EE 441 Fall 2016: Tadigadapa
Physical Vapor Deposition
EVAPORATION
SPUTTERING
• Typically used for metallization of semiconductors.
• Both Evaporation & Sputtering are done in vacuum environments.
• Typically:
– Evaporation Pressures are < 10-6 Torr (1Torr = 133.32 Pa)
– Sputtering Pressures are ~ 10-3 Torr.
• Optimization of Physical Vapor Deposition process includes:
–
–
–
–
–
–
Film Quality
Film Uniformity (Thickness)
Film Stress & Adhesion’
Film Stoichiometry (For Multi-component Films)
Film Step Coverage (Conformality)
Film Deposition Rates
EE 441 Fall 2016: Tadigadapa
Why Metallization Quality is Important?
• Metallization is used for Gates & Interconnection in IC’s. It can
easily affect the speed of a circuit by virtue of the RC-Time
constant of the signal line.
EXAMPLE:
• For a 1 cm long Polysilicon interconnection runner on 1mm thick
SiO2 (eox/e0 = 3.9), Polysilicon thickness of 5000Å and resistivity r
of 1000 mW-cm.
L
W .L
r e ox L2
RC  Rs e ox

W
d ox d poly d ox
• Since e0 = 8.86 x 10-12 F/m
1x105 3.9x8.86x10 12  2 2

RC 
10 
10
6
5x10
1x10
 7x10-8 s  70 ns
• Reduction in Rs is essential for high-speed circuits.
EE 441 Fall 2016: Tadigadapa
Vacuum & Vacuum Pumps
Rough Vacuum:
0.1 Torr –760 Torr
Medium Vacuum:
10-4 Torr – 10-1 Torr
High Vacuum:
10-8 Torr – 10-4 Torr
Ultrahigh Vacuum:
< 10-8 Torr
Rotary Vane Pump
Diaphragm Pump
Diffusion Pump
Turbomolecular Pump
Cryopump
• This division is based upon the pumps and technology
required to attain this vacuum.
• Terminology:
– The mass flow rate of a gas is given by:
dG d
qm 
 ( rV ) r = mass density, V=Volume
dt
dt
– The throughput of a gas Q is given by:
Q  qm
P
r
And has the units of
Pressure - Volume
Time
EE 441 Fall 2016: Tadigadapa
Vacuum Basics
• Gas flows are measured typically in standard volume i.e.,
volume that an equivalent amount of gas will occupy @ 273K
and 1 atmosphere pressure.
• 1 Standard Liter is the amount of gas that would occupy 1
liter @ 1 atm. and 273K. Since 1 mole of a gas occupies 22.4
liters at standard conditions,
• 1 Standard liter = (22.4)-1 moles
• 1 Standard liter/minute = a throughput of 760 (Torr-liter)/min.
• The conductance C of a vacuum component:
C
Q
Q

P1  P2 P
P1
P2
EE 441 Fall 2016: Tadigadapa
Vacuum Basics
• Conductance in parallel simply add i.e.,
C = C1 + C2 + C3 +……
• Conductance in Series add as inverse:
1
1
1
1



 ......
CSeries C1 C2 C3
Chamber
C1 C2 C3
• For a tube of diameter D and length L in viscous flow regime (<1mTorr)
4
D
C  1.8x105
Pav (Volume/second)
L
• Pav is the average pressure of P1 & P2 in Torr.
• Pumps are usually specified in terms of pumping speeds Sp as:
Q dV p
Sp 

Pp
dt
• where Pp is the inlet pump pressure. A pump of 1000 slm pumps 1000 slm
at an inlet pressure of 1 atmosphere.
EE 441 Fall 2016: Tadigadapa
Vacuum System
Pressure
Gage
Vacuum Chamber
Vent
Baffle Valve
Roughing
Line
Rough Pump
High Vacuum Liquid N2
Trap
Pump
Backing Line
• A vacuum station of volume V
with no leaks, the chamber
pressure P at any time t after
pump down has been initiated is
given by:
 S p t  QOutgassing
P  P0 exp 

