EE311 notes/ Saraswat

Metal/Semiconductor Ohmic Contacts

Silicide
metal
Rc Poly-Si

Rch
Xj Rs Rs’ Rd’ Rd
source drain

Fig. 1 components of the resistance associated with the S/D junctions of a MOS transistor.

Year 1997 1999 2003 2006 2009 2012

Min Feature Size 0.25µ 0.18µ 0.13µ 0.10µ 0.07µ 0.05µ
Contact xj (nm) 100-200 70-140 50-100 40-80 15-30 10-20
xj at Channel (nm) 50-100 36-72 26-52 20-40 15-30 10-20
Series Resistance (ohms)

140
Series Resistance (ohms)

200 PMOS
120 NMOS Scaled by ITRS Roadmap
Scaled by ITRS Roadmap
100 Rov Rov
150
80
Rext 100
60 Rext
40 Rdp 50 Rdp
20 Rcsd Rcsd
0 0
30 nm 50 nm 70 nm 100 nm 30 nm 50 nm 70 nm 100 nm
Technology or Gate Length Technology or Gate Length

Fig. 2. Various components of the resistance associated with the shallow junctions of
NMOS and PMOS transistors for different technology nodes. (Source: Jason Woo,
UCLA)
EE311 / Saraswat Ohmic Contacts

Conduction Mechanisms for Metal/Semiconductor Contacts

I
Low doping

Ef
V
Schottky
(a) Thermionic emission

Medium doping

(b) Thermionic-field emission

Heavy doping

Ohmic
(c) Field emission.
Fig. 3. Conduction mechanisms for metal/n-semiconductor contacts as a function of the
barrier height and width. (a) Thermionic emission; (b) thermionic-field emission; (c) field
emission.

(1) Thermionic emission (TE), occurring in the case of a depletion region so wide that the
only way for electrons to jump the potential barrier is by emission over its maximum (Fig.
3a). The barrier height is reduced from its original value as a result of image force barrier
lowering.

(2) Field emission (FE), consisting in carrier tunneling through the potential barrier. This
mechanism, which is the preferred transport mode in ohmic contacts, takes place when the
depletion layer is sufficiently narrow, as a consequence of the high doping concentration in
the semiconductor (Fig. 3c).

Contact Resistance and Specific Contact Resistivity (ρ c)

Contact resistance is a measure of the ease with which current can flow across a metal-
semiconductor interface. In an ohmic interface, the total current density J entering the interface
is a function of the difference in the equilibrium Fermi levels on the two sides.

The band diagram in the Fig. 4 may be used as an aid in describing the majority current flow
in the block of uniformly heavily doped semiconductor material of length l with ohmic
contacts at each end. The applied voltage V drives a spatially uniform current I through the
semiconductor bulk and ohmic contacts of cross sectional area A. Then, under the low-current

2
EE311 / Saraswat Ohmic Contacts

assumption that the voltage drop across both metal-semiconductor contacts is identical, the
I-V relation becomes:

n+

∆V

∆V

Figure 4: Ideal contacts to a heavily doped semiconductor with uniform current density.

V = Vbulk + 2Vcontact = (Rbulk + 2Rcontact)I = (1)

dV bulk ρ l
Rbulk = =
dI A
(2)
dV ρ
Rcontact = contact = c
dI A
where ρ is the bulk resistivity and ρc specific contact resistivity that can be defined through
the component resistances.
Since the voltage required to drive current through a good ohmic contact is small we restrict
the ρc definition to zero applied voltage.

