[p. 12 modifica ]
Esiste per l’energia di un campo elettromagnetico un’equazione analoga a quelle che vanno sotto il nome di principio di Hamilton e principio della minima azione nella dinamica ordinaria: si può dimostrare cioè la relazione
δ
∫
t
0
t
1
W
e
m
d
t
=
0
,
{\displaystyle \delta \textstyle \int _{t_{0}}^{t_{1}}\mathrm {W_{em}} \,dt=0,}
quando si suppongano nulle le variazioni ai limiti.
Per vedere questo si scriva:
[1]
x
1
=
A
ε
∂
X
∂
t
−
∂
M
∂
z
+
∂
N
∂
y
{\displaystyle x_{1}=\mathrm {A} \varepsilon {\frac {\partial \mathrm {X} }{\partial t}}-{\frac {\partial \mathrm {M} }{\partial z}}+{\frac {\partial \mathrm {N} }{\partial y}}}
l
1
=
A
μ
∂
L
∂
t
−
∂
Z
∂
y
+
∂
Y
∂
z
{\displaystyle l_{1}=\mathrm {A} \mu {\dfrac {\partial \mathrm {L} }{\partial t}}-{\dfrac {\partial \mathrm {Z} }{\partial y}}+{\dfrac {\partial \mathrm {Y} }{\partial z}}}
x
2
=
A
ε
∂
Y
∂
t
−
∂
N
∂
x
+
∂
L
∂
z
{\displaystyle x_{2}=\mathrm {A} \varepsilon {\frac {\partial \mathrm {Y} }{\partial t}}-{\frac {\partial \mathrm {N} }{\partial x}}+{\frac {\partial \mathrm {L} }{\partial z}}}
l
2
=
A
μ
∂
M
∂
t
−
∂
X
∂
y
+
∂
Z
∂
x
{\displaystyle l_{2}=\mathrm {A} \mu {\dfrac {\partial \mathrm {M} }{\partial t}}-{\dfrac {\partial \mathrm {X} }{\partial y}}+{\dfrac {\partial \mathrm {Z} }{\partial x}}}
x
3
=
A
ε
∂
Z
∂
t
−
∂
L
∂
y
+
∂
M
∂
x
{\displaystyle x_{3}=\mathrm {A} \varepsilon {\frac {\partial \mathrm {Z} }{\partial t}}-{\frac {\partial \mathrm {L} }{\partial y}}+{\frac {\partial \mathrm {M} }{\partial x}}}
l
3
=
A
μ
∂
N
∂
t
−
∂
Y
∂
x
+
∂
X
∂
x
{\displaystyle l_{3}=\mathrm {A} \mu {\dfrac {\partial \mathrm {N} }{\partial t}}-{\dfrac {\partial \mathrm {Y} }{\partial x}}+{\dfrac {\partial \mathrm {X} }{\partial x}}}
[p. 13 modifica ] sarà, per le equazioni di Hertz:
x
1
=
x
2
=
x
3
=
l
1
=
l
2
=
l
3
=
0
;
{\displaystyle x_{1}=x_{2}=x_{3}=l_{1}=l_{2}=l_{3}=0;}
di più si introducano sei funzioni
X
.
Y
.
Z
.
L
.
M
.
N
{\displaystyle X.Y.Z.L.M.N}
delle coordinate e del tempo, che ci riserviamo di determinare.