Sp
 V 
• Qoutgassing is the outgassing rate
within the chamber – After ~ 1
hour of pumping the second
term dominates
EE 441 Fall 2016: Tadigadapa
Load Locked Vacuum System
Pressure
Gage
Main
Vacuum Chamber
Sample
Load Lock
Chamber
Sample Transfer
Arm
Baffle Valve
Roughing
Line
Rough Pump
High Vacuum Liquid N2
Trap
Pump
High Vacuum
Pump
Backing Line
Rough Pump
Backing Line
Rotary Vane Pump
EE 441 Fall 2016: Tadigadapa
• Rough vacuum pumps involve the positive displacement of gas
through the mechanical movement of a piston, vane, plunger etc. The
process involves:
– Capture of a volume of gas – Compression of the captured volume – Gas
expulsion
– For an ideal gas the pressure differential is just the ratio of the fully
expanded to fully compressed volumes.
– For an exhaust pressure of 1 atmosphere, and a compression ratio of 100:1
the lowest pressure that can be achieved is 0.01atm=7.6Torr.
EE 441 Fall 2016: Tadigadapa
High Vacuum Pumps
Diffusion Pump
Turbomolecular Pump
• Both these pumps work by transferring
momentum to gas molecules.
• Oil is heated at the bottom and the vapor
rises through the center stack and is
ejected through vents at very high speeds.
They then strike cooled walls of the pump
and condense. Compression ratio of 108 is
achieved.
• Turbomolecular pump blades rotate at
~20,000 rpm. The stator and rotor are
spaced by ~1mm. Compression ratios of
109 can be achieved due to the many
stages.
• In both pumps, mass of molecules play a
very important role in determining the
compression ratio
EE 441 Fall 2016: Tadigadapa
Pumping Speed of Turbomolecular Pumps
Pumping Speed is a
function of the gas being
pumped: For example
for Alcatel ATP 80, the
nominal pumping speeds
are:
Source: Alcatel
EE 441 Fall 2016: Tadigadapa
Kinetic Theory of Gases
• Treats gas molecules as hard spheres with no other interaction but
physical collisions.
• In this case the probability distribution of velocities P(v) is given by
Maxwell speed distribution
3/ 2
 m 

P(v)  4 
 2k BT 
 mv 2 
v exp 

2
k
T
B 

2
• where m is the mass of molecule, kB is Boltzmann Constant, v is the
velocity of molecule, and T is temperature.
• The average magnitude of the velocity is:

v  c   vP(v)dv 
0
8k BT
m
• Similarly the root mean squared velocity is
vrms 
3k BT
m
EE 441 Fall 2016: Tadigadapa
Kinetic Theory of Gases
• For example for molecular nitrogen at room temperature the value of
vrms = 516 m/s. Since the direction of thermal velocity is random,
under the absence of any pressure gradients, the average flow is zero.
• The primary mechanism by which gas molecules randomize their
velocity is by collision. The average distance traveled by gas
molecules before a collision occurs is called “Mean Free Path” (l).
l
1
2d m2 n
• where dm is the molecular diameter (for common diatomic molecules
as N2 and O2; dm~3Å) and n is the number of gas molecules per unit
volume.
• Using Ideal Gas Law:
n
• where P is the pressure.
N
P

