LECTURE NOTES 11
POTENTIALS & FIELDSThe Potential Formulation: Scalar and Vector PotentialsV r t
( )
, and A r t( )
,How do the instantaneous sources ρtot
( )
r t, and Jtot( )
r t, (total electric charge density andtotal electric current density) generate the electric and magnetic fields E r t
( )
, and B r t( )
, ? We seek general solutions to the full Maxwell Equations:(1) Gauss’ Law:
( )
, 1 tot( )
,o
E r t ρ r t
ε
∇i =
(2) No Magnetic Charges: ∇iB r t
( )
, =0(3) Faraday’s Law: E r t
( )
, B r t( )
,t
∂
∇× = −
∂
(4) Ampere’s Law: B r t
( )
, o totJ( )
r t, o o E r t( )
,t
μ μ ε ∂
∇× = +
∂
If ρtot
( )
r t, and Jtot( )
r t, are given (i.e. specified), then what are the corresponding fields( )
, and( )
,E r t B r t ?
⇒ For the static case {i.e. ρ, , ,J E B≠ fcns t
( )
}: Coulomb’s Law( )
1( )
2 ˆ4
tot v o
r
E r ρ dτ
πε ′
′ ′
=
∫
rr
and the Biot-Savart Law
( )
( )
2 ˆ4
tot o
v
J r
B r μ dτ
π ′
′ × ′
=
∫
rr provide the answers {n.b. r ≡ −r r′ }. ⇒ However, we’re interested in the generalization of Coulomb’s Law and the Biot-Savart Law
for time-dependent situations…
This is not an easy problem!! The approach we will adopt here is to represent the
( )
, and( )
,E r t B r t fields in terms of their corresponding potentials V r t
( )
, and A r t( )
, .In electrostatics, ∇×E r t
( )
, =0 enabled us to write: E r t( )
, = −∇V r t( )
, .In electrodynamics, we cannot do this because: ∇×E r t
( )
, ≠0. { E r t( )
, B r t( )
,t
∂
∇× = −
∂ }
However, in electrodynamics ∇iB r t
( )
, =0 (still). ⇒ B r t( )
, = ∇×A r t( )
,Insert B r t
( )
, = ∇×A r t( )
, into Faraday’s Law: E r t( )
, B r t( )
, A r t( )
,t t
⎛ ⎞
∂ ∂
∇× = − = −∇×⎜⎜ ⎟⎟
∂ ⎝ ∂ ⎠
Then we see that: E r t
( )
, A r t( )
, 0t
⎡ ∂ ⎤
∇×⎢ + ⎥=
∂
⎢ ⎥
⎣ ⎦ i.e. the curl of the quantity
( )
( )
,, A r t
E r t
t
⎡ ∂ ⎤
+
⎢ ∂ ⎥
⎢ ⎥
⎣ ⎦
does vanish for any/all
( )
r t, , so ∴ we can define: E r t( )
, A r t( )
, V r t( )
,t
⎡ ∂ ⎤
+ ≡ −∇
⎢ ∂ ⎥
⎢ ⎥
⎣ ⎦
or: E r t
( )
, V r t( )
, A r t( )
,t
∂
= −∇ −
∂ ⇐n.b. reduces to E r
( )
= −∇V r( )
for the electrostatics case.Since we used ∇iB=0 and ∇× = −∂ ∂E B t in deriving the above relation, Maxwell’s equations (2) ∇iB=0 and (3) Faraday’s Law: ∇× = −∂ ∂E B t are automatically satisfied. What about equation (1) Gauss’ Law: ∇iE= −ρ εtot o and (4) Ampere’s Law: ∇× =B μo totJ +ε μo o∂ ∂E t?? (1) Gauss’ Law:
( )
, 1 tot( )
, oE r t ρ r t
ε
∇i = with: E r t
( )
, V r t( )
, A r t( )
,t
∂
= −∇ −
∂
=
( )
,( )
, 2( )
,(
( )
,)
1( )
,tot o
A r t
V r t V r t A r t r t
t t ε ρ
⎡ ∂ ⎤ ∂
∇ −∇⎢ − ⎥= −∇ − ∇ =
∂ ∂
⎢ ⎥
⎣ ⎦
i i
Electrodynamics version of the Poisson equation: 2
( )
,(
( )
,)
1 tot( )
, oV r t A r t r t
t ε ρ
∂
∇ + ∇ = −
∂ i
(4) Ampere’s Law:
B r t
( )
, o totJ( )
r t, o o E r t( )
,t
μ μ ε ∂
∇× = +
∂
with: E r t
( )
, V r t( )
, A r t( )
,t
∂
= −∇ −
∂
and: B r t
( )
, = ∇×A r t( )
,( )
(
)
( )
( )
2( )
2, ,
, o tot , o o V r t o o A r t
A r t J r t
t t
μ μ ε ⎛∂ ⎞ μ ε ∂
∇× ∇× = − ∇⎜ ⎟−
∂ ∂
⎝ ⎠
But: ∇× ∇×
(
A) ( )
≡ ∇ ∇iA − ∇2A( )
2( )
( )
( )
( )
2
2
, ,
, o o A r t , o o V r t o tot ,
A r t A r t J r t
t t
μ ε μ ε μ
⎛ ∂ ⎞ ⎛ ∂ ⎞
∇ − − ∇ ∇ + = −
⎜ ⎟ ⎜ ⎟
⎜ ∂ ⎟ ⎝ ∂ ⎠
⎝ ⎠ i ⇐
n.b. These two equations contain all of the information
that is contained in Maxwell’s 4 equations !!!
n.b. this relation is
actually 3 separate eqns. for x, y and z!