ρc = lim  contact 
dV
Ω cm 2 (3)
V →0  dJ 

where J is the current density I/A. Alternatively (3) can be defined as

vmetal − vsemicond
J= (3a)
ρc

Thermionic Emission - Schottky Contact

For a Schottky contact the current governed by thermionic emission over the barrier is given
by
 −2φ B  qV
J S = A * T 2 exp
 kT 
e ( kT
)
−1 (4)

where A* is Richardson’s constant. The specific contact resistivity as calculated by Eq. (3)
is

3
EE311 / Saraswat Ohmic Contacts

(5)
k  2φ  kT
ρ c = * exp B  =
qA T  kT  qJ s

Tunneling - Ohmic Contacts

An ohmic contact is defined as one in which there is an unimpeded transfer of majority

carriers from one material to another, i.e., the contacts do not limit the current. The way to
achieve such a contact is by doping the semiconductor heavily enough that tunneling is
possible. It is usual to heavily dope the Si regions N + or P+ so that an ohmic contact is
insured. Suppose Nd (or Na) in the semiconductor is very large. Then the depletion region
width at the metal - semiconductor interface

2 K ε o φi
Xd =
q Nd
becomes very small. When Xd < ≈ 2.5–5nm, electrons can “tunnel” through the barrier.
This process occurs in both directions M → S and S → M so the contact shows very little
resistance and becomes ohmic.

To calculate an approximate value for the required doping,

2 K ε o φi
N d min ≈ 2
≈ 6.2 × 10 19 cm −3 for X d = 2.5 nm
q Xd
This is a relatively easy value to achieve in practice and is normally how ohmic contacts are
made in integrated circuits.
Fs Jsm Fm

For a tunneling contact the net semiconductor to metal current is given by

A*T
k ∫
J sm = Fs P(E)(1 − Fm )dE (6)

where Fs and Fm are Fermi-Dirac distribution functions in metal and semiconductor

respectively, and P(E) is the tunneling probability given by

 2Φ B ε s m* 
P( E ) ~ exp - (7)
 h N 

Where m* is the effective mass of the tunneling carrier and h is the Plank's constant. The
analysis to calculate current is more is somewhat more complicated, resulting in

4
EE311 / Saraswat Ohmic Contacts

[
J sm ∝ exp −2 x d 2 m * (qφ B − qV ) / h 2 ]
Specific contact resistivity can be calculated using equations described above and is of the
form
 2φ ε s m * 
ρ c = ρ co exp B ohm − cm 2 (8)
 h N 
Where ρco is a constant dependent upon metal and the semiconductor. Specific contact
resistivity, ρ primarily depends upon
c
• the metal-semiconductor work function, φΒ,
• doping density, N, in the semiconductor and
• the effective mass of the carrier, m*.

Fig. 5. Specific contact resistivity of metal contacts to n-type and p-type Si. Solid lines are
calculated from the model. (Ref: S. Swirhun, Electrochem. Soc., Oct. 1988)

Observations

5
EE311 / Saraswat Ohmic Contacts

1. Specific contact resistivity, ρc ↓ as barrier height ↓

3. For a given doping density contact resistance is higher for n-type Si than p-type.
This can be attributed to the barrier height
2. Specific contact resistivity, ρc ↓ as doping density ↑
• Doping density can’t be scaled beyond solid solubility.
• N type dopants have higher solid solubility than P type dopants

1022
Solid Solubility (atoms cm-3 )

As
P
1021

Sb
1020
Sn
Ga
Al
1019
800 900 1000 1100 1200 1300
Temperature (˚C)
Fig. 8. Solid solubility of dopants in Si (Ref: Plummer & Griffin, Proc. IEEE, April 2001)

Barrier Height

Figure: 9 Accumulation and depletion type contacts.

pc is the physical parameter that describes the transport of majority carriers across heavily
doped Si-metal interfaces. However, experiment and modeling of ohmic conduction is still
crude. An ohmic contact is generally modeled as a heavily doped Schottky (diode) contact.