V k BT
EE 441 Fall 2016: Tadigadapa
Kinetic Theory of Gases & Vacuum
• The Mean Free Path (l) is
l
k BT
2d m2 P
• The scattering probability can then be defined as the fraction q/n0 of
molecules that are scattered in a distance d during their travel
through the gas. n0 is the total number of molecules per unit volume
and q is the number that suffered collision. Then
q
d
 1 e l
n0
• For Example: Let us calculate the percentage of molecules that
suffer collision during travel from a source to a substrate in a
deposition system at 4 mTorr and 0.7 mTorr. Let us assume sourcesubstrate distance (d=50cm) and molecular diameter dm = 3Å.
• At 0.7mTorr, l=1.01x104 cm and at 4mTorr, l=2.02 cm.
EE 441 Fall 2016: Tadigadapa
Kinetic Theory of Gases & Vacuum
•
•
•
•
Thus:
@ 0.7 mTorr, q/n0= (1-e-(50/10100))=0.0049=0.5%
@ 4 mTorr, q/n0= (1-e-(50/2.02))=1=100%
This means that during evaporation molecular motion is more or less
non randomized and there is line of sight deposition. [Remember
evaporation occurs at P < 1mTorr]
• However during Sputtering which is done at ~ 1mTorr there is
considerable randomization of direction of travel (unless a bias is
applied to the substrate). This leads to better uniformity of deposition
on stepped surfaces or better conformality.
• Most metallization in the industry is done by sputtering.
• However from research perspective, evaporation provides a unique
flexibility by a process commonly known as “Lift-Off” process.
EE 441 Fall 2016: Tadigadapa
EVAPORATION
• At every temperature there exists an equilibrium pressure
Pe of vapor above the material.
• When the sample is below the melting temperature –
SUBLIMATION.
• When the sample is in molten state the process is called
EVAPORATION.
• Evaporation involves molten samples because most
materials of interest have high vapor pressures in this
temperature range giving rise to high evaporation rates at
high vacuum.
Evaporation
EE 441 Fall 2016: Tadigadapa
• Figure gives the Pe of common element
as determined experimentally.
• To obtain reasonable deposition rates
the sample Pe ~ 10mTorr.
• Refractory metals like W, Mo, Ta, and
Ti have very low vapor pressures and
require temperatures in excess of
2500°C to achieve reasonable rates.
• The mass evaporation rate from a
source in an evaporator can be given
by:
RMe 
M
Pe
2k BT
where RMe is the mass crossing per unit area (flux) per unit time, M is the
atomic mass, Pe the equilibrium vapor pressure of the material being
evaporated and T is the temperature in Kelvin.
EE 441 Fall 2016: Tadigadapa
Evaporation
• The mass loss rate from a crucible of constant area A can be
calculated as:
RML  
M
M Pe
Pe dA 
A
2k BT
2k B T
• In practice, to find the deposition rate on the surface of a wafer, we
need to determine the fraction of the material leaving the crucible that
accumulates on the surface of the wafer.
• The ejected molecules travel in straight lines inside the evaporator.
• If we further assume,that the material that arrives on
the wafer sticks and remains there – then the constant
of proportionality (fraction of atoms reaching the
wafer) is just the fraction of the total solid angle
subtended by the wafer as seen from the crucible and is
given by:
CosCos
k
R 2
Wafer
r2

R
r1 
Charge
EE 441 Fall 2016: Tadigadapa
Evaporation
• This means that wafers directly above the crucible will be coated
more heavily than wafers off to the side.
• One method is to place the wafers symmetrically is on the surface of
a sphere. In this case:
R
Cos  Cos 
2r
• Since for a spherical arrangement; r1=r2=r.

• Solid Angle Definition:
A
W 2
r

• Since Max A= 4r2; Max W= 4.
Mass Arrival Rate/Unit Area
Deposition Rate 
Mass Density of the Film ( r )
EE 441 Fall 2016: Tadigadapa
Evaporation
• Thus the deposition rate Rd can now be calculated as:
M
Rd 
2k B
2