These two equations may seem “ugly” at this point in time, but watch what we can do with them:
a.) Add:
( )
( )
2 2
2 2
, ,
0 o o V r t o o V r t
t t
ε μ ∂ ε μ ∂
= − +
∂ ∂ to the first equation, and group terms as follows:
1st eqn:
( )
( )
( )
( )
( )
2 2
2
, , 1
, o o , o o tot ,
o
V r t V r t
V r t A r t r t
t t t
ε μ ε μ ρ
ε
⎛∇ − ∂ ⎞+ ∂ ⎛∇ + ∂ ⎞= −
⎜ ∂ ⎟ ∂ ⎜ ∂ ⎟
⎝ ⎠
⎝ ⎠ i
2nd eqn:
( )
( )
( )
( )
( )
2 2
2
, ,
, o o A r t , o o V r t o tot ,
A r t A r t J r t
t t
ε μ ε μ μ
⎛ ∂ ⎞ ⎛ ∂ ⎞
∇ − − ∇ ∇ + = −
⎜ ⎟ ⎜ ⎟
⎜ ∂ ⎟ ⎝ ∂ ⎠
⎝ ⎠ i
b.) Define: L r t
( )
, A r t( )
, o o V r t( )
,t
ε μ ∂
⎛ ⎞
≡ ∇⎜ + ⎟
∂
⎝ i ⎠.
c.) Define the D’Alembertian (aka “Box2”) operator:
2 2
2 2 2
2 2 2
1
o o
t c t
ε μ ∂ ∂
≡ ∇ − = ∇ −
∂ ∂
Then:
1st eqn: 2
( )
,( )
, 1( )
,tot o
L r t
V r t r t
t ε ρ
∂
+ = −
∂
2nd eqn: 2
( )
,( )
,( )
,o tot
A r t − ∇L r t = −μ J r t
d.) Divide 1st eqn. by c, then use 1 2
o o c
ε μ = relation:
1st eqn: 21
( )
, 1( )
, o( )
,( )
,tot o tot
o o
L r t
V r t r t c r t
c c t cε μμ ρ μ ρ
∂
+ = − = −
∂
2nd eqn: 2
( )
,( )
,( )
,o tot
A r t − ∇L r t = −μ J r t
Thus:
1st eqn: 21V r t
( )
, 1 L r t( )
, oc tot( )
r t,c c t μ ρ
∂
+ = −
∂
“time-like”
2nd eqn: 2A r t
( )
, − ∇L r t( )
, = −μo Jtot( )
r t, “space-like”We shall see later on in the semester (when we get to relativistic electrodynamics), that
relativisticfourvectors exist {valid in anyinertialreferenceframe} for the relativistic 4-vector
potential A r tμ
( )
, ≡(
V r t c A r t( )
, ,( )
,)
{where μ=0,1, 2,3; μ =0is the temporal componentof the relativistic 4-vector, and μ =1, 2,3 are e.g. the x, y, z spatial components of the 4-vector},
the 4-current density Jtotμ
( )
r t, ≡(
cρtot( )
r t J, , tot( )
r t,)
.The covariant relativistic 4-gradient operator: 1 ,
x c t
μ μ
∂ ⎛ ∂ ⎞
∂ ≡ ≡ +⎜ ∇⎟
∂ ⎝ ∂ ⎠, whereas
the contravariant relativistic 4-gradient operator: 1 ,
x c t
μ ∂ ⎛ ∂ ⎞
∂ ≡ ≡ −⎜ ∇⎟
∂ ⎝ ∂ ⎠.
Do you see/canyou see any interesting parallels between these two equations???
The relativistic D’Alembertian / “Box2” operator can thus be written in 4-vector notation as:
2
2 2
2 2
1
x x c t
μ
μ μ
μ
∂ ∂ ∂
∂ ∂ = = = ∇ −
∂ ∂ ∂
We can write: L r t
( )
, A r t( )
, 12 V r t( )
, A r t( )
, A r t( )
,c t x
μ
μ μ μ
∂ ∂
⎛ ⎞
≡ ∇⎜ + ⎟= = ∂
∂ ∂
⎝ i ⎠
We can also write:
( )
, 1( )
, 1( )
,( )
,( )
, v( )
,v
L r t L r t
L r t L r t L r t A r t
c t c t x
μ μ
μ
∂ ⎛ ∂ ⎞ ∂
−∇ + = − −⎜ + ∇⎟ = − = −∂ = −∂ ∂
∂ ⎝ ∂ ⎠ ∂
Thus, the two equations:
1st eqn: 21
( )
, 1( )
,( )
,o tot
L r t
V r t c r t
c c t μ ρ
∂
+ = −
∂
2nd eqn: 2
( )
( )
( )
, , o tot ,
A r t − ∇L r t = −μ J r t
can be written as a single equation, very compactly (& elegantly!) in relativistic electrodynamics as:
( )
,( )
,( )
,v v
v A r t vA r t oJ r t
μ μ μ μ
∂ ∂ − ∂ ∂ = −
Very shortly, we will learn that as a consequence of the gauge-invariant nature of the
electromagnetic interaction, for the choice of working in the so-called Lorenzgauge, namely that:
( )
( )
2( )
( )
( )
, ,
1
, , , 0
v
v v v
V r t A r t
L r t A r t A r t
c t x
∂ ∂
⎛ ⎞
≡ ∇⎜ + ⎟= = ∂ =
∂ ∂
⎝ i ⎠
If L r t
( )
, =0 ∀( )
r t, , then ∂L r t( )
, ∂ =t 0 ∀( )
r t, . Hence, in the Lorenzgauge:( )
,( )
,v
v A r t oJ r t
μ μ μ
∂ ∂ = − or equivalently: 2A r tμ
( )
, = −μoJμ( )
r t,This single potential equation (actually 4 separate equations, since μ =0,1, 2,3) contain(s) all of the information that is contained in Maxwell’s 4 field equations !!!
(1) Gauss’ Law:
( )
, 1 tot( )
, oE r t ρ r t
ε
∇i = (3) Faraday’s Law: E r t
( )
, B r t( )
,t
∂
∇× = −
∂
(2) No Magnetic Charges: ∇iB r t
( )
, =0 (4) Ampere’s Law: B r t( )
, o totJ( )
r t, o o E r t( )
,t
μ μ ε ∂
∇× = +
Griffiths Example 10.1
Find the electric charge and current distributions ρtot
( )
r t, and Jtot( )
r t, that produce:( )
, 0V r t =
( )
(
)
2
ˆ, for 4
,
0, for
ok ct x z x ct
c A r t
x ct
μ
⎧ − <
⎪ = ⎨
⎪ ≥
⎩
Note that the parabola:
(
ct− x)
2 =( )
ct 2−2ct x + x2 =0 at ct= x .Since V r t
( )
, =0 ∀ ,( )
r t , we know that ρtot( )
r t, =0 ∀ ,( )
r t .However, A r t
( )
, zˆ tells us that there must be some kind of current present, and in the ˆz-direction.In the process of solving this problem, we will determine what kind of current is present…
Note that A A z= zˆ i.e. note that A is an explicit fcn(x) only, and points (only) in the ˆz-direction.
