10/1/2012

EE415/515 Fundamentals
of Semiconductor Devices
Fall 2012

Lecture 3:
Density of States, Fermi Level
(Chapter 3.4-3.5/4.1)

Density of States
• Need to know the density of electrons, n, and 
holes, p, per unit volume
• To do this, we need to find the density of 
permitted energy states and then the 
probability (given by the Fermi function) that 
these states are occupied by an electron
• See Neamen Sections 3.4 and 3.5 for 
derivations, and compare these with 
alternative approaches given below
J.E. Morris ECE415/515 Fall 2012 2

1
 10/1/2012

Density of States

•See figure in “k-space.”

•Also: p=ħ k=(h/2π)k, i.e.

•Direct linear relationship between (wave-vector) k-space and (momentum) p-space.

•The Neamen derivation (next slide) is based in k-space.

•See also a slightly different approach based in p-space in following slide.

J.E. Morris ECE415/515 Fall 2012 3

Density of States

 Consider electron confined to crystal (infinite potential well) of dimensions a

(volume V= a3)
 It has been shown that k=nπ/a, so ∆k=kn+1-kn=π/a

 Each quantum state occupies volume (π/a)3 in k-space.

 Number of quantum states in range k to k+dk is 4πk2.dk

and the number of electrons in this range k to k+dk is

gT(k)dk = 2(1/8).4πk2dk/ (π/a)3 = (k2dk/ π2)a3
where “2” comes from spin degeneracy, and “1/8” because E=(ħk)2/2m,
so negative k values do not add more energy states,
i.e. use kx, ky, kz >0 quadrant only
 Converting k-space to energy ………..

J.E. Morris ECE415/515 Fall 2012 4

2
 10/1/2012

Density of States

E  E C  p 2 2mn* in the conduction band, where E  EC

 p 2   2 k 2  2mn* ( E  EC )
2k .dk  2mn* .dE /  2
1 1
 k 2 dk  (2mn* ) 2 ( E  EC ) 2 mn* dE /  3
2 1 1
k 1
So writing g T (k )dk / V  g C ( E )dE    dk .a 3 / a 3  2 (2mn* ) 2 ( E  EC ) 2 mn* dE /  3
  
4
3 1 3 1
2m*
 3
(2mn* ) 2 ( E  EC ) 2 dE  4 ( 2 n ) 2 ( E  EC ) 2 dE
h h
is electron density of states/unit energy/unit vol in the conduction band
and electron density/unit energy/unit vol in the conduction band is
n( E )dE  f FD ( E ) g C ( E )dE
J.E. Morris ECE415/515 Fall 2012 5

Density of States (alternate)

Heisenberg says ∆p∆x=h,

(but see ħ in Neamen section 2.1.3 p30, and ħ/2 in Streetman)
and for crystal volume V=L3=LxLyLz
∆px∆x=h, ∆py∆y=h, ∆pz∆z=h,
and hence for a single electron
∆px ∆py ∆pz=h3/V for ∆x=Lx, etc
is the “volume” occupied in p-space by one electron.
Number of allowed conduction electron states g(p)dp from p to p+dp
[see Fig 3.26 (slide 3) in k-space with p = ħk]
is the volume of the p-space “shell” containing all states from p to p+dp
[4πp2dp]
divided by the p-space volume of each electron, i.e. ……

J.E. Morris ECE415/515 Fall 2012 6

3
 10/1/2012

Density of states

g ( p )dp  2
4 p dp   V
2
2(4 p 2 dp ) where the 2 is due to spin degeneracy
3
h V h3
To convert to energy density : - E  E C  p 2 2mn* in the conduction band, where E  EC
 p 2  2mn* ( E  EC )
2 p.dp  2mn* .dE
1 1
 p 2 dp  (2mn* ) 2 ( E  EC ) 2 mn* dE
1 1
2mn*
So writing g ( p )dp / V  g C ( E )dE  3
4 (2mn* ) 2 ( E  EC ) 2 dE
h
4 3 1
2m* 3 1
 3 (2mn* ) 2 ( E  EC ) 2 dE  4 ( 2 n ) 2 ( E  EC ) 2 dE
h h
is electron density of states/unit energy/unit vol in the conduction band
and electron density/unit energy/unit vol in the conduction band is
n( E )dE  f FD ( E ) g C ( E )dE

