3.2.

THE RADIATING ELECTRIC DIPOLE 77

ro which is typically at the center of the charge distribution [1]. If we keep only the lowest
order term we nd
j(r; t) = dtd p(t) [r , ro ] ; (3.24)
with the dipole moment X
p(t) = qn[rn(t) , ro ] : (3.25)
n
The dipole moment is identical with the denition in Eq. 3.15 for which we had ro =0. We
assume a time harmonic time-dependence which allows us to write the current density as
j(r; t)=Refj(r) exp(,i!t)g and the dipole moment as p(t)=Refp exp(,i!t)g. Eq. 3.24 can
then be written as
j(r) = ,i!p [r , ro ] : (3.26)
Thus, to lowest order, any current density can be thought of as an oscillating dipole with
origin at the center of the charge distribution.

3.2.1 Electric dipole elds in a homogeneous space

In this section we will derive the elds of a dipole representing the current density of a small
charge distribution located in a homogeneous, linear and isotropic space. The elds of the
dipole can be derived by considering two oscillating charges q of opposite sign, separated by
an innitesimal vector ds. In this physical picture the dipole moment is given by p = qds.
However, it is more elegant to derive the dipole elds using the Green's function formalism
developed in Section 1.3. There, we have derived the so-called volume integral equations (c.f.
Eq. 1.60 and Eq. 1.61)
Z
$
E(r) = Eo + i!o G (r; r0) j(r0) dV 0 ; (3.27)
Z h V
$ i
H(r) = Ho + r G (r; r0) j(r0) dV 0 : (3.28)
V
$
G denotes the dyadic Green's function and Eo ; Ho are the elds in the absence of the current
j. The integration runs over the source volume specied by the coordinate r0 . If we introduce
the current from Eq. 3.26 into the last two equations and assume that all elds are produced
by the dipole we nd
Z
E(r) = ! o G$ (r; ro) p ;
2
(3.29)
V
H(r) = ,i! r G$ (r; ro ) p :
h i
(3.30)
Hence, the elds of an arbitrarily
$ oriented electric dipole located at r = ro are determined
$
by the Green's function G (r; ro ). As mentioned earlier, each column vector of G species
the electric eld of a$dipole whose axis is aligned with one of the coordinate axes. For a
homogeneous space, G has been derived as
 
G$ (r; ro) = $I + k12 rr G(r; ro) ; G(r; ro) = exp(ik jr , ro j) : (3.31)
4jr , ro j
78 3. NEAR-FIELD INTERACTIONS
$
where I is$ the unit dyad and G(r; ro ) the scalar Green's function. It is straightforward
$ to
calculate G in the major three coordinate systems. In a Cartesian system G can be written
as  
$ exp(ikR) ikR , 1 $ 3 , 3ikR , k2R2 RR 
G (r; ro ) = 4R 1 + k2 R2 I + k2R2 R2 ; (3.32)
where R is the absolute value of the vector R = r , ro and RR denotes the outer product of
R with itself. Eq. 3.32 denes a symmetric 3  3 matrix
2 3
$ 4 Gxx Gxy Gxz 5
G = Gxy Gyy Gyz ; (3.33)
Gxz Gyz Gzz
which, together with Eqs. 3.29 and 3.30, determines the electromagnetic eld of an arbitrary
electric dipole p with Cartesian components px ; py ; pz .
$
The Green's function G has terms in (kR),1, (kR),2 and (kR),3. In the fareld, for
which R  , only the terms with (kR),1 survive. On the other hand, the dominant terms in
the near-eld, for which R  , are the terms with (kR),3 . The terms with (kR),2 dominate
the intermediate eld at R  . To distinguish these three ranges it is convenient to write
$ $ $ $
G = GNF + GIF + GFF ; (3.34)
where the near-eld (GNF ), intermediate eld (GIF ) and fareld (GFF ) Green's functions
are given by
$ exp(ikR) 1 , $I + 3 RR=R2 ;
G NF = 4R k2 R2 (3.35)

G$ IF = exp(ikR) i $
4R kR I , 3 RR=R ;
2

(3.36)
$
GFF = exp(ikR)
$ 
4R I , RR =R 2
: (3.37)

ϑ r

p y
ϕ

Figure 3.2: The elds of a dipole are most conveniently represented in a spherical coordinate
system (r; #; ') in which the dipole points along the z-axis (#=0).
3.2. THE RADIATING ELECTRIC DIPOLE 79

