Chapter 2

•
chapter2
Semiconductor
Physics and IC
Technology
This scanning electron micrograph (SEM) photograph shows the com-
plexity of modern integrated circuit interconnect technology. The inter-
/eve/ dielectrics have been etched away lo show the metallization
patterns. (Courtesy of International Business Machines Corporation.
Unauthorized use not permitted.)
Microelectronic engineers need a working knowledge of the physi-

cal processes underlying the behavior of electronic devices. This
chapter provides a short introduction to semiconductor physics,
with the goal of developing the basic concepts and terminology
needed to model electronic devices described in Chapters 3, 4, 6,
and 7. In addition a basic familiarity with IC fabrication is essential
to understand the device structures in later chapters. This chapter
includes a short overview of the essential features of this powerful
technology.
Chapter Objectives
+ The bond model for silicon-the electron and hole as mobile

charges.
+ The control of electron and hole concentrations through doping.
+ Drift and diffusion of electrons and holes.
+ IC processes and the connection between mask layout and
device cross section.
+ Sheet resistance and the integrated-circuit resistor.
:,...2.1 PURE SEMICONDUCTORS
To begin our discussion of semiconductor physics, we consider the

dominant semiconductor used in modern microelectronics-crys-
talline silicon. Silicon is in Group IV of the periodic table, with
atomic number 14. Recalling basic atomic chemistry, its 14 elec-
trons occupy the first three energy levels. The first three orbitals, ls,
2s, and 2p, are filled by 10 electrons, leaving four electrons to
occupy the 3s and 3p orbitals. In order to minimize the overall
•
22 Chapter2 • SemiconductorPhysicsand IC Technology
(ls)2(filled) energy, the 3s and 3p orbitals hybridize to form four, tetrahedral 3sp orbitals. The
electronic configuration of silicon is
2 2 6 4
(ls) (2s) (2p) (3sp)
(2s)2(filled)
Each of the four outer hybrid orbitals have one electron and are capable of form-
ing a bond with a neighboring atom. These four electrons that participate in bonding
are called valence electrons. The inset on the left shows the electronic configuration
(2p)6 (filled) of silicon.
The crystalline form of silicon has the diamond crystal structure. Each silicon ion
in the crystal has four nearest neighbors that are aligned with the tetrahedral 3sp
hybrid orbitals. Each ion contributes an electron to fill the hybrid orbital. Figure 2.1
shows the unit cell of silicon, with the tetrahedral bonds from one atom highlighted.
~
The edge of the unit cell is a0 = 5.43 A.
How many atoms should be assigned to the unit cell? The eight corner atoms are
(3sp)' shared among the eight cells touching the corner and therefore, contribute one atom
to the unit cell. The six atoms on the faces of the cube are shared with adjacent cells
4 tetrahedral hybrid and thus, contribute three atoms. Finally, there are four atoms that lie entirely within
orbitals, half-filled with 4 the cube, for a total of eight atoms in the unit cell.
valence electrons The atomic density of crystalline silicon, in atoms per cubic centimeter, is
8 22 -3
3 = 5.00 x 10 cm (2.1)
-8
(5.43 x 10 cm)
where nceuis the number of unit cells per cm3 . Since each atom has four tetrahedral
bonds and contributes one electron to each bond, the concentration of valence elec-
trons is 4 Ns; = 2 x 1023 cm- 3 •
5.43 A
3sp tetrahedral bond
>-Figure 2.1 Unit cell of crystalline silicon (After R. S. Muller and T.I. Karn ins,
Device Electronics for Integrated Circuits, znd ed., Wiley, 1986.)
2. I • Pure Semiconductors 23
2.1.1 Electrons, Holes, and the Bond Model

We now investigate the electronic properties of the bonded network of silicon atoms
in Fig. 2.1, which we refer to as the silicon lattice. In order to sketch the silicon lattice
conveniently, we use the two-dimensional collapsed view shown in Fig. 2.2. The por-
tion of the lattice shown in Fig. 2.2 is in the interior, or bulk, of the silicon crystal away
from any surfaces. Each valence electron in a bond is symbolized by a connecting
line, which means that two lines are used to represent each tetrahedral bond between
silicon atoms. Note that this convention differs from the one used in chemistry, where
a single line represents both electrons (in a single bond).
We refer to this abstract picture of the silicon crystal as the bond model. The sil-
icon ions (nucleus and 10 inner electrons) each have a charge of (+4 q ), where
q = 1.6 x 10-19 C is the magnitude of the charge of an electron. The four valence elec-
trons per ion in the surrounding four bonds contribute a charge of (-4 q ), making the
crystal electrically neutral. In Fig. 2.2, all of the bonds are shown to be complete,
which is true only for a perfect crystal at a temperature of absolute zero (0 degrees
Kelvin).
An intrinsic semiconductor is one in which there are no impurities. A perfect
crystal of intrinsic silicon at absolute zero has all of its valence electrons tied up in
bonds since this is the lowest energy configuration. Since there are no "loose" or
"free" electrons to carry current, intrinsic silicon is an insulator at absolute zero.
At temperatures above absolute zero, lattice vibrations occur that can impart
enough energy to the valence electrons to break a very small percentage of the bonds.
Each broken bond results in a mobile electron that can move around the crystal. This
process is known as thermal generation. Note that the electron is still confined to the
4 valence electrons (-4 q),

contributed by each ion
;--~--~--~ ~--~--~--~--~
I II II II II II II I
,--0-0--0-•-o-o-o-c
i ii i
1
ii II I~ 4
1~iliconion:f+ q) II
,-0=0~0-0-0-0-0-c
I II II II II II II II I
)=O=O=O=O=O=O=O=C
~--~--~--~--":~--~--~--~
I II II II
~ border of bulk silicon region

II II II II
two electrons in bond

I
~ Figure 2.2 Collapsed two-dimensional silicon lattice at absolute zero, with all
bonds complete. This portion of the crystal is in the interior, or bulk, away from
surfaces. It is electrically neutral since the four valence electrons balance the net
charge on the silicon ion.
crystal since much more energy is required for it to escape into the surrounding
ambient.
Figure 2.3 shows the silicon lattice after one bond has been broken due to ther-
mal generation. The mobile electron has a negative charge -q = -1.6 x 10-19 C and
leaves behind an incomplete bond that takes on a charge of +q. At room tempera-
ture, the electron moves quite rapidly and undergoes frequent collisions as suggested
by the trajectory sketched in Fig. 2.3.
Since electrons have particle-like behavior in a vacuum, it seems plausible that
the mobile electron resulting from the broken bond is capable of moving through the
crystal lattice. For example, a cathode-ray tube used in a computer monitor works by
steering electron trajectories with electric fields. What is surprising is that the positive
charge arising from the broken bond is also mobile with similar properties to the elec-
tron, except that it has an equal and opposite charge. Figure 2.3 depicts that the pos-
itive charge is also free to move around the lattice. This mobile positive charge is
called a hole.
2.1.2 Limitations of the Bond Model
The bond model is useful in visualizing the creation of electron-hole pairs. However,
it has some shortcomings. Electrons and holes in crystals have both particle-like and
wave-like behavior that can only be understood from the perspective of quantum
mechanics. The bond model necessarily depicts both the electron and the hole as
localized in space, which is not correct.
Fortunately, it is not necessary to take a solid-state physics course prior to study-
ing semiconductor devices and circuits. This text considers that electrons and holes
can effectively be considered as classical particles that obey Newton's Laws and have
unit negative and positive charge. Quantum mechanics has a negligible effect on the
incomplete bond (mobile hole)
> Figure 2.3 Electron and hole resulting from breaking one bond in intrinsic
silicon.
2.2 • Generation,Recombination,and Thermal Equilibrium 25
accuracy of models for microelectronic devices that have dimensions of 0.1 µm or
larger. 1
GENERATION, RECOMBINATION,
> 2.2 AND THERMAL EQUILIBRIUM
We have seen that mobile electrons and holes can be created in pure, also known as
intrinsic, silicon by the breaking of bonds due to thermal generation. The thermal
generation rate G 1h(T), with units of cm-3s-1,increases exponentially with tempera-
ture. Another type of generation process is optical generation, with rate Cw Optical
generation occurs when a bond absorbs a photon with sufficient energy to break it.
The minimum energy for bond-breaking in silicon is around 1.1 eV, corresponding to
a wavelength of around 1 µm (in the near-infrared region of the optical spectrum.) In
contrast to thermal generation, the optical generation rate is essentially independent
of temperature. The total generation rate G is the sum of the thermal and optical rates
(2.2)
If mobile electrons and holes are continuously generated, a balancing process is nec-
essary that continuously removes them. Recombination is the elimination of an elec-
tron and hole from the crystal by completing one of the broken bonds.
Recombination processes are classified according to what happens to the energy
released when the bond forms.
Thermal recombination-the energy goes to vibrations in the crystal lattice. This
is equivalent to saying the silicon crystal is heated.
Optical recombination-the energy goes to photon emission. Optical recombi-
nation is of negligible importance in silicon. In other semiconductors such as gal-
lium arsenide, optical recombination is a very significant component of the total
recombination rate. In these materials, light emission due to optical recombina-
tion is the basic principle behind light-emitting diodes and laser diodes.
For recombination to occur, both an electron and a hole are necessary. The
recombination rate will be small if there are few electrons despite an abundance of
holes or vice versa. Therefore, it is reasonable that the recombination rate is propor-
tional to the product of the electron and hole concentrations
R=k(n·p) (2.3)
where n is the electron concentration (cm- 3 ),p is the hole concentration (cm- 3), and
k is the recombination rate constant (cm3s-1). In contrast, the generation rate G is
independent of the electron and hole concentrations since the number of broken
bonds is a tiny fraction of the total number of bonds in the crystal.
If the environment of the silicon crystal (optical radiation flux, electric field
intensity, temperature, etc.) is constant, then the electron and hole concentrations
must eventually reach stable values that are set by the balance between the recombi-
1. Current research is exploring nanoelectronic devices with critical dimensions on the order of 10 nm or
less. Quantum effects arc crucial to understanding their basic operation.
nation and generation rates. We define this situation as steady-state. The total recom-
bination rate and total generation rates must be equal for steady-state conditions
R = k (n · p) = G = Gh 1 (T) + G0 P (2.4)
Given the particular environmental conditions that determine G 1h(T) and G0 p,

the product of the electron and hole concentrations is
Gth (T) + Gop

n. p = k (2.5)
An idealized, special case of the steady-state condition called thermal equilib-

rium is of great importance in semiconductor physics. Thermal equilibrium is defined
as the prolonged absence of external energy sources. The silicon sample is completely
isolated from optical or electrical sources (e.g., power supplies and light sources) by
placing it in a "black box" at a constant temperature. Some optical generation occurs
in addition to thermal generation, due to black-body radiation at the sample temper-
ature. For the case of thermal equilibrium, Eq. (2.5) reduces to
a;h(T) + G :P(T) (2.6)

no' Po = k
0
where all quantities are subscripted or superscripted with "o" to denote thermal equi-
librium. The right-hand side of this equation is a function of temperature only (the
single variable in thermal equilibrium). Defining this function as n:(T), we can re-
express Eq. (2.6) as the mass-action law
(2.7)
This important result relates the product of the thermal equilibrium electron and
hole concentrations to a constant that depends only on temperature. The unusual
name for this equation is borrowed from chemistry, since Eq. (2.7) is analogous to an
equilibrium equation for a chemical reaction. For example, the equilibrium between
hydrogen ions (concentration [H +]) and hydroxyl ions (concentration [0 H -]) react-
ing to form water is expressed as
(2.8)
where Ke/T) is the equilibrium constant.

In pure silicon in thermal equilibrium, the electron and hole concentrations are
equal since the electrons and holes are created in pairs when the bonds are broken.
Therefore, it follows from substitution in Eq. (2.7)
(2.9)
and consequently, the electron and hole concentrations in pure silicon are both equal
to the intrinsic concentration n;(T) in thermal equilibrium. Before proceeding fur-
2.3 • Dopin1; 27
ther, it is worthwhile to consider the order of magnitude of the electron and hole con-
centrations in pure silicon at room temperature
I n; = 1010 cm-3at 300 K (2.10) intrinsic concentration
Given that the number of bonds is 4 x Ns; = 2 x 1023 cm- 3 , at room temperature only
an extremely small fraction (1 in 20 trillion) of the bonds are broken. As temperature
increases, the intrinsic concentration increases exponentially-approximately dou-
bling for every 10 °C rise over room temperature .
... 2.3 DOPING

