Ideal Metal-Semiconductor Contacts

Surface-included energy band diagrams for a metal (left) and n-type semiconductor (right).
&

 The work function  is defined as the energy difference between the vacuum level E0
and the Fermi level: Metal work function:  M  E0 − EFM

Semiconductor work function:  S   + ( Ec − EFS )Flat Band =

Eo -

Eys
"Eo-Ec
where  is the electron affinity   ( E0 − Ec ) ,  = 4.03 eV for Si
surface

 For ideal metal to n-type semiconductor contacts

↑B] PB2

↓ 5-d
Schottky barrier

PBI

Energy band diagrams for ideal MS contacts between a metal and an n-type semiconductor:
 M   S system (a) an instant after contact formation and (b) under equilibrium conditions;
 M   S system (c) an instant after contact formation and (d) under equilibrium
conditions.
半導體元件物理_陳敏璋 1
Formation of a Schottky junction between a metal and an n-type semiconductor as Φm > Φn.

 The Schottky barrier: For an ideal metal to n-type semiconductor contact with  M   S ,
the energy barrier  B encountered by electrons with E = EF in the metal is
 B =  M −  . The build-in potential in the semiconductor is given by qVbi =

 B − ( Ec − EF )FB , where FB denotes the flat-band condition.

半導體元件物理_陳敏璋 2
 Current–voltage characteristics of the Schottky barrier:

Response of the  M   S (n-type) MS contact to an applied bias. (a) Definition of current

and voltage polarities. (b) Energy band diagram and carrier activity when VA > 0. (c) Energy
band diagram and carrier activity when VA < 0. (d) General form of the I-V curve.

 For ideal metal to p-type semiconductor contacts

The surface potential barrier  B encountered by holes with E = EF in the metal for
 M   S is given by  B = Eg +  −  M .

 Brief summary of electrical nature of ideal MS contacts

半導體元件物理_陳敏璋 3
I-V Characteristics
minority diffusion current
carrier

-
0-
⑤
[
nur
-

n = -

NA

It Fermi Level rik

bye-
0 -

- -

Current components in a forward-biased (a) p+-n junction diode and (b) MS diode.

Five basic transport processes under forward bias. (1) Thermionic emission. (2) Tunneling.
(3) Recombination. (4) Diffusion and drift of electrons. (5) Diffusion of holes.

 Carrier-transport mechanism
Consider a Schottky diode: an ideal metal to n-type semiconductor contact with  M   S
(1) Thermionic emission of electrons from the semiconductor over the potential barrier
into the metal: the dominant process in Schottky diodes with moderately doped
semiconductors (e.g., Si with ND < 1017 cm-3)
(2) Quantum mechanical tunneling of electrons through the barrier (important for heavily
doped semiconductors and responsible for most Ohmic contacts)
4
depletion region
半導體元件物理_陳敏璋
 (3) Recombination in the depletion region (identical to the recombination process in a pn
junction)
(4) Diffusion and drift of electrons in the depletion region
(5) Holes inject from the metal and then diffuse into the semiconductor (equivalent to
recombination in the quasi-neutral n-type region)
 Thermionic emission theory
The electron current density due to the flows of electrons from semiconductor into metal is
given by

J s →m = q
 Emin
vx dn (Please refer to J = qvn)

where Emin is the minimum energy required for the thermionic emission of electrons over
the Schottky barrier from semiconductor into the metal, vx is the carrier velocity in the
direction of transport. The differential electron concentration is given by
dn = gc ( E ) f ( E )dE
where gc ( E ) is the density of states in the conduction band and f ( E ) is the
Fermi-Dirac distribution function. Assuming that the Boltzmann approximation is valid,
i.e., E − EF 3kT , we may write

4 ( 2m* )
3/2
1
dn = E − Ec dE
h 3
1 + exp ( ( E − EF ) kT )

4 ( 2m* )
3/2
 E − EF 
= 3
E − Ec exp  −  dE
h  kT 
4 ( 2m* )
3/2
 E − Ec + qn 
= E − Ec exp  −  dE where EF = Ec − q
h3  kT 

1 m*
Because E − Ec = m*v 2 , dE = m*vdv and E − Ec = v ,
2 2
3
 m*   qn   m*v 2 
dn = 2   exp  −  exp  −   4 v dv
2

 h   kT   2kT 
The above equation gives the number of electrons per unit volume that have the speed
between v and v + dv . We may decompose the speed into its components
v 2 = vx2 + v y2 + vz2
半導體元件物理_陳敏璋 5
 We can transform the differential volume 4 v 2dv into the differential volume dvx dv y dvz .

