Chapter 3 Crystal growth, wafer fabrication and

basic properties of silicon wafers

1. Silicon crystal structure and defects.

2. Czochralski single crystal growth.
3. Growth rate and dopant incorporation for CZ method.
4. Float zone single crystal growth and doping.
5. Wafer fabrication.
6. Measurement methods.

VLSI Manufacturing Technology

Instructor: Yiming MA, Email: [email protected] 1
Textbook: Silicon VLSI Technology (by Plummer, Deal and Griffin)
 Czochralski process - crystal growth rate
We wish to find a relationship between pull rate and crystal diameter.
Heat balance:
latent heat of fusion heat heat
(i.e. crystalization) + conducted radiated
= =
heat conducted from through the away
melt to crystal crystal (C).
(A) (B)

L = latent heat of fusion

Freezing occurs between isotherms X1 and X2.
dm
= amount of freezing per unit time
dt
dm dT dT
(1) L + kL A1 = k S A2 k L = thermal conductivity of liquid
dt dx 1 dx 2 dT
= thermal gradient at isotherm x 1
dx 1
(A) (B)
k S = thermal conductivity of solid
dT
 A1,2 = Cross-sectional area = thermal gradient at x 2
dx 2
2
 Czochralski process - crystal growth rate
dm (2)
The rate of growth of the crystal is = v P AN
dt
where vP is the pull rate and N is the density.
k S dT
Neglecting the middle term in Eqn. (1) we have: v PMAX = (3)
LN dx 2


In order to replace dT/dx2, we need to consider the heat transfer processes.


Heat radiation from the crystal (C)
is given by the Stefan-Boltzmann law*:
(
dQ = (2rdx ) T 4 ) (4)
Heat conduction up the crystal (B) is
given by
( )
Q = k S r2
dT
dx
(5)

2πrdx = radiation surface.

σ = Boltzman constant (NOT kB)
kS= thermal conductivity of the solid.
=1 for perfect blackbody.
*Stephen-Boltzmann law:
http://en.wikipedia.org/wiki/Stefan%E2%80%93Boltzmann_law <1 for “grey-body”. 3
 Czochralski process - crystal growth rate

( ) ( ) ( )
2 2
dQ 2 d T 2 dT dk S 2 d T
• Differentiating (5), we have = k S r 2
+ r  k S r (6)
dx dx dx dx dx 2

d 2 T 2 4
• Substituting (6) into (4), we have − T =0 (7)

2
dx k Sr

• kS varies roughly as 1/T, so if kM is the TM

kS = kM (8)
thermal conductivity at the melting point, T

d2 T 2
 − T5 = 0 (9)
dx 2 k M rTM

• Solving this differential equation, evaluating it at x = 0 (the freezing interface) and
substituting the result into (3), we obtain (see text):

5
1 2k M TM (10)
v PMAX =
LN 3r

Vpmax = maximum pull rate, inversely proportional to the square root of crystal radius.
 4
 Crystal growth rate: example

For perfect blackbody: =1

For “grey” body: <1

4 hours to pull one meter long boule.

However, this equation is not so accurate, and in practice feedback is used to adjust
the pulling rate in order to maintain constant diameter. 5
 Dopant incorporation during crystal growth
• Dopants are added to the melt to provide a controlled N or P doping level in the wafers.
• However, the dopant incorporation process is complicated by dopant segregation.
• Generally, impurities “prefer to stay in the liquid” as opposed to being incorporated into
the solid.
• This process is known as segregation. The degree of segregation is characterized by the
segregation coefficient, ko, for the impurity. C
kO = S
CL


CS and CL are the impurity
concentration just on the either side
of the solid/liquid interface. CS
CL

6
 Dopant behavior during crystal growth

CS
kO =
CL

Most k0 values are <1 which means the impurity prefers to stay in the liquid.
Thus as the crystal is pulled, dopant concentration will increase.
In other words, the distribution of dopant along the ingot will be graded.
7
Distribution coefficient: example

=1.019×10-4mol

8
 Dopant incorporation during crystal growth
VO, IO, CO: initial volume, number of impurities,
and impurity concentration in the melt
VL, IL, CL: volume, number, and concentration of
impurities in the melt during growth
VS, CS: volume and concentration of impurities in
the solid crystal
C0 = I0/V0
By definition: CS/CL=k0

If during growth, an additional volume dV freezes, the impurities incorporated

into dV are given by:

k0
I 0 1 − S 
V
I 0 (1 − f ) 0
k
CL =
IL
=  V0 
= = C0 (1 − f ) 0
k −1

VL V0 − VS V0 1 − f

The impurity level in the crystal (Cs)

f = Vs/Vo, the ratio of solid silicon
kO − 1
 C S = C O k O (1 − f) to the initial liquid volume.
9
 Doping concentration versus position along the
grown CZ crystal for common dopants in silicon

Note the relatively flat profile produced by

boron with a ko close to 1.
Dopants with ko << 1 produce much more Doping concentration in a solid as a
doping variation along the crystal. function of the fraction solidified.
(but no dopant has ko>1)
10
Chapter 3 Crystal growth, wafer fabrication and
basic properties of silicon wafers

1. Silicon crystal structure and defects.

2. Czochralski single crystal growth.
3. Growth rate and dopant incorporation for CZ method.
4. Float zone single crystal growth and doping.
5. Wafer fabrication.
6. Measurement methods.