J.E. Morris ECE415/515 Fall 2012 7

(Compare 1D, 2D systems with 3D)

3
1  
g (k )dk  2 4k 2 dk /   in 3D
8 a
2
1  
 2 2k dk /   in 2D
4 a
1  
 2  dk /   in 1D
2 a

4
 10/1/2012

And similarly for holes:-

For electrons in the conduction band, where E  E C :

4
 
3
gC (E)  2
2mn* 2 E  EC
h
For holes in the valence band, where E  E V :
4
 
3
gV ( E ) 2
2m *p 2 EV  E
h
and g(E)  0 for E V  E  E C

Now that we have the densities of states, we

need the probabilities that these are filled.

J.E. Morris ECE415/515 Fall 2012 9

Ex 3.3 Calculate the density of quantum states (/cm3) for

a free electron over the energy range of (a) 0 ≤ E ≤ 2eV &
(b) 1 ≤ E ≤ 2eV

J.E. Morris ECE415/515 Fall 2012 10

5
 10/1/2012

Ex 3.3 Calculate the density of quantum states (/cm3)

in Si from (Ev-kT) to Ev at T=300 K

J.E. Morris ECE415/515 Fall 2012 11

Fermi-Dirac statistics (see also text “derivation”)

 First, let’s find the Maxwell-Boltzmann distribution

 Consider a system of identical particles with random collisions
 Assume 2 particles of initial energies E1 & E2 collide with energy transfer δ to
give final energies E3 & E4
 Conservation of Energy requires that: E1  E2  E3  E4  ( E1   )  ( E2   )
 If p(E)=probability that particle has energy E, then:
Probability of collision  C. p ( E1 ). p ( E2 )
 Equilibrium: equal numbers of forward & reverse collisions for constant
energy distribution, i.e. C. p( E1 ). p( E2 )  C. p( E3 ). p ( E4 )
p( E1 ). p( E2 )  p ( E1   ). p ( E2   )
p( E1   ) p ( E2   )
i.e.    (constant)
p( E1 ) p ( E2 )
i.e. p( Ei   )   . p( Ei )
J.E. Morris ECE415/515 Fall 2012 12

6
 10/1/2012

MB continued & FD intro

Result p( Ei   )  . p( Ei ) suggests an exponential distribution,

i.e. pMB ( E )  A. exp E
 A. exp E / kT
 Check: Aexp-α(Ei+δ)=λ. Aexp-αEi if λ=exp-αδ

 For Fermi-Dirac statistics, repeat as for MB, but with the

Pauli Exclusion Principle included, i.e.

 Can only get transfer of energy E1, E2 → E3, E4

if E1 & E2 are occupied AND if E3 & E4 are unoccupied

J.E. Morris ECE415/515 Fall 2012 13

Fermi function
So the collision probability is C. p ( E1 ). p( E2 ).[1  p( E3 )].[1  p( E4 )]
and applying Equilibrium and Energy Conservation again:
C. p ( E1 ). p ( E 2 ).[1  p ( E1   )].[1  p ( E 2   )]  C. p ( E1   ). p ( E 2   ).[1  p ( E1 )].[1  p ( E 2 )]
1  p ( E1 ) p ( E1   ) 1  p( E 2   ) p( E 2 )
i.e. .  .
p ( E1 ) 1  p ( E1   ) p( E 2   ) 1  p( E 2 )
1  p( Ei ) p( Ei   )
and .   (constant)  an exponential solution again,
p( Ei ) 1  p( Ei   )
1  p( Ei )
i.e.  A. exp  E (where   A 2 exp 2 E )
p( Ei )
1 1
i.e. p ( E )   exp E for  large,
1  A exp  E A
1
  ,
kT
1 E 1
and writing  exp F , f FD 
A kT E  EF
1  exp
kT
J.E. Morris ECE415/515 Fall 2012 14