Pe   R  1  1
A  
2
T   2r  R  r
Pe A
M

2k B r 2 T 4r 2
•
•
•
•
Here the first term depends on the material to be evaporated
The second term depends on the temperature.
The third term depends on the geometry of the chamber.
In practice, the deposition rate and film thickness are measured
using a Quartz Crystal – which measures the shift in the resonance
frequency due to additional mass of the material from evaporation
EE 441 Fall 2016: Tadigadapa
Example
• Aluminum charge is maintained at a temperature T=1100°C.
The wafers are on a spherical dome of radius 40cm and the
diameter of the crucible is 5cm. Mass density of Aluminum is
27000kg/m3.
• At 1100°C the vapor pressure of Al=Pe=1x10-3Torr
• Area of the crucible=(d2/4)=(.52/4)=19.6cm2
• Atomic mass of Aluminum=27
• The deposition rate of Aluminum for the above conditions is:
• Rd = 8.45x10-6x3.59x10-3x9.8x10-4 = 17.8Å/min!
• The arrival rate of aluminum atoms is just the growth rate
times the number density of aluminum i.e.
• JA1 = (R.r/M)*AV =1.78 x 1018 atoms/m2-s
• Where AV is Avogadro’s Number.
Evaporation
EE 441 Fall 2016: Tadigadapa
• There are three types of heating
methods:
– Resistive
– Inductive
– Electron Beam
• Problems with resistive evaporation
include outgassing from the filament
source.
• E-beams on the other hand only melt the
charge. The electron gun under the
crucible ejects intense, high energy
beam. The beam can be rastered across
the charge to melt a significant fraction
E-Beam Source
of the surface.
• Wide range of materials can be deposited using e-beam evaporators
– including refractory metals. X-rays are generated – which can
damage the MOS oxide and substrates. Care must be taken about
such damage.
Evaporation Filaments
Step Coverage
EE 441 Fall 2016: Tadigadapa
• Step coverage is the primary limitation of
Evaporation
• Frequently used method of improving step
coverage is by rotation of the wafers around the
source.
• A second method to improve step-coverage is to
heat the wafer. Banks of IR lamps are used for
this purpose. This mobility helps to move
material to areas of low deposition rates
• Multicomponent Films: Such as alloys and compound
materials can be deposited using different techniques:
•
•
•
•
Single Source Evaporation
Simultaneous Evaporation
Sequential Evaporation
Most important parameters
to control is the
stoichiometry of such films.
EE 441 Fall 2016: Tadigadapa
Lift-Off Process
Expose Substrate Coated with Photoresist
Develop Photoresist
Deposit Metal by Evaporation (Line of Sight
Deposition)
Dissolve Photoresist in Resist Stripper
Positive Photoresist
Metal
• Advantages of Lift-off include:
• No etching of Metal is required
• Smaller capital investment in target materials.
Glow Discharge
EE 441 Fall 2016: Tadigadapa
• A plasma is a partially ionized gas.
• There are many types of processes that occur
in a glow discharge.
• Dissociation:
e* + AB ↔ A + B + e
• Atomic Ionization: e* + A ↔ A+ + e + e
• Molecular Ionization: e* + AB ↔ AB+ + e + e
• Atomic Excitation: e* + A ↔ A* + e
• Molecular Excitation: e* + AB ↔ AB* + e
* Refers to excited energy states.
• Dissociated atoms or molecular fragments are
called RADICALS. Radicals have incomplete
outer shells and are therefore extremely
reactive.
• Ions are charged atoms or molecules such as
A+ or AB+.
EE 441 Fall 2016: Tadigadapa
Glow Discharge
• At a pressure of ~1 Torr, a dc power supply connected to the
parallel plate assembly starts a plasma. At this pressure we need
~800 V for 10 cm electrode separation for electrical breakdown.
• Due to the electric field electrons are accelerated to the anode and
positively charged ions are accelerated towards the cathode.
• Electrons due to their small mass are accelerated more rapidly
than ions.
• When ions strike the cathode they release secondary electrons
from the material of the cathode.
• These electrons gain energy and in turn collide inelastically with
neutral atoms to create more ions.
• The processes of secondary electron generation and ion creation
sustains the plasma!
EE 441 Fall 2016: Tadigadapa
Glow Discharge
• In the bulk of Plasma, the densities of ions and electrons are equal.
• Since electrons are very rapidly accelerated from the cathode there
is a net positive charge near the cathode.
• As electrons accelerate towards the anode, they collide
inelastically with atoms and create ions. The field from the ions
shields the cathode – thus the ion density peaks and falls to a
constant value.
• Excited atoms have a very short lived state ~ 10-11 seconds! And
they decay by releasing photons – causing the glow discharge.
Moderate energy electrons are required for this process (<15 eV).
• The electric field in Crook’s dark space is most important from
fabrication view point – ions are rapidly accelerated to the cathode
due to the high electric field in this region.
EE 441 Fall 2016: Tadigadapa
RF Discharge
• Required in processes where the electrode is an insulating
material.
• As ions strike the insulating cathode they become charged
until eventually the plasma is extinguished.
• AC source at a RF frequency of 13.56 MHz is typically used
along with a tuning network to match the impedance between
the plasma and the power source.
• At f > 10 kHz, the slow ions
cannot follow the voltage
change. Electrons strike both
electrodes giving them a net
negative charge with respect to
the plasma.
RF Discharge
EE 441 Fall 2016: Tadigadapa
• Conserving current through the plasma, it can be shown for an
asymmetric chamber,
V1  A2 
  