Therefore,
3 
 m*   m*vx2 

 q 
J s →m = 2q  exp  − n  vx exp  − dv
 h 
 kT   2kT  x
  vmin  
  m*v 2y  
 m*vz2 

 −
exp  −


2 kT
 dv y



−
exp  −
 2kT  z
 
dv

The velocity vmin is the minimum velocity required in the x direction to go over the
potential barrier:
1 * 2
m vmin = q (Vbi − VA )
2
The change of variables leads to

m*vx2 q m*v 2y m*vz2

=2 + (Vbi − VA ) , = , 2
=2
2kT kT 2kT 2kT
3
 m*   2kT 2  q (Vbi − VA )  


 qn 
J s →m = 2q  
 h   m* 
exp 

−
kT


exp −
 kT


( )
 exp − 2 d
  0
    


−
(
exp −  2
)d  
−
exp −( ) d
2
 Note that


−
( )
2
exp − x dx =  


 4 qm*k 2  2  qn   q (Vbi − VA ) 
=  T exp  −  exp − 
 h 3   kT  kT
   
 q   qV 
= A*T 2 exp  − Bn  exp  A  where qBn = q (Vbi + n )
 kT   kT 

where A* = 4 qm*k 2 h3 is called the effective Richardson’s constant.

→ The electron current density from the metal to the semiconductor J m→s is exactly the
same as J s →m when the applied bias is zero. Then

 q 
J m→s (VA = 0 ) = J s →m (VA = 0 ) = A*T 2 exp  − Bn 
 kT 
The electron current density from the metal to the semiconductor J m→s always see the
same energy barrier qBn . Consequently,

J m→s (VA  0 ) = J m→s (VA = 0 )

→ The net current density in the metal-semiconductor contacts can be written as

半導體元件物理_陳敏璋 6
 J = J s →m − J m → s
which is defined to be positive in the direction from the metal to the semiconductor. We
then have
 q    qV     qV  
J = A*T 2 exp  − Bn  exp  A  − 1 = J s exp  A  − 1
 kT    kT     kT  
which is of the form identical to that of a pn junction.

Unified Theory of Diffusion and Thermionic Emission in pn Junction

 The thermionic emission current of electrons over the potential barrier from the n- to
p-region in a pn junction at forward bias is given by
 4 qm*k 2  2  q (Vbi + n )   qVA   qVA 
JTE =  T exp  −  exp  = J
 TE 0 exp  
 h3  kT  kT   kT 
   
where JTE0 can be factorized to be
 4 qm*k 2  2  q (Vbi + n ) 
JTE 0 =   T exp − 
 h 3  kT
   
3/2
1  2 m*kT   q (Vbi + n )  2kT n0 (− x p ) n p0
= q 2  exp  −  = q v = q vn
2  h 2 * n
  kT  m 2 2

where vn = 2kT  m* is the unidirectional thermal velocity, and n0 (− x p ) = n p 0 is the

electron concentration at x = − x p at thermal equilibrium (VA=0). The balance of the

current requires that the thermionic emission current of electrons over the potential barrier
半導體元件物理_陳敏璋 7
 from the n- to p-region is equal to the electron diffusion current in the p-region. Thus the
electron current density at x = − x p is

 qV  n( − x p )
( )
J n x = − x p = JTE 0 exp  A  − q
 kT  2
vn

 qVA 
= JTE 0 exp 
(n p 0 + n(− x p ) )
−q vn
 kT  2
 qV  n p0 n(− x p )
= JTE 0 exp  A  − q vn − q vn
 kT  2 2
  qV   n ( − x p )
= JTE 0  exp  A  − 1 − q vn
  kT   2
d n p D
= qDN = q N n(− x p )
dx x =− x LN
p

where the term q n(− x p )vn 2 is electron current density from the p-region back to the