6
EE311 / Saraswat Ohmic Contacts

The Schottky model predicts that upon bringing in contact Si with electron affinity X, and a
metal of work function φ m , a barrier of height φ b = ( φ m − χ ) which is independent of
semiconductor doping will be formed. Since measured φ m values for a variety of metals
range from about 2.0 to 5.5 eV, and χ Si ≈ 4.15 eV, this model should predict both
accumulation and depletion (Fig. 9) metal-semiconductor contacts. This is generally not
seen with Si; there is little evidence for the existence of any accumulation type metal to
heavily doped Si contact. The reason is poorly understood but related to the restructuring
of the metal-silicon surface. All practical n and p type ohmic contacts to Si are depletion
type.
The barrier heights that are used in modeling ohmic contact to Si are empirical values,
usually measured by capacitance-voltage, current-voltage or photoemission techniques.
Metal and silicide barrier heights to both n and p type Si as a function of metal work function
are illustrated in Fig. 10. The thin vertical lines connect data points for the same metal. The
stronger φ b dependence of metallic suicides on φ m has led to the postulation that some
interface cleanliness or the presence of an interfacial layer affects barrier height. Silicides are
known to make more intimate contact to Si.

Figure 10: Metal-semiconductor barrier height to n- and p-type Si ( φ bn - hollow symbols and φ bp
solid symbols) vs. metal work function. (Ref: S. Swirhun, PhD Thesis, Stanford Univ. 1987)

It can be noticed that the Fermi level pinning is roughly at the same energy within the
forbidden gap for both n and p type Si (i.e. the sum of φ bn and φ bp , is approximately Eg
suggesting that interface and structural factors pin the Fermi level because of a very high
density of interface states (Fig. 11). Note that for ohmic contacts we never need worry
about the occupancy of these states changing, because of very small potential drop across
the contact.

7
EE311 / Saraswat Ohmic Contacts

φBN + φBP = Eg
Figure. 11 Metal-semiconductor barrier height to n-type and p-type Si

Accurate Modeling of Contact Resistance

In practice it is difficult to construct a practical sized contact that passes a uniform current
density over its area so this definition is usually considered in the limit as the elemental
contact area approaches zero.

For a uniform current density ρc can be defined as contact resistance per unit area.
However the situation becomes complicated in real device structures as the current
distribution is non-uniform. Fig. 5 illustrates the current crowding to the front edge of a planar
metal to semiconductor resistor contact. In such situation we can’t use Eq. (2) to calculate
contact resistance.
Contact

I
I

Metal Silicon
I Silicide

Current I

Figure : Non-uniform current distribution in a contact.

8
EE311 / Saraswat Ohmic Contacts

Generalized Contact Model

Contact

Fig. 12. General topology of the contact system. The contact surface is located at Z = 0; Zj
is the effective thickness of the semiconductor layer.

In general, the contact system can only be adequately described by the three basic
transport equations, namely the Poisson and the two carrier continuity equations in 3-D.
Under most circumstances, the equations can be simplified, and 2-D and 1-D models might
be sufficient.

A. 3-Dimensional Model

The three-dimensional contact system has no restriction in the topology. Both metal
potential vm and semiconductor potential v are functions of the spatial coordinate x, y, z.. In
the heavily doped semiconductor region normally used in VLSI contacts, the following
approximations can be made:

(1) The effect of minority carriers is neglected. This assumption is equivalent to neglecting
the depletion depth or band bending in the semiconductor region at the contact
interface with respect to the depth of the semiconductor layer. (The depletion region is
where minority-carrier effects such as recombination become significant.) The total
current density J is then approximately the same as the majority carrier current density
because the metal-semiconductor interfaces inject far more majority carriers than minority
carriers.

(2) By quasi-neutrality, the majority carrier concentration is equal to the active dopant
density. Therefore, only the majority-carrier continuity equation requires solving in the
semiconductor region beneath the contact.

The majority carrier continuity equation outside the contact becomes

∂J x ∂J y ∂J z
∇⋅ J = + + =0 (9)
∂x ∂y ∂z
The current density J in the semiconductor is given by
J = − σE = σ∇v (10)

where v is the potential at coordinate (x, y, z). By combining these two equations we
obtain an equation similar to Ohms law
∇ ⋅ σ∇V = 0 (11)

This formulation also applies to the metal region with a similar set ofexpressions. If the
metal conductivity is much larger than of the semiconductor, which is generally the case, vm,
becomes constant over the entire interface. The entire contact system will then be

9
EE311 / Saraswat Ohmic Contacts

governed by the semiconductor potential v, the only variable that needs to be determined
for a given metal-semiconductor system. The total current can be evaluated over any
sectioned surface A by

Itot = − ∫ J ⋅ dA (12)

Solution of the above equations with appropriate boundary conditions will give the
necessary information about contact resistance. The 3-D model is simple in concept, but
difficult in its computation and generalization. Therefore it is advantageous to simplify the
equations and boundary conditions to 2-D, which are much more tractable and still produce
useful insights.