First, we determine E r t
( )
, and B r t( )
, from A r t( )
, :( )
,( )
,(
)
ˆ2 o
A r t k
E r t ct x z
t
μ
∂
= − = − −
∂ for x <ct, and: E r t
( )
, =0 for x ≥ct.Thus:
0
z
A A
y =
∂ ∇× =
∂
0
y
A z =
∂ −
∂
0
ˆ Ax
x z =
⎛ ⎞
∂
⎜ ⎟ +
⎜ ⎟ ∂
⎜ ⎟
⎝ ⎠
0
ˆ y
z A
A y
x x
=
⎛ ⎞ ∂
∂
⎜ − ⎟ +
⎜ ∂ ⎟ ∂
⎜ ⎟
⎝ ⎠
0
x
A y =
∂ +
∂ ˆ z ˆ
A
z y
x
⎛ ⎞
∂
⎜ ⎟ = −
⎜ ⎟ ∂
⎜ ⎟
⎝ ⎠
since Ax = Ay =0
Hence:
For x > 0
( )
( )
(
)
2(
)
ˆ ˆ
, ,
4 2
ok ok
B r t A r t ct x y ct x y
c x c
μ ∂ μ
= ∇× = − − = ± −
∂ for x <ct, and B=0 for x ≥ct.
For x < 0
n.b. For x ≥ct, E r t
( )
, and B r t( )
, =0 n.b. B x ty( )
, has a discontinuity for t > 0 at x = 0 !!( )
,z
E x t
x ct ct
−
@ t =t
@ t =0
1 2μokct −
( )
, yB x t
x ct ct
−
@ t =t
@ t =0
1 2μokt −
@ t =t 1 2μokt +
@ t =0 where k = constant
and c=1 ε μo o
( )
, zA x t
x
@t t=
@t=0
ct
− ct
1 4μokt
It can be shown that indeed ∇iE r t
( )
, =0, ∇iB r t( )
, =0 {and ∇iA r t( )
, =0} {n.b. please explicitly check/work these out yourselves!}, and that indeed:( )
, ˆ2 ok
E r t μ y
∇× =∓
( )
, ˆ2 ok
B r t z
c μ
∇× = −
( )
, ˆ2 o
B r t k
y t
μ
∂
= ±
∂
( )
,ˆ 2 o
E r t kc
z t
μ
∂
= − ∂
⇒Maxwell’s eqn’s all satisfied, with ρtot
( )
r t, =0 and Jtot( )
r t, =0.• However, we have seen before {déjà vu!} that the discontinuity in B r t
( )
, @ x = 0 for t>0heralds/signals the presence of a surfacecurrent Kfree
( )
r t, .• In our case {here}, the surface current Kfree
( )
r t, lies in the y-z plane {n.b. A Kfree zˆ!} The boundary condition associated with a free surface current (see P435 Lect. Notes 24, p. 13-17 and/or Griffiths equation 7.63 (iv), p. 333) is:1 2 1 2
1 2
1 1
ˆ free
H H B B K n
μ μ
− = − = × with: μ1 =μ2 =μo {here}
⇒ B1 −B2 =μoKfree×nˆ at x=0
i.e.:
(
0,)
0(
0,)
0 ˆ 0y x y x o free x
B x< t = −B x> t = =μ K ×n = n.b. nˆ points from medium 2 → 1 (nˆ= −xˆ) See Griffiths p. 331-2, figures 7.46 and 7.47 ˆz − =x nˆ ˆ
medium 1 (x < 0)
Kfree
ˆy ⇐ n.b. Kfree
( )
r t, lies in the y-z plane Kfreexˆ= −nˆ medium 2 (x > 0) ∴ ˆ ˆ ˆ
2 2
o o
o free
kt kt
y y K x
μ μ μ
− − = − ×
But: ˆ ˆx y z× = ˆ, ˆy z x× =ˆ ˆ, z x yˆ× =ˆ ˆ ⇒ ˆkty K= free×xˆ ⇒ Kfree
( )
t =ktzˆ at x = 0 in y-z plane. Physically, this corresponds to a uniform, but time-dependent free surface current Kfree( )
t =ktzˆ which flows in the ˆ+z direction at x = 0 in the y-z plane. It starts up from zero at time t = 0, its strength Kfree( )
t =kt increases linearly with time t.n.b. ⇒ “News” travels outward at speed c, note that E and B are still zero for points x ≥ct!!! Region 1: x < 0 Region 2: x > 0
Gauge Transformations
Using v
( )
, v( )
,( )
,v A r t vA r t oJ r t
μ μ μ μ
∂ ∂ − ∂ ∂ = − enables us to reduce the problem of finding the six components associated with the two vectors E E x E y E z= xˆ+ yˆ+ zˆ and B B x B y B z= xˆ+ yˆ+ zˆ
down to four components – the scalar potential V and the vector potential A A x A y A z= xˆ+ yˆ+ zˆ.
As we saw last semester in P435, B r t
( )
, = ∇×A r t( )
, and E r t( )
, = −∇V r t( )
, − ∂A r t( )
, ∂tdo not enable us to uniquely define / specify / determine the scalar and vector potentials V r t
( )
,and A r t
( )
, ; only potential differencesV2 – V1 and A2−A1 are physically meaningful…We are free to impose extra conditions on V r t
( )
, and A r t( )
, aslongas E r t( )
, and B r t( )
,remain unchanged by the imposition of these extra conditions…
This freedom to impose extra conditions on V r t
( )
, and A r t( )
, without changing( )
, and( )
,E r t B r t is known as gaugefreedom or (more formally) as gaugeinvariance.
