σ
m
(
x
,
ξ
+
λ
η
)
=
0
{\displaystyle \sigma _{m}(x,\ \xi \ +\ \lambda \,\eta )\ =\ 0}
sont toutes réelles. Si, de plus, les
m
{\displaystyle m}
racines réelles sont toutes distinctes, l'opérateur
D
{\displaystyle {\mathfrak {D}}}
est dit strictement hyperbolique dans la direction
η
{\displaystyle \eta }
.
D
{\displaystyle {\mathfrak {D}}}
est dit (strictement) hyperbolique dans la direction
η
{\displaystyle \eta }
dans
Ω
{\displaystyle \Omega }
s'il est strictement hyperbolique dans la direction
η
{\displaystyle \eta }
pour tout point
x
∈
Ω
{\displaystyle x\ \in \ \Omega }
.
La physique théorique fait un usage abondant de trois opérateurs d'ordre 2 :
L'opérateur laplacien est un opérateur elliptique, qui s'écrit :
en coordonnées cartésiennes dans
R
n
{\displaystyle \mathbb {R} ^{n}}
:
Δ
=
∑
k
=
1
n
∂
2
∂
x
k
2
{\displaystyle \Delta \ =\ \sum _{k=1}^{n}\ {\frac {\partial ^{2}~~}{\partial x_{k}^{2}}}}
soit en coordonnées cartésiennes tridimensionnelles :
Δ
=
∂
2
∂
x
2
+
∂
2
∂
y
2
+
∂
2
∂
z
2
{\displaystyle \Delta \ =\ {\frac {\partial ^{2}~~}{\partial x^{2}}}\ +\ {\frac {\partial ^{2}~~}{\partial y^{2}}}\ +\ {\frac {\partial ^{2}~~}{\partial z^{2}}}}
Cet opérateur est notamment utilisé en mécanique newtonienne , en électromagnétisme , et en mécanique quantique non relativiste.
L'opérateur d'alembertien est un opérateur hyperbolique, qui s'écrit en coordonnées cartésiennes
(
x
,
t
)
{\displaystyle (x,t)}
dans
R
n
+
1
{\displaystyle \mathbb {R} ^{n+1}}
:
◻
=
1
c
2
∂
2
∂
t
2
−
Δ
{\displaystyle \Box \ =\ {\frac {1}{c^{2}}}\ {\frac {\partial ^{2}~~}{\partial t^{2}}}\ -\ \Delta }
où
Δ
{\displaystyle \Delta }
est le laplacien à
n
{\displaystyle n}
variables d'espace,
t
{\displaystyle t}
est le temps, et
c
{\displaystyle c}
une constante positive, homogène à une vitesse . Cet opérateur est utilisé pour décrire la propagation des ondes à la vitesse
c
{\displaystyle c}
dans l'espace-temps. Il est notamment utilisé en acoustique , en électromagnétisme , et en théorie quantique des champs .
L'opérateur de la chaleur, qui s'écrit en coordonnées cartésiennes
(
x
,
t
)
{\displaystyle (x,t)}
dans
R
n
+
1
{\displaystyle \mathbb {R} ^{n+1}}
:
∂
∂
t
−
D
~
Δ
{\displaystyle {\frac {\partial ~}{\partial t}}\ -\ {\tilde {D}}\ \Delta }
où
Δ
{\displaystyle \Delta }
est le laplacien à
n
{\displaystyle n}
variables d'espace,
t
{\displaystyle t}
est le temps, et
D
~
{\displaystyle {\tilde {D}}}
est ici une constante, appelée coefficient de diffusion . Cet opérateur est dit parabolique .