Facciamo le variazioni di queste funzioni: moltiplichiamo la prima delle [1] per
δ
X
{\displaystyle \delta \,X}
, la seconda per
δ
Y
{\displaystyle \delta \,Y}
...... la sesta per
δ
N
{\displaystyle \delta \,N}
; sommiamo membro a membro, moltiplichiamo per
d
t
{\displaystyle dt}
e integriamo fra
t
0
{\displaystyle t_{0}}
e
t
1
{\displaystyle t_{1}}
otterremo:
∫
t
0
t
1
d
t
∭
d
v
[
A
ε
∂
X
∂
t
δ
X
+
A
ε
∂
Y
∂
t
δ
Y
+
⋯
+
A
μ
∂
N
∂
t
δ
N
{\displaystyle \int _{t_{0}}^{t_{1}}\,dt\iiint \,dv\left[\mathrm {A} \,\varepsilon {\frac {\partial \mathrm {X} }{\partial t}}\delta \,X+\mathrm {A} \,\varepsilon {\frac {\partial \mathrm {Y} }{\partial t}}\delta \,Y+\cdots +\mathrm {A} \,\mu {\frac {\partial \mathrm {N} }{\partial t}}\delta \,N\right.}
−
∂
M
∂
z
δ
X
−
∂
N
∂
x
δ
Y
−
⋅
⋅
⋅
⋅
⋅
−
∂
Y
∂
x
δ
N
{\displaystyle -{\frac {\partial \mathrm {M} }{\partial z}}\delta \,X-{\frac {\partial \mathrm {N} }{\partial x}}\delta \,Y-\quad \cdot \quad \cdot \quad \cdot \quad \cdot \quad \cdot \quad -{\frac {\partial \mathrm {Y} }{\partial x}}\delta \,N}
+
∂
N
∂
y
δ
X
+
∂
L
∂
z
δ
Y
+
⋅
⋅
⋅
⋅
⋅
+
∂
X
∂
y
δ
N
]
=
0.
{\displaystyle \left.+{\frac {\partial \mathrm {N} }{\partial y}}\delta \,X+{\frac {\partial \mathrm {L} }{\partial z}}\delta \,Y+\quad \cdot \quad \cdot \quad \cdot \quad \cdot \quad \cdot \quad +{\frac {\partial \mathrm {X} }{\partial y}}\delta \,N\right]=0.}
Ora è evidente che si può scrivere:
A
ε
∂
X
∂
t
δ
X
=
A
ε
∂
∂
t
(
X
δ
X
)
−
A
ε
X
∂
∂
t
(
δ
X
)
{\displaystyle \mathrm {A} \,\varepsilon {\frac {\partial \mathrm {X} }{\partial t}}\delta \,X=\mathrm {A} \,\varepsilon {\frac {\partial }{\partial t}}\left(\mathrm {X} \,\delta \,X\right)-\mathrm {A} \,\varepsilon \,\mathrm {X} {\frac {\partial }{\partial t}}\left(\delta \,X\right)}
e che 5 eguaglianze analoghe si potrebbero formare considerando i termini in
∂
Y
∂
t
⋯
∂
N
∂
t
{\displaystyle {\frac {\partial \mathrm {Y} }{\partial t}}\cdots {\frac {\partial \mathrm {N} }{\partial t}}}
, si ha dunque:
t
0
t
1
[
∭
d
v
(
A
ε
X
δ
X
+
A
ε
Y
δ
Y
+
⋯
+
A
μ
N
δ
N
)
]
{\displaystyle \textstyle _{t_{0}}^{t_{1}}\left[\iiint \,dv(\mathrm {A} \,\varepsilon \,\mathrm {X} \,\delta \,X+\mathrm {A} \,\varepsilon \,\mathrm {Y} \,\delta \,Y+\cdots +\mathrm {A} \,\mu \,\mathrm {N} \,\delta \,N)\right]}
−
∫
t
0
t
1
d
t
∭
d
v
[
A
ε
X
δ
⋅
∂
X
∂
t
+
A
ε
Y
δ
⋅
∂
Y
∂
t
+
⋯
+
A
μ
N
δ
⋅
∂
Z
∂
t
{\displaystyle -\int _{t_{0}}^{t_{1}}\,dt\iiint dv\left[\mathrm {A} \,\varepsilon \,\mathrm {X} \,\delta \cdot {\frac {\partial X}{\partial t}}+\mathrm {A} \,\varepsilon \,\mathrm {Y} \,\delta \cdot {\frac {\partial Y}{\partial t}}+\cdots +\mathrm {A} \,\mu \,\mathrm {N} \,\delta \cdot {\frac {\partial Z}{\partial t}}\right.}
+
∂
M
∂
z
δ
X
+
∂
N
∂
x
δ
Y
+
⋅
⋅
⋅
⋅
+
∂
Y
∂
x
δ
N
{\displaystyle +{\frac {\partial \mathrm {M} }{\partial z}}\delta \,X+{\frac {\partial \mathrm {N} }{\partial x}}\delta \,Y+\quad \cdot \quad \cdot \quad \cdot \quad \cdot \quad +{\frac {\partial \mathrm {Y} }{\partial x}}\delta \,N}
−
∂
N
∂
y
δ
X
−
∂
L
∂
z
δ
Y
−
⋅
⋅
⋅
⋅
−
∂
X
∂
y
δ
N
]
=
0.