Suppose we have e.g. two sets of potentials
{
V r t A r t( ) ( )
, , ,}
and{
V r t A r t′( ) ( )
, , ′ ,}
thatcorrespond to the same physical fields E r t
( )
, and B r t( )
, . These two sets of potentials mustbe related to each other by:
( )
,( )
,( )
,A r t′ =A r t +α r t and: V r t′
( )
, =V r t( )
, +β( )
r t,Because: B r t
( )
, = ∇×A r t( )
, = ∇×A r t′( )
, = ∇×(
A r t( )
, +α( )
r t,)
⇒ ∇×α( )
r t, ≡0 But if:(
∇×α( )
r t,)
=0, then since(
∇×∇f r t( )
,)
≡0 (always!)⇒ we can always write α
( )
r t, as the gradient of a scalar function λ( )
r t, , e.g. α( )
r t, ≡ ∇λ( )
r t, .And if:
( )
( )
( )
( )
( )
( )
( )
( )
( )
,
, ,
, , ,
, , ,
A r t
E r t V r t
t
A r t A r t r t
V r t V r t r t
t t t
α β
∂
= −∇ −
∂ ′
∂ ∂ ∂
′
= −∇ − = −∇ − ∇ − −
∂ ∂ ∂
Then we see that the scalar function β
( )
r t, must be related to α( )
r t, by:( )
r t,( )
r t, 0t α
β ∂
∇ + =
∂
But: α
( )
r t, ≡ ∇λ( )
r t, ⇒( )
r t,(
( )
r t,)
0t λ
β ∂ ∇
∇ + =
∂ ⇒
( )
( )
,, r t 0
r t
t
λ
β ∂
⎛ ⎞
∇⎜ + ⎟=
∂
⎝ ⎠
which must hold for arbitrary/any/all space-time points
( )
r t, ⇒( )
r t,( )
r t, 0t λ
β +∂ =
Note that
( )
r t,( )
r t, 0t
λ
β ∂
⎛ ⎞
∇⎜ + ⎟=
∂
⎝ ⎠ can also be satisfied if
( )
( )
,( )
, r t
r t t
t
λ
β ∂ κ
⎛ ⎞
+ =
⎜ ∂ ⎟
⎝ ⎠ ,
i.e. the scalar function κ
( )
t depends only on time, t.Thus we see that:
( )
r t,( )
r t,( )
t tλ
β = −∂ +κ
∂ .
But we can always “absorb” κ
( )
t into λ( )
r t, by adding( )
0
t t
t κ t dt
′=
′= ′
∫
to λ( )
r t, , i.e.( )
,( )
, t t0( )
tr t r t t dt
λ′ =λ +
∫
′=′=κ ′ .We can then redefine the “new” λ′
( )
r t, . ⇒ “old”λ( )
r t, .Note also that since the scalar function κ
( )
t depends only on time t, this will not affect thegradient of ∇λ
( )
r t, in any way, and hence α( )
r t, = ∇λ( )
r t, is completelyunaffected by this!Thus: A r t′
( )
, = A r t( )
, +α( )
r t, = A r t( )
, + ∇λ( )
r t, or: ∇λ( )
r t, =A r t′( )
, −A r t( )
, ≡ ΔA r t( )
, And: V r t( )
, V r t( )
,( )
r t,t λ
∂
′ = −
∂ or:
( )
,( )
( )
( )
, , ,
r t
V r t V r t V r t
t λ
∂
′
− = − ≡ Δ
∂
Thus, for any scalar function λ
( )
r t, , we can always add ∇λ( )
r t, to A r t( )
, provided thatwe simultaneouslysubtract ∂λ
( )
r t, ∂t from V r t( )
, .⇒ This will leave the E r t
( )
, and B r t( )
, -fields unchanged / invariant under this so-calledgaugetransformation!
• Gauge transformations can be exploited to adjust ∇iA r t
( )
, .• In magnetostatics, we chose ∇iA r t
( )
, =0 ( = the Coulomb gauge).• In electrodynamics, the situation is not always so clear cut!! The most convenient “gauge”
choice depends on the detailed nature of the problem! ∃ Many gauges to work with…. Coulomb Gauge: ∇iA r t
( )
, =0 (Most useful in electro/magnetostatics)Lorenz Gauge: A r t
( )
, o o V r t( )
,t
μ ε ∂
∇ = −
∂
i (Most useful in electrodynamics)
Ludwig Valentin Lorenz, Danish physicist – a contemporary of J.C. Maxwell, ca. 1867 – not to be confused with Hendrick A. Lorentz, Dutch physicist & contemporary of Albert Einstein… {see/read J.D. Jackson & L.B. Okun’s “Historical Roots of Gauge Invariance” Feb. 22, 2001; On the web at: http://arxiv.org/vc/hep-ph/papers/0012/0012061v2.pdf}.
The two most popular gauges for use in E&M
• Gauge Transformation(s) are intimately connected to the choice of an {inertial} reference
frame. “Hints” of this connection, e.g. in the context of special relativity – length contraction and time-dialation between two {inertial} reference frames are the ∇λ
( )
r t, and( )
r t, t λ′∂ ∂ {where
( )
( )
( )
0
, , t t
t
r t r t t dt
λ λ ′=κ
′=
′ = +
∫
′} terms that can be added to{subtracted from} the vector potential A r t
( )
, {scalar potential V r t( )
, }, respectively.The Coulomb Gauge: ∇iA r t
( )
, =0Here, the scalar potential V r t
( )
, is “easy” to calculate, but in comparison, the vector potential( )
,A r t is “difficult” to calculate. The inhomogeneous differential equation for V r t
( )
, is:( )
( )
2V r t, A r t,
t
∂
∇ + ∇
∂
(
i)
( )
1
, tot o
r t ρ ε
= − ⇒ 2
( )
, 1( )
,tot o
V r t ρ r t
ε
∇ = −
Then for: 2
( )
, 1( )
,TOT o
V r t ρ r t
ε
∇ = − if we set: V r
(
= ∞,t)
=0The solution to Poisson’s equation is:
( )
, 1( )
, 4tot v o
r t
V r t ρ dτ
πε ′
′ ′
=
∫
r
But V r t
( )
, alone does not determine E r t( )
, in electrodynamics – we mustalso know A r t( )
,as well! In the Coulomb gauge, the differential equation for the vector potential A r t
( )
, is:
( )
( )
( )
2 2
2
,
, o o ,
A r t
A r t A r t
t
μ ε ∂
∇ − − ∇ ∇
∂ i
( )
,( )
,
o o o tot
V r t
J r t
t
μ ε ∂ μ
⎛ ⎞
+ = −
⎜ ∂ ⎟
⎝ ⎠
becomes:
( )
( )
( )
( )
2 2
2
, ,
, o o A r t o tot , o o V r t
A r t J r t
t t
μ ε ∂ μ μ ε ⎛∂ ⎞
∇ − = − + ∇⎜ ⎟
∂ ⎝ ∂ ⎠ ⇐
In the Coulomb Gauge, the scalar potential at time t, V r t
( )
, is determined by the distribution of electric charge at the “rightnow” time t – which is acausal, because EM signals/information cannotpropagate faster than the speed of light c! However, changes in the electric charge density distribution
( )
tot r
ρ ′ at the source point r′ take a finite time to be observed at the observation point r !!!
