Lecture 21: Silicon wafer

manufacturing

Contents
1 Introduction 1

2 Poly Si manufacture 2

3 Single crystal Si manufacture 5

3.1 Czochralski crystal growth technique . . . . . . . . . . . . . . 5
3.2 Float zone technique . . . . . . . . . . . . . . . . . . . . . . . 7

4 Wafer manufacturing 9

1 Introduction
The first step in integrated circuit (IC) fabrication is preparing the high
purity single crystal Si wafer. This is the starting input to the fab. Typically,
Si wafer refers to a single crystal of Si with a specific orientation, dopant
type, and resistivity (determined by dopant concentration). Typically, Si
(100) or Si (111) wafers are used. The numbers (100) and (111) refers to
the orientation of the plane parallel to the surface. The wafer should have
structural defects, like dislocations, below a certain permissible level and
impurity (undesired) concentration of the order of ppb (parts per billion).
Consider the specs (specifications) of a 300 mm wafer shown in table 1 below.
The thickness of the wafer is less than 1 mm, while its diameter is 300 mm.
Also, the wafers must have the 100 plane parallel to the surface, to within 2◦
deviation, and typical impurity levels should be of the order of ppm or less
with metallic impurities of the order of ppb. For doped wafers, there should
be specific amounts of the desired dopants (p or n type) to get the required
resistivity.

1
 MM5017: Electronic materials, devices, and fabrication

Table 1: Specs of a typical 300 mm wafer used in fabrication. The specifica-

tions include the dimensions, orientation, resistivity, and oxygen and carbon
impurity content.
Specs Value
Diameter 300 ± 0.02 mm
Thickness 775 ± 25 µm
Orientation 100 ± 2◦
Resistivity > 1 Ω−m
Oxygen concentration 20-30 ppm
Carbon concentration < 0.2 ppm

Table 2: Impurities in MGS, after the submerged arc electrode process.

Element Concentration (ppm)
Al 1000-4350
B 40-60
Ca 245-500
Fe 1550-6500
P 20-50
Cu 15-45

2 Poly Si manufacture
The starting material for Si wafer manufacture is called Electronic grade Si
(EGS). This is an ingot of Si that can be shaped and cut into the final wafers.
EGS should have impurity levels of the order of ppb, with the desired doping
levels, so that it matches the chemical composition of the final Si wafers. The
doping levels are usually back calculated from resistivity measurements. To
get EGS, the starting material is called Metallurgical grade Si (MGS). The
first step is the synthesis of MGS from the ore.
The starting material for Si manufacture is quartzite (SiO2 ) or sand. The
ore is reduced to Si by mixing with coke and heating in a submerged elec-
trode arc furnace. The SiO2 reacts with excess C to first form SiC. At high
temperature, the SiC reduces SiO2 to form Si. The overall reaction is given
by
SiC (s) + SiO2 (s) → Si (l) + SiO (g) + CO (g) (1)
The Si(l) formed is removed from the bottom of the furnace. This is the MGS
and is around 98% pure. The schematic of the reducing process is shown in
figure 1. Typical impurities and their concentrations in MGS is tabulated in
2. MGS is used for making alloys. From table 2 it can be seen that the main

2
 MM5017: Electronic materials, devices, and fabrication

Figure 1: Schematic of the submerged arc electrode process. SiO2 is mixed

with coke and heated. It first forms SiC, which further reacts with the
remaining SiO2 forming silicon. The temperature is maintained above the
melting point of silicon so that the molten semiconductor is removed from
the bottom. Adapted from Synthesis and purification of bulk semiconductors
- Barron and Smith

metallic impurities are Al and Fe. Further purification is needed to make

EGS since the impurity concentration must be reduced to ppb levels.
One of the techniques for converting MGS to EGS is called the Seimens
process. In this the Si is reacted with HCl gas to form tricholorosilane,
which is in gaseous form.

Si (s) + 3HCl (g) → SiHCl3 (g) + H2 (g) (2)

This process is carried out in a fluidized bed reactor at 300◦ C, where the
trichlorosilane gas is removed and then reduced using H2 gas.