{\displaystyle \left.-{\frac {\partial \mathrm {N} }{\partial y}}\delta \,X-{\frac {\partial \mathrm {L} }{\partial z}}\delta \,Y-\quad \cdot \quad \cdot \quad \cdot \quad \cdot \quad -{\frac {\partial \mathrm {X} }{\partial y}}\delta \,N\right]=0.}
Ora si osservi, in modo simile a quanto si è fatto più su, che:
∂
M
∂
z
δ
X
=
−
M
δ
⋅
∂
X
∂
z
+
∂
∂
z
(
M
δ
X
)
{\displaystyle {\frac {\partial \mathrm {M} }{\partial z}}\delta \,X=-\mathrm {M} \,\delta \cdot {\frac {\partial X}{\partial z}}+{\frac {\partial }{\partial z}}\left(\mathrm {M} \,\delta \,X\right)}
11 equazioni analoghe si potrebbero scrivere, sicchè, sostituendo, si ottiene:
t
0
t
1
[
∭
d
v
(
A
ε
X
δ
X
+
A
ε
Y
δ
Y
+
⋯
+
A
μ
N
δ
N
)
{\displaystyle _{t_{0}}^{t_{1}}\textstyle [\iiint \,dv(\mathrm {A} \,\varepsilon \,\mathrm {X} \,\delta \,X+\mathrm {A} \,\varepsilon \,\mathrm {Y} \,\delta \,Y+\cdots +\mathrm {A} \,\mu \,\mathrm {N} \,\delta \,N)}
−
∫
t
0
t
1
d
t
∭
d
v
[
A
ε
X
δ
⋅
∂
X
∂
t
+
A
ε
Y
δ
⋅
∂
Y
∂
t
+
⋯
+
A
μ
N
δ
⋅
∂
Z
∂
t
{\displaystyle -\int _{t_{0}}^{t_{1}}\,dt\iiint dv\left[\mathrm {A} \,\varepsilon \,\mathrm {X} \,\delta \cdot {\frac {\partial X}{\partial t}}+\mathrm {A} \,\varepsilon \,\mathrm {Y} \,\delta \cdot {\frac {\partial Y}{\partial t}}+\cdots +\mathrm {A} \,\mu \,\mathrm {N} \,\delta \cdot {\frac {\partial Z}{\partial t}}\right.}
[p. 14 modifica ]
−
M
δ
⋅
∂
X
∂
z
−
N
δ
⋅
∂
Y
∂
x
−
⋯
−
Y
δ
⋅
∂
N
∂
x
{\displaystyle -\mathrm {M} \,\delta \cdot {\frac {\partial X}{\partial z}}-\mathrm {N} \,\delta \cdot {\frac {\partial Y}{\partial x}}-\cdots -\mathrm {Y} \,\delta \cdot {\frac {\partial N}{\partial x}}}
+
∂
∂
z
(
M
δ
X
)
+
∂
∂
x
(
N
δ
Y
)
+
⋯
+
∂
∂
x
(
Y
δ
N
)
{\displaystyle +{\frac {\partial }{\partial z}}\left(\mathrm {M} \,\delta \,X\right)+{\frac {\partial }{\partial x}}\left(\mathrm {N} \,\delta \,Y\right)+\cdots +{\frac {\partial }{\partial x}}\left(\mathrm {Y} \,\delta \,N\right)}
+
N
δ
⋅
∂
X
∂
y
+
L
δ
⋅
∂
Y
∂
z
+
⋯
+
X
δ
⋅
∂
N
∂
y
{\displaystyle +\mathrm {N} \,\delta \cdot {\frac {\partial X}{\partial y}}+\mathrm {L} \,\delta \cdot {\frac {\partial Y}{\partial z}}+\cdots +\mathrm {X} \,\delta \cdot {\frac {\partial N}{\partial y}}}
−
∂
∂
y
(
N
δ
X
)
−
∂
∂
z
(
L
δ
Y
)
−
⋯
−
∂
∂
y
(
X
δ
N
)
]
=
0.