Silicon would be of little use if the hole and electron concentrations were both equal
to 1010 cm- 3 . Fortunately, the hole and electron concentrations can be controlled by
adding minute quantities of specific elements to the silicon crystal. There are a vari-
ety of ways to incorporate these impurities into the silicon lattice using a process
called doping. We will first consider the case of donors, which are elements in Group
V of the periodic table.
2.3. 1 Donors
The bond model is useful for visualizing how donor atoms control the concentration
of electrons in silicon. Typical donors for silicon are phosphorus (P), arsenic (As),
and antimony (Sb), all of which are in Group V of the periodic table. Figure 2.4 shows
the bond model of silicon with the addition of an arsenic atom on a substitutional lat-
tice site. Since arsenic has five valence electrons, it is able to form bonds with all four
neighboring silicon atoms. The fifth valence electron is left after the bonds are
y~-~~-~~-~~-~~-~--~~-~~-~
I II II II II I
)--®-®-®-®-- @-(
I II II II 11 II II .e II I
)-@-@-@- As+ -@-E>\@-(
I 11 11 11 11 11 11 ~~bile elec~ron
)=@=®=®=® -®=®=@=(
I II II II II II II II I
~~-aD\~-aD\~-aD\~-aD\~- -~aD\~-aD\~-a
border of bulk silicon region / immobile ionized donor
~ Figure 2.4 Bond model with arsenic incorporated into a substitutional site in a
bulk region of the silicon crystal. Due to thermal excitation, arsenic loses (donates)
its fifth valence electron as a mobile electron resulting in an immobile positive
charge.
28 Chapter2 • SemiconductorPhysicsand TCTechnology
formed and is very loosely bound to the arsenic atom. For temperatures T? 100
K = -173° C, there is enough thermal energy for this fifth electron to leave and move
through the lattice, leaving behind an immobile, positively charged arsenic ion.
The charges in the bulk silicon region in Fig. 2.4 are the key to quantifying the
electron and hole concentrations for arsenic-doped silicon. The region was electri-
cally neutral before the arsenic was incorporated into the lattice. After the arsenic is
added and ionizes, which occurs for T? 100 K = -173° C, the region is still neutral
since the negative electron charge is balanced by the positive charge on the arsenic
ion. Adding up all the charge densities (units: CI cm 3) with their proper signs, we
require that the net charge density p is zero:
(2.11)
The concentration of dopants is extremely low and their presence in the lattice
does not significantly alter the mass-action law in thermal equilibrium. Solving Eq.
(2.7) for the hole concentration and substituting into Eq. (2.11), we find that
2
n;
-+N-n
n a o
= 0 (2.12)
0
By multiplying both sides by the electron concentration n 0 and rearranging, the result
is a quadratic equation
(2.13)
Using the quadratic formula, we find that the electron concentration in silicon
doped with donors is
(2.14)
where we have taken the positive root since the electron concentration must be pos-
itive. The hole concentration p O can be found from the electron concentration using
the mass-action law in Eq. (2.7).
In practice, the donor concentration is much greater than the intrinsic concentra-
tion Nd» n;. The square root in Eq. (2.14) is approximately one and the electron con-
centration n 0 is
(2.15)
Silicon doped with Nd » n; is called extrinsic. The electron concentration n 0 in

extrinsic silicon is determined by the donor concentration and is not a function of
temperature-dependent thermal generation. The result in Eq. (2.15) can be easily
remembered since it corresponds to "one electron per donor," which can be seen
from the bond model in Fig. 2.4. The relatively large donor concentration means the
thermally generated electrons can be neglected, as quantified by Eq. (2.14).
The thermal equilibrium hole concentration in silicon doped with donors can be
found from the mass-action law
2.3 • Doping 29
n;
2 2
n; n-type ·hore
1020 -3
Po= no - Po= Nern (T= 300K) (2.16) concentration
Nd d
For example, for a donor concentration of Nd= 1014 cm- 3 , the hole concentration
at room temperature is only p 0 =10 6 cm- 3-a factor of 100 million lower than the elec-
tron concentration! By adding donor impurities to the silicon sample, the mass-action
law in Eq. (2.7) depresses the hole concentration. Few holes remain in equilibrium
since recombination with the large number of electrons removes most of the them.
Silicon doped with donors is referred to as n-type, in which electrons are the majority
carriers (by far the most numerous) and holes are the minority carriers.
The lower limit of useful donor concentration is set by the requirement that Nd»
n; in Eq. (2.14), so Nd . =10 12 cm- • On the opposite extreme, there is a maximum
3
1
amount of substitutio~ ~l impurities that the silicon lattice can absorb without form-
ing a new material phase with a different crystal structure and electronic properties.
This maximum varies among the donor elements, but a typical value is around
Nd =10 cm- 3 . This corresponds to about 1 in 5000 lattice sites occupied by a
19
do~~r since the atomic density of silicon is Ns; = 5 x 1022 cm- 3 .
,..:;EXAM~LE2.1
-- . -
\" . Electron and Hole Concentrations in n,type Silicon
Find the concentration of holes and electrons for a bulk silicon sample in thermal
equilibrium at room temperature that has a donor concentration Nd= 1014 cm- 3 .
ISOLUTION I
Using Eq. (2.14), the electron concentration is
1014 1014 4 . 1020 14

no = +2 1 + ~ = 1.00000001 x 10 cm- 3
2
The hole concentration p 0 can be found from the mass-action law in Eq. (2.7)
since the sample is in thermal equilibrium
oa1o)
2
5
------ -
14
= 9.9999999 x 10 cm- 3
1.00000001 x 10
Verify that the bulk silicon region is electrically neutral by adding the charge
densities of the holes, ionized donors, and electrons
5 14 14
p = q (p 0 + Nr n 0 ) = q (9.9999999 x 10 + lxlO - 1.00000001 x 10 ) = 0
Note that since Nd exceeds n; by four orders of magnitude, we could have used
Eq. (2.15) to find n 0 =Nd= 1014 cm- 3 , and then used the mass-action law to find
the hole concentration to be p 0 = 106 cm- 3 .
2.3.2 Acceptors
A substitutional impurity from Group III of the periodic table that has only three
valence electrons is called an acceptor. The usual acceptor atom for silicon is boron
(B). Figure 2.5 is the bond model's picture of doping by acceptors.
Since the Group III boron atom only has three valence electrons, it can complete
only three of the four bonds with its silicon neighbors. However, an electron in an
adjacent bond requires only a small amount thermal energy to jump over and com-
plete the fourth bond. This process of "accepting" an electron accounts for the name
of this class of dopants. The resulting hole is free to move around the crystal, whereas
the immobile acceptor takes on a negative charge due to the extra electron as shown
in Fig. 2.5.
Given that the acceptor concentration is Na ( cm- 3), the equilibrium hole and elec-
tron concentrations can be found using the same approach used for donors. The bulk
silicon region is electrically neutral and the hole, electron, and ionized acceptor
charge densities must therefore sum to zero:
(2.17)
The electron concentration can be eliminated by substituting n = n/ I p

0 0 from
the mass-action law in Eq. (2.7)
2
n;
p - - -N = 0 (2.18)
o Po a
Using the quadratic formula to solve this equation yields an expression for the
hole concentration in silicon doped with acceptors
immobile negatively ionized acceptor
> Figure 2.5 Bond model showing boron on a substitutional lattice site in a bulk
region of the silicon crystal. A hole is created when the incomplete fourth bond is
filled by acceptance of an electron from a neighboring bond.
2.3 • Doping 31
2
4n;
P,, 1+- (2.19)
N2
a
For the typical case where N 0 is orders of magnitude larger than n;, the intrinsic
hole concentration, the square root term is approximately one and the hole concen-
tration is
(2.20)
By applying the mass-action law, we find that the equilibrium electron concen-
tration at room temperature is given by
1020
-3
Nern at300K (2.21)
a
Silicon doped with acceptors is called p-type, in which holes are the majority carriers
and electrons are the minority carriers. P-type silicon is complementary ton-type sil-
icon, in which the relationship between holes and electrons is reversed.
Due to the mass-action law, doping with Na= 1015cm- 3 of boron results in a hole
concentration p 0 = 1015cm-3 and electron concentration n 0 = 105 cm- 3 . As with donors,
the minimum useful acceptor concentration to have extrinsic, p-type silicon is deter-
12
mined by the intrinsic concentration n;. At room temperature Namin ""10 cm- 3 for Na
» n;. The maximum concentration of substitutional boron acceptors in silicon is on the
19
order of N0 "" 10 cm-3 since a boron-silicon compound begins to precipitate at
higher concen'trations.
2.3.3 Donors and Acceptors: Compensation
Silicon is often doped with both donors and acceptors. To find the equilibrium elec-
tron and hole concentrations, first consider the bond model in Fig. 2.6. Using the
same reasoning used for the donor-only and acceptor-only cases, it can be seen that
charge neutrality applies to the bulk silicon region in Fig. 2.6. The charge densities
from electrons, holes, ionized donors, and ionized acceptors must therefore sum to
zero
(2.22)
When Nd > Na, we can substitute for the hole concentration in thermal equilib-
rium p O = n;2
I n0 from the mass-action law and find
(2.23)
This quadratic equation is identical to Eq. (2.13) with the net donor concentra-
tion Nd - Na replacing the donor concentration. The solution can be found by substi-
tution into Eq. (2.14)
> Figure 2.6 Bond model showing a bulk section of silicon that is doped with
both arsenic (donors) and boron (acceptors) with Nd> Na. Donated electrons ionize
the acceptors by completing their half-filled bonds, with the excess Nd - Na
becoming mobile electrons. This process is called compensation.
2
(~ -~) (~ -~) 4n;
+ 1+ 2 (2.24)
2 2
(~ -~)
where we assume that the donor concentration is greater than the acceptor concen-
tration Nd> Na. In practice, the excess of donors over acceptors Nd - Na will usually
greatly exceed the intrinsic concentration since separately Nd » n; and Na » n;. For
example, if the donor concentration exceeds the acceptor concentration by only 5%
at a typical doping level of about 1015 cm- 3 , then
15 -3 15 -3 13 -3
~ - ~ = 1.05 x 10 cm - 1.0 x 10 cm = 5 x 10 cm » n; (2.25)
Assuming Nd - Na » n;, we find that
I no=~-~ (~ >~) (2.26)
The physical interpretation of Eq. (2.26) is that electrons donated by the donors
first ionize all the acceptors, with only the net donor concentration in excess of the
acceptor concentration contributing mobile electrons to the crystal. This process is
called compensation.
Since the sample is in thermal equilibrium, the minority hole concentration can
be found by the mass-action law
(2.27)
2.3 • Doping 33
At first glance, the fact that there are negligible holes in a sample doped with accep-
tors seems counter-intuitive. However, the mass-action law depresses the hole con-
centration since the (majority) electron concentration far exceeds n;.
For the case where the acceptor concentration is greater than the donor concen-
tration, similar reasoning shows that the sample is p-type with
compensa1ed p-type
carrier concentration
(2.28)
In summary, a silicon crystal that is doped with one type of impurities may be com-
pensated by doping with an excess of impurities of the opposite type. In section 2.5,
we will see that compensation is a cornerstone of integrated device fabrication.
Electron and Hole Concentrations

for Compensated Silicon
a) Find the thermal equilibrium electron and hole concentrations at room tempera-
ture for a bulk silicon region doped with donors at Nd= 1014cm- 3 and with acceptors
at Na= 5 x 1013cm- 3 .
I SOLUTION I
The sample is n-type since the donor concentration exceeds the acceptor concen-
tration. Since the net donor concentration Nd - Na= 1014 cm- 3 - 5 x 1013 cm- 3 =
5 x 1013 cm- 3 » n;, we can equate the electron concentration to the net donor con-
centration (Eq. 2.26)
The mass-action law in Eq. (2.7) determines that the minority hole concentration
is
2
n; 6 -3
p0 =N _ N = 2 x IO cm
d a
To verify that the region is electrically neutral, we add the contributions of the
positive holes and ionized donors with the negative electrons and negative ion-
ized acceptors
6 14 13 13
P = q (p 0 + ~ - ~ - f0 = q (2 X 10 + 10 - 5 X 10 - 5 X 10 ) "'0
Note: The charge density would have been exactly zero if the exact expression
for the compensated electron concentration in Eq. (2.24) had been used.
(b) Now consider a bulk n-type silicon sample with Nd= 1014 cm- 3 . We would like to
add acceptors to make the hole concentrationp 0 = 5 x 1013 cm- 3 . What acceptor dop-
ing concentration Na is needed? What is the electron concentration?
34 Chapter 2 • SemiconductorPhysicsand IC Technology
I SOLUTION I
The desired hole concentration is much greater than the intrinsic concentration;
therefore, the compensated sample is p-type. Using Eq. (2.28), we can solve for
the acceptor concentration
The electron concentration follows from the mass-action law

1020 6
13
= 2x 10 cm- 3
5 x 10
:,...·~o CARRIER TRANSPORT

At room temperature, the holes and electrons in silicon are in agitated, random
motion. Their average velocity is called the thermal velocity v1h, and is about 107 cm/s-
only a factor of 3000 less than the speed of light! Holes and electrons collide frequently
by scattering off ionized impurities, crystal defects, surfaces, and each other. A typical
value for the average collision time 'tc at room temperature in bulk silicon is about
1 x 10-13 s = 0.1 picoseconds. In between collisions, carriers on average will travel a
distance
'A,=v 1h 'tc = 10 nm= 0.01 µm (2.29)
where 'A,is defined as the mean free path. State-of-the-art silicon microelectronic
devices have dimensions that are at least 0.1 µm, but are typically much larger. There-
fore, holes and electrons undergo many collisions while traversing a device structure.
In the next section, this important fact will allow us to find simple equations that
describe the average transport of charge carriers due to electric fields or to concen-
tration gradients.
2.4.1 Drift Velocity

In thermal equilibrium, the electrons and holes in a region of the silicon lattice have
no net motion. In other words, the average position of all the electrons ( or separately,
all of the holes) does not change with time. Figure 2.7(a) illustrates this situation for
electrons. The random trajectories of three electrons along the x axis are shown in
Fig. 2.7(a) for a time interval M = 7 1 0 during which seven collisions occur. To show
clearly the electron motion, the x-axis trajectories are spread out and all trajectories
are shown originating at x = X;. The average change in position after the interval llt is
llx = (xf, I -x;) + (xf,2 -x;) + (xf,3-x;) "'0

(2.30)
3
using the approximate values for the position changes in Fig. 2.7(a). Of course, the
average position change for electrons in a region of the silicon lattice will be taken
2.4 • Carrier Transport 35
over a huge number of electrons. Since the trajectories are random, Lix will be zero
after any time interval Lit» 'tc-
The situation with an applied electric field Eis different, as shown in Fig. 2.7(b ). Sign Convention:
It is important to first be sure of the sign for the electric field and other directional positive:
variables. By convention, the electric field is positive when it points in the same direc- E
tion as the reference direction for the x axis. Thus, the electric field E in Fig. 2.7(b) is
positive. The electrons are subject to an electrostatic force x
negative
F'_1= -qE (2.31)
that points in the -x direction. During their "free flight" in between collisions, the
electrons with mass mn experience a constant acceleration a= Fel I mn = -qE/mn that
deflects their trajectories toward the -x direction as shown in Fig. 2.7(b ).
Averaging the position changes over the interval M, we find
(2.32)
(a) Thermal Equilibrium, E = 0 (b) Electric Field E > 0
I Electron # 1I I Electron # 1 I
X; x X; Xj,l x
I Electron #2 I I Electron # 2 I
~
x
==*~ X; x
I Electron # 31 I Electron # 31
x x
* x; = initial position * x1; n = final position of electron n after 7 collisions
> Figure 2. 7 (a) Trajectories for three electrons in thermal equilibrium (zero
applied electric field) for seven collisions. (b) Trajectories for the same three
electrons for an electric field E > 0. All of the trajectories are drawn with the same
initial position X;.
using the three sample trajectories in Fig. 2.7(b). Dividing the average position
change by the time interval, we define the drift velocity of the electrons as
(2.33)
which is negative for E > 0. Note that the drift velocity is the average net velocity for
a huge number of electrons in a region of the silicon lattice. Experimentally, the drift
velocity vdn is found to be proportional to the electric field
(2.34)
where µn is the electron mobility with units of cm 2 /(Vs).