B. 2-dimensional Model

The transformation from 3-D to 2-D involves a few simplifications. First, the contact interface
is regarded as a 2-D surface perpendicular to the z axis as illustrated in Fig. 12. The
semiconductor or diffusion layer is located below the contact surface with an effective
thickness of zj. The conductivity is assumed to be independent of the x and y spatial
variables, i.e., σ = σ(z). This simplification is valid for most of the modern VLSI
technologies with planar diffusion layers. The aim of the 2-D model is to lump all the effects
of the z-axis into just one single parameter Rs, the sheet resistance of the diffusion layer is
given by
( ∫ σ (z)dz )
−1
Rs = (13)

The metal plane potential vm, seen by the contact will be essentially constant because the
metal layer is usually much more conductive than the semiconductor layer. If this constant
metal potential is set at zero then 3D equations in the contact region can be simplified to the
Helmholtz equation (see the paper by Loh et al. for details)

Rs V V
∇2V = = 2 (14)
ρc lt

with lt, as the transfer length defined as lt = ρ c Rs . In the other bulk region where there is
no contact surface on top, the Laplace equation describes the potential by
∇ 2V = 0 (15)

A solution of these equations gives the I-V relationship at the contact interface. By
comparing the experimental data with the 2-D model an accurate value of ρc can be
extracted. This accurate value can then be used for further calculation of the contact
resistance for the appropriate structure.

C One-dimensional Model

One more spatial variable can be eliminated if the potential changes only slightly, and not
affecting other potentials along the variable axis. The contact system is oriented such that
the y-axis variation is neglected. The Helmholtz equation becomes

∇ 2 V ( x) V ( x)
= 2 (16)
∂2x lt

10
EE311 / Saraswat Ohmic Contacts

The Laplace equation also reduces to Ohm's law. All the boundary conditions become
trivial: V = Vi at the contact leading edge (x = 0) and ∂V/∂x=0 at the contact trailing edge (x
= I).
The potential can be shown as
 l − x
cosh  
 lt 
V ( x) = V i (17)
l
cosh  
 lt 
and total current is simply
W ∂V
Itot = (18)
Rs ∂ x x =0

Historically, this is the approach taken to model the distributive effect of current entering a
contact window. A solution of Eq. (17) and (18) will give contact resistance.

D. Zero-Dimensional Model

Under very special circumstances, such as in large contact windows, an extremely high
value of ρc or when very small contacts are encountered, the I-D model can b e
degenerated into the O-D model or the one-lump model. This is the simplest of all existing
models. Although its validity is scarce, it is the most common model used as a first pass to
estimate the upper bound of ρc. In other words, its accuracy is poor, but it offers a very
intuitive "feel" of the contact resistance. This model states that the potential is constant in the
semiconductor layer and the current density entering the contact window is uniform. From
(3a) the macroscopic "definition" of ρc appears as
ρc
Rc = (19)
A
Intuitively, the contact resistance Rc of a contact will approach ρc /A as the contact sizes
decreases below the transfer length It.

Measurement of Contact Resistance and Specific Contact

Resistivity (ρ c )
The contact resistance Rc is a measured V/I ratio of a structure, that is controlled by the
contact size, structure layout, semiconductor doping density, and specific contact resistivity
ρc. Whereas ρc is a fundamental quantity governed by the interface, Rc is accounts for the
layout dependent non-uniform current flow pattern. A more accurate 2D or 3D analysis is
required for accurate calculations of Rc.

11
EE311 / Saraswat Ohmic Contacts

Fig. 13. 1D transmission line contact model.