{\displaystyle \left.-{\frac {\partial }{\partial y}}\left(\mathrm {N} \,\delta \,X\right)-{\frac {\partial }{\partial z}}\left(\mathrm {L} \,\delta \,Y\right)-\cdots -{\frac {\partial }{\partial y}}\left(\mathrm {X} \,\delta \,N\right)\right]=0.}
Se si ammette che
X
.
Y
.
.
.
N
{\displaystyle X.Y...N}
siano nulle sulla superficie che racchiude il campo si possono cancellare i 12 termini dell’integrale che sono derivate parziali rispetto alle coordinate.
Ordinando ciò che resta rispetto ad
X
.
Y
.
.
.
N
{\displaystyle X.Y...N}
si trova:
t
0
t
1
[
∭
d
v
(
A
ε
X
δ
X
+
A
ε
Y
δ
Y
+
⋯
+
A
μ
Z
δ
Z
)
]
{\displaystyle _{t_{0}}^{t_{1}}\textstyle [\iiint \,dv(\mathrm {A} \,\varepsilon \,\mathrm {X} \,\delta \,X+\mathrm {A} \,\varepsilon \,\mathrm {Y} \,\delta \,Y+\cdots +\mathrm {A} \,\mu \,\mathrm {Z} \,\delta \,Z)]}
−
∫
t
0
t
1
d
t
∭
d
v
[
X
δ
⋅
(
A
ε
∂
X
∂
t
−
∂
M
∂
z
+
∂
N
∂
y
)
{\displaystyle -\int _{t_{0}}^{t_{1}}\,dt\iiint dv\left[\mathrm {X} \,\delta \cdot \left(\mathrm {A} \,\varepsilon {\frac {\partial X}{\partial t}}-{\frac {\partial M}{\partial z}}+{\frac {\partial N}{\partial y}}\right)\right.}
+
Y
δ
⋅
(
A
ε
∂
Y
∂
t
−
∂
N
∂
x
+
∂
L
∂
z
)
{\displaystyle +\mathrm {Y} \,\delta \cdot \left(\mathrm {A} \,\varepsilon {\frac {\partial Y}{\partial t}}-{\frac {\partial N}{\partial x}}+{\frac {\partial L}{\partial z}}\right)}
(*)
+
⋯
⋯
{\displaystyle +\cdots \cdots }
+
N
δ
⋅
(
A
μ
∂
L
∂
t
−
∂
Y
∂
x
+
∂
X
∂
z
)
]
=
0.
{\displaystyle \left.+\mathrm {N} \,\delta \cdot \left(\mathrm {A} \,\mu {\frac {\partial L}{\partial t}}-{\frac {\partial Y}{\partial x}}+{\frac {\partial X}{\partial z}}\right)\right]=0.}
Le
X
.
Y
.
.
.
N
{\displaystyle X.Y...N}
sono ancora in nostro arbitrio, le assoggetteremo alle equazioni seguenti.