The electron mobility is a function of both the temperature and total doping con-
centration. Mobility decreases with increasing temperature since thermal agitation
causes lattice vibrations to scatter electrons. Ionized dopants also scatter electrons
and, therefore, reduce the collision time. This means the mobility is a decreasing
function of Nd+ Na. Figure 2.8 gives experimental measurements of electron mobility
for Nd+ Na= 1013 to 1020 cm- 3 at room temperature. It is useful to have a rough num-
ber for electron mobility in silicon for approximate calculations. Substituting a typi-
cal total dopant concentration of 2.5 x 1016cm- 3 , Fig. 2.8 gives an electron mobility of
2
µn"" 1000 cm /(Vs).
1400
1200
,,___
1000
>"'
N'
a 800
~
;E
:E
0 600
a
400
200
0
1013 1014 1015 1016 1017 1018 1019 1020
Nd+ Na total dopant concentration (cm-3)
> Figure 2.8 Linear-log plot of electron and hole mobilities at room temperature,
as functions of the total doping concentration Nd+ N 8 • After R. S. Muller and T. I.
Kamins, Device Electronics for Integrated Circuits, 2"d ed., Wiley, 1986.
2.4 • Carrier Transport 37
Holes also drift in an electric field-the holes drifting with the field on average,
because the positive charge causes them to accelerate in the same direction as E
between collisions. The drift velocity for holes is also proportional to the electric field
(2.35) hole .drift v:e,lpcify
where µPis the hole mobility. In silicon, hole mobilities are lower than electron mobil-
ities. They range from 100-500 cm2 /(Vs) at room temperature as shown in Fig. 2.8.
For a typical dopant concentration of 2.5 x 1016 cm~3 , the "round number" for hole
mobility is µP ""400 cm 2 /(Vs).
The linear relationship between drift velocity and electric field breaks down when
the electric field is high enough that the drift velocity approaches vsat = 107cm/s.
Figure 2.9 shows Ivdi as a function of electric field for electrons and holes and illus-
trates that both carriers have a maximum drift velocity of about 107 cm/s. This phe-
nomenon is known as velocity saturation.
Advances in integrated-circuit technology are enabling ever-smaller electronic
devices, a trend that is increasing the internal electric fields. Velocity saturation is an
important phenomenon for accurate modeling of VLSI device characteristics. In this
book, we will assume that the linear relationships between electric field and drift
velocity in Eqs. (2.34) and (2.35) are valid unless velocity saturation is specifically
mentioned.
-vdn• vdp
(cm/s)
108
10 104 105 E
(V/cm)
>-Figure 2.9 Log-log plot of the drift velocity as a function of electric field for
electrons and holes, showing the linear region where the velocity is proportional to
the field and velocity saturation at high fields. After R. S. Muller and T. I. Kam ins,
Device Electronics for Integrated Circuits, 2nd ed., Wiley, 1986.
2.4.2 Drift Current Density
Drift of charged particles results in the transport of charge through the silicon lattice.
We quantify the transport of charge by the current density J, which is defined as the
rate at which charge passes through a reference plane in the +x direction per unit
area. The units for current density are C/(s cm2 ) or A/cm 2 . The sign convention for
current density is illustrated in Fig. 2.10. J is positive when positive charge passes in
the +x direction. Since the charge carriers can be positive (holes) or negative (elec-
trons), there are two ways that J can be positive.
1. Holes can pass through the reference plane in the +x direction.
2. Electrons can pass through the reference plane in the -x direction.
Note that either case results in a net transport of positive charge through the refer-
ence plane in the +x direction making the current density J positive.
The hole current density due to drift in an electric field is expressed as J:, - sub-
script p designates holes and superscript dr, drift. This can be expressed in terms of
the electric field by directly applying the definition of current density, as shown in
Fig. 2.ll(a).
Over a time interval At, holes in the volume AV=A(vdpAt) will drift across the
reference plane in the +x direction. The number of holes ANP that cross the plane in
the interval At is given by the product of the concentration p of holes and the volume
AV:
(2.36)
The hole charge drifting across the reference plane AQP is the product of ANP and
the charge per hole q. According to the definition of current density
qpAV qpAvdpAt
AAt = AAt = qpvdp (2.37)
reference reference
plane plane
x x
]>0 ]>0
> Figure 2.10 Sign convention for current density J. Both cases represent
positive values of J since there is a net transport of positive charge in the +x
direction.
2.4 • CarrierTransport 39
reference plane reference plane

(area;;) ea A)
volume AV= A(vdpAt)
- --
volume AV= A(-vdn At)
E- \ ..,._ £-
(sign:+) (sign:+)
vdp - -vdn
(sign:+) (sign:-)
x, x x
x, x,- vdnAt
(a) (b)
> Figure 2.11 (a) The definition of current density applied to drifting holes, and
(b) to drifting electrons. The reference plane is located at x= x, for this example. For
an electric field pointed in the +x direction, holes drift through the reference plane
from left to right, whereas electrons drift from left to right. The current density Jd' > O
for both cases.
Substituting for the drift velocity from Eq. (2.35), we find that
(2.38)
"e:le
driftcurrent deAsity
For electrons, the drift velocity is opposite in direction to the electric field as ------- ..................
0
· Jdr
• ,,
shown in Fig. 2.ll(b ). On average, electrons in a volume~ V ==A(-v dn At) drift across
the reference plane in the time interval At. The number of electrons crossing the
plane is therefore ~Nn ==n~ V, and the electron charge is equal to
(2.39)
The electrons drift across the reference plane in the -x direction as shown in Fig.
2.ll(b ). Therefore, the drift current density is given by
(2.40)
Substituting the electron charge from Eq. (2.39) and the drift velocity from Eq.
(2.34), the electron drift current density is given by
dr [ (qnAvdnM)]
l n == - A~t == -qnvdn = -qn (-µnE) = qnµn E (2.41) electron drift current
density
,..,------Jdr p
Therefore, the electron drift current points in the same direction as the electric
field. Restating Eq. (2.41)
I I« vsat .;.._
_____ .,.x
J ndr -- qnµn E ' ,or
+ vdn ' (2.42)
> EXAMPLE 2.3 Drift Current Density

Figure Ex2.3A shows a slab of p-type silicon (Na= 1016 cm- 3) under the influence of
a uniform electric field E. Plot the drift current density as a function of E over the
range -5 x 105 V/cm < E < 5 x 105 V/cm. Note: Velocity saturation occurs for holes
drifting in electric fields with magnitudes in excess of 105 V/cm. Find the voltage
across the slab at which the current reaches its maximum value.
I SOLUTION I
For small electric fields, we can use Eq. (2.38) to find the relationship between E
and the drift current density
where we have substituted p = Na for the hole concentration. Even though the
holes are drifting with the electric field and the slab is not in thermal equilibrium,
their concentration remains approximately equal to the acceptor concentration.
As we will see later in this chapter, metal coatings on the left and right faces of
the silicon slab can supply and remove the drifting carriers.
From Fig. 2.8, the hole mobility is µP = 425 cm 2 I (Vs) and the low-field drift cur-
rent density is 1:" E
= (0.68) (units: J: A/cm 2 , E: V/cm).
According to Fig. 2.9, this linear relationship is valid until the hole drift velocity
saturates, which occurs at electric fields greater than Ecrit"" 105 V/cm = 100 kV/
cm. Since we don't have a model for velocity saturation, we limit the drift current
2
density at 1:~ax = (0.68) · Ecrit = 68 kA/cm as shown in Fig. Ex2.3B
Substituting for the thickness of the slab, the voltage needed to saturate the drift-
ing holes is Ecrit x 1 µm = (10 5 V/cm) (10--4cm)= 10 V. The maximum current is
given by the product of 1:~ax and the area of the slab: !max= (68 kA/cm 2 )
(5 x 5 x 10-8 cm 2 ) = 17 mA.
~ -------E
1
5µm
j x
lµm
> Figure Ex2.3A P-type silicon slab with uniform applied electric field E.
2.4• CarrierTransport 41
J dr (kA/cm2)
p
80
60
40
I/'= (0.68) E
-140 -100 -60

40 60 80 100 140 E (kV/cm)
-40
-60
-80
> Figure Ex2.3B Drift current density through slab as a function of the applied
electric field.
2.4.3 Diffusion Current Density
Diffusion is the process where gradients in concentration cause a net transport of par-
ticles. For example, suppose an ammonia bottle is opened in one corner of a room in
which the air is at room temperature. Even if the air in the room is perfectly still, the
random motions (with frequent collisions) of the air and ammonia molecules enable
the transport of ammonia throughout the room. The net transport will be from
regions of high concentration in the corner with the open bottle, to regions of low
concentration in the opposite corner. If the bottle is capped, the concentration of
ammonia will equalize throughout the room after a period of time. Diffusion reduces
and eventually eliminates all gradients in ammonia concentration throughout the
room.
Since electrons and holes also undergo random motion with frequent collisions,
gradients in the electron or hole concentrations result in diffusion transport. In order
to model diffusion, we consider holes having a concentration p(x), which decreases
with x as shown in Fig. 2.12. On the average, holes crossing the reference plane at
x = xr started within one mean free path A of x,. The diffusion current density 1;itf of
holes across the reference plane has its origin in the difference in the number of holes
in the two regions of volume L1V = AA to the left and right of the reference plane.
Only half of the holes starting in either region will cross the reference plane in the
time interval 'tc, while the other half are moving away from the reference plane.
Therefore, the hole diffusion current density at x = xr is given by
J
diff
(x=x) = q
[1P(x = Xr-A)AA-1p (x = xr+ A)AA.
(2.43)
P r A't c
where AA is the volume. Note that holes crossing from the right to the left, opposite
to the +x direction, enter into Eq. (2.43) with a negative sign.
reference plane
p(x)
_____ /ea=A)
ho!.......
diffusion ....,..
-----+-~ Jp diff (positive)
volume AA:
volume AA:
holes moving
holes moving
in + x direction cross
reference plane within in - x direction cross
reference plane within
!J.t=tc
/:,.t= 'tc
x
x,- A x, x,+ A
> Figure 2.12 One-dimensional picture for quantifying the hole diffusion current
density as a function of the gradient in hole concentration.
For Eq. (2.43) to be a good approximation, the mean free path (about 0.01 µmat
room temperature) must be much smaller than the distance over whichp(x) is vary-
ing. This requirement is usually satisfied for microelectronic device structures. Note
that we have highly exaggerated the gradient in the hole concentration in Fig. 2.12 to
make the difference in the hole population across the reference plane clearly visible.
Using first-order Taylor expansions for p(x = x, + A) and p(x = x, - A) and simplify-
ing, we find that
(2.44)
Canceling the hole concentration and collecting terms, Eq. (2.44) reduces to
diff
JP (x =x,) = -q
l A2Jdp
~ c
dx I
x=x r
dp
= -qDPdx I
x=x r
(2.45)
In Eq. (2.45), the term A.2/'tcis identified as the hole diffusion coefficient Dp, with units
of cm 2 /s. The diffusion current density for holes is
diff = dp
JP -qDP dx (2.46)
The proportionality of the diffusion current density to the concentration gradient

is known as Fick's Law and is a fundamental result for diffusion processes. The neg-
ative sign in Eq. (2.46) reflects the fact that particles diffuse from regions of high con-
centration toward regions of low concentration. The concentration gradient dpldx in
Fig. 2.12 is negative, so the holes diffuse in the +x direction and the diffusion current
density is positive.
For the case of the electron concentration gradient shown in Fig. 2.13, we find the
net transport of charge across the reference plane in the +x direction as
diff [1n(x = xr - A)AA -1n(x = xr +A)AA1 electron·aiffus,~n.