2SiHCl3 (g) + 2H2 (g) → 2Si (s) + 6HCl (g) (3)

The process flow is shown in figure 2. A Si rod is used to nucleate the

reduced Si obtained from the silane gas, as shown in figure 3. During the
conversion of silicon to trichlorosilane impurities are removed and process can
be cycled to increase purity of the formed Si. The final material obtained is
the EGS. This is a polycrystalline form of Si, like MGS, but has much smaller
impurity levels, closer to what is desired in the final single crystal wafer. The
impurities in EGS are tabulated in 3. EGS is still polycrystalline and needs
to be converted into a single crystal Si ingot for producing the wafers.

3
 MM5017: Electronic materials, devices, and fabrication

Figure 2: Schematic of the process to purify MGS to obtain EGS. The process
involves conversion of silicon to trichlorosilane gas, which is purified, and then
reduced to obtain silicon. Adapted from Synthesis and purification of bulk
semiconductors - Barron and Smith

Figure 3: The Seimens deposition reactor where the purified Si is condensed.

This is the electronic grade Si, same purity level as Si wafers, but polycrys-
talline. Adapted from Synthesis and purification of bulk semiconductors -
Barron and Smith

4
 MM5017: Electronic materials, devices, and fabrication

Table 3: Impurities in EGS, after purification from MGS. Compared to table

2, the concentration levels of the metals have dropped to ppb levels.
Element Concentration (ppb)
As <0.001
Sb <0.001
B <0.1
C 100-1000
Cu 0.1
Fe 0.1-1
O 100-400
P <0.3

3 Single crystal Si manufacture

There are two main techniques for converting polycrystalline EGS into a
single crystal ingot, which are used to obtain the final wafers.

1. Czochralski technique (CZ) - this is the dominant technique for

manufacturing single crystals. It is especially suited for the large wafers
that are currently used in IC fabrication.

2. Float zone technique - this is mainly used for small sized wafers.
The float zone technique is used for producing specialty wafers that
have low oxygen impurity concentration.

3.1 Czochralski crystal growth technique

A schematic of this growth process is shown in figure 4. The various compo-
nents of the process are

1. Furnace

2. Crystal pulling mechanism

3. Ambient control - atmosphere

4. Control system

The starting material for the CZ process is electronic grade silicon, which
is melted in the furnace. To minimize contamination, the crucible is made
of SiO2 or SiNx . The drawback is that at the high temperature the inner
liner of the crucible also starts melting and has to replaced periodically. The

5
 MM5017: Electronic materials, devices, and fabrication

Figure 4: Schematic of the Czochralski growth technique. The polycrystalline

silicon is melted and a single crystal seed is then used to nucleate a single
crystal ingot. The seed crystal controls the orientation of the single crystal.
Adapted from Microchip fabrication - Peter van Zant.

6
 MM5017: Electronic materials, devices, and fabrication

Figure 5: Single crystal Si ingot. This is further processed to get the

wafers that are used for fabrication. Source http://www.chipsetc.com/silicon-
wafers.html

furnace is heated above 1500 ◦ C, since Si melting point is 1412 ◦ C. A small

seed crystal, with the desired orientation of the final wafer, is dipped in the
molten Si and slowly withdrawn by the crystal pulling mechanism. The seed
crystal is also rotated while it is being pulled, to ensure uniformity across
the surface. The furnace is rotated in the direction opposite to the crystal
puller. The molten Si sticks to the seed crystal and starts to solidify with
the same orientation as the seed crystal is withdrawn. Thus, a single crystal
ingot is obtained. To create doped crystals, the dopant material is added
to the Si melt so that it can be incorporated in the growing crystal. The
process control, i.e. speed of withdrawal and the speed of rotation of the
crystal puller, is crucial to obtain a good quality single crystal. There is a
feedback system that control this process. Similarly there is another ambient
gas control system. The final solidified Si obtained is the single crystal ingot.
A 450 mm wafer ingot can be as heavy as 800 kg. A picture of a such an
ingot is show in figure 5.

3.2 Float zone technique

The float zone technique is suited for small wafer production, with low oxygen
impurity. The schematic of the process is shown in figure 6. A polycrystalline
EGS rod is fused with the single crystal seed of desired orientation. This is
taken in an inert gas furnace and then melted along the length of the rod
by a traveling radio frequency (RF) coil. The RF coil starts from the fused
region, containing the seed, and travels up, as shown in figure 6. When
the molten region solidifies, it has the same orientation as the seed. The
furnace is filled with an inert gas like argon to reduce gaseous impurities.

