ch. 4 basics of device fabrication reference: s. m. sze...
TRANSCRIPT
KOREA UNIVERSITY
Photonics Laboratory
Ch. 4 Basics of device fabrication
Reference: S. M. Sze, Semiconductor Devices, ISBN 0-471-87424-8
: We will understand the common techniques for growing single crystals. (a) The starting
materials (e.g., silicon dioxide for a silicon wafer) are chemically processed to form a high-purity
polycrystalline semiconductor from which single crystals are grown. (b) The single-crystal ingots
are shaped to define the diameter of the material and sawed into wafers. (c) These wafers are
etched and polished to provide smooth, specular surfaces on which devices will be made.
A technology closely related to crystal growth involves the growth of single-crystal
semiconductor layers upon a single-crystal semiconductor substrate. This is called epitaxy.
(1) Crystal growth from the melt
There are basically two techniques for crystal growth from the melt: the Czochralski technique
and the Bridgman technique. A substantial percentage (~90%) of the silicon crystals for the
semiconductor industry are prepare by the Czochralski technique; virtually all the silicon used for
fabricating integrated circuits is prepared by this technique. Most gallium arsenide, on the other
hand, is grown by the Bridgman technique. However, the Czochralski technique is becoming more
popular for the growth of large-diameter gallium arsenide.
1
KOREA UNIVERSITY
Photonics Laboratory 2
KOREA UNIVERSITY
Photonics Laboratory
(2) Starting materials
The starting material for silicon is a relatively pure form of sand (SiO2) called quartzite. This is
placed in a furnace with various forms of carbon (coal, coke, and wood chips). This process
produces metallurgical-grade silicon with a purity of about 98%.
).()()()()( 2 gasCOgasSiOsolidSisolidSiOsolidSiC
Next, the silicon is pulverized and treated with hydrogen chloride (HCl) to form
trichlorosilane (SiHCl3):
).()()(3)( 23
300
gasHgasSiHClsolidHClsolidSiCo
The trichlorosilane is a liquid at room temperature (boiling point 32oC). Fractional
distillation of the liquid removes the unwanted impurities. The purified SiHCl3 is then
used in a hydrogen reduction reaction to prepare the electronic-grade silicon (EGS):
).(3)()()( 23 gasHClsolidSigasHgasSiHCl
This reaction takes place in a reactor containing a resistance-heated silicon rod, which
serves as the nucleation point for the deposition of silicon. The EGS, a polycrystalline
material of high purity, is the raw material used to prepare device quality, single-crystal
silicon. Pure EGS generally has impurity concentrations in the parts-per-billion
range.
3
KOREA UNIVERSITY
Photonics Laboratory
(3) The Czochralski technique
The Czochralski technique for silicon crystal growth uses an apparatus called a puller. The puller
has three main components: (1) a furnace, which includes a fused-dilica (SiO2) crucible, a
graphite susceptor, a rotation mechanism, a heating element, and a power supply; (2) a crystal-
pulling mechanism, which includes a seed holder and a rotation mechanism; and (3) an ambient
control, which includes a gas source, a flow control, and a exhaust system.
In the crystal-growing process, polycrystalline silicon is placed in the crucible and the furnace is
heated above the melting temperature of silicon. A suitably oriented seed crystal (e.e., <111>) is
suspended over the crucible in a seed holder. The seed is inserted into the melt. Part of it melts,
but the tip of the remaining seed crystal still touches the liquid surface. It is then slowly
withdrawn. Progressive freezing at the solid-liquid interface yields a large single crystal. A typical
pull rate is a few millimeters per minute.
4
KOREA UNIVERSITY
Photonics Laboratory 5
KOREA UNIVERSITY
Photonics Laboratory
(4) Unitary diagrams
These diagrams show the phase change in a single element as a function of temperature and
pressure.
The common point, referred to as a triple point, is invariant for the system and defines the
temperature and pressure at which solid, liquid, and gaseous phases are all in equilibrium with one
another.
6
KOREA UNIVERSITY
Photonics Laboratory
(5) Binary diagrams
These phase diagrams show the relationship between two components as a funciton of
temperature. The second variable, pressure, is usually set at 1 atm. In this way a relatively
complex three-dimensional representation is avoided.