V2  A1 
4
• Where A1 and A2 are the areas of the two electrodes
DC voltage as function of position
– in RF Plasma
V1 = VPlasma – VTop Electrode
V2 = VPlasma – VBottom Electrode
Sputtering
Paschen’s Law:
VBreakdown
PL

log 10 ( PL )  b
EE 441 Fall 2016: Tadigadapa
• Sputtering provides better step coverage
• Less radiation damage
• Better layers of multi-component materials
• In sputtering high energy ions strike the
target containing the material to be
Where P is the gas pressure, L is
deposited.
the electrode spacing and b is a
• An inert gas is used at a gas pressure of
constant.
0.1 Torr. (Mean free path ~ 100mm)
• When energetic ion strikes the surface
of a material:
Parallel plate plasma reactor in
vacuum chamber
• For Energy < 10eV ion may adsorb on the
surface giving up the energy as heat.
• For Energy: 10eV-10keV ion may penetrate
several atomic layers depositing energy into
material causing structural changes .
• For Energy >10keV ions penetrate deep into
the material and the process is called ionimplantation.
EE 441 Fall 2016: Tadigadapa
• In sputtering:
•
•
•
•
• Part of the energy goes to heat the target material.
• Part of the energy goes into the physical rearrangement of top few
atomic layers.
When this happens, substrate atoms and clusters of atoms are ejected
from the surface of the target. These atoms and clusters escape with
energies ~10 – 50 eV (100 times energy of evaporated atoms).
This additional energy provides sputtered atoms surface mobility for
improved step coverage.
At typical sputtering energies, 95% of the ejected material is atomic
and the remainder is diatomic molecules.
The ionization potential of some common gases are:
Gas
1st Ionization (eV)
2nd Ionization (eV)
Argon
15.759
27.629
Oxygen
13.618
35.116
Nitrogen
14.534
29.601
High Density Plasma
EE 441 Fall 2016: Tadigadapa
• Typical densities of ions in normal sputtering
systems are 0.0001%. While in a magnetron
(high density plasma: HDP) system this
Electrons ejected from the cathode
approaches 0.03% (1011 ions/cm3)!
are confined by Lorentz force to
stay in Cathode dark space.
• In addition a magnetron allows the formation
of plasma at lower pressure ~ 10-5 –10-3 Torr.
• Application of a magnetic field (B) causes a
Lorentz force(F):
 
F  q(v xB)
• If the velocity v of the ion is constant it will
cause a circular motion of radius (r):
mv
r
qB
• This orbital motion increases the probability of
collisions – i.e. creation of more ions!
Magnetron Target
EE 441 Fall 2016: Tadigadapa
Ions
S
N
Areas of maximum target erosion
N
S
 
v xB 
S
N
Target is the cathode
electron
Ions are heavier than electrons such that
their v varies slowly perpendicular to
the magnetic field and continue to be
rapidly accelerated towards the target.
ion