( )
, 1( )
,4
tot v o
r t
V r t ρ dτ
πε ′
′ ′
=
∫
r
Thus in the Coulomb Gauge, the scalar potential V r t
( )
, instantaneously reflects all changesinρtot
( )
r t′, . However, recall that V r t( )
, byitself is not a physically measurable quantity – only potential differences such as ΔV t21( )
≡V r t( )
2, −V r t( )
1, or ΔV r(
1,Δ ≡t)
V r t(
1, 2)
−V r t(
1 1,)
are!The “usual” Poisson’s equation
with: r ≡ −r r′ and:
2 2
r r′
= = −
r r
n.b. Also depends
An astronaut standing on the surface of the moon can only e.g. directly measure E r t
( )
, , which manifestly involves both V r t( )
, and A r t( )
, : E r t( )
, = −∇V r t( )
, − ∂A r t( )
, ∂t, thus inthe Coulomb gauge, while the scalar potential V r t
( )
, instantaneously reflects all changes in( )
, tot r tρ ′ , the combination −∇V r t
( )
, − ∂A r t( )
, ∂t does not have this instantaneous,“right-now” behavior in the Coulomb gauge, i.e. the combination −∇V r t
( )
, − ∂A r t( )
, ∂t is causal in the Coulomb Gauge, and thus E r t( )
, changes in a causally-connected manner only after EM“news” / information arrives at the appropriate / causally-related time interval Δt, as a direct consequence of the propagation speed (= c in free space/vacuum) of this EM “news” /
information. Thus, in the Coulomb gauge, the causal behavior is carried by / encrypted into the
vector potential A r t
( )
, , in satisfying the above differential equation for A r t( )
, !The Lorenz Gauge:
( )
, o o( )
,V r t A r t
t
μ ε ∂
∇ = −
∂
i ⇒
( )
,( )
, o o( )
, 0V r t
L r t A r t
t
μ ε ∂
≡ ∇ + =
∂ i
Here we obtain:
(a) 2V r t
( )
, L r t( )
, t∂ +
∂
1 tot
( )
, or t
ρ ε
= − ⇒ 2
( )
, 1 tot( )
,o
V r t ρ r t
ε
= −
and:
(b) 2A r t
( )
, − ∇L r t( )
,( )
,o totJ r t μ
= − ⇒ 2A r t
( )
, = −μo totJ( )
r t, where:2
2 2
2 2
1
c t
∂ ≡ ∇ −
∂
.
Thus, we see from 2
( )
,( )
,tot o
V r t = −ρ r t ε
and 2
( )
,( )
,o tot
A r t = −μ J r t
that the
Lorenz gauge puts the {time-like} scalar potential V r t
( )
, and the {space-like} vector potential( )
,A r t on an equal footing! (i.e. a “democratic” treatment of V r t
( )
, and A r t( )
, ).In relativity, V r t c
( )
, and A r t( )
, respectively are the temporal and spatial components of the relativistic 4-vector potentialA r tμ( )
, and cρtot( )
r t, and Jtot( )
r t, respectively are the temporal and spatial components the relativistic 4-vector current Jμ( )
r t, , where the index μ =0 : 3 {0,1, 2,3}= . In tensor notation: 2A r tμ( )
, =Jμ( )
r t, ⇐ 4-D {space-time} Poisson Equation in the Lorenz gauge!where: Aμ ≡
(
V c A A A, x, y, z)
and Jμ ≡(
c J J Jρ, ,x y, z)
and2
2 2
2 2
1 v
v
x x c t
∂ ∂ ∂
≡ = ∇ −
∂ ∂ ∂
n.b. Convention: repeated indices (here = v) are summedover (i.e. summed over v = 0:3).
n.b. Griffiths claims that he will use the Lorenz gauge for the remainder of his book {we will also!}. The whole of electrodynamics thus reduces to solving the inhomogeneous 4-D Poisson’s
Continuous Electric Charge and Current Distributions Advanced and Retarded Potentials
We derived {see above}the following relations for the potentials in the Lorenz gauge:
L r t
( )
, A r t( )
, o o V r t( )
, 0t
μ ε ∂
≡ ∇ + =
∂
i 2A r tμ
( )
, =Jμ( )
r t, ⇐4-D Poisson Equation with: A r tμ
( )
, ≡(
V r t c A r t( )
, ,( )
,)
=(
V r t c A r t A r t A r t( )
, , x( )
, , y( ) ( )
, , z ,)
and: Jμ( )
r t, ≡(
cρ( ) ( )
r t J r t, , ,)
=(
cρ( ) ( ) ( ) ( )
r t J r t J r t J r t, , x , , y , , z ,)
Or: 2
( )
, 1( )
,tot o
V r t ρ r t
ε
= −
where:
2 2
2 2 2
2 2 2
1
o o
xγ xγ c t μ ε t
∂ ∂ ∂ ∂
≡ = ∇ − = ∇ −
∂ ∂ ∂ ∂
( )
( )
2 , ,
o tot
A r t = −μ J r t
Or:
( )
( )
( )
2
2 , , 1 ,
o o tot
o
V r t
V r t r t
t
μ ε ρ
ε
∂
∇ − = −
∂
( )
( )
( )
2
2 , , ,
o o o tot
V r t
A r t J r t
t
μ ε ∂ μ
∇ − = −
∂
For static situations (no time dependence) these reduce to the familiar “4-D” Poisson Equation:
( )
( )
2 1
tot o
V r ρ r
ε
∇ = − ⇒ solution is:
( )
1( )
4
tot v o
r
V r ρ dτ
πε ′
′ ′
=
∫
r
( )
( )
2
o tot
A r μ J r
∇ = − ⇒ solution is:
( )
( )
4
tot o
v
J r
A r μ dτ
π ′
′ ′
=
∫
r
( )
(
)
(
)
(
)
2 2
2
2 2
r
2 2 2
r
c t
c t t
r r c t t
+ Δ
= + −
′
= − + −
r r
(
r)
c t
c t t
Δ
= −
r
Field/Observation Point P r t
( )
,Source Point S r t
(
′, r)
ct
r ct
The Light-Cone in Space-Time:
c time⋅
space
( )
0,0r
Now consider what happens if the sources ρtot
( )
r t′, and Jtot( )
r t′, in the volume v’ aretime-dependent:
An observer at the field point P r t
( )
, detects changes in the potentials and/or the EM fieldsat the time t. However, those changes observed at the field point P r t
( )
, result from changes in(
, r)
tot r tρ ′ and / or Jtot
(
r t′, r)
that occurred at earlier time(s), because it takes a finite amount oftime for EM “news” (i.e. changes) that occur at a source point S r t
(
′, r)
to propagate to the field / observation point P r t( )
, .In terms of a space-time light-cone diagram (for propagation of EM news in free space = vacuum):
(
r)
c t c t tΔ = − =r
or: Δ = −t
(
t tr)
= r cor: t t= +r r c
or: tr = −t r c = retardedtime
with: r = r t
( )
−r t′( )
rand: r =r t
( )
−r t′( )
rDue to causality, it takes a finite time Δ = − =t t tr r c= −r r c′ for a changee.g. in the
electric charge density ρtot
(
r t′, r)
that occurs at the source point S r t(
′, r)
at the earlier, retardedtime trto propagate to the field/observation point P r t
( )
, at the later time, t > tr: t t= +r r c.In free space (the vacuum) this EM “news” / information propagates with speed c = 3 × 108m/sec.