Si les coefficients
a
α
{\displaystyle a_{\alpha }}
sont indépendants des
n
{\displaystyle n}
variables d'espace
x
k
{\displaystyle x^{k}}
, le symbole de l'opérateur différentiel
D
{\displaystyle {\mathfrak {D}}}
d'ordre
m
{\displaystyle m}
est seulement une fonction
σ
(
ξ
)
{\displaystyle \sigma (\xi )}
des
n
{\displaystyle n}
variables
ξ
{\displaystyle \xi }
polynomiale en
ξ
{\displaystyle \xi }
:
σ
(
ξ
)
=
∑
|
α
|
=
0
m
a
α
ξ
α
{\displaystyle \sigma (\xi )=\sum _{|\alpha |=0}^{m}\ a_{\alpha }\ \xi ^{\alpha }}
de telle sorte que :
(
D
f
^
)
(
ξ
)
=
σ
(
ξ
)
f
^
(
ξ
)
{\displaystyle ({\widehat {{\mathfrak {D}}\,f}})(\xi )\ =\ \sigma (\xi )\ {\hat {f}}(\xi )}
Le symbole principal de l'opérateur différentiel
D
{\displaystyle {\mathfrak {D}}}
d'ordre
m
{\displaystyle m}
à coefficients constants est la fonction des
n
{\displaystyle n}
variables
ξ
{\displaystyle \xi }
:
σ
m
(
ξ
)
=
∑
|
α
|
=
m
a
α
ξ
α
{\displaystyle \sigma _{m}(\xi )=\sum _{|\alpha |=m}\ a_{\alpha }\ \xi ^{\alpha }}
On a vu que plus haut :
(
D
f
)
(
x
)
=
∫
R
n
d
ξ
~
e
+
i
⟨
ξ
,
x
⟩
σ
(
x
,
ξ
)
f
^
(
ξ
)
{\displaystyle ({\mathfrak {D}}\,f)(x)\ =\ \int _{\mathbb {R} ^{n}}\mathrm {d} {\tilde {\xi }}\ \mathrm {e} ^{+\,\mathrm {i} \,\langle \,\xi \,,\,x\,\rangle }\ \sigma (x,\xi )\ {\hat {f}}(\xi )}
Pour un opérateur différentiel dont les coefficients
a
α
(
x
)
{\displaystyle a_{\alpha }(x)}
ne sont pas constants, le symbole
σ
(
x
,
ξ
)
{\displaystyle \sigma (x,\xi )}
dépend des coordonnées d'espace
x
{\displaystyle x}
, et on a :
(
D
f
^
)
(
ξ
)
≠
σ
(
x
,
ξ
)
f
^
(
ξ
)
{\displaystyle ({\widehat {{\mathfrak {D}}\,f}})(\xi )\ \neq \ \sigma (x,\xi )\ {\hat {f}}(\xi )}
Partons de la relation générale :
(
D
f
)
(
x
)
=
∑
|
α
|
=
0
m
a
α
(
x
)
∫
R
n
d
ξ
~
e
+
i
⟨
ξ
,
x
⟩
ξ
α
f
^
(
ξ
)
{\displaystyle ({\mathfrak {D}}\,f)(x)\ =\ \sum _{|\alpha |=0}^{m}\ a_{\alpha }(x)\ \int _{\mathbb {R} ^{n}}\mathrm {d} {\tilde {\xi }}\ \mathrm {e} ^{+\,\mathrm {i} \,\langle \,\xi \,,\,x\,\rangle }\ \xi ^{\alpha }\ {\hat {f}}(\xi )}
Si l'on introduit la transformée de Fourier des coefficients :
a
α
(
x
)
=
∫
R
n
d
η
~
e
+
i
⟨
η
,
x
⟩
a
^
α
(
η
)
{\displaystyle a_{\alpha }(x)\ =\ \int _{\mathbb {R} ^{n}}\mathrm {d} {\tilde {\eta }}\ \mathrm {e} ^{+\,\mathrm {i} \,\langle \,\eta \,,\,x\,\rangle }\ {\hat {a}}_{\alpha }(\eta )}
on obtient :
(
D
f
)
(
x
)
=
∑
|
α
|
=
0
m
∫
R
n
d
η
~
e
+
i
⟨
η
,
x
⟩
a
^
α
(
η
)
×
∫
R
n
d
ξ
~
e
+
i
⟨
ξ
,
x
⟩
ξ
α
f
^
(
ξ
)
{\displaystyle ({\mathfrak {D}}\,f)(x)\ =\ \sum _{|\alpha |=0}^{m}\ \int _{\mathbb {R} ^{n}}\mathrm {d} {\tilde {\eta }}\ \mathrm {e} ^{+\,\mathrm {i} \,\langle \,\eta \,,\,x\,\rangle }\ {\hat {a}}_{\alpha }(\eta )\ \times \ \int _{\mathbb {R} ^{n}}\mathrm {d} {\tilde {\xi }}\ \mathrm {e} ^{+\,\mathrm {i} \,\langle \,\xi \,,\,x\,\rangle }\ \xi ^{\alpha }\ {\hat {f}}(\xi )}
soit :
(
D
f
)
(
x
)
=
∑
|
α
|
=
0
m
∫
R
n
d
η
~
∫
R
n
d
ξ
~
e
+
i
⟨
ξ
,
x
⟩
a
^
α
(
η
)
ξ
α
f
^
(
ξ
)
{\displaystyle ({\mathfrak {D}}\,f)(x)\ =\ \sum _{|\alpha |=0}^{m}\ \int _{\mathbb {R} ^{n}}\mathrm {d} {\tilde {\eta }}\ \int _{\mathbb {R} ^{n}}\mathrm {d} {\tilde {\xi }}\ \mathrm {e} ^{+\,\mathrm {i} \,\langle \,\xi \,,\,x\,\rangle }\ {\hat {a}}_{\alpha }(\eta )\ \xi ^{\alpha }\ {\hat {f}}(\xi )}