Notice that the intermediate eld is 90 out of phase with respect to the near- and fareld.
Because the dipole is located in a homogeneous environment, all three dipole orientations
lead to elds which are identical upon suitable frame rotations. We therefore choose a
coordinate system with origin at r = ro and a dipole orientation along the dipole axis, i.e.
p = pnz (c.f. Fig. 3.2). It is most convenient to represent the dipole elds in spherical
coordinates r =(r; #; ') and in spherical vector components E =(Er ; E# ; E'). In this system
the eld components E' and Hr ; H# are identical to zero and the only non-vanishing eld
components are
 
p cos # exp(ikr) 2 2i
Er = 4" " r k k2 r2 , kr ; (3.38)
2
o
 
E# = p4" sin # exp(ikr) k2 1 , i , 1 ; (3.39)
o" r k2 r2 kr
 r
p
H' = 4" " sin # exp(ikr) 2 i
k , kr , 1 "o " : (3.40)
o r o
The fact that Er has no fareld term ensures that the eld in the fareld is purely transverse.
Furthermore, since the magnetic eld has no terms in (kr),3 the near-eld is dominated by
the electric eld. Thus, close to the origin of the dipole the magnetic eld strength is much
smaller than the electric eld strength which justies a quasi-electrostatic consideration.

1000 1000
transverse field longitudinal field

-3
100 -3
100
r
r
field amplitude

10 10

1 1

-1
r -2
r

0.1 0.2 0.5 1 2 5 10 0.1 0.2 0.5 1 2 5 10

kr kr

Figure 3.3: Radial decay of the dipole's transverse and longitudinal elds. The curves corre-
spond to the absolute value of the expressions in brackets of Eq. 3.38 and Eq. 3.39, respec-
tively. While both the transverse and the longitudinal eld contribute to the near-eld, only
the transverse eld survives in the fareld. Notice, that the intermediate eld with (kr),2
does not really show up for the transverse eld. Instead the near-eld dominates for (kr)<1
and the far-eld for (kr)>1.
80 3. NEAR-FIELD INTERACTIONS

So far we have considered a dipole which oscillates harmonically in time, i.e. p(t) =
Refp exp(,i!t)g. Therefore, the electromagnetic eld is monochromatic and oscillates at the
same frequency. Although it is possible to generate any time-dependence by a superposition
of monochromatic elds (Fourier transformation), it is of advantage for ultrafast applications
to have the full time dependence available. The elds of a dipole p(t) with arbitrary time
dependence can be derived by using the time-dependent Green's function. In a nondispersive
medium it is easier to introduce the explicit time dependence by using the substitutions
 m  m m
exp(ikr) k p = exp(ikr) c (,i!) p ! in
m in m d , n

c dt m p t , r /c ; (3.41)
where n denotes the (dispersionfree) index of refraction and (t , nr=c) is the retarded time.
With this substitution, the dipole elds read as
 
cos # 2 n 2 d ,
Er (t) = 4" " r3 + c r2 dt p t , r n/c ;


(3.42)
o
 
E# (t) = 4"sin # 1 + n 1 d + n2 1 d2 p,t , r n
; (3.43)
/c
o " r3 c r2 dt c2 r dt2
 
H'(t) = 4" sin # r "o " n 1 d + n2 1 d2 p,t , r n
: (3.44)
o " o  c r2 dt c2 r dt2 /c
We see that the fareld is generated by the acceleration of the charges which constitute the
dipole moment. Similarly, the intermediate eld and the near-eld are generated by the
speed and the position of the charges, respectively.

3.2.2 Dipole radiation

It can be shown (see Problem 3.3) that only the fareld of the dipole contributes to the net
energy transport. The Poynting vector S(t) associated with the fareld can be calculated by
retaining only the r,1 terms in the dipole elds. We obtain
 2 2
S(t) = E(t)  H(t) = 1612" " sinr2 # nc3 dtd 2 p,t , r n/c
nr :
2 3
(3.45)
o
The radiated power P can be determined by integrating S(t) over a closed spherical surface
as Z  2 2
P (t) = S
n da = 4"1 " 23 nc3 d dtp(t)
3
; (3.46)
@V o 2