Jn (x=xr) = (-q) kc (2.47)
I.·..::o.·.....
c current <;le:i:tE!iiY
·•
n(x}
This is identical in form to Eq. (2.43) for holes, except we have accounted for the neg-
ative charge on the electrons crossing the reference plane. Again, use first-order Tay-
lor expansions for the concentrations at xr - A and xr + A ~
.•. ·.·.·
.•....•.
J"tfi ()
· .• 'ttf"!'/
.
. .. ·. x
diff ( -q) 'A[[
n(xr)-::1
AJ-[n(xr)+!:I
Al]
x, x,
Jn (x=x,) = -.- 2 (2.48)
n(~xJ d .· .•.
c
. o ...· •..·.
Collecting terms, the electron diffusion current density can be expressed by .. f":>!( ·•
o ", o'O:
. K
J!,iff (x = xr) (2.49)
reference plane
• (area =A)
n(x) II
electron· diffusion
..--t---Jn diff (negative)
volume AA.:
volume AA.:
in + x direction cross electrons moving
reference plane within in - x direction cross
l",.t = 'tc. reference plane within
ti.t= 'tc.
I
x
xr-A xr x,+ A
~ Figure 2.13 One-dimensional picture for quantifying the electron diffusion

current density as a function of the gradient in electron concentration. Note that the
electrons diffuse left-to-right (in the +x direction), but that Jtff points in the -x
direction and is negative.
where the electron diffusion coefficient is identified as DwThe negative gradient in
electron concentration (dnldx < 0) shown in Fig. 2.12 means the electron diffusion
current density is negative (i.e., it points in the -x direction). The physical reason
behind this result is that the electrons diffuse from regions of high concentration to
regions of low concentration in the +x direction in Fig. 2.13. However, each electron
carries a charge -q, which results in J;!iff pointing in the -x direction.
The diffusion coefficients for holes and electrons are functions of doping concen-
tration and temperature since these are determined by the mean free path and colli-
sion time. The mobility µ and diffusion coefficient D were derived from physical
models based on large numbers of carriers undergoing random thermal motion with
frequent collisions. In a solid-state physics course, it is shown that the two constants
are proportional and their ratio is equal to
D kT
(2.50)
µ q
where k = 1.38 x 10-23 J /K is Boltzmann's Constant. This important result is called

the Einstein Relation-named after Albert Einstein, who was active in solid-state
physics in addition to his other interests. Eq. (2.50) holds separately for holes and for
electrons
I~:= kT
q
and (2.51)
One implication of the Einstein Relation is that only the mobility need be mea-
sured. The diffusion coefficient can be found directly from Eq. (2.51). Using room-
temperature values for the electron and hole mobilities in silicon (1000 and 400 cm 2/
(Vs)), typical round numbers for the electron and hole diffusion coefficients are
(2.52)
The thermal voltage is defined as the quantity kT/q:
vth -- -kTq (2.53)
The thermal voltage has units of Volts and will appear in many device physics equa-
tions. It ranges from 25 m Vin a cool room in which the temperature is 17 °C or 62 °F
to 26 m V in a somewhat warm room at 28.5 °C or 83 °F. In this book, we will choose
either V 1h = 25 mV or 26 mV-whichever results in "round numbers."
2.4.4 Total Current Density

As we have seen in the previous two sections, electrons and holes drift in an electric
field and diffuse due to concentration gradients. The total current density J is the sum
of all four components
J = J n+ J
p
= Jdrn +
Jdiff
n +
Jdr
p
+ Jdiff
p
(2.54)
The total electron and total hole current densities Jn and JP are important results,
therefore they are given separately
dr diff dn
Jn =Jn + Jn = qnµnE + qDn dx (2.55) .1:1«1ctrcm
and·hote total
~urrentdensities
(2.56)
In most cases, we will be able to simplify Eqs. (2.55) and (2.56) for a particular device
structure and focus on the dominant components of the current densities.
)I,,-IXAM'111:
·2.4. Diffusion Current Density
Figure Ex2.4 shows a 1 µm-thick slab of p-type silicon having Na= 1017 cm- 3 . By con-
tinuously injecting minority electrons on the left side at x = 0 and extracting them
from the right side at x = 1 µm, we have set up a linear gradient in the electron con-
centration n(x). What is the electron diffusion current density in the slab?
I SOLUTION I
On the linear plot of n(x), we note that the electron concentration at x = 1 µm,
103 cm-3 , is indistinguishable from zero. For a total doping concentration of
1017cm- 3 , Fig. 2.8 indicates that the electron mobility is µn = 750 cm 2 /(Vs). Substi-
tuting into the Einstein Relation from Eq. (2.51), we find that the electron diffu-
sion coefficient is
-3 2 2
Dn = '{h · µn = (25 x 10 V) · (750cm /Vs) = 18.75 cm /s
n(x=O)= n(x= 1 µm)

1014 cm-3 = 103 cm-3
5 x 1013
0 O -+--~--'--~-~-x [µm]
o 0.5
~ Figure Ex2.4 Electron concentration through a slab of p-type silicon.

We apply Eq. (2.49) to calculate the electron the diffusion current density in the
silicon slab
J
diff
= qDn
dn
-d = ( 1.6 x
-19
10 ) (18.75)
(I03-1QI4)
_ = -3 A/cm
4
2
n X 10
Note that the current density is negative since the electrons are diffusing from left
to right and carry a negative charge. The magnitude of the diffusion current den-
sity in this example is also small in comparison with typical majority carrier drift
current densities. A voltage of less than 1 m V between the left and right faces of
the slab is all that is needed to cause a drift current of majority holes of 3 A/cm 2 .
,> 2.5 SILICON INTEGRATED CIRCUIT TECHNOLOGY

The focus of this text is silicon integrated devices and circuits, in which an electronic
functional block, subsystem, or entire system is implemented on a single integrated
circuit (IC). A basic appreciation of the manufacturing processes used to make I Cs
will be helpful in later chapters since it will give you an appreciation for the dimen-
sional scales and physical structures of integrated devices. As you will see, the origin
of the important "parasitic" circuit elements in I Cs is obvious from the device struc-
tures.
This section provides a basic introduction to IC fabrication technology. This fas-
cinating discipline draws from materials science and engineering, chemical engineer-
ing, and applied physics, as well as electrical engineering. By many measures, silicon
IC technology is the crowning technological achievement of the late 20th century.
Few areas of life have been unaffected by the penetration of silicon I Cs such as micro-
processors, microcontrollers, and modems. Since the fundamental physical limits of
these products are at least two decades away, the early 21st century is certain to see
further advances in the core fabrication processes that will have great impact on com-
puters and other electronic systems. In addition, "spin-offs" from IC technology are
catalyzing new fields such as micro electromechanical systems (MEMS). Silicon
MEMS have display, sensing, or actuating functions. They may incorporate 106
micromechanical elements on the silicon chip along with a complex electronic inter-
face. In summary, although the first silicon IC dates to around 1960, the end of the IC
revolution is by no means in sight.
Figure 2.14 shows the fundamental unit of IC manufacturing-a silicon wafer.
The typical wafer diameter is 20 cm in the mid-1990's. The wafer is a nearly perfect
crystal that has been polished to be extremely flat. Wafers of this diameter cost about
US$50.00 in large quantities. A group of around 24 wafers, called a lot, undergoes the
same sequence of fabrication steps together. In a complex IC such as a microproces-
sor or dynamic random-access memory (DRAM), the process has around 200 steps
and takes about three weeks to complete for a typical manufacturing schedule.
Figure 2.14 shows that when the process is complete, the wafer is tiled with an
array of I Cs interspersed with a few diagnostic test chips for monitoring the fabrica-
tion process. A typical complex IC has an area of about 1.5 cm x 1.5 cm, or about 2
cm 2 , on which are several million electronic devices. For a 20 cm-diameter wafer,
there will be around 150 chips of this area. The cost of the completed wafer is roughly
US$2,000, which is a function of the fabrication process. How many of the chips work
2.5 • SiliconIntegratedCircuit Technology 47
~ Figure 2.14 Silicon wafer and inset showing a single IC chip (die), with bond
pads, and a side view of a dual in-line (DIP) package, shown with the lid removed and
only a single bond wire. Wafer diameter and chip area are typical for mid-1990's
large-volume, complex ICs.
(the yield) will affect the price at which the manufacturer sells each chip. The yield is
a function of the quality control in the manufacturing plant, as well as the design mar-
gins built into the process and circuit. For example, a high-performance IC may have
dimensions that approach the technological limits with the result that the yield is low.
However, the high performance may justify a very high price and make the process
profitable despite the low yield. As a particular IC design matures in production, the
yield can be predicted to improve to around 90% or higher, which allows the selling
price to drop steadily. The IC industry is extremely competitive. Companies aggres-
sively track the yield curve of most products by dropping prices as yields increase.
This maximizes volume and makes it more difficult for other companies to enter into
direct competition on the product.
General-use I Cs are packaged in plastic dual in-line (DIP) packages as shown in
Fig. 2.14. In recent years, there has been great progress in economical packaging of
I Cs into high-density multichip modules for use in applications such as lap-top com-
puters and cellular phones. Instead of the DIP shown in Fig. 2.14, chips are often
encapsulated in low-profile surface mount packages or mounted directly on the cir-
cuit board using a "flip-chip" technique. Electrical connections between the board
and inverted chip are made through electroplated solder bumps, rather than through
bond wires, allowing for many more interconnects per chip.
2.5.1 Photolithography and Pattern Transfer

The essential process in IC fabrication is the imaging, alignment, and transfer of com-
plex patterns onto the silicon wafer. A modern fabrication process uses around 15-
20 separate patterning steps to define the transistors, diodes, and several levels of
electrical interconnections (i.e., thin-film wires) that make up the IC. A wafer stepper
is used to transfer patterns from an optical plate called a mask, to a photosensitive
film coating the top surface of the wafer called photoresist. The process shown sche-
matically in Fig. 2.15 depicts the exposure of a single die to the image of the mask.
ultraviolet light illumination
exposed
region of
silicon
wafer
(coated with
photoresist)
> Figure 2.15 Schematic diagram showing imaging of the mask pattern on one
die in a wafer stepper. The mask pattern is shown in gray on exposed dice.
Unexposed dice are shown in outline only. In reality, the image in the photoresist is
not visible until after development.
After completion of the exposure, the wafer stepper indexes to the next die, automat-
ically aligns to special features previously etched on it, and repeats the exposure.
Since the image of the mask pattern must be distortion-free and the error allowed in
the alignment is less than 0.1 µm, it is not surprising that a high-volume wafer stepper
costs well over US$2 million in the mid-1990s.
Figure 2.16 shows the step-by-step process of how the transfer from the mask to
the wafer is accomplished through the photolithography and pattern transfer process.
Figure 2.16 also introduces the customary way of visualizing three-dimensional IC
structures. The layout consists of the series of masks used to create the IC structure.
Lines across the layout denote the locations of cross sections through the silicon
wafer. Typically, only the top few µm of the wafer are needed to show the device
structure. In order to draw the cross sections, we must know the sequence of masking
and fabrication steps. This sequence is called the fabrication process, or process for
short. In analogy with food preparation, the fabrication process is sometimes called
the recipe.
The single-mask photolithography and pattern transfer process in Fig. 2.16
involves five steps.
2.5 • Silicon IntegratedCircuit Technology 49
1. A 500 nm-thick film of silicon dioxide, Si0 2 , is grown on the silicon wafer. In
order to grow the thermal Si0 2 film, the wafer is placed in a furnace at an ele-
vated temperature. Silicon surfaces react with oxygen to form a very high quality
insulating film.
2. The wafer is uniformly coated with a 1 µm-thick film of photoresist by applying
a mixture of photoresist and solvent, spinning the wafer at several thousand rpm
for 30 seconds, and then baking to drive off the solvent.
3. Areas of this film that are exposed to ultraviolet light (UV) during the exposure
step dissolve in an alkaline solution-a process known as development. Note
carefully the differences in cross sections A-A and B-B after development of the
mask image in the photoresist.
Process and Cross Sections Oxide Mask
j 1. Grow 500 nm of Si0 2 j ~ thermal Si0

2
Silicon substrate I
j 2. Coat wafer with 1 µm of photoresist I
k
I "-::::photoresist I
0 2 3 4 5
r
6 x (µm)
0 2 3 4 5 6
IB x(µm)
0 2 3 4 5 6 x (µm)
j 4. Etch Si0 2 in fluorine plasma I j s. Strip photoresist in 0 2 plasmaj
A~
0
~
2 3 4 5
IA
6 x (µm)
A~
0 2
I
3
--------
4
thermal Si0
I
5
r
2
6 x (µm)
B~ IB B~ IB
0 2 3 4 5 6 x (µm) 0 2 3 4 5 6 x(µm)
>-Figure 2.16 Cross sections for transferring the oxide mask pattern to a thermal oxide film. A
shorthand description of this process is "grow 500 nm of thermal Si0 2 and pattern using the oxide
mask."
4. After baking the photoresist to drive off residual solvents, the wafer is immersed
in a vacuum chamber containing a fluorine plasma. A typical example of a
plasma is the glow discharge in a fluorescent light. The plasma reacts with Si0 2
and etches it much faster than it etches the photoresist film. Once the silicon
dioxide is etched away from the exposed areas, the underlying silicon is also
etched, but very slowly. This enables the etch to be terminated before significant
silicon erosion occurs.
5. The final step is to remove the photoresist in an oxygen plasma. This plasma
aggressively attacks organic materials, such as photoresist, but does not etch sil-
icon or silicon dioxide. Comparing the cross sections for steps 3 and 5 in Fig. 2.16,
we see that the mask pattern has been successfully transferred to the silicon diox-
ide film.
As we mentioned, an IC process involves many photolithography and pattern trans-
fer steps on deposited films. In order to avoid repeating steps 1-5 each time we use
this procedure, we will summarize Fig. 2.16 by the statement: "Grow 500 nm of Si0 2
and pattern using the oxide mask." The photolithography and pattern transfer
sequence can be thought of as a subroutine of the IC fabrication process.
2.5.2 Doping by Ion Implantation
Acceptors and donors can be incorporated into selected regions of the top surface of
a silicon wafer by means of ion implantation. Ions of boron, arsenic, or phosphorus
are extracted from a plasma, accelerated to energies from 20 ke V to 3 Me V
(eV = electron volt), and formed into a tightly aligned ion beam. The ion beam is
scanned over the surface of the silicon wafer. The beam is scanned until the desired
dose of ions is implanted into the wafer. The ion dose has units of ions per unit area
(cm- 2). After entering the silicon, the bombarding ions are slowed and then stopped
by collisions with the silicon lattice.
Ion implantation creates a damaged region near the surface, since the collisions
knock silicon atoms out of their positions in the lattice. The damaged region loses its
crystal structure entirely and becomes amorphous. By heating the silicon wafer
above 900° C, the damaged region recrystallizes, a process which is known as anneal-
ing. Remarkably, most of the dopant ions in the recrystallized region end up on sub-
stitutional lattice sites. During annealing, the implanted dopant ions also diffuse
further into the silicon wafer.
Figure 2.17 illustrates how ion implantation can be used to dope specific regions
of the wafer. We start with the patterned oxide film from Fig. 2.16, grown on a p-type
wafer with a background donor concentration Na= 1015 cm-3. A dose of Qd = 1014 cm-2
phosphorous ions is implanted into the wafer. Their energy is such that the phospho-
rus ions do not penetrate the Si0 2 film as shown in Fig. 2.17. After a thermal cycle, the
phosphorus concentration Nix) is found to exceed the background donor concentra-
tion up to a depth xj = 500 nm, which is called the junction depth. Over the past 20
years, extensive research has investigated and modeled the details of dopant concen-
trations after implantation and annealing. Further discussion of this subject will be left
to advanced courses on IC fabrication technology.
In this book, we will simply use the average donor concentration in the layer Nd
as defined in the ion profile in Fig. 2.17. Since the dose Qd of donors (units: cm-2 ) is
located in a layer of depth xj after annealing, the average concentration is
2 .5 • SiliconIntegratedCircuit Technology 51
11.Implant phosphorus ions j I3. Phosphorus profile after annealing I

phosphorus-doped
phosphorus ions, dose Qd = 1014cm-2
/ n-type \
F-;_;~;R-r-5 x,
Nix)
T
,_,
12.Initial implant profile I
P-implanted Si0 2 stops
I layer
r:::::,
/ phosphorus
F xj (500 nm) x
> Figure 2.17 Phosphorus implantation into the patterned oxide film shown in
Fig. 2.16. The average concentration of phosphorus is defined in the plot of Nd(x)
after annealing.
(2.57)
For the case shown in Fig. 2.17, Nd= (1014 cm- 2 )/(5 x 10-5 cm)= 2 x 1018 cm- 3 .
2.5.3 Deposited Conducting and Insulating Films

