La demostración se realiza a partir del principio de Fermat .
Tomamos dos trayectorias distintas separadas infinitesimalmente,
[
A
B
]
{\displaystyle [AB]}
y
[
A
B
′
]
{\displaystyle [AB']}
, donde
A
{\displaystyle A}
es el foco y
B
{\displaystyle B}
y
B
′
{\displaystyle B'}
son los puntos de llegada separados por caminos ópticos iguales. Entonces definimos los respectivos caminos ópticos como:
L
A
B
=
∫
A
B
n
(
r
)
d
s
{\displaystyle L_{AB}=\int _{A}^{B}n({\textbf {r}})ds}
L
A
B
′
=
∫
A
B
′
n
(
r
′
)
d
s
′
{\displaystyle L_{AB'}=\int _{A}^{B'}n({\textbf {r}}')ds'}
Siendo
r
{\displaystyle {\textbf {r}}}
y
r
′
{\displaystyle {\textbf {r}}'}
los respectivos vectores de posición,
d
s
{\displaystyle ds}
y
d
s
′
{\displaystyle ds'}
los respectivos diferenciales de espacio y
n
(
r
)
{\displaystyle n({\textbf {r}})}
el índice de refracción .
Admitimos que el índice de refracción es derivable.
Emplearemos:
Desarrollo en serie de Taylor de primer orden:
f
(
r
)
=
f
(
r
0
)
+
∇
→
f
(
r
0
)
⋅
(
r
−
r
0
)
{\displaystyle f({\textbf {r}})=f({\textbf {r}}_{0})+{\vec {\nabla }}f({\textbf {r}}_{0})\cdot ({\textbf {r}}-{\textbf {r}}_{0})}
Ecuación de la trayectoria de un rayo luminoso (deducida a partir del principio de Fermat ):
d
d
s
(
n
u
)
=
∇
→
n
{\displaystyle {d \over ds}(n{\textbf {u}})={\vec {\nabla }}n}
de lo que
(
n
u
)
=
∇
→
n
d
s
{\displaystyle (n{\textbf {u}})={\vec {\nabla }}nds}
.
La relación
r
′
=
r
+
ε
b
{\displaystyle {\textbf {r}}'={\textbf {r}}+\varepsilon {\textbf {b}}}
de la cual
dr
′
=
dr
+
ε
db
{\displaystyle {\textbf {dr}}'={\textbf {dr}}+\varepsilon {\textbf {db}}}
.
Admitimos que el índice de refracción admite un desarrollo en serie de Taylor de orden 1. Entonces se obtiene que
n
(
r
′
)
=
n
(
r
)
+
∇
→
n
(
r
)
⋅
(
r
′
−
r
)
=
n
(
r
)
+
∇
→
n
(
r
)
⋅
ε
b
{\displaystyle n({\textbf {r}}')=n({\textbf {r}})+{\vec {\nabla }}n({\textbf {r}})\cdot ({\textbf {r}}'-{\textbf {r}})=n({\textbf {r}})+{\vec {\nabla }}n({\textbf {r}})\cdot \varepsilon {\textbf {b}}}
.
Por otro haremos lo mismo con el módulo del vector posición:
∣
r
′
∣=∣
r
∣
+
∇
→
∣
r
∣
⋅
(
r
′
−
r
)
=∣
r
∣
+
∂
∂
r
r
u
^
r
⋅
(
r
′
−
r
)
=∣
r
∣
+
u
⋅
(
r
′
−
r
)
=∣
r
∣
+
u
⋅
r
′
−
u
⋅
r
=∣
r
∣
+
u
⋅
r
′
−
∣
r
∣=
u
⋅
r
′
{\displaystyle \mid {\textbf {r}}'\mid =\mid {\textbf {r}}\mid +{\vec {\nabla }}\mid {\textbf {r}}\mid \cdot ({\textbf {r}}'-{\textbf {r}})=\mid {\textbf {r}}\mid +{{\partial } \over {\partial r}}r{\hat {u}}_{r}\cdot ({\textbf {r}}'-{\textbf {r}})=\mid {\textbf {r}}\mid +{\textbf {u}}\cdot ({\textbf {r}}'-{\textbf {r}})=\mid {\textbf {r}}\mid +{\textbf {u}}\cdot {\textbf {r}}'-{\textbf {u}}\cdot {\textbf {r}}=\mid {\textbf {r}}\mid +{\textbf {u}}\cdot {\textbf {r}}'-\mid {\textbf {r}}\mid ={\textbf {u}}\cdot {\textbf {r}}'}
.