where we have shrunk the radius of the sphere to zero to get rid of the retarded time. The
average radiated power for a harmonically oscillating dipole turns out to be
P = 4"p2 n3 !4 ; (3.47)
o " 3 c3
which could have been also calculated by integrating the time averaged Poyning vector hSi =
(1=2) Re fE  H g, E and H being the dipole's complex eld amplitudes given by Eqs. 3.38-
3.40. We nd that the radiated power scales with the fourth power of the frequency. To
3.2. THE RADIATING ELECTRIC DIPOLE 81
 ') radiated into an
determine the normalized radiation pattern we calculate the power P(#;
innitesimal unit solid angle d
=sin #d#d' and divide py the total radiated power P
 ')
P(#; 3 sin2 # :
P = 8 (3.48)
Most of the energy is radiated perpendicular to the dipole moment and there is no radiation
at all in direction of the dipole.
Although we have considered an arbitrary time-dependence for the dipole we will restrict
ourselves in the following to the time harmonic case. It is straightforward to account for
dispersion when working with time-harmonic elds and arbitrary time-dependences can be
introduced by using Fourier transforms.

3.2.3 Rate of energy dissipation in inhomogeneous environments

According to Poynting's theorem (c.f. Eq. 1.36) the radiated power of any current distribution
with a harmonic time dependence in a linear medium has to be identical to the rate of energy
dissipation dW=dt given by
dW = , 1 Z Refj
Eg dV; (3.49)
dt 2 V
V being the source volume. It is important to notice that j is not the total current density.
Instead, it represents the source current js which generates the elds, or, a loss current jc
which is associated with thermal losses. In either way, j represents both energy sources
and energy sinks. If we introduce the dipole's current density from Eq. 3.26 we obtain the
important result
dW = ! Im fp
E(r )g ; (3.50)
dt 2 o

where the eld E is evaluated at the dipole's origin ro . At rst sight this seems not possible
to evaluate since exp(ikR)=R appears to be innite at r = ro . As we shall see this is not the
case. We rst note that due to the dot product between p and E we need only to evaluate
the component of E in direction of p. Since we chose p =pnz we calculate Ez as
kR h
Ez = 4"p " e R k 2 sin2# + R12 (3 cos2 # , 1) , iRk (3 cos2# , 1) :
i i
(3.51)
o
Since the interesting part is the eld at the origin of the dipole, the exponential term is
expanded into a series [exp(i kR) = 1 + i kR + (1=2)(i kR)2 + (1=6)(i kR)3 + :: ] and the
limiting case R ! 0 is considered. Thus,
dW = lim ! p ImfE g = ! p 2 lim n 2 k 3 + R2 (::) + :: o = p 2 ! k 3 ; (3.52)
dt R!0 2 z 8"o " R!0 3 12 "o "
which is identical with Eq. 3.47. Thus, Eq. 3.50 leads to the correct result despite of the
apparent singularity at R=0.
82 3. NEAR-FIELD INTERACTIONS

The importance of Eq. 3.50 becomes obvious if we consider an emitting dipole in an

inhomogeneous environment such as an atom in a cavity or a molecule in a superlattice.
The rate at which energy is released can still be calculated by integrating the Poynting
vector over a surface enclosing the dipole emitter. However, to do this, we need to know the
electromagnetic eld everywhere on the enclosing surface. Because of the inhomogeneous
environment, this eld is not equal to the dipole eld alone! Instead, it is the self-consistent
eld, i.e. the eld E generated by the superposition of the dipole eld Eo and the scattered
eld Es from the environment. Thus, to determine the energy dissipated by the dipole we rst
need to determine the electromagnetic eld everywhere on the enclosing surface. However,
by using Eq. 3.50 we can do the same job by only evaluating the total eld at the dipole's
origin ro . It is convenient to decompose the electric eld at the dipole's position as
E (ro ) = Eo (ro ) + Es (ro ) ; (3.53)
where Eo and Es are the primary dipole eld and the scattered eld, respectively. Introducing
Eq. 3.53 into Eq. 3.50 allows us to split the rate of energy dissipation P = dW=dt into two
parts. The contribution of Eo has been determined in Eq. 3.47 and Eq. 3.52 as
p2 ! k3 ;
Po = 12 (3.54)
"o "
which allows us to write for the normalized rate of energy dissipation
P = 1 + 6"o " 1 Imfp
E (r )g : (3.55)
Po p2 k 3 s o

Thus, the change of energy dissipation depends on the secondary eld of the dipole. This
eld corresponds to the dipole's own eld emitted at a former time. It arrives at the position
of the dipole after it has been scattered in the environment.