We also point out {here, for completeness’ sake} that both the source point S r t
(
′, r)
and field/observation point P r t
( )
, are at rest (e.g. in the labframe – an inertial {i.e. non-accelerating} referenceframe). The electrodynamics of this situation will be different e.g. if the observer is movingrelative to the source, or if both the source and the observer are moving with respect toa chosen reference frame (e.g. the lab frame). Special relativity deals with these situations… Thus, for non-static source volume charge density/current density distributions ρtot
(
r t′, r)
and Jtot
(
r t′, r)
, the scalar and vector potentials V r t( )
, and A r t( )
, at the observation/field point( )
,P r t at the later, causal time t t= +r r c (t > tr) are related to the sources ρtot
(
r t′, r)
and(
, r)
totJ r t′ at the source point S r t
(
′, r)
at the earlier, so-called retarded time, tr = −t r cby the following relations:
( )
(
r)
r
, 1
, 4
tot v o
r t
V r t ρ dτ
πε ′
′
′
=
∫
r with tr = −t r c
( )
(
r)
r
, ,
4
tot o
v
J r t
A r t μ dτ
π ′
′
′
=
∫
r and r = r t
( )
−r t′( )
rThese expressions for the potentials are known as retardedpotentials because changes in the
source volume charge density and/or current density distributions ρtot
(
r t′, r)
and Jtot(
r t′, r)
at source point(s) S r( )
′ occurring at the (earlier) “retardedtime”, tr < t, take a time intervalr
t t t c r r c′
Δ = − =r = − to propagate from the source point S r t
(
′, r)
to the observation/field point P r t( )
, arriving there at the later, causal time t t= +r r c, where r = r t( )
−r t′( )
r .• This is exactly analogous to the situation where an observer is looking out into the night sky. Light (= EM radiation) from a star a distance rstar = robs
( )
t −rstar′( )
tr away, arriving on Earth at time t had to have left the surface of that star at an earlier time: tr = −t rstar c .The transit/propagation time of the light from the star to the Earth is:Δ = − =t t tr rstar c.
• From our own star (the sun), this time interval is:
11 8
1.496 10
3 10 /
Earth Sun m
t
c c m s
⊕− − ×
Δ = = =
×
r r 500 seconds = 8.3 minutes
Thus, we see that causality over astronomical distances is significant, but is also important even for laboratory/everyday distance scales.
Retarded Scalar and
Vector Potentials:
In the previous semester’s course (P435) we saw that, for static sources:
( )
( )
1( )
4
tot v o
r
E r V r ρ dτ
πε ′
′ ′
= −∇ = −∇
∫
r where: r = −r r′ and ∇ = fcn r
( )
only( )
1( )
4
tot v
o
r
E r ρ dτ
πε ′
′
⎛ ⎞
′
= −∇⎜ ⎟
⎝ ⎠
∫
r where: ρtot( )
r′ = fcn r( )
′ only( )
( )
21
ˆ 4
tot v o
r
E r ρ dτ
πε ′
′ ′
= +
∫
r r
( )
( )
( )
4
tot o
v
J r
B r A r μ dτ
π ′
′ ′
= ∇× = ∇×
∫
r where: Jtot
( )
r′ = fcn r( )
′ only( )
( )
4
tot o
v
J r
B r μ dτ
π ′
⎛ ′ ⎞
′
= ∇×⎜⎜ ⎟⎟
⎝ ⎠
∫
r( )
( )
2ˆ 4
tot o
v
J r
B r μ dτ
π ′
′ × ′
=
∫
rr
Now we cannot simply “generalize” these to the time-dependent case merely by adding t and tr
arguments to the
{
E and B}
and{
ρtot and Jtot}
(field and source) variables respectively!!i.e.
( )
(
r)
r 2
, 1
ˆ ,
4
tot v o
r t
E r t ρ dτ
πε ′
′
′
≠
∫
r r Nyet !!!
( )
(
r)
r 2
ˆ , ,
4
tot o
v
J r t
B r t μ dτ
π ′
′ ×
′
≠
∫
rr Nyet
2!!!
The reason why these expressions are not correct is simple: The causal connection between
t and tr has not been properly taken into account in the above two formulae: tr = −t r c with
( )
( )
rr t r t′
= −
r .