7
KOREA UNIVERSITY
Photonics Laboratory
(6) The lever rule
At any temperature, the equilibrium composition of the two single phases that make up a two-
phase region may be determined as follows. Consider a melt of initial composition CM (the
percentage weight of B in the melt). Let this melt be cooled from some temperature T1 to a
tempeerature T2, corresponding to a point in the two-phase region:
WL=weight of liquid at this temperature
Ws=weight of solid (in the β phase, for this example)
CL, Cs=composition of the liquid and solid, respectively (percentage amount of B by weight)
Then WLCL and WsCs are the weights of B in the liquid and solid, respectively. But total weight of
B is (WL+Ws)CM. Hence by the conservation of matter
.rulelever theasknown is sregion.Thi phase- twothe
of boundaries the toordinate C thefrom measured lines two theoflength theare s and l where M
s
l
CC
CC
W
W
Ms
LM
L
s
8
KOREA UNIVERSITY
Photonics Laboratory
(7) The phase rule
The correct interpretation of phase diagrams is greatly helped by knowledge of the phase rule.
This rule, which is based on thermodynamic consideration, states that for any system in thermal
equilibrium, the sum of the number of phases P and the number of degrees of freedom F is related
to the number of components C by
P + F = C + 2.
Here the degrees of freedom are the number of variables that can be independently changed while
still preserving a specific phase.
Ex) For a single-component diagram of the type shown in Fig. 2.1
P + F = 3
(a) For water in its liquid phase (P=1), F=2; we have 2 degrees of freedom, i.e., both pressure and
temperature can be independently changed, and still maintain water in this liquid phase.
(b) Along 0B, however, P=2 (liquid and vapor) so that we have 1 degree of freedom (F=1). Now
either pressure or temperature (but not both) can be independently varied while this two-phase
coexistence is still preserved.
(c) At 0, P=3, and F=0. i.e., there is a unique temperature-pressure combination where water
coexists in all three phases.
For a binary phase diagram, F + P = C + 2 = 4.
9
KOREA UNIVERSITY
Photonics Laboratory
(8) Distribution of dopant
In crystal growth, a known amount of dopant is added to the melt to obtain the desired doping
concentration in the grown crystal. For silicon, boron and phosphorus are the most common
dopants for p- and n-type materials, respectively.
As a crystal is pulled from the melt, the doping concentration incorporated into the crystal (solid)
is usually different from the doping concentration of the melt (liquid) at the interface. The ratio of
these two concentrations is defined as the equilibrium segregation coefficient k0:
l
s
c
ck 0
where Cs and Cl are respectively the equilibrium concentrations of the dopant in the solid and
liquid near the interface.
Consider a crystal being grown from a melt having an initial weight M0 with an initial doping
concentration C0 in the melt (i.e., the weight of the dopant per 1 gram melt). A given point of
growth when a crystal of weight M has been grown, the amount of dopant remaining in the melt
(by weight) is S.
10
KOREA UNIVERSITY
Photonics Laboratory
1
0
00
00
0
0
0
0
000
0
00
000
0
0
0
00
00
00
0
0
0
0
0
0
0
0
00
)1(
)1(/1
/1)(
)1(
)1ln(lnln
)(
k
s
ks
sl
k
MS
MC
l
s
l
s
M
MCkC
M
M
C
MM
k
C
C
MM
k
C
MC
MMC
MC
S
M
M
MC
S
M
Mk
M
MMk
MC
S
MM
dMk
S
dS
MM
dMk
MMC
dMC
S
dS
MM
SC
dMCdS where Cs and Cl are respectively the equilibrium
concentrations of the dopant in the solid and
liquid near the interface.
Consider a crystal being grown from a melt
having an initial weight M0 with an initial
doping concentration C0 in the melt (i.e., the
weight of the dopant per 1 gram melt). A given
point of growth when a crystal of weight M has
been grown, the amount of dopant remaining in
the melt (by weight) is S.
11
KOREA UNIVERSITY
Photonics Laboratory 12
KOREA UNIVERSITY
Photonics Laboratory
Problem 1) A silicon ingot, which should contain 1016 boron atoms/cm3, is to be grown by the
Czochralski technique. What concentration of boron atoms should be in the melt to give the
required concentration in the ingot? If the initial load of silicon in the crucible is 60 kg, how many
grams boron (atomic weight 10.8) should be added?