3.2.4 Classical lifetimes and decay rates

Let us consider an undriven harmonically oscillating dipole. As the dipole oscillates it radiates
energy according to Eq. 3.46. As a consequence, the dipole dissipates its energy into radiation
and its dipole moment decreases. We are interested in calculating the time  after which the
dipole's energy decreases to 1=e of its initial value.
Homogeneous environment
The equation of motion for an undriven harmonically oscillating dipole is2
d 2 p(t) + d p(t) + ! 2 p(t) = 0 : (3.56)
dt 2 o dt o
The natural frequency of the oscillator is !o and its damping constant is o . The solution
for p is n p o
p(t) = Re po e,i!o 1,( o2=4 !o2) t e, o t=2 : (3.57)
2 We ignore radiation reaction, i.e. the eect of the dipole's radiated eld on itself.
3.2. THE RADIATING ELECTRIC DIPOLE 83

Because of losses introduced through o the dipole forms a non-conservative system. The
damping rate not only attenuates the dipole strength put also produces a shift in resonance
frequency. In order to be able to dene an average dipole energy W at any instant of time we
have to make sure that the oscillation amplitude stays constant over one period of oscillation.
In other words, we require
 !o : (3.58)
The average energy of a harmonic oscillator is the sum of the average kinetic and potential
energy. At time t this average energy reads as 3
 = m2 !o2 p2(t) + p_2 (t) = m!2o jpo j2 e, o t;
2
W(t) 2q 2q (3.59)
where m is the mass of the particle with charge q. For a system with several particles we
have to sum the individual energies. The lifetime o of the oscillator is dened as the time
for which the energy decayed to 1=e of its initial value at t=0. We simply nd
o = 1= o : (3.60)
We now turn to the rate of energy loss due to radiation. The average radiated power Po
at time t is (c.f. Eq. 3.47)
p2(t) n3 !o4 :
Po (t) = 4" (3.61)
o " 3 c3
Energy conservation requires that the decrease in oscillator energy must equal to energy
losses, i.e. Z t

W(t=0) , W (t) = qi Po (t0 ) dt0 ; (3.62)
0

where we introduced the so-called intrinsic quantum yield qi. This parameter has a value
between zero and one and indicates the fraction of energy losses associated with radiation.
For qi = 1, all the oscillator energy is transformed to radiation. It is now straight forward
to solve for the decay rate. We introduce Eq. 3.59 and Eq. 3.61 into the last equation and
obtain
o = qi 4" n3 2q!o2 (classical) : (3.63)
o " 3 m c3
This is the classical formula for the atomic decay rate and through Eq. 3.60 also for the
atomic lifetime. It depends on the oscillation frequency and the oscillating particle's mass
and charge. We see that the higher the index of refraction of the surrounding medium is,
the shorther the lifetimes of the oscillator will be. o can easily be generalized to multiple
particle systems by summing over the individual charges qn and masses mn . The quantum
mechanical analog of the decay rate (c.f. Appendix ??) can be arrived at by replacing the
oscillator's initial average energy m!o2 jpo j2=(2q2) by the lowest energy of a quantum oscillator
h!o =2. Thus,
o = qi 4" n3 4 jpoj2 !o3 (quantum) : (3.64)
o " 3 h c3
3 This is easily derived by setting p = qx, !o2 = c=m and using the expressions mx_ 2=2 and cx2=2 for the
kinetic and potential energy, respectively.
84 3. NEAR-FIELD INTERACTIONS

Here, po refers to the transition dipole matrix element between two atomic states. While the
last equation is correct for empty space n=1 it requires some correction in dielectric media.
This correction is related to the local eld at the dipole's position. Depolarization of the
dipole's microscopic environment in uences its emission properties. The resulting correction
is similar to the Clausius-Mossotti relation.
Inhomogeneous environment
In an inhomogeneous environment, a harmonically oscillating dipole left to itself will expe-
rience its own eld as a driving force. This driving eld is the eld that arrives back to the
oscillator after it has been scattered in the environment. If we ignore radiation reaction, the
equation of motion is
d 2 p(t) + d p(t) + ! 2 p(t) = q 2 E (t) ; (3.65)
dt 2 o dt o m s
with Es being the secondary local eld, respectively. We expect that the interaction with
Es will cause a shift in resonance frequency and a modication of the decay rate. Therefore,
we use the following trial solutions for dipole moment and driving eld
n o n o
p(t) = Re po e,i! t e, t= ; Es (t) = Re Eo e,i! t e, t= :
2 2
(3.66)
and ! are the new decay rate and resonance frequency, respectively. The two trial solutions
can be inserted into Eq. 3.65. As before, we assume that is much smaller than ! (c.f.
Eq. 3.58) which allows us to reject terms in 2 . Furthermore, we assume that the interaction
with the eld Es is weak. In this limit the last term on the left hand side of Eq. 3.65 is
always larger than the driving term on the right hand side, or
! 2 po  qm Eo :
2
(3.67)
With these approximations and the classical expression for the `undisturbed' decay rate o
from Eq. 3.63 we obtain
= 1 + q 6"o " 1 Imfp
E (r )g : (3.68)
o i jp j 2 k 3 o s o
o