7
 10/1/2012

MB/FD/(BE) comparisons

Maxwell-Boltzmann Bose-Einstein Fermi-Dirac

Particles Particles indistinguishable (wave functions
distinguishable overlap – cannot find individual particle)
Any number of particles can be in the same Only one particle in
energy state any energy state
(Pauli) but note spin
degeneracy
Particles are statistically independent
M

 Ni  N
M

N E
i 1
i i E
i 1

Atoms, molecules, etc Bosons Fermions

Photons, phonons, Electrons, protons,
4He nuclei neutrons, 3He nuclei
(integral spins) (½-odd-integral spins)

J.E. Morris ECE415/515 Fall 2012 15

Fermi distributions

 Electrons in solids obey Fermi-Dirac statistics due to

the Pauli exclusion principle (each state can have
only one electron – but remember spin!)
 Probability density function gives the ratio of filled to
total allowed states at a given energy.
 Using statistical mechanics to count states we find
the Fermi-Dirac distribution function:
• f(E) = {1 + exp[(E-Ef)/kT]}-1
• k is Boltzmann’s constant = 8.62x10-5 eV/K
= 1.38x10-23J/K
• EF is the Fermi level, [where f(EF)=½]

J.E. Morris ECE415/515 Fall 2012 16

8
 10/1/2012

The Fermi-Dirac distribution function.

1
T  0 K : E  EF  0
1  exp 
1
E  EF  1
1  exp 

1
For all T f (EF ) 
2

J.E. Morris ECE415/515 Fall 2012 17

At T = 0 K

J.E. Morris ECE415/515 Fall 2012 18

9
 10/1/2012

At T > 0 K

J.E. Morris ECE415/515 Fall 2012 19

Fermi levels
 The Fermi distribution gives the probability that an available state will
be occupied.
• Within the band gap there are no available states.
• Even with non-zero f(E) there is no occupancy in the gap. (An
energy state at the Fermi level would have a probability of ½ of
being occupied, if there was one.)
 At 0K every available state up to the Fermi level is occupied and every
state above the Fermi level is empty.
 Fermi function is “symmetric” about the Fermi level
 For holes:
E  EF
exp
1 kT 1
1  f (E)  1   
E  EF E  EF E E
1  exp 1  exp 1  exp F
kT kT kT

J.E. Morris ECE415/515 Fall 2012 20

10
 10/1/2012

Fermi function symmetry:

electrons and holes

J.E. Morris ECE415/515 Fall 2012 21

Approximate FD → MB for (E-EF)>>kT

See Example 3.6:

f(EF+3kT) ≈ 5% for 300K
So f(E) varies quickly

Example 3.8:
(fMB-fFD)/FFD=5%
at E-EF=kT ln(1/0.05) ≈ 3kT

Note f(E) = {1 + exp[(E-Ef)/kT]}-1

≈ exp -[(E-Ef)/kT]
for
“(E-EF)>>kT” condition
where
exp[(E-Ef)/kT] >> 1

J.E. Morris ECE415/515 Fall 2012 22

11
 10/1/2012

Ex 3.6 Assume EF=EC-0.3eV and T=300 K.

Determine probability of state occupied at
(a) E=EC+kT/4 & (b) E=EC+kT

J.E. Morris ECE415/515 Fall 2012 23

Ex 3.7 Assume EF=EC-0.3eV. Determine T at

which the probability of state at
E=EC+0.025eV is occupied is 8x10-6.

J.E. Morris ECE415/515 Fall 2012 24

12
 10/1/2012

Ex.3.8 Calculate the energy, in terms of kT and EF, at

which the difference between the FD and MB
distributions is 2% of the FD value.