In order to fabricate integrated circuits, we need conducting films that can be pat-
terned into the microscopic wiring called interconnects. The interconnects enable
electrical signals to be propagated around the IC. In addition, insulating films are
needed to enable crossovers between two or more levels of interconnects. In the mid-
1990s, three to five levels of interconnect are used for efficient packing and for min-
imum interconnect lengths in high-density digital ICs. As we will see in Chapter 5,
short interconnects are important for achieving high switching speed.
Our first conductor is polycrystalline silicon or polysilicon. This material is an
aggregate of crystallites or grains of single crystal silicon. Thin films of polysilicon are
deposited on the surface of a silicon wafer at temperatures around 600 °Cina process
known as chemical vapor deposition (CVD). Polysilicon is usually deposited with a
maximum doping level of phosphorus. As we will see in the next section, polysilicon
is a mediocre conductor; however, it has the virtues of being thermally stable and free
of contaminants.
Aluminum is the most commonly used conducting film for interconnects. It is
deposited by sputtering, a process in which material is deposited on the wafer from a
nearby target that is bombarded by ions from a plasma discharge. One drawback of
aluminum interconnects is that they cannot tolerate temperatures higher than about
450 °C due to degradation of the Al-Si junctions. Other metals used increasingly in
I Cs are tungsten and copper, both of which can be deposited by CVD processes.
There are several processes for depositing insulating films using CVD. The most
common inter-metal insulator is silicon dioxide, Si0 2 . Phosphorus and boron are
often added during growth in order to allow the film to soften and flow slightly at
temperatures around 900° C. This process, called reflow, is important in smoothing
(or planarizing) the surface of the wafer after several depositions and masking steps
have created too rough a surface. Silicon nitride, Si3N 4, is another CVD insulating
film that is able to mask a high-temperature thermal oxidation since it is imperme-
able to oxygen. In plasma-enhanced CVD processes, Si0 2 and Si3N 4 can be grown at
low temperatures (around 300° C) for use as a final protective layer over the IC.
2.5.4 Layouts and Process Flovvs
You are now familiar with the lithography and etching processes and the basic mate-
rials that are the foundation of silicon IC technology. In this section, we will first
describe a three-mask process for fabricating an ion-implanted resistor and then
describe a more complex process to make an MOS field-effect transistor. You need
not understand anything at this point about how these structures function as elec-
tronic devices. In following these sample processes, focus on the connection between
the layout and the evolving device cross sections.
Figure 2.18(a) depicts the layout of the resistor, as it would be viewed in a com-
puter-aided design (CAD) program. At first glance, you would expect that these pat-
terns would correspond exactly to the actual masks used for photolithography. In
fact, a mask that is nearly opaque (referred to as dark field mask) will obscure the
other masks. Even with the use of color and a variety of "fill" and "border" designs,
a nearly opaque mask will make it difficult to see features on the other masks in a lay-
out. Fortunately, there is a simple solution: we display the negative, or complement,
of the dark field mask in depicting the layout. Clear field masks have few opaque
regions and are displayed directly. Of course, we must keep a careful record of which
are clear field and which are dark field, so that the correct masks can be made. Many
wafer lots have been ruined due to miscommunication on this minor, but critically
important point.
In the resistor layout shown in Fig. 2.18(a), the oxide mask is identified as a dark
field mask. Since the negative of the oxide mask is shown in the layout, the rectangle
is clear on the actual mask. You should note carefully from cross sections A-A and B-
B after step 1 that the oxide has been etched from the region of the wafer exposed by
the clear rectangle. If this mask had been clear field, the oxide would have been
removed everywhere except the rectangle, which would have been dark on the phys-
ical mask.
We now describe the process sequence in Fig. 2.18(b ). The first step is to grow
500 nm of thermal oxide, which is done by heating the wafer in a furnace with oxygen
flowing through it. This film is patterned using the oxide mask and the photolithog-
raphy and etching process shown in Fig. 2.16. The wafer is then implanted with a dose
of phosphorus ions that is sufficiently heavy to compensate the background acceptor
concentration in the p-type silicon wafer. The thermal Si0 2 is thick enough to stop
the phosphorus ions from penetrating into the underlying silicon substrate and is,
therefore, effective at masking the implant.
oxide mask
D (dark field)
•
~
contact mask
(dark field)
metal mask
(clear field)
A A
B
(a)
11.Grow 500 nm of thermal oxide and pattern using oxide mask I

thermal Si0 2
B
p-type silicon wafer p-type silkon w,~•
12.Implant phosphorus and anneal I
I ,
AR~nty~pe
ptype
qAI
./
Bl ( ntype )
ptype
13.
Deposit 600 nm of CVD oxide and pattern using contact mask\
~CVDSiO, Ii contact
window
A~ n ty
\~A
p type ~CJ p type 1·
I4. Sputter 1 µm of aluminum and pattern using metal mask\
p type p type
(b)
~ Figure 2.18 (a) Layout and (b) process flow with cross sections A-A and 8-8 for
fabrication of an ion-implanted resistor.
After annealing the wafer at high temperature, the phosphorus diffuses further
into the substrate and creates a rectangular n-type region as shown in the cross sec-
tions after Step 2 in Fig. 2.18. Note that the phosphorus diffuses laterally under the
oxide during the anneal, with the result that the dimensions of the n-type region are
somewhat larger than the rectangle on the oxide mask. A chemical-vapor deposition
(CVD) process is then used to deposit 600 nm of CVD Si0 2 . Contact windows are
patterned in this film using a second photolithography and etching step. Note that the
contact mask is also a dark field mask. A 1 µm-thick film of aluminum is sputtered
and patterned using the metal mask, which is a clear field mask. In order for the alu-
minum to form good electrical contacts to then-type silicon region, the wafer must
be heated to around 425 °C. It is helpful to observe the differences in the A-A and the
B-B cross sections in the completed structure and to connect these to the layout.
With only a slightly more complicated process, we can fabricate a "starter" ver-
sion of an MOS field-effect transistor-the core device in digital ICs. Figure 2.19
gives the four mask patterns and the sequence of cross sectional views of the struc-
ture. After patterning a window in the initial 500 nm thick Si0 2 film in the first mask-
ing step, a 15 nm-thick Si0 2 "gate oxide" layer is grown in the bare silicon regions by
a high-temperature oxidation process. Polysilicon is deposited by CVD with n-type
doping and is patterned in the second masking step. The polysilicon film serves as the
mask for a high-dose arsenic implant that penetrates the gate oxide. After annealing,
n-type regions are formed on each side of the polysilicon "gate" that are seen only in
the B-B cross section perpendicular to the gate. There is some lateral spreading of the
arsenic implant during the anneal as seen in the B-B cross section. The next step
involves the deposition of a CVD Si0 2 film, etching contact windows to the gate and
to the two n-type regions adjacent to it. Finally, the MOSFET is completed by depo-
sition and patterning of the aluminum interconnect layer.
An IC technology is often specified by "A,its minimum feature size. In an MOS
technology, "Ais the etched polysilicon gate length as shown in the final cross section
in Fig. 2.19(b ). The trend has been to ever-smaller dimensions, with state-of-the-art
production processes at A< 0.5 µmin the mid-1990s.
2.5.5 Geometric Design Rules
In Fig. 2.19(a), the contact nests inside both the oxide and metal patterns. The over-
laps between patterns on different masks are specified in the geometric design rules
for the process. These rules allow for systematic changes in the dimensions of the
device structures, due to lithography, diffusion, or etching. They also incorporate
knowledge of the random variations in these factors and the misalignment tolerances
of the mask making and lithography equipment. There are "proximity" rules that
specify the separation between the edges of patterns on different masks. Finally,
there are rules that define the minimum line and space widths for each mask.
The yield of a particular product in a given process technology will be heavily
influenced by the margins built into the design rules. With an IC containing 106 MOS-
FETs, each with at least three contacts, carefully chosen design rules for the contact
mask nesting are essential for achieving a balance between high yield and a reason-
able die size. Systematic experimentation and modeling of statistical variations are
required to optimize the design rules for a particular technology and product family.
2.5 • Silicon IntegratedCircuit Technology 55
oxide mask
D (dark field) (a) Layout
polysilicon mask
[]TI]
. (clear field)
•
~
contact mask
(dark field))
metal mask
(clear field)
A
(b) A -A Cross Sections
1. Grow 500 nm of thermal Si0 2

and pattern using oxide mask
\ I
p-type silicon wafer
12.Grow 15 nm of thermal Si0 2 I 15 nm-thick gate oxide
/ I
3. Deposit 500 nm of CVD polysilicon

and pattern using polysilicon mask
polysilicon gate
,...........................
8\, / 1@!··················1
:::::::::::::::'.::::::::\0:::::::::::::::::::::'.:::::::::::00::::::::::::::::::::::::::iij:;:,:::::::::::::::::
I 4. Implant arsenic and anneal I - no effect on A-A cross section
5. Deposit 600 nm of CVD Si0 2

and pattern using contact mask
....................
......................
································· ......................
......................
.
····················································································
···················································································
.................................................................................
.................................................................................
~ Figure 2.19 (a) Layout and (b) A-A cross sections for a simple MOSFET
oxide mask
polysilicon mask
[ill]
. (clear field)
•
~
contact mask
(dark field)
metal mask
(clear field)
6. Sputter 1 µm of aluminum and

pattern using metal mask
aluminum
interconnect
,,_ _____________ ...................

.....................
.....................
_..·······················
....................................................................................
..................................................................................
.................................................................................
A I-----------'·=··,,,···,,,··,,,···,,,··,,,···,,,··,,,···,,,···,,,··,,,···,,,··,,,··,,,···,,,··,,,···,,,··,,,···,,,··,,,···,,,···,,,··,,,···,,,··,,,···,,,··,,,···
------1 A
(b) Selected B-B Cross Sections
3. Deposit 500 nm of CVD polysilicon polysilicon gate

and pattern using polysilicon mask 15 nm-thick gate oxide
14.Implant arsenic and anneal I

n-type (As)
\_~>~ \
p-type
> Figure 2.19 (cont.) (a) Layout and final A-A section and (b) selected 8-8 cross
sections for a MOSFET.
oxide mask
F:::J polysilicon mask
Ed (clear field)
•
~
contact mask
(dark field)
metal mask
(clear field)
(b) Selected B-B Cross Sections
5. Deposit 600 nm of CVD Si0 2