(9) Wafer shaping and material characterization
After a crystal is grown, (a) the first shaping operation is to remove the seed and the other end of
the ingot, which is last to solidify. (b) The next operation is to grind the surface so that the
diameter of the material is defined. (c) After that, one or more flat regions are ground along the
length of the ingot. These regions, or flats, mark the specific crystal orientation of the ingot and
the conductivity type of the material. The largest flat, the primary flat, allows a mechanical locator
in automatic processing equipment to position the wafer and to orient devices relative to the
crystal in a specific manner. Other smaller flats, called secondary flats, are ground to identify the
orientation and conductivity type of the crystal.
13
KOREA UNIVERSITY
Photonics Laboratory 14
KOREA UNIVERSITY
Photonics Laboratory 15
KOREA UNIVERSITY
Photonics Laboratory
1) Once these operations have been done, the ingot is ready to be sliced by diamond saw into
wafers. Slicing determines four wafer parameters: surface orientation, thickness, taper (i.e., wafer
thickness variations from one end to another), and bow (i.e., surface curvature of the wafer,
measured from the center of the wafer to its edge).
2) After slicing, both sides of the wafer are lapped using a mixture of Al2O3 and glycerine to
produce a typical flatness uniformity within 2μm. The lapping operation usually leaves the surface
and edges of the wafer damaged and contaminated. The damaged and contaminated regions can
be removed by chemical etching. The final step of wafer shaping is polishing. Its purpose is to
provide a smooth, specular surface where device features can be defined by lithographic
processes.
3) The oxygen and carbon concentrations are substantially higher in Czochralski crystals than in
float zone crystals due to the dissolution of the silica crucible (for oxygen) and transport to the
melt from the graphite susceptor (carbon) during crystal growth. Especially, the precipitates of
oxygen due to the solubility effect, can be used for gettering. Gettering is a general term meaning
a process that removes harmful impurities or defects from the region in a wafer where devices are
fabricated.
16
KOREA UNIVERSITY
Photonics Laboratory
Table 2 Table 3
17
KOREA UNIVERSITY
Photonics Laboratory 18
KOREA UNIVERSITY
Photonics Laboratory
(10) Vapor-phase epitaxy
Epitaxial processes are differentiated from the melt growth processes in that the epitaxial layer
can be grown at a temperature substantially below the melting point (typically 30 to 50% lower).
Four silicon sources have been used for vapor phase epitaxial growth. They are silicon
tetrachloride (SiCl4), dichlorosiliane (SiH2Cl2), trichlorosilane (SiHCl3), and silane (SiH4).
Silicon tetrachloride has been the most studied and has the widest industrial use. The typical
reaction temperature is 1200oC.
)(2)()(
)(4)()(2)(
24
24
gasSiClsolidSigasSiCl
gasHClsolidSigasHgasSiCl
19
KOREA UNIVERSITY
Photonics Laboratory 20
KOREA UNIVERSITY
Photonics Laboratory
The dopant is introduced at the same time as the silicon tetrachloride during epitaxial growth.
Gaseous diborane (B2H6) is used as the p-type dopant, while phosphine (PH3) and arsine
(AsH3) are used as n-type dopants. Gas mixtures are ordinarily used with hydrogen as the diluent
to allow reasonable control of flow rates for the desired doping concentration. The dopant
chemistry shows dopant being adsorbed on the surface, decomposing, and being incorporated into
the growing layer.
21
KOREA UNIVERSITY
Photonics Laboratory
(11) Molecular-beam epitaxy (MBE)
MBE is an epitaxial process involving the reaction of one or more thermal beams of atoms or
molecules with a crystalline surface under ultrahigh vacuum conditions(~10-10Torr). MBE can
achieve precise control in both chemical compositions and doping profiles. Single crystal
multilayer structures with dimensions of the order of atomic layers can be made using MBE.