A
ε
∂
X
∂
t
−
∂
M
∂
z
+
∂
N
∂
y
=
−
ε
4
π
X
{\displaystyle \mathrm {A} \,\varepsilon {\frac {\partial X}{\partial t}}-{\frac {\partial M}{\partial z}}+{\frac {\partial N}{\partial y}}=-{\frac {\varepsilon }{4\pi }}\mathrm {X} }
A
ε
∂
Y
∂
t
−
∂
N
∂
x
+
∂
L
∂
z
=
−
ε
4
π
X
{\displaystyle \mathrm {A} \,\varepsilon {\frac {\partial Y}{\partial t}}-{\frac {\partial N}{\partial x}}+{\frac {\partial L}{\partial z}}=-{\frac {\varepsilon }{4\pi }}\mathrm {X} }
⋅
⋅
⋅
⋅
⋅
⋅
⋅
⋅
⋅
⋅
⋅
{\displaystyle \cdot \quad \cdot \quad \cdot \quad \cdot \quad \cdot \quad \cdot \quad \cdot \quad \cdot \quad \cdot \quad \cdot \quad \cdot }
A
μ
∂
N
∂
t
−
∂
Y
∂
x
+
∂
X
∂
y
=
−
μ
4
π
Z
.
{\displaystyle \mathrm {A} \,\mu {\frac {\partial N}{\partial t}}-{\frac {\partial Y}{\partial x}}+{\frac {\partial X}{\partial y}}=-{\frac {\mu }{4\pi }}\mathrm {Z} .}
Ciò posto l'equazione (*) diventa:
t
0
t
1
[
∭
d
v
(
A
ε
X
δ
X
+
A
ε
Y
δ
Y
+
⋯
+
A
μ
Z
δ
Z
)
]
{\displaystyle _{t_{0}}^{t_{1}}\textstyle [\iiint \,dv(\mathrm {A} \,\varepsilon \,\mathrm {X} \,\delta \,X+\mathrm {A} \,\varepsilon \,\mathrm {Y} \,\delta \,Y+\cdots +\mathrm {A} \,\mu \,\mathrm {Z} \,\delta \,Z)]}
+
δ
∫
t
0
t
1
d
t
∭
d
v
[
ε
8
π
(
X
2
+
Y
2
+
Z
2
)
+
μ
8
π
(
L
2
+
M
2
+
N
2
)
]
=
0
;
{\displaystyle +\delta \int _{t_{0}}^{t_{1}}\,dt\iiint dv\left[{\frac {\varepsilon }{8\pi }}\left(\mathrm {X^{2}+Y^{2}+Z^{2}} \right)+{\frac {\mu }{8\pi }}\left(\mathrm {L^{2}+M^{2}+N^{2}} \right)\right]=0;}
se le variazioni ai limiti sono nulle la parte integrata rispetto al tempo si annulla da sè, quindi bisogna che sia
[p. 15 modifica ]
δ
∫
t
0
t
1
d
t
∭
d
v
[
ε
8
π
(
X
2
+
Y
2
+
Z
2
)
+
μ
8
π
(
L
2
+
M
2
+
N
2
)
]
=
0
;
{\displaystyle \delta \int _{t_{0}}^{t_{1}}\,dt\iiint dv\left[{\frac {\varepsilon }{8\pi }}\left(\mathrm {X^{2}+Y^{2}+Z^{2}} \right)+{\frac {\mu }{8\pi }}\left(\mathrm {L^{2}+M^{2}+N^{2}} \right)\right]=0;}
vale a dire
δ
∫
t
0
t
1
W
e
m
d
t
=
0
,
{\displaystyle \delta \,\int _{t_{0}}^{t_{1}}\mathrm {W_{\mathrm {em} }} \,dt=0,}
ciò che si voleva dimostrare.
Questo teorema si deve al prof. Vito Volterra 1 .