and pattern using contact mask
CVDSi0 2
p-type
6. Sputter 1 µm of aluminum and

pattern using metal mask
aluminum
interconnect
p-type
> Figure 2.19 (cont.) (a) Layout and (b) selected 8-B cross sections for the simple
MOSFET.
IC RESISTORS
Now that you have a basic understanding of IC technology, we are prepared to model
our first IC device-a silicon resistor. First we must derive Ohm's Law from the fun-
damental concept of drift current density. We will then re-express the resistance in a
form that will be more useful for IC design.
2.6.1 Ohm's Lavv
Figure 2.20 is a schematic view of a slab of n-type silicon with two sides having met-
allized contacts. A DC voltage Vis applied across the contacts. What is the magnitude
of the resulting current I? Our first observation is that there is an electric field E in
the silicon with a magnitude given by
v (2.58)
E=L
where L is the length of the slab in Fig. 2.20.
The electric field points in the direction of the x coordinate since it points from
the contact with the higher potential (the nearer one in Fig. 2.20) to the contact with
the lower potential. This electric field will cause electrons and holes in the silicon slab
to drift, with the resulting drift current density being given by
(2.59)
in which we have assumed that the electric field is low enough that velocity saturation
is not an issue. There is no concentration gradient along the slab, since the mobile
carriers are supplied and removed by the metal contacts. As a result, diffusion is
ignored in Eq. (2.59). Since the current I [units: A] is the product of the current den-
sity J [units: A/cm 2] and the cross sectional area of the slab A= Wt, we can find an
equation relating current and voltage
I = J · A = (qnµn + qpµP) (f)(Wt)
metal contact 1-
\
v
+
,_____ w-------- ----1
>-Figure 2.20 n-type silicon slab with length L, width W, and thickness t. A voltage
Vis applied across two metal contacts at opposite ends.
2.6 • IC Resistors 59
Ohm's Law states that I= V/R and therefore, we can identify the resistance R of
the slab from Eq. (2.60)
(2.61)
The first term on the right-hand side of Eq. (2.61) is defined as the resistivity, p
[units: Q cm]:
(2.62)
Substituting for the resistivity, Eq. (2.61) reduces to a simple expression
(2.63)
In order to evaluate the resistivity, we must substitute the electron and hole con-
centrations n andp. The slab in Fig. 2.20 was assumed to be n-type with a donor con-
centration of Nd. Although the slab is not in thermal equilibrium, the carrier
concentrations are not significantly perturbed from their equilibrium values by their
constant drift across the slab. Substituting n= Nd and p = n/I Nd for then-type slab in
Eq. (2.62), the resistivity Pnof n-type silicon is given by
(2.64)
In other words, we can neglect the contribution of minority carriers in the drift
current density since their concentration is orders of magnitude smaller than the
majority carriers. For a p-type silicon region, holes are the majority carriers and the
resistivity Ppis
(2.65)
For doping concentrations of 1013 cm-3 to 1019 cm-3 , the resistivity of silicon
ranges from about 500 Q-cm to 5 mQ-cm.
2.6.2 Sheet Resistance
Integrated-circuit resistors are fabricated from regions of one doping type in a sub-
strate of opposite type. Figure 2.21 shows the layout and cross section of an n-type,
ion-implanted IC resistor, that is fabricated using the process illustrated in Fig. 2.18.
Metal interconnects make electrical contact at each end of the implanted region
through contact windows etched through the deposited oxide film. The rectangular
central portion of the implanted region has length L, width W, and thickness t. The
design rules require that the ends of then-type region are widened to allow room for
the contact windows.
If a voltage is applied across the contacts to the resistor, the resulting electric
field will cause a lateral electron drift current density as shown in Fig. 2.21(b ). As we
will see in Chapter 3, ifwe apply positive voltages to both contacts with respect to the
substrate, there is negligible current across the n-p junction between the n-type
region and the p-type substrate. Hence, then-type region is electrically isolated from
the substrate. This simple structure hints at some of the difficulties in integrated cir-
cuit design, in that a p-n junction is the unwanted by-product of making an ion-
implanted resistor. We will revisit the IC resistor in Example 3.6 in Chapter 3 and
develop a more accurate model for it, after we have studied the reverse biased p-n
junction.
From the previous section, the resistance of the rectangular region is given by
(2.66)
In IC design, it is very convenient to re-express Eq. (2.66) in a form that separates

those parameters determined by the fabrication process (mobility µn, donor concen-
tration Nd, and thickness t of the implanted layer) from those that are determined by
D oxide mask
(dark field)
• contact mask
(dark field) ~
metal mask
(clear field)
r~
1· 2• 3• 4,
t
--5:w6: -7: 8•
A ' '
I
I '
l '
'
'I
L
(a)
t contact
deposited oxide
drift current - n-type region

thermal oxide p-type substrate
(b)
),,,,Figure 2.21 (a) Layout (b) cross section of an n-type resistor having LI W= 9 that
is fabricated using the process sequence described in Fig. 2. 18. Arrows show
direction of electron drift current density for the case of a higher voltage on the left-
hand contact.
the layout (length L and width W of the resistor). Collecting terms in Eq. (2.66)
accordingly
(2.67)
r
in which we have defined the sheet resistance R 0 = (q µn Nd t 1. The ratio oflength to
width is also defined as N 0 = (L/W), the number of "squares." The units for sheet
resistance are by convention Q/o, pronounced "ohms per square." Of course, the
number of "squares" is dimensionless. The layout in Fig. 2.21(a) illustrates the
"square" concept by dividing the central portion into 9 segments of length W. In prac-
tical cases, the length L is often an integer multiple of the width Wand N 0 will be an
integer.
For average doping levels of 1015 cm- 3 to 1019 cm- 3 and a typical layer thickness of
t = 1 µm, the sheet resistance ranges from 50 kQ /D to 10 Q /D. For other IC materials,
heavily phosphorus-doped (n+) gate polysilicon films are typically 500 nm thick and
have R 0 approximately 20 Q /D. Aluminum interconnections are 1 µm in thickness
and have a sheet resistance of around 30 mQ /D-far lower than for the polysilicon or
implanted silicon layers.
Why is sheet resistance such a useful parameter to the IC designer? The answer
is that the parameters of the fabrication process such as the resistor thickness, dop-
ing concentration, and mobility are all "givens" for an IC designer. The layout
shown in Fig. 2.21(a) is all that is under the IC designer's control in a particular pro-
cess technology.
Thus far, we have neglected the contributions of the contact regions at the ends of
the resistor to the total resistance. The details of how the current spreads out from the
metal-silicon contact determine the resistance for a particular contact layout. We will
neglect any additional resistance due to the metal-silicon junction not being an ideal
ohmic contact. For the "dogbone" style contact shown in Fig. 2.22(a), each contact
region adds 0.65 squares. It is convenient to express the additional resistances in units
of squares, so that they can be counted into the total N 0 • In Fig. 2.22(b ), the effect of
the 2-D distribution of current density at a right angle bend is that the "corner square"
contributes only 0.56 squares to the total N 0 • Accurate accounting for corners is
important, since large-value resistors are typically folded in order to fit compactly into
the layout. Example 2.5 illustrates the concept of sheet resistance for resistor layout.
-I 3W 1-
-lwl- i w
w
~3W
~T 0.56 squares
(a) (b)
~ Figure 2.22 Effective fractional square for (a) a "dogbone" contact area and (b)
a corner. After R. C. Jaeger, Introduction to Microelectronic Fabrication, Addison-
Wesley, 1988.
> EXAMPLE 2:.li IC Resistor Layout

(a) Given that a dose Qd = 1012 cm-2 of arsenic is implanted into p-type silicon, with
a junction depth of xj = 400 nm, estimate the sheet resistance of this n-type layer.
I SOLUTION I
The average arsenic concentration in the implanted layer is the thickness of the
layer (in cm) divided into the dose (per cm 2)
Q 12
Nd = __.!!.= I x10_5 = 2.5 x 1016cm-3
xj 4 x 10 ·
From Fig. 2.8, the electron mobility for the average concentration is µn = 1000
cm 2/Vs. Substituting into the definition of sheet resistance in Eq. (2.67), we find
that
-----------.,-----
( 1.6 x 10-
19
1000) (2.5 x 10
) (
16
) (4 x 10-
5
)
=6 kQ/D
In letting xj = t in finding R 0 , we assume there is no penetration of the n-type
region by the charged layer that forms at the p-n junction. This approximation is
reasonable so long as the background acceptor concentration in the substrate,
Nm is much less than the donor concentration Nd in the implanted layer. We will
quantify this approximation in Chapter 3. We have also made this same assump-
tion in letting the electron concentration n = Nd, rather than Nd - N 0 • For a typi-
cal substrate doping, these approximations lead to small errors.
(b) For this n-type layer, find the length L for a straight resistor (no corners) of value
R = 56 kQ. The minimum width W = 2.5 µm and the two contact areas have the layout
shown in Fig. 2.22( a)
I SOLUTION I
First, the total number of squares needed is
N 0 = R!R 0 = 56kQ I (6 kQ/D) = 9.3 D
Subtracting off the contributions of the contacts, the rectangular portion of the
resistor has
(L!W) = N 0 - 2 (0.65) = 9.3 - 1.3 = 8

Therefore, the length L = 8 x 2.5 µm = 20 µm.
2.6.3 Statistical Variations in IC Resistors

In order to use the IC resistor as a circuit element, its variation from lot to lot and
over temperature must be understood. We will introduce a simple way to quantify the
effect of manufacturing variations on the resistance, but will leave the subject of tern-
perature effects to a more advanced course. This section is applicable to other device
structures since the sources of variation are fundamental to IC manufacturing.
From Eq. (2.66), the resistance is the product of the sheet resistance and the
number of squares, each of which contains parameters or dimensions that vary from
chip to chip:
R cN:µnr)(t)
= (2.68)
Since a background in probability and statistics is not assumed, the variation of a

particular quantity, such as the doping concentration Nd, will be expressed simply as
(2.69)
where Nd is the average doping and EN is defined as the normalized uncertainty. It

is convenient to use the normalized uncfertainty rather than the absolute uncertainty
in the particular quantity. We can determine the range of variation by multiplying the
average by the normalized uncertainty. For example, a doping concentration defined
17 3
by an average of Nd = 10 cm- and a normalized uncertainty of EN = 0.025 means
that d
(2.70)
Returning to Eq. (2.68), we substitute for each term that is subject to manufac-
turing variations:
(2.71)
It is reasonable that the average resistance should be
(2.72)
However, how do we predict the normalized uncertaintyfR given the normalized

uncertainties in the parameters and dimensions? Applying an important result from
statistics, the normalized uncertainty in the resistance is the square root of the sum of
the squares of the normalized uncertainties of each variable
(2.73)
This result is valid when the variation of one variable is independent of those of
the other variables and when all variables follow the Gaussian, or normal, probability
distribution. In practice, Eq. (2.73) is a good first-cut estimate for the case where the
normalized uncertainties are relatively small (< 0.1). This "statistical" approach is
convenient and we will use it to estimate the effects of process and layout varia-
tions-even though we are not going to verify the underlying assumptions.
We now turn to the origin of the variations in the terms in Eq. (2.71). The ion
implanter will largely determine the run-to-run control on the doping concentration.
Mobility variations are due to random local variations in crystal defects. Note that
there is also coupling between the mobility and the doping concentration, as shown
in Fig. 2.8. This coupling will be ignored in our first-cut estimate of uncertainty in Eq.
(2.73). The thickness t of the resistor is subject to variations in the time and local
wafer temperature during the furnace annealing process.
The lateral dimensions of the resistor are affected by random linewidth varia-
tions from the pattern transfer processes, as shown in Figure 2.23. Defining the uncer-
tainty on one edge as 0/2, the width and length of the resistor are
(2.74)
where we have added the uncertainties from each edge. The width and length can be
rewritten in the form
-(
W = W 1± W b) = W- (1 ± Ew) and (2.75)
-( fb)= L- ( 1 ±
L = L 1± EL) . (2.76)
From Eqs. (2.75) and (2.76), the normalized uncertainties in the width and length
are inversely proportional to their average values:
Ew = b!W and EL = b/L. (2.77)
Therefore, the designer has control over the normalized uncertainties in width
and length and can make trade-offs between the resistor area (the product of width
and length) and its normalized uncertainty. It is common for the length to greatly
l_o12
> Figure 2.23 Highly magnified top view of the resistor showing the uncertainty
o/2 in the edges due to random variations in the photolithography and etching
processes. Note that the sketch only shows the ends of the relatively long and narrow
resistor.
exceed the width, in which case the normalized uncertainty in the length becomes
negligibly small, according to Eq. (2.77)
What are the magnitudes of the process uncertainties that enter into Eq. (2.73)?
To answer this question, we must first specify the collection of samples used to eval-
uate the average and variation. The variation is smallest for structures that are nearby
on the same chip, since they are subjected to nearly identical processes. If an entire
wafer is considered, then the spread in parameters and dimensions increases. The
uncertainties increase further when a wafer lot is evaluated and are largest when
measurements are made on samples from many lots. Finally, the IC process equip-
ment and the organization and discipline of the workforce are major factors in setting
the baseline levels of process uncertainties.
The statistics on process variations are highly proprietary for each manufacturer,
needless to say. However, some rough numbers are useful for gaining an appreciation
for the limits of IC technology. Nominally identical, adjacent ion-implanted resistors
can have a normalized uncertainty as low as 10-3.With careful attention to the layout,
resistor ratios can have a similar level of precision. A further benefit of adjacent resis-
tors is that they are at nearly the same temperature. As a result, their resistances track
with changing ambient temperature. When the set of resistors is expanded to include
samples from a wafer, a lot, or a series of lots, the normalized uncertainty increases
by a factor of 10 to 100-not including the significant variation over temperature. As
a result, the precision is too poor to allow a designer to use an IC resistor for circuits
where its absolute value must be tightly controlled. The absence of precision inte-
grated resistors has been a major motivator of new approaches to circuit design.
)ii-. IXAMPU:
2~6 IC Resistor Process Variations
An ion-implanted resistor with the cross section shown in Fig. 2.21 is designed to
have a resistance R = 20 kQ. The process parameters and their normalized uncertain-
ties over a series of lots are: Nd= (4 x 1017 cm-3 )(1 ± 0.04), µn = (450 cm2Ns)(l ±
0.02), t = 550 nm(l ± 0.035).
(a) Find the average sheet resistance and its normalized uncertainty.
I SOLUTION I
Substituting the average values for the parameters into the definition of sheet
resistance in Eq. (2.67), we find that
R D -- -19 17 -3 2 -4
630Q/D.
( 1.6 x 10 C) ( 4 x 10 cm )( 450 cm Ns)(0.55 x 10 cm)
The normalized uncertainty in the sheet resistance is given by
2
EN
d
2
+ Eµ + E 1
n
2
= J (0.04) 2 + 2
(0.02) + (0.035)
2
= 0.06
(b) Over several wafer lots, the linewidth edge variation is o= 0.15 µm. For nom-
inal resistor widths of W = 2 µm and W = 4 µm, find the normalized uncertainty
for the resistor.
I SOLUTION I
The number of squares required is:
20kQ
= 630 Q/sq. = 31.75
Due to the large number of squares, the uncertainty in the length can be
neglected. According to Eq. (2.77), the normalized uncertainty in the resistor
width is
0.15 µm
Ew = W = 0.075 (W = 2 µm) and Ew = 0.038 (W = 4 µm)
Therefore, the normalized uncertainties of the resistor for the two widths are
J(0.06) 2
+ (0.075)
2
0.096 (W = 2 µm) and
ER = Jc!0
+ £~ = J(0.06) 2
+ (0.038)
2
= 0.0625 (W = 4 µm)
The nominally 20 kQ IC resistor has an absolute uncertainty of nearly 2 kQ for