De modo que:
d
s
′
=∣
dr'
∣=
u
⋅
dr
′
=
u
⋅
dr
+
u
⋅
ε
db
=
d
s
+
ε
u
⋅
db
{\displaystyle ds'=\mid {\textbf {dr'}}\mid ={\textbf {u}}\cdot {\textbf {dr}}'={\textbf {u}}\cdot {\textbf {dr}}+{\textbf {u}}\cdot \varepsilon {\textbf {db}}=ds+\varepsilon {\textbf {u}}\cdot {\textbf {db}}}
.
Se reemplaza en el camino óptico:
L
B
′
=
∫
A
B
′
n
(
r
′
)
d
s
′
=
∫
A
B
′
[
n
(
r
)
+
∇
→
n
⋅
ε
b
]
(
d
s
+
ε
u
⋅
db
)
{\displaystyle L_{B'}=\int _{A}^{B'}n({\textbf {r}}')ds'=\int _{A}^{B'}[n({\textbf {r}})+{\vec {\nabla }}n\cdot \varepsilon {\textbf {b}}](ds+\varepsilon {\textbf {u}}\cdot {\textbf {db}})}
L
B
′
=
∫
A
B
′
n
(
r
)
d
s
+
∫
A
B
′
∇
→
n
⋅
ε
db
+
∫
A
B
′
n
(
r
)
ε
u
⋅
db
+
∫
A
B
′
∇
→
n
⋅
ε
b
ε
u
⋅
db
{\displaystyle L_{B'}=\int _{A}^{B'}n({\textbf {r}})ds+\int _{A}^{B'}{\vec {\nabla }}n\cdot \varepsilon {\textbf {db}}+\int _{A}^{B'}n({\textbf {r}})\varepsilon {\textbf {u}}\cdot {\textbf {db}}+\int _{A}^{B'}{\vec {\nabla }}n\cdot \varepsilon {\textbf {b}}\varepsilon {\textbf {u}}\cdot {\textbf {db}}}
Eliminamos los elemento de orden de
ε
2
{\displaystyle \varepsilon ^{2}}
:
L
B
′
=
∫
A
B
′
n
(
r
)
d
s
+
∫
A
B
′
∇
→
n
⋅
ε
db
+
∫
A
B
′
n
(
r
)
ε
u
⋅
db
{\displaystyle L_{B'}=\int _{A}^{B'}n({\textbf {r}})ds+\int _{A}^{B'}{\vec {\nabla }}n\cdot \varepsilon {\textbf {db}}+\int _{A}^{B'}n({\textbf {r}})\varepsilon {\textbf {u}}\cdot {\textbf {db}}}
Calculamos
Δ
L
{\displaystyle \Delta L}
:
Δ
L
=
∫
A
B
′
n
(
r
)
d
s
+
∫
A
B
′
∇
→
n
⋅
ε
b
d
s
+
∫
A
B
′
n
(
r
)
ε
u
⋅
db
−
∫
A
B
n
(
r
)
d
s
{\displaystyle \Delta L=\int _{A}^{B'}n({\textbf {r}})ds+\int _{A}^{B'}{\vec {\nabla }}n\cdot \varepsilon {\textbf {b}}ds+\int _{A}^{B'}n({\textbf {r}})\varepsilon {\textbf {u}}\cdot {\textbf {db}}-\int _{A}^{B}n({\textbf {r}})ds}
Factorizamos los términos por potencias de
ε
{\displaystyle \varepsilon }
:
Δ
L
=
∫
A
B
′
n
(
r
)
d
s
−
∫
A
B
n
(
r
)
d
s
+
ε
[
∫
A
B
′
∇
→
n
⋅
b
d
s
+
∫
A
B
′
n
(
r
)
u
⋅
db
]
{\displaystyle \Delta L=\int _{A}^{B'}n({\textbf {r}})ds-\int _{A}^{B}n({\textbf {r}})ds+\varepsilon \left[\int _{A}^{B'}{\vec {\nabla }}n\cdot {\textbf {b}}ds+\int _{A}^{B'}n({\textbf {r}}){\textbf {u}}\cdot {\textbf {db}}\right]}
Por la ecuación de las trayectorias tenemos que
d
(
n
u
⋅
b
)
=
d
(
n
u
)
⋅
b
+
n
u
⋅
db
=
∇
→
n
⋅
b
d
s
+
n
u
⋅
db