.T asely approximat
re temperatuabsolute with increase and /s1cm oforder on the are re temperaturoomat gases of iesdiffusivit The
.weightmolecular
:ionconcentrat its andweight molecular its ofproduct by thegiven is gas a of density The Torr.in is P where
cmmolecules/ 1065.9
:e)unit volumper molecules ofnumber (then ion concentratmolecular thecalculate tolaw ideal theusecan
welowered, is pressure theas gas ideal thelike more and more behave gases real Since J/K). 10(1.38constant Boltzmann
theisk and mole),molecules/10(6.02constant Avogadro theN K,in re temperatuabsolute theis T
K),-/molecm-atm 82or K,-cal/mol (1.98constant gas theis R gas, of mole one of volume theis V pressure, theis P where
thatstates law gas ideal he
2
2
d
318
23-
23
A
3
kT
P
T
P
kT
P
V
Nn
kTNRTPV
T
d
A
A
22
KOREA UNIVERSITY
Photonics Laboratory
32).(Moxygen for s-cmmolecules/ 103.6 is ratet impingemen the
pressure,Torr 10 and300K at Therefore, weight.molecular theis M andTorr in pressure theis P where
1051.3)2(22
exp2
bygiven is ratet impingemenmolecular The
.2
exp2
1
unit time.per areaunit aon
impinge moleculesmany how is, that rate,t impingemenmolecular theisy technologfor vacuumparameter important
.22
v
is speed average The dv. vand between v
speed a having molecules be will there volume,in the molecules are thereIf molecule. a of mass theis m where
2exp
2
41
v,speedgiven afor that states which law,n distributiBoltzmann -Maxwell by the described is
s velocitieofon distributi The dependent. re temperatuare ocities their velandmotion constant in are molecules gas The
214
6
222/1
0
22
2/1
0
22
2/1
0
0av
22
2/3
-
MT
PmkTP
m
kTndv
kT
mvv
kT
mnvdnv
kT
mvv
kT
mf
dv
dn
n
An
m
kT
dvf
dvvf
dnn
kT
mvv
kT
mf
dv
dn
n
xx
xxxx
xxv
x
x
v
v
v
x
23
KOREA UNIVERSITY
Photonics Laboratory
Torr 10at hr 6
Torr10at s 43.2
Torr 1at 21043.2
2Nt
:ratet impingemen thefrom obtained is sticking) 100% (assumingmonolayer a form torequired timeThe
.cm108.6 isoxygen for areaunit per mlecules of
Nnumber thepacking, close Assuming .A3.64 isoxygen ofdiameter molecular effictive The
Torr.10 and ,10 1, of pressuresat oxygen ofmonolayer a form torequired time theFind Ex)
10-
6-
6
s
2-14
s
o
-10-6
st
P
mkTN s
24
KOREA UNIVERSITY
Photonics Laboratory
Another important parameter is the mean free path. During their motion, the molecules suffer
collisions among themselves. The average distance traversed by all the molecules between successive
collisions with each other is defined as the mean free path. A molecule having a diameter d and a
velocity v will move a distance vδt in the time δt. The molecule suffers a collision with another
molecule if its center is anywhere within the distance d of the center of another molecule. Therefore, it
sweeps out (without collision) a cylinder of diameter 2d. The volume of the cylinder is
e. temeraturroomat )A3.7 ofdiameter molecular t (equivalen moleculesair for cm ) (
105
2
gives derivation rigorous more
1
thenis path freemean theand
1
,collisionsbetween timeaverage theist If occurred. hascollision a
thus,molecule;other one average on thecontain must it ,1/ toequal is V volumeWhen the
.cm1/n average on the is molecule one with associated volume the,cmmolecules/ are thereSince
.)2(4
o3
2
22
2
33
2
TorrinPPd
kT
A
dT
Pn
d
kT
ndv
vdn
n
n
tvdV
25
KOREA UNIVERSITY
Photonics Laboratory 26
KOREA UNIVERSITY
Photonics Laboratory
Problem 2. Assume an effusion oven geometry of area A=5 cm2 and a distance L between the top of
the oven and the gallium arsenide substrate of 10 cm. Calculate the MBE growth rate for the effusion
oven filled with gallium arsenide at 900oC.