the narrow width layout and about 1.25 kQ for the wider layout. The penalty for
the wider layout is an increase in area by a factor of 4 since the length must also
be doubled to keep the number of squares at 31.75.
For large resistor values, it is desirable to fold the resistor in order to make a
more compact IC structure. We conclude the chapter with a design example which
explores the geometry involved in designing for a square layout.
Given an n-type layer with a sheet resistance R 0 = 1 kQ/o, we want to lay out a folded
resistor with R = 75 kQ. The geometric design rules for the n-type region specify that
the minimum width is W, = 2 µm and that the minimum gap between n-type regions
is Wg= 3 µm, as defined in the sample layout in Fig. Ex2.7. The contact regions each
contribute 0.80. In order to make the resistor easy to pack into the IC layout, we want
to fold the resistor.
What is the segment length Ls (defined in the sample layout) that results in a nearly
square resistor layout? How much area does the resistor require? You may neglect
the portions of the contact areas outside the folded segments in finding the resistor
area.
I SOLUTION I
Assuming that the resistor consists of Ns segments, the total number of squares
N 0 is
l_
intermediate segment ws
T
intermediatesegment
'ic{• neglect in finding area
firnt~·-·· Ls - 2Ws ---------i~1

comer
square
~
0.8 squares
> Figure Ex2. 7 Layout of a folded IC resistor with offset contacts.
N 0 =2(0.8) +2 ( Ls-li~J
W, + (Ns-2) (Ls]
Ws + (Ns-1) (2(0.56) +;,WJ
in which the first two terms are for the first and last segments (with the contacts),
the third term is for the intermediate segments of length Ls, and the final term is
for the folds (with the two corner squares contributing 0.56 squares each). Using
the minimum width W, = 2 µm and the minimum gap ~ = 3 µm, we find
N 0 =l.6+Ls-4+ (
(Ns-2))
2 Ls+ (Ns-l) (2.62) =
(L
-:f+2.62 )Ns-5.02
In order to achieve a square layout, Fig. Ex2.7 indicates that
L 5 +2W,= (N 5 -l)(W,+~)+W,= (N5 -1)(5µm)+2µm
in which the left hand side is the total length of one segment and the right hand
side is the length of the side with the folds. Solving for the number of segments,
we find that
L s + 2Ws + Wg Ls + 7
Ns =
W,+wg 5
The total number of squares needed is N 0 = RI R 0 = 75 kQ I (l kQ/D) = 75. Substi-

tuting for N, in the expression for the total number of squares, we find a qua-
dratic equation for the segment length
Ls
75 = [ 2 + 2.62
](L
~
+7)
- 5.02, which simplifies to
L~ + ( 12.24) Ls - 763.32 = 0
Taking the positive root, we find Ls = 22.2 µm. The number of segments needed
IS
N s = Ls+ 7 - 22.2 + 7 = 5 84
5 - 5 .
One approach is to use an integer number of folds so we round up to Ns = 6. The

length of each segment must be modified in order to maintain N 0 = 75 using the
new number of folds
N0 = 75 = [ ~s + 2.62 )JV,
- 5.02 = [ ~s + 2.62) 6 - 5.02 ~ Ls= 21.4 µm
The final layout of the folded resistor with Ls= 21.4 µmis not a perfect square,
since the segments are 25.4 µm long (counting the folds) and the edge with the 6
folds is 27 µm long. An alternative to forcing Ns to be an integer would be to
adjust the lengths of the first or the last segments, in order to make the overall
layout closer to a square. In practice, the integer number of folds is close enough
to a square layout. The area of the resistor is
A= (Ls+2Ws) [(Ns-1) (Ws+Wg) +Ws] = (21.4+4) (5·5+2) 686µm 2
~ SUMMARY
This chapter has provided a foundation in two subjects that are essential for inte-
grated device and circuit modeling: the physics of holes and electrons in silicon and
the technology of IC fabrication. Key concepts that you should understand are:
+ The bond model and its description of an electron and a hole in silicon.
+ The concept of thermal equilibrium and the mass-action law.
+ Hole and electron concentrations for silicon with donors, with acceptors, or with
both donors and acceptors.
+ Drift velocity and drift current density.
+ Diffusive transport and diffusion current density.
+ The total electron and hole current densities.
+ Lithography and pattern transfer.
+ Masked ion implantation of impurities.
+ Cross sections through device structures from the layout and the sequence of fab-
rication steps (process).
+ Resistivity and sheet resistance.
+ Ion-implanted IC resistors and their layout, including equivalent squares for con-
tact regions and corners.
Chapter2 • Problems 69
+ Normalized uncertainty and its calculation for a function of several variables,
given their uncertainties.
> FURTHER READING
Semiconductor Physics
1. C. G. Fonstad, Microelectronic Devices and Circuits. McGraw-Hill, 1994, Chapters 1 and
2. About the same level.
2. R. F. Pierret, Semiconductor Fundamentals, 2nd Ed. Modular Series on Solid State
Devices, Vol. I, G. W. Neudeck and R. F. Pierret, eds., Addison-Wesley, 1988,
Chapters 1 and 3. Somewhat more advanced.
IC Technology
1. W. Maly, Atlas of VLSI Technologies, Addison-Wesley, 1986. Several step-by-step pro-
cess sequences provide a good intuitive understanding of the connection between layout,
process, and device cross sections.
2. R. C. Jaeger, Introduction to Microelectronic Fabrication, Modular Series on Solid State
Devices, Vol. V, G. W. Neudeck and R. F. Pierret, eds., Addison-Wesley, 1988, Chapters
1 and 2. Early chapters provide a good overview of IC fabrication.
> PROBLEMS
Note: Assume that the sample temperature is 27 °C , unless otherwise noted in the
problem statement.
EXERCISES
E2.l A silicon wafer is doped with donors at a concentration of Nd= 1015 cm- 3•
(a) What is the electron concentration n 0 (cm- 3) at room temperature?
(b) What is the hole concentrationp 0 (cm- 3 ) at room temperature?
E2.2 A silicon wafer is doped with acceptors at a concentration of Na= 1014 cm-3 •
(a) What is the electron concentration n 0 (cm- 3) at room temperature?
(b) What is the hole concentrationp 0 (cm- 3 ) at room temperature?
E2.3 Given that a silicon wafer is doped with acceptors with Na= 1014 cm- 3 . We
add a donor concentration of Nd= 7.5 x 1013 cm 3 to a particular region.
(a) What type is the region (nor p)?
(b) What is the electron concentration n 0 ( cm- 3) in the region?
(c) What is the hole concentrationp 0 (cm- 3) in the region?
E2.4 Given that a silicon wafer is doped with donors with Nd= 5 x 1017 cm- 3 • We
add an acceptor concentration of Na= 5.5 x 1017 cm 3 to a particular region.
(a) Which carrier is the majority carrier?
(b) What is the hole concentrationp 0 (cm- 3) in the region?
(c) What is the electron concentration n 0 (cm- 3) in the region?
E2.5 For a wafer doped with acceptors at Na= 1014 cm- 3 , what concentration of
donors must be added in order to reduce the hole concentration by 90%?
E2.6 In a subatomic particle detector, extremely low doped silicon substrates are
needed. What donor concentration is required in order to have an electron concen-
tration n0 = 8 n;?You will need to account for thermal generation for this problem.
E2.7 In modern VLSI processes, the dimensions of device structures are
approaching 1 µm 3 . For a donor concentration Nd= 5 x 1014 cm-3 , find the number of
electrons and the number of holes in this volume. Comment: your answer will show
one reason why there is a tendency for doping concentrations to increase as device
dimensions are scaled into the submicron region.
E2.8 In a silicon region with donor concentration Na= 1016 cm- 3 , we apply an elec-
tric field in the +x direction with a magnitude of 103 V/cm.
(a) What is the electron drift velocity (magnitude and sign)?
(b) What is the electron drift current density (magnitude and sign)?
(c) How long is required for an electron to drift 1 µm, on average?
(d) How many collisions occur while it is drifting? You can assume that the
collision time for the drifting electron is 'Cc= 0.1 ps.
E2.9 Repeat E2.8 with a donor concentration of Nd= 1015 cm- 3 .
E2.10 In a silicon region with acceptor concentration Na= 1018 cm- 3 , we apply an
electric field in the +x direction with a magnitude of 2 x 103 V/cm.
(a) What is the hole drift velocity (magnitude and sign)?
(b) What is the hole drift current density (magnitude and sign)?
(c) How long is required for a hole to drift 1 µm, on average?
(d) How many collisions occur while it is drifting? You can assume that the
collision time for the drifting hole is 'Cc= 0.05 ps.
E2.11 A silicon region is doped with acceptors at Na= 1015 cm- 3 and with donors at
Nd= 8 x 1014 cm- 3 .
(a) What type is the silicon (nor p)?
(b) What electric field must be applied for the magnitude of the drift veloc-
ity of the majority carriers to be 4 x 106 cm/s?
(c) What is the drift current density of the majority carriers for the case of
the electric field in part (b )?
E2.12 By optical means, we set up a minority hole concentration gradient across a
2 µm-wide silicon region that is given by:
where x is the coordinate in the direction of the concentration gradient. The donor
concentration in the region is Nd= 1016 cm-3 .
(a) Find the hole diffusion current density in the region.
(b) How long does a hole take to diffuse across the 2 µm-wide region?
E2.13 The minority hole concentration gradient across a 2.5 µm-wide silicon region
is given by:
p(x) = (5 x 1018 cm- 4 ) (2.5 µm - x)

where x is the coordinate in the direction of the concentration gradient. The donor
concentration in the region is Nd= 5 x 1017 cm-3 •
(a) Find the hole diffusion current density in the region.
(b) How long does a hole take to diffuse across the 2.5 µm-wide region?
E2.14 In a silicon sample, we establish a minority electron concentration gradient
over the region x :2:0 that is given by
The acceptor concentration is Na= 1017 cm-3 •

(a) Find the magnitude and sign of the electron diffusion current density at
x=O.
(b) Plot the electron diffusion current density over the interval O< x < 10
µm.
E2.15 The gradient in minority hole concentration in a silicon sample over the
region x :2:0 is given by
The donor concentration is Nd= 5 x 1015 cm-3 •

(a) What is the hole concentration at x = O?
(b) What is the hole diffusion current density at x = O? Provide an explana-
tion for the apparent discrepancy between parts (a) and (b).
(c) Plot the hole diffusion current density over the interval O~ x ~ 5 µm.
E2.16 By continuous improvements in fabrication technology, three generations of
a microprocessor IC have areas that are given in the table below. The cost of process-
ing a 20 cm-diameter wafer increases with each generation, due to the need for more
expensive process equipment.
,, .... :!!:!
1st Generation 2 x 2 cm 2 $1750
2
2"d Generation 1.5x1.5cm $2000
2
3'd Generation 1 x 1 cm $2250
The yield of good chips for each generation is 75%.

(a) Find the number of good chips on a 20-cm wafer for each generation.
Do not consider the complication of dice that intersect the edge of the
wafer-an estimate of the chip count per wafer is adequate.
(b) What is the manufacturing cost per good microprocessor chip for each
generation?
E2.17 For an implant dose Qa = 1018 cm-2 of acceptors and a post-anneal thickness
oft= 0.5 µm, what is the acceptor concentration in the layer?
E2.18 Given a silicon wafer with an acceptor concentration of Na= 1016 cm- 3 . We
would like to use ion implantation to dope a surface region so that the average accep-
tor concentration is Na= 1017 cm- 3 . If the post-anneal thickness of the region is t = 0.1
µm, what dose is needed?
E2.19 Starting with a silicon wafer with a donor concentration of Nd= 1017 cm- 3 , we
would like to use ion implantation to convert a surface region top-type, with an aver-
age acceptor concentration Na= 3 x 1017 cm- 3 . What implant dose is needed, if the
post-anneal thickness of the implanted region is 0.25 µm.
E2.20 The five-step process for a two-layer structure made from silicon dioxide and
polysilicon is as follows:
1. Grow 0.5 µm of silicon dioxide (Si0 2)
2. Pattern using the oxide mask.
3. Grow 500 A of Si0 2 .
4. Deposit 1 µm of polysilicon.
5. Pattern using the polysilicon mask.
The CAD layout for this structure is shown in Fig. E2.20.
(a) Sketch the cross section A-A from the process sequence, assuming that
the oxide mask is dark field and the polysilicon mask is clear field.
(b) Repeat part (a), with the oxide mask clear field and the polysilicon mask
dark field.
(c) Repeat part (a), with the oxide mask dark field and the polysilicon mask
dark field.
(d) Repeat part (a), with the oxide mask clear field and the polysilicon mask
clear field.
E2.21 The layout in Fig. E2.21 is for the ion-implanted resistor process outlined in
Fig. 2.18.
(a) Sketch the cross section A-A from the layout in Fig. E2.21.
(b) Sketch the cross section B-B from the layout in Fig. E2.21.
f-3µm-l
oxide
~ polysilicon
Ed
A
>-Figure E2.20
oxide mask
D (dark field)
•
~
contact mask
(dark field) A
metal mask
(clear field)
A
> Figure E2.21
E2.22 The layout in Fig. E.2.22 is for the simple MOS transistor process described
in Fig. 2.19 (b).
(a) Sketch the cross section along A-A.
(b) Sketch the cross section along B-B.
(c) Sketch the cross section along C-C.
(d) Sketch the cross section along D-D.
E2.23 The parameters for an ion-implanted region are: arsenic dose Qd = 1012 cm-2 ,
thickness after anneal t = 1.1 µm. The background doping of the p-type substrate is
1014cm-2.
(a) What is the average arsenic concentration in the ion-implanted layer?
(b) What is the sheet resistance in units of 0/D.
c
oxide mask
D (dark field)
polysilicon mask
Lill(clear field)
•
~
contact mask
(dark field)
metal mask
(clear field)
c
> Figure E2.22
E2.24 After discovering that boron was implanted at a dose of Qa = 1011 cm- 2
instead of phosphorus, you try to salvage the lot by compensating the layer back to
n-type by means of implanting a heavier dose of phosphorus.
(a) Assuming that the layer thicknesses for the phosphorus and boron are
the same at t = 1 µm after annealing, what phosphorus dose is necessary
to achieve the desired electron concentration?
(b) What is the sheet resistance for the compensated layer? Compare your
answer to what the sheet resistance would have been if the original
mistake had not been made.
(c) From your answer in part (b ), what phosphorus dose is needed to
achieve the original sheet resistance?
E2.25 Given a 1 µm-thick implanted layer and a maximum average dopant concen-
tration of 1 x 1020 cm- 3 in this layer.
(a) What implant dose is needed to reach the maximum concentration?
Neglect the background doping in the substrate.
(b) For an n-type layer, what is the minimum sheet resistance? Note that
the mobility degrades with increasing doping concentration, as shown in
Fig. 2.8.
(c) For a p-type layer, what is the minimum sheet resistance? Note that the
mobility degrades with increasing doping concentration, as shown in
Fig. 2.8.
E2.26 Given a sheet resistance of 100 Q/D in a 1 µm-thick, n-type implanted layer.
(a) What is the average donor concentration in the layer? Neglect the back-
ground doping in the substrate.
(b) Sketch the layout for a 2.5 kQ resistor, using the "dogbone" style con-
tacts shown in Fig. 2.22(a).
E2.27 Given a sheet resistance of 250 Q/D in a 1 µm-thick, p-type implanted layer.
(a) What is the average acceptor concentration in the layer? Neglect the
background doping in the substrate.
(b) Sketch the layout for a 2 kQ resistor, using the "dogbone" style contacts
shown in Fig. 2.22(a).
E2.28 Given that the normalized uncertainties are: £N = 0.025 in the donor con-
centration and £ 1 = 0.02 in the layer thickness. The un~ertainty in the mobility can
be neglected.
(a) For an average donor concentration of 1016 cm- 3 , find the variation in
Nd.
(b) For an average layer thickness of 500 nm, find the average sheet resis-
tance and its normalized and absolute uncertainties.
E2.29 Given that the uncertainty in the lateral dimension of an ion-implanted resis-
tor is 8 = 500 A and that the sheet resistance is 100 Q/o.
(a) Neglecting the uncertainty in the sheet resistance, what are the normal-
ized uncertainty and the variation in a 5 kQ resistor for a resistor width
W=2µm.
(b) How wide must the resistor be made in order to have the normalized
uncertainty less than 0.01 (1 % )?
E2.30 For a doped polysilicon film, the sheet resistance is 25 Q/o for a 5000 A-thick
film. If the thickness of the film is controlled to ±2.5% over the wafer, what is the vari-
ation in the resistance of a 20 kQ resistor?
PROBLEMS
P2.1 A sample of silicon is doped with phosphorus with a concentration of