{\displaystyle d(n{\textbf {u}}\cdot {\textbf {b}})=d(n{\textbf {u}})\cdot {\textbf {b}}+n{\textbf {u}}\cdot {\textbf {db}}={\vec {\nabla }}n\cdot {\textbf {b}}ds+n{\textbf {u}}\cdot {\textbf {db}}}
de modo que:
Δ
L
=
∫
A
B
′
n
(
r
)
d
s
−
∫
A
B
n
(
r
)
d
s
+
ε
[
∫
A
B
′
d
(
n
b
⋅
u
)
]
{\displaystyle \Delta L=\int _{A}^{B'}n({\textbf {r}})ds-\int _{A}^{B}n({\textbf {r}})ds+\varepsilon \left[\int _{A}^{B'}d(n{\textbf {b}}\cdot {\textbf {u}})\right]}
Si suponemos que hemos escogido
B
≡
B
′
{\displaystyle B\equiv B'}
entonces:
Δ
L
=
∫
A
B
′
n
(
r
)
d
s
−
∫
A
B
′
n
(
r
)
d
s
+
ε
[
∫
A
B
′
d
(
n
b
⋅
u
)
]
=
0
{\displaystyle \Delta L=\int _{A}^{B'}n({\textbf {r}})ds-\int _{A}^{B'}n({\textbf {r}})ds+\varepsilon \left[\int _{A}^{B'}d(n{\textbf {b}}\cdot {\textbf {u}})\right]=0}
Δ
L
=
+
ε
[
∫
A
B
′
d
(
n
b
⋅
u
)
]
=
0
{\displaystyle \Delta L=+\varepsilon \left[\int _{A}^{B'}d(n{\textbf {b}}\cdot {\textbf {u}})\right]=0}
Por lo que:
∫
A
B
′
d
(
n
b
⋅
u
)
=
0
{\displaystyle \int _{A}^{B'}d(n{\textbf {b}}\cdot {\textbf {u}})=0}
n
(
r
B
′
)
b
(
r
B
′
)
⋅
u
(
r
B
′
)
−
n
(
r
A
)
b
(
r
A
)
⋅
u
(
r
A
)
=
0
{\displaystyle n({\textbf {r}}_{B'}){\textbf {b}}({\textbf {r}}_{B'})\cdot {\textbf {u}}({\textbf {r}}_{B'})-n({\textbf {r}}_{A}){\textbf {b}}({\textbf {r}}_{A})\cdot {\textbf {u}}({\textbf {r}}_{A})=0}
Como el punto A es el foco la separación es siempre nula, en consecuencia:
n
(
r
B
′
)
b
(
r
B
′
)
⋅
u
(
r
B
′
)
=
0
{\displaystyle n({\textbf {r}}_{B'}){\textbf {b}}({\textbf {r}}_{B'})\cdot {\textbf {u}}({\textbf {r}}_{B'})=0}
b
(
r
B
′
)
⋅
u
(
r
B
′
)
=
0
{\displaystyle {\textbf {b}}({\textbf {r}}_{B'})\cdot {\textbf {u}}({\textbf {r}}_{B'})=0}
b
(
r
B
′
)
⊥
u
(
r
B
′
)
{\displaystyle {\textbf {b}}({\textbf {r}}_{B'})\perp {\textbf {u}}({\textbf {r}}_{B'})}
Como B' es un punto arbitrario se tiene que
b
(
r
)
⊥
u
(
r
)
{\displaystyle {\textbf {b}}({\textbf {r}})\perp {\textbf {u}}({\textbf {r}})}
.
Teníamos que
r
′
=
r
+
ε
b
{\displaystyle {\textbf {r}}'={\textbf {r}}+\varepsilon {\textbf {b}}}
, por lo que
ε
b
(
r
)
{\displaystyle \varepsilon {\textbf {b}}({\textbf {r}})}
une los puntos de la superficie. De lo que la superficie formada es ortogonal a cada rayo. Podemos justificar que los puntos forman una superficie por continuidad.