26
KOREA UNIVERSITY
Photonics Laboratory
12) Thermal oxidation
Semiconductors can be oxidized by various methods. These include thermal oxidation,
electrochemical anodization, and plasma reaction. Among these methods thermal oxidation is by far
the most important for silicon devices. It is the key process in modern silicon integrated-circuit
technology. The silicon-silicon dioxide interface moves into the silicon during the oxidation process.
).(2)()(2)(
)()(
:aporor water voxygen in silicon ofoxidation thermal thedescribe reactions chemical following The
222
22
gasHsolidSiOgasOHsolidSi
solidSiOgasOSi(solid)
Problem 3) If a silicon oxide layer of thickness x is grown from the thermal oxidation, what is the
thickness of silicon being consumed?
KOREA UNIVERSITY
Photonics Laboratory
KOREA UNIVERSITY
Photonics Laboratory
KOREA UNIVERSITY
Photonics Laboratory
13) Metallization
Metallization refers to the formation of metal films used for interconnections, ohmic contacts, and
rectifying metal-semiconductor contacts. Metal films can be formed by various methods, the most
important being physical vapor deposition and chemical vapor deposition.
We shall now consider the thickness of the film deposited from an evaporation source.
KOREA UNIVERSITY
Photonics Laboratory
KOREA UNIVERSITY
Photonics Laboratory
.cos
stream, vapor the
ofdirection the to anglean at inclined is normal whosesurface aon dS2 area small aat arrives material theIf
. cos
:law cosine by thegiven is unit timeper
surface the tonormal with the anglean formingdirection ain d angle solid a through passing material ofamount The
(g/s). ofm rate aat
side one from evaporated is material which thefrom dS1 area of source plane a fromn evaporatio theis case related
.4
thenis unit timeper direction any in d angle solid a through passing material ofamount The source.point a called
is source gevaporatinan Such (g/s). m of rate aat directions allin materialn evaporatio dS1 sphere small aConsider
2
2
r
dSd
dm
dm
A
dm
dm
KOREA UNIVERSITY
Photonics Laboratory
source. plane for the ](L/H)[1
1
)(
and sourcepoint for the (L/H)][1
1
4)(4
surface, receiving plane-parallel a and source theoft arrangemen practical a devices, discretefor ,Especially
source. plane for the coscos
and sourcepoint for the cos
4
bygiven then is dS area the toingcorrespond locationsat film theof thicknessThe
.
.dS bemust dSon deposited material of volumethe
(cm/s), is unit timeper deposited film theof thickness theand )(g/cm density a has material theAssuming
source. plane for the coscos
and sourcepoint for the cos
4
222222
2
3/222/322
2
2
2
2
22
3
d
22
22
H
m
LH
mHl
H
m
LH
mHl
r
ml
r
ml
ldSdm
l
l
dSr
mdm
dSr
mdm
dd
dd
d
d
d
KOREA UNIVERSITY
Photonics Laboratory
Problem 4) A silicon wafer is placed at a perpendicular distance of 30cm from a plane source. If the
total deposited mass is 1 g and the density is 2.7 g/cm3, what is the film thickness at L=0? If the
variation in film thickness must be less than 10%, how large can the wafer be?
Aluminum Metallization
Aluminum and its alloy are used extensively for metallization in integrated circuits. Because
aluminum and its alloys have low resistivities (i.e., 2.7μΩ-cm for Al and up to 3.5 μΩ-cm its alloys),
these metals satisfy the requirements of low resistance. However, the use of aluminum in integrated
circuits at shallow junctions often creates problems such as spiking and eletromigration. Hence, during
annealing silicon will dissolve into the aluminum. The amount of silicon dissolved will depend not
only on the solubility at the annealing temperature but also on the volume of aluminum to be saturated
with silicon.
KOREA UNIVERSITY
Photonics Laboratory
KOREA UNIVERSITY
Photonics Laboratory
.2
is consumed be uldsilicon wo which depth to theA, areacontact over theuniformly
place n takesconsumptio theIf re. temperatuannealing at the aluminumin silicon of solubility theis S where
,)(2
is consumedsilicon of volume thesilicon, and aluminum of densities theare and Assuming SilA
Si
Al
Si
Al
SA
HZDtb
SHZDtVol
Problem 5) For T=500oC, t=30min, ZL=16μm2, Z=5μm, and H=1μm. Find the depth b, assuming
uniform dissolution.