n-region. One can obtain

  qV  
JTE 0  exp  A  − 1
  kT  
n(− x p ) =
D v
q N +q n
LN 2
Therefore, the electron diffusion current density is given by
  qV  
JTE 0  exp  A  − 1
D D   kT  
J n (x = − x p ) = q N n(− x p ) = q N
LN LN D v
q N +q n
LN 2
  qV     qV     qV  
JTE 0  exp  A  − 1 JTE 0  exp  A  − 1 JTE 0  exp  A  − 1
  kT     kT     kT  
= = =
v 2 qn p 0vn 2 J
1+ n 1+ 1 + TE 0
DN LN qn p 0 DN LN J D0
where JTE0 and JD0 are correlated with the thermionic emission and diffusion current,
respectively. It is indicated that the electron current is determined by which one, thermionic
emission or diffusion current, is the “bottleneck”.
  qV  
If J D0 JTE 0 , J n  JTE 0  exp  A  − 1 → thermionic emission dominates
  kT  

  qV  
If JTE 0 J D0 , J n  J D 0  exp  A  − 1 → diffusion dominates
  kT  
半導體元件物理_陳敏璋 8
Practical Contact Consideration
 Real metal-semiconductor contacts are invariably nonideal. A Schottky barrier junction
includes a termination of the semiconductor crystal. The semiconductor surface contains
surface states due to incomplete covalent bonds and other effects, which can lead to the
presence of charges at the metal-semiconductor interface. Furthermore, the contact is
seldom an atomically sharp discontinuity between the semiconductor crystal and the
metal. There is typically a thin interfacial layer, which is neither semiconductor nor metal.
For example, silicon devices are likely to contain a thin oxide layer (5Å~25 Å) between
the metal and the semiconductor. A native oxide layer forms almost instantly when silicon
is exposed to the atmosphere. Therefore, deposition of a metal on such a Si surface leaves
a glassy interfacial layer at the junction. Although electrons can tunnel through this thin
layer, it does affect the barrier to current transport through the junction. Because of the
emergence of surface states, an interfacial layer, microscopic clusters of
metal-semiconductor phases, and other effects at the metal-semiconductor interface, it is
difficult to fabricate junctions with barriers near the ideal values predicted from the work
functions of the two isolated materials.
 Fermi-level pinning:

Fermi level pinning by interface states in compound semiconductors: EF is pinned near EC

-0.8 eV in n-type GaAs, regardless of the choice of metal.

→ The contact of metal and semiconductor usually results in the formation of a lot of
surface states at a specific energy within the bandgap. The surface states tend to fix or
“pin” the equilibrium Fermi level. Because of this Fermi level pining effect, the observed
Schottky barrier height varies only slightly with the work function of the metal used in the
半導體元件物理_陳敏璋 9
 metal-semiconductor contact. For example, regardless of the metal employed in forming a
GaAs metal-semiconductor contact,  B = 2 Eg / 3 for n-type GaAs and  B = Eg / 3 for

p-type GaAs.
 Ohmic Contacts:

(a) Heavy doping of the semiconductor beneath the MS contact to facilitate ohmic contact
formation. (b) Emission currents across a barrier-type contact as a function of doping. The
emission is shown varying from solely thermionic emission at low semiconductor doping to
predominantly tunneling through the barrier at high semiconductor doping.

→ Ohmic contacts are usually produced in practice by heavily doping the surface region
of the semiconductor immediately beneath the contact. The width of depletion region,
and hence the width of the potential barrier, decreases with increased semiconductor
doping. When the semiconductor doping exceeds ~1018cm-3, significant tunneling takes
place through the potential barrier.
→ Annealing the metal-semiconductor contacts is necessary to minimize the contact
resistance.
→ Spiking: aluminum is annealed to above 475oC

Visualization of spiking beneath an Al-Si contact.