11
 Float-zone crystal growth: overview

• For CZ-grown Si, impurities (O and C) can be introduced from the melt contacting
the SiO2 crucible and from graphite susceptor/supporter.
• This limits the resistivity to 20Ωcm, while intrinsic Si is 230kΩcm.
• These crystals are more expensive and have very low oxygen and carbon and thus,
are not suitable for the majority of silicon IC technology.
• Carrier concentrations down to 1011 atoms/cm3 have been achieved.
• It is far less common, and is reserved for situations where oxygen and carbon
impurities cannot be tolerated.
• Float-zone does not allow as large Si wafers as CZ does (200mm and 300mm) and
radial distribution of dopant in FZ wafer is not as uniform as in CZ wafer.
• It is good for solar cells, power electronic devices (thyristors and rectifiers) that
use the entire volume of the wafer not just a thin surface layer, etc.

12
 Float-zone crystal growth process
• Polycrystalline silicon is converted into single-crystal by zone heating (zone melting).
• The entire poly-Si rod from the EGS process is extracted as a whole.
• The rod is clamped at each end, with one end in contact with a single crystal seed.
• An RF heating coil induces eddy currents (power I2R) in the silicon, heating it beyond its
melting point in the vicinity of the coil.
• The "floating" melt zone is about 2cm wide/high.
• The seed crystal touches the melt zone and is pulled away, along with a solidifying Si boule
following the seed. The crystalline direction follows that of the seed single crystal.
• Limited to about a 4" wafer, as the melt zone will collapse - it is only held together by
surface tension (and RF field levitation).

Melt is not held in a

container, it is
“float”, thus the
name “float zone”.

13
 Float-zone: zone refining

• Dopants/impurities prefer to stay in the liquid than

in the solid.
• Thus, the impurities generally stay in the melt zone,
and don't solidify in the boule.
• That is, segregation (and evaporation) of impurities
in the melt zone help purify the Si further.
• One can "purify" FZ wafers further by successively
passing the coil along the boule. The impurities then
segregate towards the end of the boule.
• Of course, if neglecting impurity evaporation, the
total amount of impurity is the same. Yet the
impurity at the lower part is much lower than, and
at the upper part approaches to, the original
impurity concentration.
• Thermal instability in the melt zone can cause micro-
variations in composition and doping. Difficult in
making a uniform dopant concentration.

14
 Doping in FZ growth

(Doping can of course be achieved if the starting material poly-crystalline Si rod is

doped. But due to zone-refining, the doping is not uniform along the boule)

Gas doping:
Dopants are introduced in gaseous form during FZ growth.
n-doping: PH3 (Phosphine), AsCl3
p-doping: B2H6 (Diborane), BCl3
Good uniformity along the length of the boule.

Pill doping:
Drill a small hole in the top of the EGS rod, and insert the dopant.
If the dopant has a small segregation coefficient, most of it will be carried with the
melt as it passes the length of the boule.
Resulting in only a small non-uniformity.
Ga and In doping work well this way.

15
 Zone refining
Zone length is L. The rod has initial uniform impurity
concentration of C0.
If the molten zone moves upwards by dx, the number of
impurities in the liquid (=I) will change since some will be
dissolved into the melting liquid at the top (=C0dx) and
some will be lost to the freezing solid on the bottom
(=CSdx=k0CLdx). Thus: (assume cross-sectional area =1)

dI = (C0 -k 0C L )dx, but C L = I / L

x dI
I
0 dx = 
I0 k0 I
C0 −
L
C0 L  C0 L  −k 0 x L
I = −  − I 0 e
k0  k0 
I
C S = k 0C L = k 0 ; I 0 = C0 L
L
 k0 x

C S = C0 1 − (1 − k 0 )e
−
L

  16
 Floating zone crystal growth – zone refining

0.05

Impurity during float-zone growth or zone Zone refining with multiple

refining. One pass of the molten zone passes, k0=0.1. L is the
through the solid. length of the molten zone.
L is the length of the molten zone (2cm)
17
Chapter 3 Crystal growth, wafer fabrication and
basic properties of silicon wafers

1. Silicon crystal structure and defects.

2. Czochralski single crystal growth.
3. Growth rate and dopant incorporation for CZ method.
4. Float zone single crystal growth and doping.
5. Wafer fabrication.
6. Measurement methods.

18
Steps for wafer preparation

Trimming and
Grinding

19
Ingot grinding

20
 Wafer slicing

Traditional method of slicing Wire saw for large wafers

The saw blade itself is about 400m thick, together

with the loss at the seed and tail end of the crystal,
only 50% of the boule ends up in wafer form.