B

v
Into the plane of
the transparency!
Whereas electrons move
in spiral paths constantly
creating new ions
Sputter Yield
EE 441 Fall 2016: Tadigadapa
• Sputter deposition depends upon:
• Ion flux to the Target
• Probability that an impact will eject target
atoms (Sputter Yield)
• Transport of the sputtered atoms across the
plasma to the substrate.
• Ion flux in dc plasma is given by LangmuirChild relationship as:
Sputter yield as a function
of ion-energy for normal
incidence of argon ions
1 V 3/ 2
J ion 
mion d 2
• Where mion is the mass of the ion, V is the
voltage difference between the target and the
substrate wafer and d is the dark space
thickness.
• Sputter yield (S) is the ratio of the number of
target atoms ejected from the target to the
number of ions incident on the target.
EE 441 Fall 2016: Tadigadapa
Sputtering
• For each target atom, threshold energy
(ETh) exists below which no sputtering
occurs.
• This threshold energy is ~10-30 eV.
• For ions with energy slightly greater than ETh, the yield
increases as square of the energy up to about 100 eV.
• Thereafter the increase is linear till about 750 eV.
• Above 750 eV, the yield increases very slightly until the onset
of ion implantation.
• The maximum sputter yield occurs at ~1 keV.
• Sputter yield is a weak function of the plasma composition and
increase with ion-mass. see fig. below
EE 441 Fall 2016: Tadigadapa
Sputter yield as a function of the
bombarding ion atomic number
for Silver, Copper and Tantalum
at an energy of 45 keV is shown.
Notice the maximum in sputter
yield for ions with full or close
to full valence shells such as Ar,
Kr, and Xe.
Angular Dependence of Sputtering:
At low energies, a minimum exists at near
normal incidence.
At high energies,
S
Angular Dependence of Sputter yield.  is the
angle between target normal and incident ion
velocity vector.
M gas ln E 1
M T arg et E Cos
Sputtering
EE 441 Fall 2016: Tadigadapa
• Morphology of Films:
• At low substrate temperatures and ion energies, the material is
amorphous, highly porous with low mass density.
• At somewhat raised temperatures (~200C) or low pressure, films are
highly specular with very small grains.
• Increasing temperature increases grain size (tall columnar grains or
large 3D grains). The surface of these films are moderately rough and
appear milky or hazy!
• Step Coverage:
• On the top surface and near the upper
corner the deposition rates are high.
Sidewall thickness tapers towards the
bottom. At the bottom corner a
pronounced notch/crack exists
EE 441 Fall 2016: Tadigadapa
Stress in Deposited Layers
• A thin film deposited on a substrate can be either tensile (i.e. will
relax by contracting) or may be in compressive stress.
• If the stress in the film is too large, the film may peel from the
surface of the wafer.
• Large stresses may cause void formation during subsequent
annealing.
• One component of stress is caused by thermal expansion
mismatch of the film with the substrate – when deposition is not
at room temperature. If E is the Young’s modulus, n is the
Poisson’s ratio, and a is temperature coefficient of expansion,
then the thermal stress sth is given by:
s Th 
E film
TDeposition
1 n film
T0 is the room temperature.
 (a
T0
film
 a substrate )dT
EE 441 Fall 2016: Tadigadapa
Intrinsic Stress in Thin Films
• Intrinsic stress in films is hard to quantify and depends
on the deposition conditions as well as any subsequent
high-temperature process. If the deposition is not
performed at very high temperatures, the intrinsic stress
is small.
• Stress in a film is measured using the change in bow of
the wafer before and after film deposition. The film
stress is given by:
Deposited Film
d E T2
s
t 1 n 3R 2
Wafer
d
• Where d is the change in the deflection of the Compressive Stress
center of the wafer, t is the film thickness, R is Tensile Stress
the radius of the wafer and T is the thickness
d
of the wafer.
Wafer
EE 441 Fall 2016: Tadigadapa
Chemical Vapor Deposition (CVD)
In CVD the constituents
of vapor phase react at
a hot surface to deposit
a solid film.
•
•
•
•
•
•
•
•
Mass transport of reactant & diluent gases in the bulk gas flow region.
Gas phase reactions (homogeneous) leading to film precursors.
diffusion of the
Mass transport of film precursors & reactants to the growth surface. -->
precursors
Adsorption of film precursors and reactants on the growth surface.
Surface reactions (heterogeneous) of adatoms occurring selectively on heated
surfaces. (Surface diffusion)
Surface migration of film formers to the growth sites & nucleation
Desorption of by-products of the surface reactions.
Mass transport of the by-products in the bulk gas flow region away from the
substrates.
https://www.youtube.com/watch?v=hkYb35e5JGo
EE 441 Fall 2016: Tadigadapa
Chemical Vapor Deposition (CVD)
Transport in gas phase takes place,
through diffusion proportional to the
diffusivity of the gas D, and the
concentration gradient across the
boundary layer that separates the bulk
flow (source) and substrate surface
(sink). Assuming laminar flow and the
establishment of a equilibrium boundary
layer, the flux of the depositing material
F1 can be given by:
c
F1  D 3 Re  hG (CG  CS )
2L
where c is the concentration difference across d, L is the length of the tube
where deposition occurs, and Re is Reynolds number. Since Reynolds number
is proportional to square root of velocity, so does film growth rate in this mass
flow controlled regime.
At high flow rates, the reaction rate controls the deposition rate.
EE 441 Fall 2016: Tadigadapa
Chemical Vapor Deposition
• F2 is the flux of reactant consumed by the reaction at the surface. Assuming a
first order reaction kinetics, this can be written as:
F2  k S CS
• Where ks is the surface reaction rate and CS is the surface concentration of the
reacting species. Assuming steady-state deposition conditions, these two
processes must be equal. Thus F=F1=F2.
• This leads to the equation:
1
•
 k 
CS  CG 1  S 
 hG 
The growth rate of the film is now given by:
v
F
k h CG
 S G
N k S  hG N
• Where v is the deposition rate and N is the number of atoms incorporated per
unit volume into the film.
EE 441 Fall 2016: Tadigadapa
Chemical Vapor Deposition
• If we define Y as the mole fraction of the incorporating species in the gas
phase:
Y
CG
CT
• Where CT is the concentration of all molecules in the gas phase. Y is also equal
to the partial pressure of the incorporating species, PG, divided by the total
pressure in the system.
Y
CG
CT

