Lecture 5

Classical Theory of Disperson (From here, we will understand the microscopic picture of light-matter
interactions)

Up to this point, we have taken the linear constitutive relations D =  E and B =  H for granted ,with 𝜀 and
𝜇 experimental parameters .We have seen that the index 𝑛 = √𝜀𝑟 𝜇𝑟 is determined by 𝜀 and 𝜇, but we have no
microscopic picture of why 𝜀𝑟 might be ≠ 1.(As usual, we are ignoring magnetic materials and considering 𝜇𝑟 =
1, as we can for most of optics .)

Also, we have considered the propagation of light in free space and linear isotropic media , but we have no
picture of why light can be emitted (i.e. generated ).The flip side of that is that we also have no microscopic
picture of optical absorption .

The so-called Lorentz model, or classical electron oscillator (CEO) model, gives a purely classical picture of
both the index of refraction and the emission + absorption of light .This classical model turns out to still have a
few “free parameters“ which cannot be calculated from first principles , but must be obtained from experiments,
as we shall see. Those free parameters (“resonant frequency” and “oscillator strength”) can only be predicted on
the basis of a quantum-mechanical theory of matters.

The starting point if the model is the recollection that electromagnetic waves are generated by accelerating charge
or, equivalently, by the time-varying currents. We all know that antennas (e.g. dipole antennas) are useful for
generating e.m. waves in the radio frequency region of the spectrum .However, there are no antennas for optical-
frequency e.m. waves!! Light is generated by atoms or molecules, not tiny metal antennas!

Lorentz proposed that atoms be considered to be small (<<λ in size) harmonically oscillating dipoles.

Dipole moment P = er , r = electron-nucleus separation, and resonant frequency, 0 determined from

experiment.

Armed with these results, we can now consider an electromagnetic wave incident on a single atom, which we
assume has a resonant frequency  0 . The field will induce a polarization in the atom, and the atom will therefore
radiate.
Def. mass +charge 𝑚𝑒 , -e for electron ; 𝑚𝑛 , +e for nucleus

ren = re − rn = relative position vector

Fne = − Fen ; Fen = force on e − due to n (i.e. Harmonic restoring force)
The e.m. field exerts a force on each charge
𝐹⃑ = 𝑞(𝐸⃑⃑ + 𝑉
⃑⃑ × 𝐵
⃑⃑) ≅ 𝑞𝐸⃑⃑

Since the 𝑉 ⃑⃑ term is negligible at nonrelativistic speeds (is smaller by V / c => negligible for optical fields
⃑⃑ × 𝐵
in linear response regime. It is important in experiments with terawatt lasers, but in that regime the Lorentz
model is nonsensical
Newton’s law:
d 2 re
me 2 = −eE ( re, t ) + Fen( ren )
dt
d 2 rn
mn 2 = eE ( rn , t ) + Fne( ren )
dt
Define center-of-mass coordinates:
mere +mnrn
R= ; x = ren
me +mn
mn
re = R + x
M
me
rn = R- x
M
M = me + mn
Substitute these into Newton’s equation:
d 2R d 2x mn
me 2
+ m 2
= −eE ( R + x , t ) + Fen ( x )
dt dt M
memn
m=
Where me +mn = reduced mass

d 2R d 2x me
mn 2
− m 2
= eE ( R − x , t ) + Fne( x )
dt dt M
We can add and subtract these two equations to separate R and x :

Add => Lead to equation (1):

d 2R me mn
M 2
= e[ E ( R − x, t) − E ( R + x , t )]
dt M M

Subtract => Lead to equation (2):

Discussion equation (1)

th
Consider the J component of the field, and Taylor expand to first orders:
dEi ( x )
Ei ( x + x ) = Ei ( x ) + x
dx
Now in 3-D, the slope along x is x̂  E , so in general we have
𝐸⃑⃑(𝑅⃑⃑ + ∆𝑟⃑) ≅ 𝐸⃑⃑(𝑅⃑⃑) + ∆𝑟⃑ ∙ ∇𝐸⃑⃑

Thus =>
The justification for keeping only the first term is that for optical fields 𝜆 ≅ 5000 Å
but 𝑥 ≅ 1Å，so the approximation is equivalent to the dipole approximation.
The center-of-mass eqn. now becomes
d 2R me mn
M 2
= e[ − x  E ( R , t ) − x  E ( R, t )]
dt M M
d 2R
M 2 = −exE ( R, t )
Or dt

The center of mass of the atom is thus accelerated by the gradient of the field

This force exerted by spatially varying fields is actually quite useful for trapping and cooling atoms.
In fact, we shall see shortly that the dipole moment of a Lorentz atom is proportional to the applied field, so

P = −ex =  E

Where the proportionality constant  is called the polarizability.

d 2R 
M 2
= P  E =  E   E =  ( E  E )
Then dt 2

But
( )
( E  E ) = E  E + E  E = 2 E  E

F= E
2

Or 2

Which is the force on the center of mass of the atom. It just follows the gradient of the intensity!
Now P =  E for any kind of small (sub-  ) dielectric particle in a weak electric field
E,g. For a dielectric sphere with index n in a medium with index nm , def. nr = n / nm
nr 2 − 1
 = nm 2a 3 ( ）
nr 2 + 2
So if nr  1,   0 , and this particle can be trapped

E.g. Gaussian beam at focus

“Optical tweezers”
Discussion equation (2)

But what happens if we subtract equation (2) from equation (1)? We will focus on this and this is very
important!!

