EE 223

Applied Quantum Mechanics II

David Miller
Optical absorption in
semiconductors
Optical absorption in semiconductors

Optical absorption in semiconductors

is a very important practical process
e.g., digital cameras, telecommunications
photodetectors, solar cells
With our good model for crystalline semiconductors
and knowing Fermi’s Golden Rule from time-
dependent perturbation theory
we are ready to calculate a simple case of this
absorption process
and to illustrate the use of Fermi’s Golden Rule
Direct optical absorption
conduction
The transition rate for absorption band

from an initial electron state  i

with energy Ei Ef f
to a final state  f with energy Ef E ?
with an oscillating perturbation of i valence
angular frequency  Ei band

is, from Fermi’s Golden Rule

2
  E f  Ei   
2
wabs   f Hˆ po  i  0 
  k
a a
Optical absorption in
semiconductors

Perturbing Hamiltonian
Direct optical absorption
2
 f Hˆ po  i   E f  Ei   
2
In Fermi’s Golden Rule wabs 

Hˆ po is the amplitude of a perturbation oscillating in
time at (angular) frequency , such as is defined in
Hˆ p  t   e E  t  z  Hˆ po exp  it   exp  it  
where E is the electric field in the z direction
We are also now interested in the spatial variation of Hˆ po
through the spatial dependence of the
electromagnetic wave amplitude, as explicitly in
Hˆ p  r, t   Hˆ po  r  exp  it   exp  it  
Direct optical absorption

The matrix element in Fermi’s Golden Rule,  f Hˆ po  i

can now be written explicitly as
 Hˆ    *  r  Hˆ  r   r  d 3r
f po i  f po i

where  i (r ) and  f (r ) are, respectively,

the wave functions of the initial and final states
and we explicitly allow Hˆ po  r  to vary in space
because the electromagnetic field at
(angular) frequency  will vary in space
Perturbing Hamiltonian for electromagnetic field
Though we could continue with the “electric dipole”
version of the perturbing Hamiltonian Hˆ p  r, t   e E  r, t  z
we can connect to other semiconductor phenomena
if we switch to using the magnetic vector potential
A instead of the electric field, and write
e
Hˆ p  r, t    A  pˆ
mo
where mo is the usual free electron mass and
pˆ  i is the momentum operator
(Note: both the electric and magnetic fields can be
deduced from the vector potential A)
Perturbing Hamiltonian for electromagnetic field

Now we write the vector potential of a wave at

frequency 
 A0 
A  e  exp i  k op  r  t    exp  i  k op  r  t   
A0
   
2 2 
Here kop is the wave vector of the optical field inside
the material, and
we take the field to be linearly polarized
with its electric vector in the direction of the unit
vector e
Perturbing Hamiltonian for electromagnetic field
 A0 
Now in A  e  exp 
 i  k op  r   t  
 
A0
exp 
 i  k op  r   t  
 
2 2 
we are going to keep only the term in exp  it 
because we know from our previous discussion of
Fermi’s Golden Rule that this term
corresponds to absorption rather than emission
So the spatial part of our perturbing Hamiltonian is
Hˆ po  r     e / 2mo  A0 exp  ik op  r  e  pˆ
I. e., the total perturbing Hamiltonian is
Hˆ p (r, t )  Hˆ po  r  exp(it )
Optical absorption in
semiconductors

Direct transitions
Direct optical absorption
conduction
To proceed, we need to know band
the initial and final wavefunctions
We are most interested in the Ef f
transitions between E ?
an initial valence band state  i and i valence
a final conduction band state  f Ei band

We presume that we can write them

as “single-electron” Bloch states
and to calculate matrix elements  0 
 k
we must normalize them a a
Normalizing Bloch functions

With normalization constants Bi and Bf, we have

 i  r   Bi uv  r  exp  i k v  r 
and
 f  r   B f uc  r  exp  i k c  r 
Here, and below, for simplicity
we presume uv and uc do not depend on k
a good approximation for an “allowed” process
so we omit the “k” subscripts on uv and uc
We do, however, now explicitly allow the conduction (uv)
and valence (uc) unit cell functions to be different
 Normalizing Bloch functions

We choose uv (r ) and uc (r ) to be normalized over a unit cell

 u *
v  r
unit cell
 u v  r  d 3
r  1 and similarly for uc (r )
Hence, normalizing  i (r ) and  f (r ) , we have, e.g., for  i (r )

 i

V
*
 r  i  r  d 3
r  1  Bi 
2
u *
v
V
 r  exp  i k v  r  u v  r  exp  i k v  r  d 3
r

 Bi2  uv*  r  uv  r  d 3r  Bi2 N  uv*  r  uv  r  d 3r  Bi2 N

V unit cell

where N is the number of unit cells and

V is the volume of the crystal
Hence we have Bi  B f  1/ N
 Matrix element for Fermi’s Golden Rule

