δ
S
=
δ
(
∫
a
b
(
−
m
c
d
s
−
q
A
~
⋅
d
x
~
)
)
=
∫
a
b
(
−
m
c
δ
(
d
s
)
−
q
δ
(
A
~
⋅
d
x
~
)
)
{\displaystyle \delta S=\delta \left(\int _{a}^{b}(-mc~ds-q{\tilde {A}}\cdot d{\tilde {x}})\right)=\int _{a}^{b}(-mc~\delta (ds)-q~\delta ({\tilde {A}}\cdot d{\tilde {x}}))}
Avec
δ
(
d
s
)
=
δ
(
d
x
~
⋅
d
x
~
)
=
δ
(
d
x
μ
d
x
μ
)
=
d
x
μ
δ
(
d
x
μ
)
d
s
=
u
μ
d
(
δ
x
μ
)
{\displaystyle \delta (ds)=\delta ({\sqrt {d{\tilde {x}}\cdot d{\tilde {x}}}})=\delta ({\sqrt {dx^{\mu }dx_{\mu }}})={\frac {dx_{\mu }\delta (dx^{\mu })}{ds}}=u_{\mu }d(\delta x^{\mu })}
et
δ
(
A
~
⋅
d
x
~
)
=
δ
(
A
μ
)
d
x
μ
+
A
μ
d
(
δ
x
μ
)
{\displaystyle \delta ({\tilde {A}}\cdot d{\tilde {x}})=\delta (A_{\mu })dx^{\mu }+A_{\mu }d(\delta x^{\mu })}
On a
δ
S
=
∫
a
b
(
−
m
c
u
μ
−
q
A
μ
)
d
(
δ
x
μ
)
+
∫
a
b
−
q
δ
(
A
μ
)
d
x
μ
{\displaystyle \delta S=\int _{a}^{b}(-mc~u_{\mu }-qA_{\mu })d(\delta x^{\mu })+\int _{a}^{b}-q\delta (A_{\mu })dx^{\mu }}
La première intégrale s'intègre par parties (comme dans la démonstration générale des équations d'Euler-Lagrange) :
∫
a
b
(
−
m
c
u
μ
−
q
A
μ
)
d
(
δ
x
μ
)
=
∫
a
b
(
m
c
d
u
μ
+
q
d
A
μ
)
δ
x
μ
−
[
(
m
c
u
μ
+
q
A
μ
)
δ
x
μ
]
a
b
{\displaystyle \int _{a}^{b}(-mc~u_{\mu }-qA_{\mu })d(\delta x^{\mu })=\int _{a}^{b}(mc~du_{\mu }+q~dA_{\mu })\delta x^{\mu }-\left[(mc~u_{\mu }+qA_{\mu })\delta x^{\mu }\right]_{a}^{b}}
et, comme
δ
x
μ
(
a
)
=
δ
x
μ
(
b
)
=
0
{\displaystyle \delta x^{\mu }(a)=\delta x^{\mu }(b)=0~}
, le terme intégré est nul.
On a ainsi :
δ
S
=
∫
a
b
(
m
c
d
u
μ
+
q
d
A
μ
)
δ
x
μ
−
q
δ
(
A
μ
)
d
x
μ
=
∫
a
b
(
m
c
d
u
μ
+
q
∂
A
μ
∂
x
ν
d
x
ν
)
δ
x
μ
−
q
∂
A
μ
∂
x
ν
δ
x
ν
d
x
μ
=
∫
a
b
[
m
c
d
u
μ
+
q
(
∂
A
μ
∂
x
ν
−
∂
A
ν
∂
x
μ
)
d
x
ν
]
δ
x
μ
{\displaystyle {\begin{aligned}\delta S&=\int _{a}^{b}(mc~du_{\mu }+q~dA_{\mu })\delta x^{\mu }-q~\delta (A_{\mu })dx^{\mu }\\&=\int _{a}^{b}(mc~du_{\mu }+q~{\frac {\partial A_{\mu }}{\partial x^{\nu }}}dx^{\nu })\delta x^{\mu }-q~{\frac {\partial A_{\mu }}{\partial x^{\nu }}}\delta x^{\nu }dx^{\mu }\\&=\int _{a}^{b}\left[mc~du_{\mu }+q\left({\frac {\partial A_{\mu }}{\partial x^{\nu }}}-{\frac {\partial A_{\nu }}{\partial x^{\mu }}}\right)dx^{\nu }\right]\delta x^{\mu }\\\end{aligned}}}
car en échangeant les indices muets,
∂
A
μ
∂
x
ν
δ
x
ν
d
x
μ
=
∂
A
ν
∂
x
μ
δ
x
μ
d
x
ν
{\displaystyle {\frac {\partial A_{\mu }}{\partial x^{\nu }}}\delta x^{\nu }dx^{\mu }={\frac {\partial A_{\nu }}{\partial x^{\mu }}}\delta x^{\mu }dx^{\nu }}
.
On obtient la forme finale de
δ
S
{\displaystyle \delta S~}
est factorisant
d
s
{\displaystyle ds~}
:
δ
S
=
∫
a
b
[
m
c
a
μ
+
q
(
∂
A
μ
∂
x
ν
−
∂
A
ν
∂
x
μ
)
u
ν
]
δ
x
μ
d
s
{\displaystyle \delta S=\int _{a}^{b}\left[mc~a_{\mu }+q\left({\frac {\partial A_{\mu }}{\partial x^{\nu }}}-{\frac {\partial A_{\nu }}{\partial x^{\mu }}}\right)u^{\nu }\right]\delta x^{\mu }~ds}