J.E. Morris ECE415/515 Fall 2012 25

Electron & hole densities

n(E) = gC(E).f(E)

p(E) = gV(E).[1-f(E)]

J.E. Morris ECE415/515 Fall 2012 26

13
 10/1/2012

Effective density of states & Fermi integral

Electron density in the conduction band
EC' EC'
n0   n( E )dE   f FD ( E ) g C ( E ).dE where EC' is the top of the conduction band
EC EC
3 1
  2m *  E  E C  2
2
  4  2 n  dE for E C/  
 h  1  exp E  E F
EC

kT
3
 2m *  2
 4  2 n  F1 2  F ,
 h 
1
  2 d E  EC E  EC
where F1  F   0 , , F  F
1  exp   F 
2
kT kT
This is the (tabulated) Fermi Integral

A similar treatment gives a similar result for g V ( E ) and p 0 for holes in the valence band

J.E. Morris ECE415/515 Fall 2012 27

Effective density of states

For conduction electrons with E  E F  kT (usually so)

E  EF
f FD ( E )  f MB ( E )  exp
kT
3 3
 2m *
 2 1 E  EF  2m *  2 E  EF  1 1
so n0   4  2 n  E  EC  2 exp dE  4  2 n  exp C  kT  2 x 2 e  x kTdx
EC
 h  kT  h  kT 0

1
E  EC E  EF E  EF E  EC  
[writing x  , dE  kTdx, exp  exp C exp , and using  x 2 e  x dx  ]
kT kT kT kT 0 2
3 1 3
 2m * kT  2 E  EF  2  2mn* kT  2 E  EF E  EF
 4  n2  exp C .  2 2
 exp C
  N C exp C
 h  kT 2  h  kT kT
3
 2mn* kT  2
where N C  2 2

 h 
NC is the effective density of states as if all electrons at conduction band edge EC. Only
limiting assumption is that EC-EF>>kT; if so, result holds for intrinsic and extrinsic.
(Note: EF→EC for N-type, but approx usually still OK.)

J.E. Morris ECE415/515 Fall 2012 28

14
 10/1/2012

Electron Concentrations
 Simplify: represent states throughout the conduction band by
an effective density of states Nc located at Ec.
n0 = Ncf(Ec)
 If Ec – EF > 3kT then
f(Ec) ~ e-(Ec-EF)/kT (Maxwell-Boltzmann)
 kT = 0.026 eV at 300K, so usually OK
n0 = Nc e-(Ec-EF)/kT ; Nc = 2[2πmn*kT/h2]3/2
 Termed non-degenerate.
 If EF within 3kT of either band edge or inside band 
degenerate and we need to use the full Fermi function.

J.E. Morris ECE415/515 Fall 2012 29

Hole Concentrations
 Concentration of holes in valence band
p0 = Nv [1 – f(Ev)]
 If EF – Ev > 3kT then Fermi function reduces to
Maxwell Boltzmann and
p0 = Nv exp[ -(EF – Ev)/kT ]
Nv = [2π mp*kT/h2]3/2
 p0 = ∫gV(E)[1-f(E)]dE
 1-f(E)=1-[1+exp((E-EF)/kT)]-1
=[exp (E-EF)/kT)]/[1+exp((E-EF)/kT)]
=1/[1+exp((EF-E)/kT)] ≈ exp-(EF-E)/kT

J.E. Morris ECE415/515 Fall 2012 30

15
 10/1/2012

Ex 4.1 Determine the probability that a quantum

state at E=EC+kT is occupied by an electron, and
calculate the electron concentration in GaAs at
T=300 K if EF=EC-0.25eV.

(Nc from Table 4.1)

J.E. Morris ECE415/515 Fall 2012 31

Ex 4.2 (a) Calculate the equilibrium hole concentration in

Si at T=250K if EF=EV+0.27eV. [NV(300K)=1.04x1019/cm3]
(b) What is p0(250K)/p0(400K)? (See Example 4.2)