The simple 1D model assumes a semiconductor modeled by a distributed sheet

resistance Rs and no vertical extent. The metal sheet resistance is assumed negligible, i.e.,
uniform vm. Assuming a long transmission line the contact resistance is given by the
characteristic impedance of the line which can be obtained by a solution of Eq. (17) and
(18). The transmission line like model gives current density that decreases roughly
exponentially from the leading edge of the contact to the trailing edge.

 
I1 exp( − x lt )
x
I( x) = I1 exp  − =
ρ c Rs 
(20)


The characteristic length of the transmission line lt = ρ c Rs is the distance at which 63%
of the current has transferred into the metal. This model is valid only for an electrically long
contact (d >> lt).

Contact resistance test structures are usually fabricated with other conventional test devices
on the same die or wafer to monitor a particular process. Therefore, the most commonly
used contact test structures for extraction of p, are planar devices: the cross bridge Kelvin
resistor (CBKR), the contact end resistor (CER), and the transmission line tap resistor
(TLTR). In all of these structures, a current is sourced from the diffusion level up into the
metal level via the contact window. A voltage is measured between the two levels using
two other terminals. The contact resistance for each structure is simply this voltage divided
by the source current. It is important to realize that each device measures the voltage at a
different position along the contact as shown in Fig. 14; hence the resistance values
measured are different, and must be clearly defined and distinguished from one another.

12
EE311 / Saraswat Ohmic Contacts

Fig. 14

The front resistance Rf is defined as the ratio of the voltage drop Vf across the interfacial
layer at the front edge of the contact, where the current density is the highest, to the total
current I1 flowing into the contact. For a bounded structure (I2 = 0) it can be shown by solving
of Eq. (17) and (18) that
Rs ρ c
R f = V f / I1 = coth ( d / lt ) (21)
w
For a very large value of lt or for d << lt
ρc
Rf ≈ (22)
wd
which is the equation for uniform current density. A convenient structure to make the
measurements is the transmission line tap resistor in which several contacts are made to a
long diffused line.

Fig. 15 Transmission line tap resistor for Rf measurement

V24 = V f + IRSi + V f

V24
Rt = = 2 R f + Rs ls w (23)
I

13
EE311 / Saraswat Ohmic Contacts

Fig. 16. Characteristic plot of the total resistance as a function of the contact separation in a
TLTR structure. The nominal dimensions were l1 = 30, l2 = 80, l3 = 120, l4 = 150, l5 = 180, w
= 40, W = 50 µm.

By varying the value of ls the value of Rf can be determined as shown in the figure
above. However, since Rf is a smaller number in comparison to the RSi the relative
accuracy of this method is depends upon the measurement accuracy. Slight error in the
measurement of Rt can result in a large error in the calculated value of Rf .

Fig 17. (a) Test pattern for Re contact end resistance measurement. Where current I is forced through pads
I and 2 and voltage V across pads 3 and 4 is sensed. RE = V/I. (b) Cross-sectional view of (a)

Similarly the end resistance Re is defined as the ratio of the voltage drop Ve across the
interfacial layer at the back edge of the contact, where the current density is the lowest, to the
total current I flowing into the contact and is given by

14
EE311 / Saraswat Ohmic Contacts

Vm − V4 Rs ρ c
Re = Ve / I = = (24)
I w sinh(d / lt )
The most popular structure is the cross bridge Kelvin resistor. It assumes a uniform current
density and uses Eq (2) for calculations.

1 2 Metal

l
. l
l
l
N+ Diffusion

Rk =
Vk V14 ρ c
= = 2
. Vk
I I23 l
3 4
Metal .
N+ Diffusion I
Contact

Fig. 18 cross bridge Kelvin resistor

In all of these structures we have assumed that current flows only in one direction. In
reality the current flow is highly 2D.

N+ diffusion
region

Fig. 19 Schematic of 2D current flow in the contacts.

This leads to overestimation of ρc. For example in a Kelvin structure the measured
resistance does not scale with the contact area (Fig. 20).