半導體元件物理_陳敏璋 10
Appendix: unidirectional thermal velocity
 Consider a uniform homogeneous semiconductor at equilibrium with an electron
concentration n .

 d k f ( E ) = 4  d k f ( E )
2
1 2 2 k2
n= 3 3
E =EC +
abc 2 2 2 3
2m*
a b c
where abc is the volume of the crystals, the factor 2 takes two allowed spin states (spin up
2 2 2
and spin down) into account, and a volume contains one allowed state in the
a b c
k-space. Transform to the spherical coordinate

 dkkd k sin d f ( E )
1
n=
4 3
  2 

4    
1 4
= dkk f ( E ) d sin  2
d = 3 dkk 2 f ( E )
3
0 0 0 4 0

Replace dk in terms of dE
 

 
1
n= dkk f ( E ) =
2
dEg ( E ) f ( E )
2 0 EC

The density of state g ( E ) is given by

k 2 dk k2
g ( E )dE = , g (E) =
2  2 ( dE dk )
2
k2 dE 2
k
E = EC + → = *
2m* dk m
m*k m*
g (E) = = 2m* ( E − EC )
 2 2
 2 3

For a nondegenerate semiconductor,

1  E − EF 
f (E) =  exp  − 
 E − EF   k BT 
1 + exp  
 k BT 

( 2m ) * 3/2 

 
E − EC
n= dEg ( E ) f ( E ) = dE
EC 2 2 3
EC 1 + exp ( ( E − EF ) k BT )

( 2m ) * 3/2  3/2


 E − EF   2 m*kT   EF − EC 
= dE E − EC exp  −  = 2  exp  
2 2 3
EC  k BT   h
2
  k BT 
半導體元件物理_陳敏璋 11
 Similarly, the current density in the positive z-direction can be expressed as ( J z = qnvz )


2
1 k k k2
Jz = q  2 d k *z f ( E ) ,
3
vz = *z E =EC +
8 3 k z 0 m m 2m*

where k z = k cos , and transform to the spherical coordinate

 dkkd k sin dk cos f ( E )

q
Jz =
4 3m*
  2 2 

4 m   d sin cos  
q q
= dkk f ( E ) 3
d = dkk 3 f ( E )
3 *
0 0 0 4 m
2 *
0

Replace dk in terms of dE
 

 
q qm*
Jz = dkk f ( E ) = 2 3
3
dE f ( E ) ( E − EC )
4 2 m* 0 2 EC

In a nondegenerate semiconductor
 E − EF 
f ( E )  exp  − 
 k BT 

Thus we have
qm* ( k BT )
2
 E − EC  k BT
Jz = exp  F  = qn
2 2 3
 k BT  2 m*

The average velocity in the positive z-direction vz is given by

Jz 2k BT
vz = =
q ( n 2)  m*
The factor of 2 is taken into account for the fact that only one half the electrons n 2 are
traveling in the positive z-direction.

半導體元件物理_陳敏璋 12
Heterojunctions

Energy band discontinuities for a thin layer of GaAs sandwiched between layers of wider
band gap Al0.3Ga0.7As. In this case, the GaAs region is so thin that quantum states are formed
in the valence and conduction bands. Electrons in the GaAs conduction band reside on
“particle in a potential well” states such as E1 shown here, rather than in the usual conduction
band states. Holes in the quantum well occupy similar discrete states, such as Eh.
半導體元件物理_陳敏璋 13
 Heterojunctions consist of those between two lattice-matched semiconductors with
different band gaps. The interface between two such lattice-matching semiconductors may
be virtually free of defects. The availability of heterojunctions and multilayer structures in
compound semiconductors opens a broad range of possibilities for device development.
→ For example, Figure shows the spatial variation in conduction and valence bands for a
multilayer structure in which a very thin layer of GaAs is sandwiched between two layers
of Al0.3Ga0.7As, which has a wider band gap that the GaAs. A consequence of confining
electrons and holes in a very thin layer is that these particles behave according to the
particle in a potential well problem. Therefore, the conduction band electrons in the
narrow-gap material are confined to discrete quantum states. Similarly, the states in the
valence band available for holes are restricted to discrete levels in the quantum well. This
is one of the clearest demonstrations of the quantum mechanical results. From a practical
device point of view, the formation of discrete quantum states in the GaAs layer changes
the energy at which photons can be emitted. An electron on one of the discrete conduction
band states (E1) can make a transition to an empty discrete valence band state in the GaAs
quantum well (such as Eh), giving off a photon of energy Eg+E1+Eh , greater than the GaAs
band gap. Semiconductor lasers have been made in which such a quantum well is used to
raise the energy of the transition from the infrared, typical of GaAs, to the red portion of
the spectrum.
 When semiconductors of different band gaps, work functions, and electron affinities are
brought together to form a junction, we expect discontinuities in the energy bands as the
Fermi levels line up at equilibrium. The discontinuities in the conduction band Ec and
the valence band Ev accommodate the difference in band gap between the two
semiconductors Eg .