7
 MM5017: Electronic materials, devices, and fabrication

Figure 6: Schematic of the float zone technique. The polycrystalline ingot is

fused with a seed crystal and locally melted by a traveling radio frequency
coil. As the ingot melts and resolidifes it has the same orientation as the
seed. Adapted from Microchip fabrication - Peter van Zant.

8
 MM5017: Electronic materials, devices, and fabrication

Also, since no crucible is needed it can be used to produce oxygen ’free’ Si

wafers. The difficulty is to extend this technique for large wafers, since the
process produces large number of dislocations. It is used for small specialty
applications requiring low oxygen content wafers.

4 Wafer manufacturing
After the single crystal is obtained, this needs to be further processed to
produce the wafers. For this, the wafers need to be shaped and cut. Usually,
industrial grade diamond tipped saws are used for this process. The shaping
operations consist of two steps

1. The seed and tang ends of the ingot are removed.

2. The surface of the ingot is ground to get an uniform diameter across

the length of the ingot.

Before further processing, the ingots are checked for resistivity and orienta-
tion. Resistivity is checked by a four point probe technique and can be used
to confirm the dopant concentration. This is usually done along the length of
the ingot to ensure uniformity. Orientation is measured by x-ray diffraction
at the ends (after grinding).
After the orientation and resistivity checks, one or more flats are ground
along the length of the ingot. There are two types of flats.

1. Primary flat - this is ground relative to a specific crystal direction.

This acts as a visual reference to the orientation of the wafer.

2. Secondary flat - this used for identification of the wafer, dopant type
and orientation.

The different flat locations are shown in figure 7. p-type (111) Si has only one
flat (primary flat) while all other wafer types have two flats (with different
orientations of the secondary flats). The primary flat is typically longer than
the secondary flat. Consider some typical specs of 150 mm wafers, shown
in table 4. Bow refers to the flatness of the wafer while ∆t refers to the
thickness variation across the wafer.
After making the flats, the individual wafers are sliced per the required thick-
ness. Inner diameter (ID) slicing is the most commonly used technique. The
cutting edge is located on the inside of the blade, as seen in figure 8. Larger
wafers are usually thicker, for mechanical integrity.
After cutting, the wafers are chemically etched to remove any damaged and

9
 MM5017: Electronic materials, devices, and fabrication

Figure 7: Flats for the different wafer types and orientations. All orientations
and doping types have a primary flat, while there are different secondary flats
for different types (a) p(111) (b) n(111) (c) p(100) and (d) n(100). Adapted
from Microchip fabrication - Peter van Zant.

Table 4: Specs of a typical 150 mm wafer

Specs Value
Diameter 150 ± 0.5 mm
Thickness 675 ± 25 µm
Orientation 100 ± 1◦
Bow 60 µm
∆t 50 µm
Primary flat 55-60 mm
Secondary flat 35-40 mm

10
 MM5017: Electronic materials, devices, and fabrication

Figure 8: Inner diameter wafer slicing, used for cutting the ingots into indi-
vidual wafers. The thickness is slightly higher than the final required thick-
ness to account for material loss due to polishing. Adapted from Microchip
fabrication - Peter van Zant.

contaminated regions. This is usually done in an acid bath with a mixture of

hydrofluoric acid, nitric acid, and acetic acid. After etching, the surfaces are
polished, first a rough abrasive polish, followed by a chemical mechanical pol-
ishing (CMP) procedure. In CMP, a slurry of fine SiO2 particles suspended
in aqueous NaOH solution is used. The pad is usually a polyester material.
Polishing happens both due to mechanical abrasion and also reaction of the
silicon with the NaOH solution.
Wafers are typically single side or double side polished. Large wafers are
usually double side polished so that the backside of the wafers can be used
for patterning. But wafer handling for double side polished wafers should
be carefully controlled to avoid scratches on the backside. Typical 300 mm
wafers used for IC manufacture are handled by robot arms and these are
made of ceramics to minimize scratches. Smaller wafers (3” and 4” wafers)
used in labs are usually single side polished. After polishing, the wafers are
subjected to a final inspection before they are packed and shipped to the fab.

11