Subtract (1)-(2) =>

The above equation can be written as:

The second and forth terms in the right hand side are negligible in the dipole approximation (i.e. in keeping
only the first term in the Taylor expansion, we are assuming they are small)

Numerical justification:
𝑒𝑥⃑ ∙ ∇𝐸⃑⃑ ≅ 𝑒𝑎0 𝑘𝐸, where a0 = Bohr radius
𝑒𝑥⃑∙∇𝐸⃑⃑ 𝑒𝑎0 𝑘𝐸 2𝜋𝑎0 6Å
𝑒𝐸
≅ 𝑒𝐸
≅ 𝑎0 𝑘 = 𝜆
= 6000 Å = 0.001
d2R
Thus, the gradient form is 1000 times less than the direct force eE on the electron. Also, is small,
dt 2
since the force exerted on it is small (and the total mass M is big ).
d2 x e 1
Thus 2
= − E ( R, t ) + Fen ( x )
dt m m
Lorentz: assume a harmonic restoring force for small displacements x :
Fen ( x ) = −kx  −m02 x
Where 0  k / m is the oscillator’s natural frequency. (note m = reduced mass )
d2 x e
=> 2
= − E − 02 x
dt m
As we have seen, the oscillator must be damped, since the oscillator dipole radiates power
 add a “friction” term (as usual  velocity )
dx
Fdamping = −m
dt
 Lorentz eqn. of motion for a single driven atom

d2 x dx eE
2
+ + 02 x = − <- Now, we will talk about the physical meaning of this equation.
dt dt m
1. Polarization and energy decay (still this is a single dipole picture)

The free-field (homogeneous) solution is that of an oscillating atom with E = 0.

Try a solution of the form
x (t ) = x (t0 )e( +i )t
dx
= ( + i ) x (t0 )e( +i )t
dt
d2 x
= ( + i ) 2 x (t0 )e( +i )t
dt 2
Plug these into the Lorentz eqn. and divide out x (t0 )e( +i )t
( + i )2 +  ( + i ) + 02 = 0
(2 −  2 +  + 02 ) + i(2 +  ) = 0
The real and imaginary parts must separately=0:
Im:  (2 +  ) = 0   = − / 2 amplitude decay rate

Re:

 
2

Or  =  −  2
0 Oscillation frequency (renormalized by damping)
2
Thus the dipole oscillates at a frequency nearly equal to the nature frequency 0 ,but shifted slightly
downward due to the damping.

For all cases we might consider (indeed nearly all cases in optics)
𝛾
2
≪ 𝜔0
Typical numbers:
𝛾
2
≅ 106 − 108 𝑠 −1
𝜔0 ≅ 1015 𝑠 −1

Thus the frequency shift is negligible, and the solution is :


x (t ) = x (t0 ) exp[− (t − t0 ) + i0 (t − t0 )]
2

Note that the dipole moment (or atomic polarization) P = −ex decays at a rate
2
The energy decays with a rate  :
1 1
U (t ) = m02 x 2 + mx 2 = U (t0 )e − (t −t0 )
2 2
(since the energy goes as the square of the polarization .
Normally, we identify  =  rad .In many circumstances, these can also be nonradiative energy relaxation  nr ,
so  =  rad +  nr
2. Macroscopic Polarization
Our goal is to understand the propagation of light in a uniform medium. Therefore we must extend the
treatment from consideration of the polarization of a single atom to determine the macroscopic polarization
induced in an uniform medium containing N dipoles per unit volume .
The simplest approach, taken by most authors, is to write
P = − Nex N = total # dipoles /volume
However, this is not strictly true in the classical model.
The correct way to obtain P is
1 NV
P=  j (note: the vector sum )
V j =1
Why?
(1) Not all the dipoles necessarily point in the same direction ; collisions can reorient them.
(2)The dipole may not all have the same phase .
e.g. a snapshot in time


e NV e − t + i0t + j
=> P(t ) = −  x j (t ) = −  x0 j e 2
V j =1 V j
x0i direction + amplitude
i phase
Collisions or coupling to vibrations (phonons) in a solid will generally cause x0i to change directions at
random times, and i to make random jumps at random times.
Note: if phases and directions are random, then

Even though for any specific dipole x  0

Thus in general P(t )  Ne x (t )
See the next page for N=3 case.
Note: amplitudes of individual oscillators unchanged!