With Hˆ po  r     e / 2mo  Ao exp  ik op  r  e  pˆ

1 1
 i r   uv  r  exp  i k v  r  and  f  r   uc  r  exp  i k c  r 
N N
the matrix element for Fermi’s Golden Rule is
 f Hˆ po  r   i 


eAo

2mo N V

 u *
c  r  exp   i k c  r  
 exp  i k op  r  e  ˆ
p 
 u v  r  exp  i k v  r  
 d 3
r
Matrix element for Fermi’s Golden Rule
We are interested in transitions involving states near the
center of the Brillouin zone, so
k v and k c are both   / a
( /a corresponds to the edge of the zone)
Though strictly, pˆ  i operates on all of uv  r  exp  ik v  r 
with these small values of kv
exp  ik v  r  changes very slowly compared to the
rate of change of uv  r 
so, at least as a first approximation, we take
pˆ uv  r  exp  ik v  r   pˆ uv  r  exp  ik v  r    exp  ik v  r  pˆ uv  r 
 Matrix element for Fermi’s Golden Rule
For definiteness, we choose the polarization direction
(i.e., unit vector e) in the x direction
x will be one of the directions perpendicular to the
propagation of the electromagnetic wave
With this choice and our approximations so far
 f Hˆ po  r   i 


eA0
2mo N V  c
 u *
 r  exp   i k c  r  
 exp  op   v
i k  r e  ˆ
p  u  r  exp  i k v  r  
 r
d 3

exp i  k v  k c  k op   r  uc*  r  pˆ x uv  r   d 3r

eA0

2mo N V
Matrix element for Fermi’s Golden Rule

The optical wave vector kop corresponds to wavelengths

of 100's of nm or more (even inside the material)
A typical crystal unit cell size is, e.g., a ~ 0.5 nm
so exp  ik op  r  is slowly varying over a unit cell
Hence, since we have already restricted k v and k c to
being small by assumption
then the entire factor
exp i  k v  k c  k op   r 
varies slowly over the length scale, a, of a unit cell
 Matrix element for Fermi’s Golden Rule
As a result, we can approximately separate the integral
 f H po  r   i  
ˆ eAo
2mo N V
exp 
 i  k v  k c  k op   r 


 u *
c  r  ˆ
p u
x v  r  
 d 3
r

into a sum of integrals over a unit cell

treating the value of exp i  k v  k c  k op   r 
as approximately constant within a unit cell, i.e.,
exp i  k v  k c  k op   R m 
eAo
 f Hˆ po  r   i  
2mo N
c pˆ x v 
m ( i .e., unit cells )

where Rm is the position of (the center of) the mth unit cell
and c pˆ x v   uc*  r  pˆ x uv  r  d 3r  pcv
unit cell
 Matrix element for Fermi’s Golden Rule
The summation in
exp i  k v  k c  k op   R m 
eAo
 f H po  r   i  
ˆ
2mo N
c pˆ x v 
m ( i .e., unit cells )

will average approximately to zero unless k v  k c  k op  0

because otherwise
the function exp i  k v  k c  k op   r  is oscillatory
(Note this condition k v  k c  k op  0 can be seen to
correspond to conservation of crystal momentum k )
In this case, the sum becomes
m exp i  k v  k c  k op   R m   m exp(0)  N
 Fermi’s Golden Rule transition rate

Hence
exp i  k v  k c  k op   R m 
eAo
 f H po  r   i  
ˆ
2mo N
c pˆ x v 
m ( i .e., unit cells )

eAo
becomes  f H po  r   i  
ˆ pcv
2mo
and Fermi’s Golden Rule gives a transition rate
2
 f Hˆ po  i   E f  Ei   
2
wabs 

2 e 2 Ao2
 pcv   E f  Ei   
2

 4mo 2
Direct valence to conduction band absorption

Hence, for an incident optical wave of frequency ,

wavevector kop and magnetic vector potential
amplitude Ao
for an initial state of energy Ei and Bloch wavevector
kv in the valence band
and a final state of energy Ef and Bloch wavevector kc
in the conduction band
if and only if k v  k c  k op  0 , we have a “direct
transition rate” 2 e 2 Ao2
wabs  pcv   E f  Ei   
2

 4mo 2
Direct valence to conduction band absorption

To summarize so far
we have made use of the facts that
(i) the electron wavefunctions in the crystalline
semiconductor are in the Bloch form, and
(ii) the optical wavelength is much larger than the
unit cell size
Note that the concept of conservation of (crystal)
momentum k emerged automatically
and is consistent with k being the effective electron
momentum in a Bloch state
Direct valence to conduction band absorption

Note also that the transition rate

2 e 2 Ao2
wabs  pcv   E f  Ei   
2

 4mo 2

is proportional to
2
a (squared) matrix element, cv p
the optical intensity (which is proportional to Ao2 )
and hence to the average arrival rate of photons
in the semiconductor (per unit area)
Optical absorption in
semiconductors

Total transition rate

Total transition rate

2 e 2 Ao2
 pcv   E f  Ei   
2
wabs
 4mo 2

is transition rate from

an initial valence band state with Bloch wavevector kv
to the conduction band state with wave vector
k c  k v  k op
To get the total transition rate
and hence the optical absorption rate of photons
we need to sum over all the possible initial and
final states
Total transition rate

