6.

772/SMA5111 - Compound Semiconductors

Supplement 1 - Semiconductor Physics Review - Outline
• The Fermi function and the Fermi level
The occupancy of semiconductor energy levels
• Effective density of states
Conduction and valence band density of states
1. General
2. Parabolic bands

• Quasi Fermi levels

The concept and definition
Examples of application
1. Uniform electric field on uniform sample
2. Forward biased p-n junction
3. Graded composition p-type heterostructure
4. Band edge gradients as effective forces for carrier drift

Refs: R. F. Pierret, Semiconductor Fundamentals 2nd. Ed., (Vol. 1 of the Modular Series
on Solid State Devices, Addison-Wesley, 1988); TK7871.85.P485; ISBN 0-201-12295-2.
S. M. Sze, Physics of Semiconductor Devices (see course bibliography)
Appendix C in Fonstad (handed out earlier; on course web site)
C. G. Fonstad, 2/03 Supplement 1 - Slide 1
 Fermi level and quasi-Fermi Levels - review of key points
Fermi level: In thermal equilibrium the probability of finding an
energy level at E occupied is given by the Fermi function, f(E):
[E -E f ]/ kT
f (E) = 1 (1 + e )
where Ef is the Fermi energy, or level. In thermal equilibrium Ef is
constant and not a function of position.

The Fermi function has the following useful properties:

-[E -E f ]/ kT
f (E) ª e for (E - E f ) >> kT
[E -E f ]/ kT
f (E) ª 1- e for (E - E f ) << -kT
f (E f ) = 1/2 for E = E f
These relationships tell us that the population of electrons decreases
exponentially with energy at energies much more than kT above the
Fermi level, and similarly that the population of holes (empty
electron states) decreases exponentially with energy when more than
kT below the Fermi level.
C. G. Fonstad, 2/03 Supplement 1 - Slide 2
 A final set of useful Fermi function facts are the values of f(E) in the
limit of T = 0 K: lim f (E) = 1 for E < E f
T=0

f (E ) = 1/2 at E = E f
lim f (E) = 0 for E > E f
T=0

Effective densities of states: we can define an effective density of

states for the conduction band, Nc, as
•
Nc ≡ Ú Ec
r (E)e-[E -E c ]/ kT dE

and an effective density of states for the valence band, Nv, as

Ú-•
Ev [E -E v ]/ kT
Nv ≡ r (E)e dE
where r(E) is the electron density of states in the semiconductor.

C. G. Fonstad, 2/03 Supplement 1 - Slide 3

 If the energy bands are parabolic, i.e., when the density of states
depends quadraticly on the energy away from the band edge, we find
simple relationships between the densities of states and the effective
masses:
If r (E) = 2(m*e ) 3 ( E - E c ) p 2 h 3 when E > E c ,

then N c = 2 [2p m kT h *
e ]
2 3/2

and if r (E) = 2(m*h ) 3 ( E v - E ) p 2 h 3 when E < E v ,

then N v = 2 [2p m kT h *
h ]
2 3/2

When (Ec-Ef)>>kT, we can write the thermal equilibrium electron

concentration in terms of effective density of states of the conduction
band and the separation between the Fermi level and the conduction
band edge, Ec, as: n (x) = N (x)e-[ E c (x )-E f ]/ kT
o c

Similarly when (Ef-Ev)>>kT we can write: -[ E f -E v (x )]/ kT

po (x) = N v (x)e
Note: In homogeneous material Nc, and Nv do not depend on x.
C. G. Fonstad, 2/03 Supplement 1 - Slide 4
 Quasi-Fermi levels: When a semiconductor is not in thermal
equilibrium, it is still very likely that the electron population is at
equilibrium within the conduction band energy levels, and the hole
population is at equilibrium with the energy levels in the valence
band. That is to say, the population on electrons is distributed in the
conduction band states with the Boltzman factor: e-[E -E fn ]/ kT
Here Efn is the effective, or quasi-, Fermi level for electrons.
Similarly, there is a quasi-Fermi level for holes, Efp, and the holes are
distributed in the valence band states as: e-[E fp -E ]/ kT
The quasi-Fermi levels for electrons and holes, Efn and Efp, are
not in general equal. To find them we usually begin with n(x) and
p(x), and write them in terms of the conduction and valence band
densities of states and the quasi-Fermi levels:
-[E (x)-E (x )]/ kT
For example, we write n(x) = N c (x)e c fn

This then defines E fn : E fn (x) ≡ E c (x) - kT ln[N c (x) / n(x)]

We define E fp similarly : E fp (x) ≡ E v (x) + kT ln[N v (x) / p(x)]

C. G. Fonstad, 2/03 Supplement 1 - Slide 5

 Quasi-Fermi levels, cont.:
A very important finding involving quasi-Fermi levels is that we
can write the electron and hole currents in terms of the gradients of
the respective quasi-Fermi levels, at least in the low field limit where
drift mobility is a valid concept. We find:
J n (x) = m e n(x) ∂E fn (x) ∂x
and
J p (x) = m h p(x) ∂E fp (x) ∂x
Examples:
A. Uniformly doped n-type semiconductor with uniform E-field
At low to moderate E-fields, the populations are not disturbed
from their equilibrium values, and we have
n(x) ª n o ª N d and p(x) ª po = n i2 /N d
Also, E fn ª E fp ª E f - qf (x), so :
J e ª m e n o (-q ∂f ∂x) = qm e n o Fx and J e ª qm h po Fx
As expected, the currents are the respective drift currents.