Now suppose that at time t=0,all the dipoles are aligned and in phase (e.g. by a short pulse of light).
Suppose that every time a collision occurs, it randomizes the phase (and /or direction)
Thus, for all the dipoles which have undergone collisions,  P = 0
In other words, we consider
N = Nin + Ncoll =constant
Where Nin = # dipoles in phase (have not yet collided)
N coll = # dipoles which have collided
At time t=0, Nin = N . As time increases, N in  as N coll  .The macroscopic polarization is just
P = −eNin x
Terminology: these collisions are called dephasing collisions, since they randomize the individual dipoles’
phases, thereby reducing the macroscopic polarization.
Note that collisions might (and usually do) cause the dipoles to dephase without actually changing the
amplitude (and hence the energy) of any individual dipole. We must discriminate between dephasing
collisions and energy-changing collisions.

1
Suppose dephasing collisions occur with a rate
T2
Then the number of collisions dNc occurring in a time interval dt is
dt dt
dNc = Nc  dNin = − Nin
T2 T2
dNin 1
=> = − Nin  Nin (t ) = N 0e−t /T2
dt T2
 1
− ( + )t
=> P(t ) = − Nexe 2 T2 i0t
e  − Nexe−t /T2 ei0t
Therefore the macroscopic polarization decays at a rate
1  1
= +
T2 2 T2
A common notation in the literature defines
1
=  =  rad +  nr = energy decay rate
T1
1
= polarization decay rate (Remember: no energy decay here)
𝑇2′
In general, 𝑇2′ < 2𝑇1

Typical numbers 𝜔0 ≅ 1015 𝑠 −1

Atoms in a gas:
𝑇1 = 𝛾 −1 = 10−6 𝑡𝑜 10−6 𝑠
𝑇2′ = 10−8 𝑡𝑜 10−10 𝑠

Atoms in a solid:
𝑇1 = 𝛾 −1 = 10−3 𝑡𝑜 10−12 𝑠
𝑇2′ = 10−13 𝑡𝑜 10−16 𝑠

Now we can go back to the Lorentz eqn, for a single dipole

d2 x dx eE
2
+ + 02 x = −
dt dt m
The equation of motion for the macroscopic polarization is obtained by multiplying through by − Ne
(from P = − Nex ) and putting in the correct form of the damping:
d 2P 2 dP Ne2
+ ( + ) + 0 P =
2
E
dt 2 T2 dt m
Now we can solve to find the polarization induced in the material by an applied electromagnetic field.
3. Steady – State Solution (nonhomogeneous equation)

Assume a simple harmonic incident field:

E (t ) = ˆ E0eit
Where ˆ = polarization vector
E0 = complex amplitude
(as usual ,the real part is understood; E0 is complex => determines the phase of the field )
We therefore expect a solution of the form
P(t ) = ˆ P0eit ( P0 also complex; may have a different phase from E0 !)
dP d 2P
= i P 2
= − 2 P
dt dt
 after dividing out ˆ0eit on both sides ,we get the scalar equation for the amplitudes:
2 Ne2
− 2 P0 + i ( + ) P0 + 02 P = E0
T2 m

( )
P0 [ 02 − 2 + i ( +
2
)] =
Ne2
E0
T2 m
Ne 2 1
Or P0 = [ ]E0
( )
m  2 − 2 +i ( + 2 )
0
T2
Note:
(1) The amplitude of the induced polarization is linearly proportional to the driving field
amplitude
(2) The proportionality constant is complex, so there is a phase shift between the field
and polarization due to the damping of the polarization (i.e. finite T2 ).
The usual procedure is to define
Ne2 1
P =  0  E or  =
 0 m  2 − 2 +i ( + 2 )
0( ) T2
Where  = electric susceptibility (dimensionless)

Now we have half the story. We have determined the polarization induced by the field (atomic response to
the field ), but now we must find self-consistently the effect of that polarization back on the field – i.e. the
polarization is a source term in the wave eqn. Thus to understand the propagation of light in matter, we
must simultaneously solve the Lorentz eqn. and the wave eqn.

Wave eqn. with source:

B
As usual, begin with  E = −
t
  
(
  E = − )
t
(
  B = − 0 )
t
 H ( )
  D  2
( )
   E −  2 E = − 0  
t  t 
= −  0
t 2
0E + P ( )
2 E 2 P
2 E − 0 0 = 
t 2 t 2
0

This shows that a time-dependent polarization acts as a source for electromagnetic waves.
For weak fields (linear response) P =  0  E
For a harmonic driving field E =  (r )eit we have the Helmholtz eqn.
1
2 E (r ) − 0 0 (1 +  ) 2 E (r ) = 0  0 0 =
c2
The form of this equation is familiar, so we can try plane wave solutions
E = ˆ E 0 ei (t −k r )
2
Plug in => −k 2 + (1 +  ) = 0
c2
2
Or k =
2
(1 +  )
c2
As usual, one can define a complex index of refraction n so that
n2 2
k2 = where n2 = 1 + 
c2
Just as we saw with metals, a complex wavevector (or complex index, or complete dielectric constant
– those are all just different ways of saying the same thing ) leads to damping .We will come back to consider
absorption in some detail , but first we should take a look at the traditional ,familiar real index of
refraction .This is important when light is propagating through nonabsorbing, i.e., transparent media (e.g.
glass !)