→ In an ideal case, Ec would be the difference in electron affinities q(χ2 - χ1), and Ev
would be found from Eg − Ec . This is known as the Anderson affinity rule.

→ In practice, the band discontinuities are found experimentally for particular

semiconductor pairs. For example, in the commonly used system GaAs/AlxGa1-xAs, the
band gap difference between the wider band gap AlxGa1-xAs and the narrower band gap
GaAs is apportioned approximately 2/3 in the conduction band and 1/3 in the valence
半導體元件物理_陳敏璋 14
 band for the heterojunction.
→ The built-in contact potential is divided between the two semiconductors as required to
align the Fermi levels at equilibrium. The resulting depletion region on each side of the
heterojunction and the amount of built-in potential on each side (making up the contact
potential V0) are found by solving Poisson’s equation with the boundary condition of
continuous electric flux density, 1 1 = 2 2 at the junction. The barrier that electrons
must overcome in moving from the n side to the p side may be quite different from the
barrier for holes moving from p to n.
→ To draw the band diagram for any semiconductor device involving homojunctions or
heterojunctions, we need material parameters such as the band gap and the electron
affinity which depend on the semiconductor material but not on the doping, and the work
function which depends on the semiconductor as well as the doping. The electron affinity
and work function are referenced to the vacuum level Evac.

半導體元件物理_陳敏璋 15
An ideal heterojunction between a p-type, wide band gap semiconductor and an n-type
narrower band gap semiconductor: (a) band diagrams before joining; (b) band discontinuities
and band bending at equilibrium.

→ Example: For heterojunctions in the GaAs/AlxGa1-xAs system, the band gap difference
Eg is accommodated approximately 2/3 in the conduction band and 1/3 in the valence

band. For an Al composition of x = 0.3 in the GaAs/AlxGa1-xAs system with

Eg = 0.42 eV , Ec = 0.28 eV and Ev = 0.14 eV . Accordingly, the band diagrams for

two heterojunction cases: N+-Al0.3Ga0.7As on n-type GaAs, and N+-Al0.3Ga0.7As on

p+-GaAs are given as follows:

半導體元件物理_陳敏璋 16
 A heavily n-type AlGaAs grown on lightly doped GaAs is a particularly important
example of a heterojunction. In this example the discontinuity in the conduction band
allows electrons to spill over from the N+-AlGaAs into the GaAs, where they become
trapped in the potential well. As a result, electrons collect on the GaAs side of the
heterojunction and move the Fermi level above the conduction band in the GaAs near the
interface. These electrons are confined in a narrow potential well in the GaAs conduction
半導體元件物理_陳敏璋 17
 band. If we construct a device in which conduction occurs parallel to the interface, the
electrons in such a potential well form a two-dimensional electron gas with very
interesting device properties. The electron conduction in such a potential well can result in
very high mobility electrons. This high mobility is due to the fact that the electrons in this
well come from the AlGaAs, and not from doping in the GaAs. As a result, there is
negligible impurity scattering in the GaAs well, and the mobility is controlled almost
entirely by lattice scattering (phonons). At low temperatures, where phonon scattering is
low, the mobility in this region can be very high. If the band-bending in the GaAs
conduction band is strong enough, the potential well may be extremely narrow, so that
discrete states such as E1 and E2 are formed.

A heterojunction between N+-AIGaAs and lightly doped GaAs, illustrating the potential well
for electrons formed in the GaAs conduction band.

→ Another obvious feature is that the concept of a contact potential barrier qV0 for both
electrons and holes in a homojunction is no longer valid for the heterojunction. The barrier
for electrons qVn is smaller than the barrier for holes qVp. This property of a heterojunction
can be used to alter the relative injection of electrons and holes.

半導體元件物理_陳敏璋 18