Since Es is proportional to po , the dependence on the magnitude of the dipole moment

cancels out. Besides the introduction of qi, Eq. 3.68 is identical with Eq. 3.55 for the rate of
energy dissipation in inhomogeneous environments. Thus, for qi = 1 we nd the important
relationship
P
=P :
o o
(3.69)
This equation can be veried by expressing the radiated power in terms of the average energy
W (t) given in Eq. 3.59. With the help of Eq. 3.62 we nd
P ,( , o ) t ;
Po = o e (3.70)
3.2. THE RADIATING ELECTRIC DIPOLE 85

where we assumed that the intrinsic quantum yield remains unaected by the inhomogeneous
environment. Besides the exponential time dependence the last two equations are the same.
The exponential function can be expanded in powers of its argument and for suciently
small times only the lowest order term, which is equal to one, is retained. This veries, that
Eq. 3.69 is fullled for suciently small times.
Eq. 3.68 can be adapted to describe the (normalized) spontaneous emission rate of a
quantum system (c.f. Appendix ??). In this case the classical dipole represents the quan-
tum mechanical transition dipole matrix element from the excited to the ground state (see
Appendix ??). The decay rate of the excited state is equal to the spontaneous emission
rate P=(h!), where h! is the photon energy. Eq. 3.55 provides a simple means to calculate
lifetime variations of atomic systems in arbitrary environments. In fact, this formula has
been used by dierent authors to describe uorescence quenching near planar interfaces and
the achieved agreement to experiment is excellent (c.f. Fig. 3.4).

Frequency shifts
The inhomogeneous environment not only in uences the lifetime of the oscillating dipole but
also causes a frequency shift
! =! , !o of the emitted light. An expression for
! can be
derived by inserting Eqs. 3.66 into Eq. 3.65. The resulting expression for
! reads as
" s  #

! = ! 1 , 1 , !12 m jqp j 2 Re fpo
Esg + 2 o ,
2

4 : (3.71)
o

Figure 3.4: Comparion of classical theory and experimental measurements of molecular life-
times in inhomogeneous environments. In the experiment, a layer of Eu3+ ions is held by
fatty acid spacers of variable thickness close to a silver surface (data after Drexhage [6]). The
calculted curve is due to Chance et al. [7] (used with no permission yet!).
86 3. NEAR-FIELD INTERACTIONS

After expanding the square root to rst order and neglecting the quadratic terms in , the
expression for the normalized frequency shift reduces to

! = q 3"o " 1 Re fp
E g : (3.72)
o i jp j 2 k 3 o s
o

For molecular uorescence the frequency shift is very small, in the range of the radiative
linewidth. The latter follows from Heisenberg's uncertainty principle as follows: The average
time t that is available to `measure' the excited state is hti =o . From the width in energy of
the transition
E =
!h and Heisenberg's uncertainty
E hti h it follows that
!  o,1.
Thus, a typical radiative lifetime of o  10nsec corresponds to a width of
  2 10,3 nm.
For molecules close to planar interfaces, the frequency shift varies as h ,3, h being the
height of the molecule, and reaches its maximum near the surface plasmon frequency. The
dependence on h ,3 suggests that the observation of the frequency shift should be possible
for small h. Yet, this is not the case because for small h also the linewidth increases. A shift
in the range of
  20nm was experimentally observed for small dipolar scatterers (silver
islands) close to a silver layer [8]. In this conguration the dipolar scatterers were excited
close to their resonance frequency leading to a highly enhanced polarizability. At cryogenic
temperatures, the linewidths are very narrow and the frequency shifts can be well resolved.
Notice again, that since Es is proportional to po , the dependence on the magnitude of
the dipole moment in Eq. 3.72 cancels out.