After slicing, mechanical lapping and wet chemical

etching is performed before final chemical
mechanical polishing. The wet etching is typically:
3Si + 4HNO3 + 18HF → 3H2SiF6 + 4NO + 8H2O
21
 Wafer polishing
Chemical mechanical polishing

Slurry consists The rotation and

of nano- pressure generates
particles heat that drives a
(10nm SiO2 or chemical reaction in
Al2O3) and which OH- radicals
chemicals from the NaOH oxidize
(NaOH). the silicon. The SiO2
particles abrade the
oxide away.
22
 Common (not always) wafer surface orientation

Should
be {110} Another flat configuration
plane for {100} n-type wafer

{100} wafer usually breaks along {110} plane (actually Si cleaves naturally along {111}
plane, which meet the surface at an angle of 54.7o, the angle between <001> and <111>).
Sometimes (not often) {100} wafers break along {100} plane. ({100} = (100)+(010)+(001)) 23
Advantage of larger diameter wafers

More chips per wafer for larger wafer.

24
Chapter 3 Crystal growth, wafer fabrication and
basic properties of silicon wafers

1. Silicon crystal structure and defects.

2. Czochralski single crystal growth.
3. Growth rate and dopant incorporation for CZ method.
4. Float zone single crystal growth and doping.
5. Wafer fabrication.
6. Measurement methods.

25
 Hot probe measurement of carrier type

Vm

Cold Hot

e-
n-type wafer

• The hot probe technique is used to determine the type of dopant in a wafer.
• It relies on the generation of the Seebeck voltage (open circuit), i.e. the diffusion
of free carriers (electrons or holes) as a result of a temperature gradient.
• The sign of the voltage tells the carrier type.
• Alternatively, one can measure the current direction (short circuit).
• The current that flows due to the majority carrier is given by J = qn p dT
n n n
dx
Pn is thermoelectric power, negative for electrons, positive for holes. 26
Electrical measurement techniques – four point probe

t S

=(qnn+qpp)-1, n(p) is mobility, n(p) is carrier density.

The four point probe method measures the resistance/resistivity of a wafer.

Using values of carrier mobility, one can calculates the carrier concentration.
Use four points (rather than two) to eliminate the effect of contact resistance.

27
 Hall measurement of carrier type & mobility

Thickness of the slab is t

(along B direction), width is
w (along VH direction). VH is
measured, and used to
calculate RH. Then carrier
concentration p, n=1/qRH.

Hall voltage (VH) measurement

Silicon is placed inside a magnetic field and the resultant motion of charge will
experience a force perpendicular to the electric field. F=qvB
In equilibrium this results in a transverse potential difference known as the Hall
voltage VH. (such that F=qE=qVH/w=qvB, i.e. forces are balanced, so VH=vBw)
The magnitude and sign of the Hall voltage leads directly to the carrier concentration,
type, and mobility (if the resistivity is known).
Current density j=I/tw=E/=qnv, VH=vBw=(I/twqn)Bw=IB/tqn.
Hall coefficient RH  VHt/BI=1/qn=μH, μH is Hall mobility (we know that =1/qn).
(μH is slightly lower than conductivity mobility, due to more scattering when
magnetic field is present.  can be measured from four-point probe) 28
 Crystal CZ growth: GaAs

GaAs Liquid Encapsulated CZ (LEC):

• GaAs single crystal is more difficult to grow than Si.
• At 1238oC, the vapor pressure of As is 10
atmospheres, while Ga is only 0.001 atmospheres.
• So As is rapidly lost to evaporation, resulting in Ga-
rich melt (non-stoichiometric).
• A cap is used to encapsulate the melt and stop
evaporation.
• This cap is typically B2O3 that melts at 400oC.
• This allows the seed crystal to be lowered through
the liquid cap and pulled out of the cap.

29
 Horizontal Bridgman GaAs growth
Historically, high defect density with LEC GaAs limits its use from electronic applications.
Most GaAs for optoelectronics is produced by Bridgman method.
The GaAs charge is held in a sealed ampoule with excess arsenic.
Thus, higher As pressure can be reached that limits As evaporation.

30
 Homework 1

Problem 3.3&3.4 in the textbook

A Czochralski crystal is pulled from a melt containing 1015 cm-3 boron and
2×1014 cm-3 phosphorus. Initially the crystal will be P type but as it is pulled,
more and more phosphorus will build up in the liquid because of segregation.
At some point the crystal will become N type. Assuming kO = 0.32 for
phosphorus and 0.8 for boron, calculate the distance along the pulled crystal at
which the transition from P to N type takes place.

A Czochralski crystal is grown with an initial Sb concentration in the melt of

1×1016 cm-3. After 80% of the melt has been used up in pulling the crystal, pure
silicon is added to return the melt to its original volume. Growth is then
resumed. What will the Sb concentration be in the crystal after 50% of the new
melt has been consumed by growth? Assume ko = 0.02 for Sb.

31