A
ξ
{\displaystyle \xi }
fixé, on fait le changement de variable :
η
→
t
=
ξ
+
η
{\displaystyle \eta \to t=\xi +\eta }
, ce qui donne :
(
D
f
)
(
x
)
=
∑
|
α
|
=
0
m
∫
R
n
d
t
~
e
+
i
⟨
ξ
,
x
⟩
∫
R
n
d
ξ
~
a
^
α
(
t
−
ξ
)
ξ
α
f
^
(
ξ
)
{\displaystyle ({\mathfrak {D}}\,f)(x)\ =\ \sum _{|\alpha |=0}^{m}\ \int _{\mathbb {R} ^{n}}\mathrm {d} {\tilde {t}}\ \mathrm {e} ^{+\,\mathrm {i} \,\langle \,\xi \,,\,x\,\rangle }\ \int _{\mathbb {R} ^{n}}\mathrm {d} {\tilde {\xi }}\ {\hat {a}}_{\alpha }(t-\xi )\ \xi ^{\alpha }\ {\hat {f}}(\xi )}
On reconnait le produit de convolution :
(
a
^
α
∗
ξ
α
f
^
)
(
t
)
=
∫
R
n
d
ξ
~
a
^
α
(
t
−
ξ
)
ξ
α
f
^
(
ξ
)
{\displaystyle \left(\,{\hat {a}}_{\alpha }\ *\ \xi ^{\alpha }\ {\hat {f}}\,\right)(t)\ =\ \int _{\mathbb {R} ^{n}}\mathrm {d} {\tilde {\xi }}\ {\hat {a}}_{\alpha }(t-\xi )\ \xi ^{\alpha }\ {\hat {f}}(\xi )}
d'où :
(
D
f
)
(
x
)
=
∑
|
α
|
=
0
m
∫
R
n
d
t
~
e
+
i
⟨
ξ
,
x
⟩
(
a
^
α
∗
ξ
α
f
^
)
(
t
)
{\displaystyle ({\mathfrak {D}}\,f)(x)\ =\ \sum _{|\alpha |=0}^{m}\ \int _{\mathbb {R} ^{n}}\mathrm {d} {\tilde {t}}\ \mathrm {e} ^{+\,\mathrm {i} \,\langle \,\xi \,,\,x\,\rangle }\ \left(\,{\hat {a}}_{\alpha }\ *\ \xi ^{\alpha }\ {\hat {f}}\,\right)(t)}
qu'on peut réécrire :
(
D
f
^
)
(
ξ
)
=
∑
|
α
|
=
0
m
(
a
^
α
∗
ξ
α
f
^
)
(
ξ
)
{\displaystyle ({\widehat {{\mathfrak {D}}\,f}})(\xi )\ =\ \sum _{|\alpha |=0}^{m}\ \left(\,{\hat {a}}_{\alpha }\ *\ \xi ^{\alpha }\ {\hat {f}}\,\right)(\xi )}
(en) Lars Hörmander , The Analysis of Linear Partial Differential Operators , Springer-Verlag , 1983-1985. Traité de référence en quatre volumes, par le récipiendaire de la médaille Fields 1962. Le volume I est sous-titré : Distribution theory and Fourier analysis , et le volume II : Differential operators with constant coefficients . Les volumes III et IV sont consacrés à la théorie moderne via les opérateurs pseudo-différentiels .
(en) Lars Hörmander, Linear Partial Differential Operators , Springer-Verlag, 1963. Ce livre contient les travaux pour lesquels l'auteur a obtenu la médaille Fields en 1962.
(en) Yu. V. Egorov et M. A. Shubin (en) , Foundations of the Classical Theory of Partial Differential Equations , Springer-Verlag, 2e éd., 1998 (ISBN 3-540-63825-3 ) . Premier volume d'une série qui en comporte neuf, écrits pour l'Encyclopaedia of Mathematical Sciences . Les volumes suivants sont consacrés à la théorie moderne via les opérateurs pseudo-différentiels .
(en) Michael E. Taylor (en) , Partial Differential Equations - Basic Theory , coll. « Texts in Applied Mathematics » (no 23), Springer-Verlag, 2e éd., 1999 (ISBN 0-387-94654-3 ) . Premier volume d'une série qui en comporte trois. Les volumes suivants sont consacrés à la théorie moderne via les opérateurs pseudo-différentiels .
classification
quelques types
attributs des variables
relations de processus
solutions
existence et unicité
à propos des solutions
méthodes de solution
exemples
mathématiciens