Properly taking into account this causal connection we need to realize that:
(
, r)
(
,)
tot r t tot r t c
ρ ′ =ρ ′ −r
(
, r)
(
,)
tot tot
J r t′ =J r t′ −r c
Thus, in order to correctly / properly calculate E r t
( )
, and B r t( )
, we need to back up and calculate these relations much more carefully!!!i.e. ρtot and Jtot are now also implicit functions of r = −r r′
via the causal relation tr = −t r c and hence are implicit
In calculating e.g. the Laplacian of the retarded scalar potential V r t
( )
, , it is criticalto realize that the integrand in
( )
(
r)
r , 1 , 4 tot v o r t
V r t ρ dτ
πε ′
′
′
=
∫
r depends on r in two places: Explicitly, in the denominator of the integrand, because: r = r t
( )
−r t′( )
r , and also Implicitly, in the numerator of the integrand, because: tr = −t r c t r r c= − − ′ .Thus:
( )
(
r)
(
)
r , , 1 1 , 4 4 tot tot v v o o
r t r r c r t
V r t d d
r r ρ
ρ
τ τ
πε ′ πε ′
′ − − ′ ′ ′ ′ = = ′ −
∫
r∫
Then since ∇ = fcn r
( )
only, then:r
t = −t r c t r r c= − − ′
( )
(
)
(
)
(
)
r r r , 1 , 4 , , 1 1 4 4 tot v o tot tot v v o o r tV r t d
r t r r c r t d d r r ρ τ πε ρ ρ τ τ πε πε ′ ′ ′ ′ ⎛ ⎞ ′ ∇ = ∇⎜ ⎟ ⎝ ⎠ ⎛ ′ − − ′ ⎞ ′ ⎛ ⎞ ′ ′ = ∇⎜ ⎟ = ∇⎜⎜ − ′ ⎟⎟ ⎝ ⎠ ⎝ ⎠
∫
∫
∫
r rBut ρtot
(
r t′, r)
is an explicit fcn r( )
′ and also an implicit fcn r( )
because tr = −t r c t r r c= − − ′ .By the chain rule:
(
r)
(
(
)
)
(
)
r r
,
1 1 1 1
, , 4 4 tot tot tot v v o o r t
d r t r t d
ρ
τ ρ ρ τ
πε ′ πε ′
′ ⎛ ⎞ ′ ⎡ ′ ′ ⎛ ⎞⎤ ′ ∇⎜ ⎟ = ⎢ ∇ + ∇⎜ ⎟⎥ ⎝ ⎠ ⎣ ⎦ ⎝ ⎠
∫
r∫
r rBut:
(
)
(
r)
(
)
r r r
r
, ,
, , , tot tot r
tot tot tot
r r r t r t
r t r t r t t t
c c t t
ρ ρ
ρ ′ ρ ⎛ ′ ⎞ ρ ⎛ ′ − ⎞ ∂′ ′ ∂ ′
∇ = ∇ ⎜ − ⎟= ∇ ⎜ − ⎟= ∇ = ∇
∂ ∂
⎝ ⎠ ⎝ ⎠
r
n.b. In the last step, note that we have used the fact that since tr = −t r c with r = −r r′ ≠ fcn t
( )
(because {here} the source and the observer are notmovingrelative to each other – i.e. they are at fixed/stationary points, e.g. in the lab reference frame), therefore ∂ = ∂tr t and thus: ∂ ∂ = ∂ ∂tr t.
Since∇ = fcn r
( )
only: tr t tc ⎛ ⎞ ∇ = ∇ −⎜ ⎟= ∇ ⎝ ⎠ r 0
1 1 1 ˆ
c c c
≡
− ∇ = − ∇ = −r r r
where: ∇ = ∇ − =r r r′ rˆ and: 2
ˆ 1 1 r r ⎛ ⎞ ∇⎜ ⎟= ∇ = − ′ − ⎝ ⎠ r r r
Thus:
(
)
(
r)
(
r)
(
)
(
)
r r r r
, , ˆ 1
ˆ
, tot tot , ,
tot tot tot
r t r t
r t t r t r t
t t c c c
ρ ρ
ρ ′ ∂ ′ ∂ ′ ⎛−∇ ⎞ ρ ′ ⎛ ⎞ ρ ′
∇ = ∇ = ⎜ ⎟= ⎜− ⎟= −
∂ ∂ ⎝ ⎠ ⎝ ⎠
r r r
where:
(
)
(
r)
r , , tot tot r t r t t ρρ ′ ≡∂ ′
∂
See Appendices A & B
Thus: r
( )
(
r)
(
r)
1 1 1
, , ,
4 v tot tot
o
V r t ρ r t ρ r t dτ
πε ′
⎡ ′ ′ ⎛ ⎞⎤ ′
∇ = ⎢ ∇ + ∇⎜ ⎟⎥
⎝ ⎠
⎣ ⎦
∫
r rBut:
(
r)
(
r)
1
ˆ
, ,
tot r t tot r t
c
ρ ′ ρ ′
∇ = − r and ∇⎛ ⎞⎜ ⎟1 = − ˆ2
⎝ ⎠ r
r r {from above}
Therefore: r
( )
(
r)
(
r)
2, ˆ 1 ˆ , , 4 tot tot v o r t
V r t r t d
c ρ ρ τ πε ′ ′ − ⎡ ⎛ ⎞⎤ ′ ′ ∇ = ⎢ − ⎜ ⎟⎥ ⎝ ⎠ ⎣ ⎦
∫
r rr r
If we now take the divergence of ∇V r tr
( )
, , i.e. the Laplacian of V r tr( )
, :
( )
(
( )
)
(
)
(
)
(
)
(
)
r 2
r r r 2
r r , ˆ 1 ˆ , , , 4 ˆ ˆ 1 1 , , 4 tot tot v o tot tot v o r t
V r t V r t r t d
c
r t r t
c ρ ρ τ πε ρ ρ πε ′ ′ ′ ⎡ ′ ⎛ ⎞⎤ ′ ∇ = ∇ ∇ = ∇ −⎢ − ⎜ ⎟⎥ ⎝ ⎠ ⎣ ⎦ ⎡ ⎤ ⎧ ⎛ ⎞ ′ ′ ⎛ ⎞ = ⎨− ⎢⎜ ⎟∇ + ∇ ⎜ ⎟⎥ ⎩ ⎣⎝ ⎠ ⎝ ⎠⎦
∫
∫
i i i i r r r r r r r r(
r)
(
)
2 2
ˆ ˆ
−⎡⎢⎜⎛ ⎟⎞∇ρtot r t′, +ρtot r t′, r ∇⎜⎛ ⎞⎟⎥⎤⎬⎫dτ′
⎝ ⎠ ⎝ ⎠
⎣ i i ⎦⎭
r r
r r
Since: tot
(
r t, r)
1 tot(
r t, r)
ˆc
ρ ′ ρ ′
∇ = − r {from above}
Then:
(
r)
(
r)
(
r)
1 1 ˆ
, , , , ,
tot tot tot tot tot
r r
r t r t r t r t r t
c c c c
ρ ′ ρ ⎛ ′ ⎞ ρ ⎛ ′ − ⎞′ ρ ′ ρ ′
∇ = ∇ ⎜ − ⎟= ∇ ⎜ − ⎟= − ∇ = −
⎝ ⎠ ⎝ ⎠
r r r
where:
(
)
(
)
(
)
2
r r
r 2
, ,
, tot tot
tot
r t r t
r t
t t
ρ ρ
ρ ′ ≡∂ ′ =∂ ′
∂ ∂ and: ∇ =r rˆ
and: ∇ ⎜ ⎟⎛ ⎞ˆ = ∇⎛⎜r rˆ ˆ− ′⎞⎟= 12
⎝ ⎠ ⎝ ⎠
i r
r r r and:
( )
3 2 ˆ 4πδ ⎛ ⎞ ∇ ⎜ ⎟= ⎝ ⎠
i r r
r
( )
3
δ r = 3-D delta fcn for r = −r r′
Thus:
( )
(
)
(
) ( )
(
)
( )(
) ( )
( ) r r 2 3r 2 r
2 r 3 2 2 , , , 1 1
, 4 ,
4
,
1 1 1
, 4 tot tot tot v o tot tot r v v o o r t V r t
r t
V r t r t d
c
r t
d r t d
c t
ρ ρ
πρ δ τ
πε
ρ
τ ρ δ τ
πε ε ′ ′ ′ = ⎡ ⎛ ′ ⎞ ⎤ ′ ′ ∇ = ⎢+ ⎜ ⎟− ⎥ ⎢ ⎝ ⎠ ⎥ ⎣ ⎦ ′ ⎡ ⎤ ∂ ′ ′ ′ = ⎢ ⎥− ∂ ⎣ ⎦
∫
∫
∫
r r r rOr:
( )
( )
( )
( )
2 r
2 2
r r 2 2
,
1 1