15
EE311 / Saraswat Ohmic Contacts

Fig. 20 Contact resistance as a function of contact area for Kelvin structure

A detailed 3D or at least 2D simulation should be used to determine the correct values of

specific contact resistivity, ρc.

By comparing the experimental data with the 2-D model an accurate value of ρc can be
extracted. This is illustrated in Fig. 21 for the case of Kelvin structure. This accurate value can
then be used for further calculation of the contact resistance for the appropriate structure.

Fig. 21. Kelvin resistance vs contact area. Diffusion width(w) is larger than contact window size (l)
by 5 µm in the top set of curves. The ideal case is w = l . Sheet resistance of the diffusion is 11
Ω/sq. the simulation parameterρc is varied from 2.33 x 10-7 to 2.33 x 10-9 Ωcm2.

16
EE311 / Saraswat Ohmic Contacts

Requirements of Ohmic Contacts

1. Low contact resistance to both N+ and P+ regions

2. Ease of formation (deposition, etching)

3. Compatibility with Si processing (cleaning etc.)

4. No diffusion of the contact metal in Si or SiO2

5. No unwanted reaction with Si or SiO2 and other materials used in

backend technology.

6. No impact on the electrical characteristics of the shallow junction

7. Long term stability

Aluminum Contacts to Si

Al has been used for a long time to make contacts to Si because it meets
many of the above requirements.

Aluminum

Oxide
Oxide
N+

Silicon

For shallow junctions it suffers with the problem of junction spiking.

17
EE311 / Saraswat Ohmic Contacts

• Silicon has high solubility in Al ~ 0.5% at 450ºC

• Silicon has high diffusivity in Al. At 450°C D = 10-8 cm2/sec

• Si diffuses into Al. At 450°C Dt ≈ 40µm

18
EE311 / Saraswat Ohmic Contacts

Si surface after etching Al shows spiking

• Voids form in Si which fill with Al: “Spiking” occurs.

• Pure Al can’t be used for junctions < 2-3 µm
• By adding 1-2% Si in Al to satisfy solubility requirement junction spiking
is minimized

Aluminum + 1-2% Si

Oxide
Oxide
N+

Silicon

• But Si precipitation can occur when cool down to room temperature ⇒

+
bad contacts to N Si as the Si precipitates are saturated with Al which is
a p type dopant

19
EE311 / Saraswat Ohmic Contacts

Silicide Contacts

Barrier
TiW
TiN Aluminum

Oxide
Oxide
N+

Contact TiSi2
PtSi Silicon

• Silicides like PtSi, TiSi2 make excellent contacts to Si

• However, they react with Al

• A barrier like TiN or TiW prevents this reaction

20
EE311 / Saraswat Ohmic Contacts

Barriers

Structure Failure Failure Mechanism

Temperature (Reaction products)
(˚C)
Al/PtSi/Si 350 Compound formation
(Al 2Pt, Si)
Al/TiSi2/Si 400 Diffusion
(Al5Ti7Si12, Si at 550˚C)
Al/NiSi/Si 400 Compound formation
(Al 3Ni, Si)
Al/CoSi2/Si 400 Compound formation
Al 9Co2, Si)
Al/Ti/PtSi/Si 450 Compound formation
(Al3Ti)
Al/Ti30W70/PtSi/Si 500 Diffusion
(Al2Pt, Al12W at 500˚C)
Al/TiN/TiSi2 /Si 550 Compound formation
(AlN, Al3Ti)

• Silicides react with Al at T < 400°C

• A barrier like TiN or TiW prevents this reaction upto 500°C

Progress of Interfacial reaction

21
EE311 / Saraswat Ohmic Contacts

Contact resistivity vs anneal temperature for a variety of

metallization systems indicating the effectiveness of various
barriers (Ref: Wittmer, JVST 1984)

ΦB (eV)

T (°C)
.

Without the barrier the contacts will be severely degraded as shown by

change in barrier height in an Al/PtSi Schottky contact without the barrier.
At higher temperature Al reacts with PtSi reducing the barrier height. If
there was a shallow junction being contacted, it would have been
damaged.

22