C. G. Fonstad, 2/03 Supplement 1 - Slide 6

 B. P-side of forward biased n+-p junction, long-base limit:
Diode diffusion theory gives us n(x) on the p-side*:
n(x) = n op [(e qv ab / kT -1)e-x / Le + 1], where n op = n i2 N Ap
When vAB >> kT, and x is not many Le, we can approximate
n(x) as : -x / L e qv ab / kT -x / L e
n(x) = n op [(e
qv ab / kT
-1)e + 1] ª n op e e
from which we find:
E fn (x) = E c + kT ln[n(x) / N c ]
ª E c + kT ln[n o /N c ] + qv ab - kT x /Le
ª E fo + qv ab - kT x /Le
We see that Efn(x) is qvAB higher than the equilibrium Fermi level,
Efo, at the edge of the depletion region on the p-side, and decreases
linearly going away from the junction. Farther away from the junction,
where x is many Le, n(x) approaches nop, and Efn(x) approaches Efo.
Finally, notice that for low-level injection, p(x) ≈ ppo, and Efp ≈ Efo.

C. G. Fonstad, 2/03 Supplement 1 - Slide 7

 Quasi-Fermi levels - Illustrating examples A and B
Figure C-8 from Fonstad, Microelectronic Devices and Circuits with
quasi-Fermi levels added:

Example A:

Efn ≈ Efp ≈ Efo

Example B:

Efn

Efn

C. G. Fonstad, 2/03
Supplement 1 - Slide 8
 C. Graded composition p-type heterostructure with uniform low
level electron injection.
Assume the grading is from Eg1, X1 @ x = 0, and Eg1, X1 @ x = L.
E g (x) = E g1 + x(E g 2 - E g1) / L; c (x) = c1 + x( c 2 - c1 ) / L
In thermal equilibrium the Fermi level, Efo, is flat, and the valence
band edge is flat:
E v = E fo - kT ln(N v /N Ap )
whereas the conduction band edge slopes: E c (x) = E v - E g (x)
With low-level electron injection, n(x) ≈ n’ (>>npo):
Hole population is changed insignificantly, and Efp(x) ≈ Efo
Electron population is now n’, and so
E fn (x) = E c + kT ln[n' / N c ] ª E v + E g (x) + kT ln[n' / N c ]
Using this to get Je(x), we find
J e (x) ª m e n' ∂E g (x) ∂x = q m e n' Fe,eff , where Fe,eff ≡ q-1∂E g (x) ∂x
From this we see that the band gap grading acts like an effective
electric field acting on the electrons (but not on the holes)!
C. G. Fonstad, 2/03 Supplement 1 - Slide 9
 Quasi-Fermi levels - Illustrating example C

Example C:
Fe,eff ≡ q-1∂E c (x) ∂x = - (E g1 - E g 2 ) /qL

Ec(x)

Eg1
L Eg2
Efo
Ev(x)

C. G. Fonstad, 2/03 Supplement 1 - Slide 10

 D. General meaning of band-edge gradings
In general we can write the electron quasi-Fermi level as:
E fn (x) = E c (x) + kT ln[n(x) / N c (x)]
and thus in general we can write the electron current as:
J e (x) = m e n(x) ∂E fn (x) ∂x
∂E c (x) ∂n(x) n(x) ∂N c (x)
= m e n(x) + m e kT + m e kT
∂x ∂x N c (x) ∂x
typically
q-1∂E c (x) Ê ∂n(x) n(x) ∂N (x) ˆ small
= se + De Á + c
˜
∂x Ë ∂ x N c (x) ∂ x ¯
Drift Term Diffusion Term
[Note: In getting this we have used the Einstien relation and definition
of conductivity:
m e kT = q De , and s e = q m e n(x) ]
From our final result we see that the gradient in the conduction band
edge is the force leading to electron drift, while the gradient in the
carrier and density of states concentrations are the diffusion force.
C. G. Fonstad, 2/03 Discussion continued for holes on next foil. Supplement 1 - Slide 11
 D. cont.
We obtain the corresponding result for holes if we similarly
substitute valence band quantities for conduction band quantities.
Begin with:
E fp (x) = E v (x) + kT ln[ p(x) / N v (x)]
and thus
J h (x) = m h p(x) ∂E fp (x) ∂x
∂E v (x) ∂p(x) p(x) ∂N v (x)
= m h p(x) + m h kT + m h kT
∂x ∂x N v (x) ∂x
typically
-1
q ∂E v (x) Ê ∂p(x) p(x) ∂N v (x) ˆ small
= sh + Dh Á + ˜
∂x Ë ∂x N v (x) ∂x ¯
Drift Term Diffusion Term
Now we see that the gradient in the valence band edge is the force
leading to hole drift, while the gradient in the carrier and density of
states concentrations are the diffusion force.
Summarizing, the conduction and valence band-edge gradients can
be viewed as effective electric fields for electrons and holes, respectively:
Fe,eff = q-1∂E c (x) ∂x and Fh,eff = q-1∂E v (x) ∂x
C. G. Fonstad, 2/03 Supplement 1 - Slide 12