, , tot ,
o
V r t
V r t V r t r t
c t ε ρ
∂
= ∇ − = −
∂
Thus, we see that the retarded scalar potential V r tr
( )
, does indeed satisfy the inhomogeneous wave equation / the 4-D Poisson’s equation!Using the chain rule: See Appendix D See Appendix C See Appendix A
The same methodology can be carried out for the retarded vector potential A r tr
( )
, with the same results {please work through this yourselves !!!}, such that:( )
( )
2 r( )
( )
2 2
r r 2 2
,
1 1
, , tot ,
o
V r t
V r t V r t r t
c t ε ρ
∂
= ∇ − = −
∂
( )
( )
2 r( )
( )
2 2
r r 2 2
, 1
, , A r t o tot ,
A r t A r t J r t
c t μ
∂
= ∇ − = −
∂
where: r
( )
(
r)
, 1
, 4
tot v o
r t
V r t ρ dτ
πε ′
′
′
=
∫
r
( )
(
r)
r
, ,
4
tot o
v
J r t
A r t μ dτ
π ′
′
′
=
∫
r
Note that because the D’Alembertian operator 2 2 2
2 2
1
c t
∂ ≡ ∇ −
∂
explicitly involves the second
derivative with respect to time, ∂ ∂2 t2 (i.e. note that it is
quadratic {not linearly } dependent in
the time variable t ), therefore the D’Alembertian operator does not distinguish past from future!
Thus, there exist equally mathematically-acceptable, but physically unacceptable , acausal
solutions (i.e. ones which violatecausality), known as the so-called advancedpotentials
(n.b. the above proof(s) are also valid for the advancedpotential solutions) where: AdvancedTime: ta = +t r c with ta > t and thus: t t= −a r c.
and: a
( )
(
a)
, 1
, 4
tot v o
r t
V r t ρ dτ
πε ′
′
′
=
∫
r
( )
(
a)
a
, ,
4
tot o
v
J r t
A r t μ dτ
π ′
′
′
=
∫
r
with:
( )
( )
( )
( )
2 a
2 2
a a 2 2
,
1 1
, , tot ,
o
V r t
V r t V r t r t
c t ε ρ
∂
= ∇ − = −
∂
( )
( )
2 a( )
( )
2 2
a a 2 2
, 1
, , A r t o tot ,
A r t A r t J r t
c t μ
∂
= ∇ − = −
∂
The advanced potentials are entirely consistent with Maxwell’s Equations, but violate
causality – because they predict potentials now (at time t) that depend on the charge and current
distributions at a future time ta = +t r c ⇐ We do not observe such things in our universe!
{n.b. this has not stopped physicists from seriously looking for such things as tachyons, etc.} • In our universe, direct/empirical observation tells us that electromagnetic influences /
changes / disturbances propagate with time going forward, not going backward in time
– i.e. the universe that we live in behaves causally.
Retarded potentials associated with
the retarded time: tr = −t r c.
Advanced potentials are associated
• The macroscopic theory of electrodynamics must be manifestlytime-reversalinvariant
(i.e. even under the operation of time reversal, t→−t) because at the microscopic/ elementaryparticlephysics level, the electromagnetic interaction manifestly obeys
time-reversal invariance. This is not a trivial point, because e.g. the microscopic weak interaction violates time-reversal invariance in certain situations!! MacroscopicT-violating effects can
also manifest themselves here too (e.g. weak interaction induced regeneration of 0
S
K in a 0
L
K
beam passing through matter; the so-called Bvs. A phase of superfluid LHe3
{where ∃ a macroscopic B-field vs. a macroscopic E-field, respectively})….
Griffiths Example 10.2:
An infinitely long straight wire carries a time-dependent current I(t) = 0 for t ≤ 0, I(t) = Io for t > 0.
I(t)
Io n.b. t = 0 is the time atthewire
⇒ thus: t = 0 is tr = 0 !!!
0 t
Find / determine the resulting E and B fields at an observation point P r t
( )
, a radial distance ρfrom the axis of the wire. We choose the current-carrying long wire to lie along the ˆz-axis as shown in the figure below:
n.b. We assume that the ∞-long line current is {always} electrically neutral, hence the retarded
scalar potential Vr
( )
ρ,t =0 everywhere ∀( )
ρ,t .The retarded vector potential Ar
( )
ρ,t is: r( )
,( )
r ˆ4 z o
z
I t
A ρ t μ dz z
π =+∞ =−∞
=
∫
r where
2 z2
ρ
= +
r
and t t= +r r c where tr = 0 is the time that the current is switched on.
ˆ
x
ˆ
z
ˆ
y
Field/Observation Point P r t
( )
, SourcePoint
(
, r)
S r t′
( )
r t
r =ρ
( )
r r tr z
′ ′ =
2 2
r r′ ρ z
= = − = +
r r
( )
( )
rr t r t′
= −
r