J.E. Morris ECE415/515 Fall 2012 32

16
 10/1/2012

Intrinsic Concentrations
 EF = EFi for intrinsic material
n0=ni = Ncexp[-(Ec-EFi)/kT] pi = Nvexp[-(EFi-Ev)/kT]
 n0 p0 = constant.
 Plug in above => n0 p0 = ni2 = [NcNv]exp[-(EC-EV)/kT]
ni = [NcNv]1/2 exp[-Eg/2kT]
= 2[2πkT/h2]3/2(mnmh)3/2 exp[-Eg/2kT]
 Remember n0p0 = ni2 and ni = 1.5x1010 cm-3 for Si
 You’ll use both often!
 Also for Si: 6 different conduction band valleys (2 each x,y,z)
3 1
(mn* ) 2  6m mt  2
*
 Note: mn varies in different sources

 See Appendix F for more detail (complications) on effective mass

J.E. Morris ECE415/515 Fall 2012 33

Ex 4.3 (a) Calculate the intrinsic carrier

concentration in GaAs at T=400 K and T=250 K,
assuming Eg=1.42eV. (b) What is ni(400K)/ni(250K)?

J.E. Morris ECE415/515 Fall 2012 34

17
 10/1/2012

Intrinsic concentrations

 In general, if EF ≠ EFi, then re-write these as

n0 = ni exp[(EF – EFi)/kT]
p0 = pi exp[(EFi – EF)/kT]
 Also used frequently and helpful for visualization
 With each problem consider what information you have and
what you are looking for.

 Law of Mass Action:

• n0p0=ni2
• Independent of EF, (i.e. of doping)
• So applies to both intrinsic and extrinsic at equilibrium

J.E. Morris ECE415/515 Fall 2012 35

Temperature
Dependence
ni  T3/2 exp[-Eg/2kT]

J.E. Morris ECE415/515 Fall 2012 36

18
 10/1/2012

Physical constants
(@300 K unless noted)

Boltzmann’s 1.38 * 10-23 J/K Planck’s constant (h) 6.63*10-34 J-s

const (k) 8.62 * 10-5 ev/K 4.14*10-15 eV-s
Room temp 0.0259 eV Electron rest mass 9.11 * 10-31 kg
value of kT (m0)
Electronic 1.60 * 10-19 C Permittivity of free 8.85*10-14 F/cm
charge (q) space (ε0)
Density of states Relative permittivity 11.9
Nc 2.8*1019 cm-3 of silicon (εr)
Nv 1.04*1019 cm-3
Speed of light (c) 3*1010 cm/s Electron volt (eV) 1.6*10-19 J
Band Gap Si Eg 1.11 ev Band gap Ge Eg 0.67 eV

J.E. Morris ECE415/515 Fall 2012 37

Effective density of states & masses

See Appendix B for m* values

J.E. Morris ECE415/515 Fall 2012 38

19
 10/1/2012

Intrinsic EF position

n 0  p 0  ni  p i
  EC  E Fi     E Fi  EV  
N C exp 
kT   N V exp  kT 
   
E Fi  EC E V  E Fi
 n ( N C )   n ( N V ) 
kT kT
1 1  NV 
E Fi  ( EC  EV )  kT .n 
2 2  NC 
1 3  m *p 
 ( EC  EV )  kT .n *. 
2 4 m 
 n 
3  m *p 
i.e. E Fi  E midgap  kT .n *.   E midgap if m *p  mn*.
4 m 
 n 

J.E. Morris ECE415/515 Fall 2012 39

Ex 4.4 Find the position of EFi at T=300 K wrt the

band-gap center for (a) GaAs, and (b) Ge.

J.E. Morris ECE415/515 Fall 2012 40

20
 10/1/2012

Summary: Fermi function

 Fermi function describes probability of a given state being
occupied at temperature T.
f(E) = {1 + exp[(E-EF)/kT]}-1
 For intrinsic semiconductors EF is mid-gap.
 If E – EF > 3kT (non-degenerate, with EF in the forbidden gap)
then
f(Ec)~e-(E-EF)/kT
 Use effective density of states Nc located at Ec or Nv located at
Ev. (still non-degenerate)
n0 = Nc e-(Ec-EF)/kT ; Nc = 2[2πmn*kT/h2]3/2
p0 = Nv e-(Ev-EF)/kT; Nv = [2π mp*kT/h2]3/2

J.E. Morris ECE415/515 Fall 2012 41

21