PG
PTotal
• The equation for the deposition velocity is now
v
k S hG CT
Y
k S  hG N
EE 441 Fall 2016: Tadigadapa
Chemical Vapor Deposition
Case 1: Surface Reaction Controlled Regime: kS<<hG
CT
v
k SY
N
• Mass transfer through the gas boundary layer is relatively fast, while the
surface reaction is sluggish. In this case CS approaches CG.
Case 2: Mass Transfer Controlled Regime: hG << kS
CT
v
hGY
N
• Here the surface reaction rate is fast with respect to the mass transfer rate. Here
the surface concentration CS approaches zero since the reactants are consumed
immediately upon arrival.
EE 441 Fall 2016: Tadigadapa
Chemical Vapor Deposition (CVD)
•
•
•
E
kS term with k S  k0 exp a k T 

B

hG term with hG=Const
•
Surface reactions are modeled by
thermally activated phenomena
proceeding at a rate R, given by:
kS = k0 exp [-Ea/kBT]
Where k0 is the frequency factor, Ea is
the activation energy in eV, and T is
the temperature in K.
In a Low Pressure CVD (LPCVD)
system, the deposition is reaction rate
limited since the diffusivity of
reactants is ~1000 greater than at
atmospheric pressure. Here the wafers
are typically vertically stacked. Mass
transport is not an issue.
In Atmospheric Pressure CVD
(APCVD) reactor, the reaction is mass
transport limited and the wafer are
place horizontally next to each other.
EE 441 Fall 2016: Tadigadapa
Chemical Vapor Deposition Systems
EE 441 Fall 2016: Tadigadapa
Reactor Design Aspects
• In APCVD (mass transfer controlled regime) wafers are placed
on an incline so that the cross sectional area of the chamber is
decreased, increasing the gas flow velocity along the length of
the susceptor. This compensates for both boundary layer and
depletion effects.
• In the LPCVD systems (reaction rate controlled regime) a
temperature gradient of 5-25°C is maintained along the length
to compensate for depletion effect when operating in this
regime.
EE 441 Fall 2016: Tadigadapa
CVD & Step Coverage
•
•
•
•
•
•
Mean free path (l) of atoms plays a crucial role in
determining the step coverage characteristics and
can be estimated to be:
0.005/PT(Torr) cm @ 300K
0.01/ PT(Torr) cm @ 600K
where PT is the total pressure.
Case A occurs when reactants after first hitting
the solid have enough energy for surface
migration before bond establishment.
In case B the mean free path is large enough that
the molecules reach the bottom but do not have
sufficient energy for surface migration.
In case C, the mean free path is too short for the
reactants to reach the bottom of the trench.
The value of the integral of the material flux F1d
and thus CVD film thickness are directly
proportional to the range of feasible angles of
arrival, , of the depositing species.
EE 441 Fall 2016: Tadigadapa
LPCVD Processes
• Most LPCVD polysilicon is done with silane SiH4 in furnaces at
temperatures ranging from 575-650C. The activation energy for poly
deposition is ~1.7eV. Typical deposition rates are 100-1000Å/min. It is
common to obtain thickness uniformities of 5% across a batch of 100 100200mm diameter wafers.
• For hot wall deposition of oxide in LPCVD reactor, silane and oxygen,
dichlorosilane (SiCl2H2) and nitrous oxide (N2O) and the decomposition of
TEOS (Tetraethyloxysilane) are commonly used. The silane oxygen
process can be run at substrate temperatures below 500C. The films
deposited are found to contain significant amount of Silanol (SiOH),
Silicon Hydride (SiH) and water. The dichlorosilane/nitrous oxide process
requires to be run at 900C. Excellent oxide quality is obtained in this
process with little hydrogen in the films.
• LPCVD Silicon nitride is most commonly deposited using dichlorosilane
and ammonia (NH3). Typical deposition temperatures are between 750900C. Ellipsometry and etch resistance in HF are typically used to
determine the quality of LPCVD films of nitride.
EE 441 Fall 2016: Tadigadapa
CVD Processes