The total transition rate WTOT is formally

2 e 2 Ao2
WTOT   wabs  pcv    E f  Ei   
2

i f  4mo 2
i f

where we presume pcv is ~ independent of k

2

We have shown above that,

for a given initial state with wave vector kv
the only final state possible is the conduction band
state with k c  k v  k op
With this knowledge
we can drop the separate sum over final states
Total transition rate

Since k op is generally a very small fraction of the size of the

Brillouin zone
we will now neglect it for simplicity, taking k c  k v
This negligible size of the optical wavevector means that
the direct optical transitions are essentially “vertical” on
the energy-momentum diagram
Hence, for a given initial state kv
only one term remains in the sum over the final states
namely, the one with k c  k v
From now on, we drop the suffixes “v” and “c”, using just k
Total transition rate

Hence the total transition rate becomes

2 e 2 Ao2
WTOT  pcv    Ec  k   Ev  k    
2

 4mo 2
k , spin

now including a sum over the two possible spin states

Now we formally rewrite (considering unit volume)
  
 2
k , spin k
 2 g (k ) d 3
k
k
where g  k  is the density of states in k space
We will next change variables in the integral
to the energy Ec  k   Ev  k 
Direct optical absorption
conduction
Assuming parabolic bands band
with meffe and meffh as the hole
and electron effective masses Ec  k 
here both positive by
EJ (k )
convention
we can define Ev  k 
EJ (k )  Ec (k )  Ev (k ) valence
band
2k 2  1 1 
     Eg  0 k 
2  meffe meffh  
a a
Joint density of states
1 1 1
Defining a “reduced effective mass” eff ,  
we can rewrite eff meffe meffh

2k 2  1 1  2k 2
EJ (k )  Ec (k )  Ev (k )      Eg   Eg
2  meffe meffh  2 eff
We can then define a “joint density of states” g J  EJ 
the number of transitions per unit (photon) energy
Like the energy density of states in a band, we can write
g J ( EJ )dEJ  2 g (k )d 3k
including a factor of 2 for spin
Joint density of states

We now follow mathematically identical arguments to

those used to deduce
the density of states in energy
from the density of states in k-space
but now for this joint density of states
obtaining, for EJ  Eg

1  2 eff 
3/2

E  Eg 
1/2
g J ( EJ )  2  2 
2   
J
 Total transition rate
2 e 2 Ao2
So, from WTOT     Ec  k   Ev  k    
2
pcv
 4mo 2
k , spin

using   2  2  g (k )d 3k   g J  EJ  Eg  dEJ
k , spin k k EJ  Eg

1  2 eff 
3/2

E  Eg 
1/2
with g J ( EJ )  2  2 
2   
J

then
1  2 eff 
3/2
2 e A
2 2

 E  Eg    EJ    dEJ
1/2
 0 2
2  2 
WTOT pcv
 4m 2
0 EJ  Eg
2   
J
 Total transition rate

Integrating to eliminate the delta function in

1  2 eff 
3/2
2 e A2 2
pcv  E  Eg    EJ    dEJ
1/2
 o 2
2  2 
WTOT
 4m 2
o EJ  Eg
2   
J

we finally obtain, for   Eg

2 1  2  eff 
3/2
2 e A 2 2

   E g 
1/2
WTOT ( )  pcv o
2  2 
 4m 2   
2
o

Hence we finally see how Fermi’s Golden Rule can be

used with an appropriate density of states
to give a transition rate
Optical absorption in
semiconductors

Absorption coefficient
Absorption coefficient

The final step is to relate the absorption coefficient, 

the probability of absorption of a photon per unit length
to the total transition rate per unit volume WTOT
With photon energy 
the number of photons incident per unit area per second
is, by definition n p  I / 
where I is the optical intensity (power per unit area)
So the probability a photon is absorbed per unit length is
W  WTOT
  TOT 
np I
Absorption coefficient

From electromagnetism, we can deduce the intensity

from the vector potential amplitude through
nr c 0 2 Ao2
I
2
where
nr is the refractive index
c is the velocity of light, and
o is the permittivity of free space (electric constant)
 Absorption coefficient
Hence

3/2
 2  2 e Ao 1  eff 
2 2

   E 
2 1/2
 (  )  WTOT 
2
2  2 
pcv
nr c 0 Ao  4mo 2   
2 2 2 g
I
2  2  eff 
2 3/2

pcv  2     Eg 
e 1 1/2

2 mo c o nr
2
  
The parameter E p   2 / mo  pcv (~ 20 eV in many
2

semiconductors) is often used in calculations

in which case we can rewrite
1 E p  2 eff 
3/2
e 2
      2     E g 
1/2

4 mo c o nr    
Optical absorption

Absorption coefficient
In direct gap semiconductors
the optical absorption in this
    E g 
1/2

model
rises as    Eg  above
1/2

the bandgap energy Eg

This model is used as the
starting point for optical
calculations in
Eg
semiconductors
Photon energy 