1015 cm-3.
(a) What is the average volume occupied by an electron in this sample, in
µm3?
(b) What is the average volume occupied by a holes in this sample, in µm 3?
P2.2 A region of a silicon wafer is doped with three impurities:
arsenic: 10 16 cm- 3, boron: 1.15 x 10 16 cm- 3, phosphorus: 2.5 x 10 15 cm-3.
Find the electron and hole concentrations n0 and p 0 •
P2.3 A sample of silicon is doped with Nd= 1.1 x 10 13 cm- 3 and Na= 1 x 10 13 cm-3.
(a) Find the electron and hole concentrations n0 and p 0 at room tempera-
ture.
(b) Show that the charge density p = 0 (exactly).
P2.4 The intrinsic concentration in silicon n; has a strong temperature depen-
dence, as shown by these measurements over the temperature range for automotive
electronic applications:
-55 2.24 x 106
0 1.42 x 109
70 2.82 x 1011
125 5.06 x 1012
A course in semiconductor physics shows that the functional dependence of

the intrinsic concentration is
3!2 -6608.7/T
ni = A · T e ,
where Tis in Kelvin.

(a) Find the constant A by fitting the function to the measurements above.
(b) How rapidly the intrinsic concentration varies near room temperature
is important for device modeling. Find the temperature increase (in
degrees Centigrade) needed to double ni and the temperature decrease
needed to halve n;.
P2.5 A sample of silicon is doped with Nd= 10 13 cm- 3. As the temperature
increases, the contribution of thermal generation to the electron and hole concentra-
tions eventually becomes significant. Using the result from P2.4(a) for the intrinsic
concentration as a function of temperature,
(a) Plot no(T) over the temperature range -55° C < T < 125° Con a log-
linear plot.
(b) What is the maximum temperature for whichpjn 0 < 10- 37
P2.6 A sample of silicon is doped with Na= 2 x 1014 cm- 3 and Nd= 9 x 1013 cm- 3 .
Using the result from P2.4(a) for the intrinsic concentration as a function of temper-
ature,
(a) Plot p 0 (T) over the temperature range -55° C < T < 125° C on a log-
linear plot.
(b) What is the maximum temperature for which n) p 0 < 5 x 10-4?
P2.7 Figure P2.7 is a top view of a silicon region which has metal electrical con-
tacts on two sides, across which a voltage ~ = 2 V is applied. The silicon layer is
doped with arsenic at a concentration of 1013 cm- 3 and is 2 µm thick and you can
assume room temperature. You should use the appropriate mobility for the given
dopant concentration.
(a) Find the value of the electric field Ex in the silicon.
(b) Find the value of the drift velocity of electrons and of holes in the sili-
con. How long does is take an electron, on average, to drift across the
silicon region.
(c) Find the drift current densities for electrons and for holes. What is the
percentage error in the total current I if the minority carrier drift current
density is neglected?
(d) What is the value of the resistance R in ohms?
P2.8 If the arsenic concentration is increased to 1018cm- 3 , repeat P2.7.
P2.9 Plot the maximum current !max across the silicon region in Fig. P2.7 as a func-
tion of doping concentration from Nd= 1013 cm- 3 to Nd= 1018 cm- 3 , using a log-log
scale. You can assume that the saturation velocity of electrons is vsat = 107 cm/s and
should include the dependence of mobility on doping concentration. Plot the applied
voltage ~,max at which !max is reached as a function of doping concentration.
P2.10 Figure P2.10 is a top view of a 1 µm-thick silicon region with variable accep-
tor concentration, across which a voltage ~ is applied through metal electrical con-
tacts. Assume that the hole saturation velocity is vsat = 107 emfs.
(a) For an applied voltage ~ = 1 V, plot the electric field E(x). Hint: you
will find it helpful to consider the silicon region as three series resistors.
(b) What is the maximum current !max through this structure?
(c) What voltage ~ax must be applied to reach this maximum current?
(d) Plot the electric field E(x) corresponding to ~ax-
(2 µm thick) metal
contact
6µm
------------- 24 µm --------
~ Figure P2. 7
(1 µm thick) 2 µm
I
-----
4 6 10 x (µm)
> Figure P2.10
P2.ll Figure P2.ll is a top view of a 1 µm-thick silicon region with variable donor
concentration. A voltage ~ is applied through metal electrical contacts.
(a) For an applied voltage~= 1 V, what is the current I? You may find it
helpful to view the silicon region as two resistors in parallel.
(b) Plot the current as a function of voltage over the range -20 V < ~ < 20
V. You can assume that the saturation velocity of electrons vsat = 107 cm/
s and should include the variation of mobility with donor concentration.
P2.12 The top view of a 1 µm-thick silicon region with a varying width is shown in
Fig. P2.12. The donor concentration is Nd= 1017 cm- 3 throughout. You can assume
that 25% of the applied voltage is dropped across the "neck," as long as the satura-
tion velocity is not reached in any part of the region.
(a) Find the ratio of the current density in the wide region to that in the nar-
row region, for the case where ~ = 2 V.
(b) Sketch the electric field E(x) along the center line of the structure, for
the case when~= 2 V. The numerical value in the regions of constant
width should be accurate. It will be helpful to view the silicon region as
three resistors in series, with the voltage drop across the "neck" region
given as 25% of~
(c) What is the maximum current !max through this structure? What is the
applied voltage when !max is reached?
(2 µm thick)
0
x(µm)
> Figure P2.11

(1 µm thick)
neck
metal
contact
x
6µm 1 µm
f-<l •f---- 5 µm --1
I• sµm --1
> Figure P2.12
P2.13 In the slab of silicon region in Fig. P2.13, the minority carrier concentration
is maintained at its thermal equilibrium value at the bottom surface (x = 0.3 µm). The
silicon is doped with acceptors at Na= 1016 cm- 3 .
(a) If the minority carrier concentration at the top surface is 1/10 the major-
ity carrier concentration and we assume a linear gradient from top to
bottom, what is the diffusion current density in the slab?
(b) How long does it take for a minority carrier to diffuse across the slab, on
average?
P2.14 When a gradient in minority carrier concentration is set up in a semiconduc-
tor region, an identical gradient in the majority carrier concentration is quickly estab-
lished, in order to keep the silicon region electrically neutral. This situation is shown
in the carrier concentration distributions in Fig. P2.14.
(a) Find the minority electron diffusion current density.
(b) Find the majority hole diffusion current density.
(c) Since the silicon region is open-circuited, the total current density must
be zero. Find the required drift current density.
(d) Assuming that the contribution of the minority carriers to the drift cur-
rent is negligible (which it is), what electric field is required in the sam-
ple to generate the drift current density needed from part (c).
slab dimensions:
W=L=25µm
t= 0.3 µm
> Figure P2.13

p(x) = (10 16cm- 3) + (10 13cm- 3)(1 -x)

(x in units of µm)
1015
carrier
concentration 1014
(cm- 3)
I~)4 r n(x) = 104 cm- 3 + (10 13cm- 3)(1 - x)

(x in units of µm)
I
103~~~---'-l~~~~I ~~---'-1~~~~1 ~~---'-~~-,~
0.2 0.4 0.6 0.8 x (µm)
~ Figure P2.14
P2.15 The process sequence and layout for making a simple polysilicon microme-
chanical bridge are listed below and in Fig. P2.15. Using IC-based processes, we can
make structures for sensing non-electrical inputs, such as acceleration.
1. Deposit 2 µm of CVD Si0 2 .
2. Pattern with the anchor mask.
3. Implant phosphorus and anneal.
3. Deposit 2 µm of phosphorus-doped polysilicon and anneal.
4. Pattern with the poly mask.
5. Etch for 4 minutes in hydrofluoric acid (Si0 2 etch rate: 1 µm per minute);
rinse wafer in water and dry.
anchor mask f::::1 poly mask

• (dark field) E:J (clear field)
c
A A
~ Figure P2. 15
(a) Draw the cross sections A-A after step 4 and after completion of the
process.
(b) Draw the cross section B-B after step 5 is half completed (2 minutes
of HF etching.)
(c) Draw the cross section C-C after completion of the process.
P2.16 This problem considers the fabrication of a two-level micromechanical struc-
ture, which is accomplished by extending the process in P2.15.
Steps 1-4: same as those in P2.15, with the masks renamed as anchor] and
poly]
5. Deposit 1 µm of Si0 2 .
6. Pattern the oxide with the anchor2 mask.
7. Deposit 1 µm of phosphorus-doped polysilicon and anneal.
8. Pattern the second polysilicon layer with the poly2 mask.
9. Etch for 12 minutes in hydrofluoric acid (Si0 2 etch rate: 1 µm per
minute); rinse wafer in water and dry.
(a) Draw the cross section A-A after step 8.
(b) Draw the cross section B-B after completion of the process.
(c) Draw the cross section C-C after completion of the process.
(d) Draw the cross section D-D after completion of the process.
P2.17 Texturing the silicon surface is a way to increase the surface area of a capac-
itor; the following process and layout are one way to achieve this goal.
• anchor 1 mask
(dark field)
F::1 polyl mask
L.;i;J (clear field)
5µm
B
anchor2 mask
(dark field)
poly2 mask
(clear field)
A ;;;;;;;;;;;;;;;;;;;;;;;;;;::'.'.'.'.'.'.'.'.'.'.'.'.'.'.'.1.11111111111111111111!1111111111iiiiiiii111111
..
.. ..... .... ....
..........
c
> Figure P2. 16

1. Etch 5 µm deep into the silicon substrate using the trench mask*.
2. After stripping the photoresist and cleaning the wafer, grow 500 A of
thermal Si0 2 .
3. Deposit 7500 A of phosphorus-doped polysilicon.
4. Pattern using the polysilicon mask.
(a) Draw the cross section A-A after step 2.
(b) Draw the cross section A-A after step 3.
(c) Draw the cross section B-B after completion of the process.
P2.18 A simple double metallization process is as follows; the mask layout is shown
in Fig. P2.18.
1. Grow 0.5 µm of thermal Si0 2 •
2. Deposit 0.5 µm of phosphorus doped polysilicon and pattern using poly-
silicon mask.
3. Deposit 0.5 µm of CVD Si0 2 and pattern using contact mask.
4. Deposit 1 µm of sputtered Al and pattern using metall mask.
5. Deposit 0.5 µm of CVD Si0 2 and pattern using via mask.
6. Deposit 1 µm of sputtered Al and pattern using metal2 mask.
(a) Accurately sketch the cross section A-A, assuming that the deposited
CVD films conformally coat the surface.
trench mask El
II (dark field) B
polysilicon mask
(clear field)
B .............
.............
.........................................................................................................
.................
.
····························································································································
rn•1rnrn1•1rnrn1•rn111111111rn111•1rn1111•11
rni1•1111 ···················
..........................................................................................................
.........................................................................................................
···················
::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::8:::::::::::::::::
.........................................................................................................................
........
...
...
...
........ ........ ...........
........
.......
......
A A
........ ........
·····························································································································
.............................................................................................................................
·····························································································································
..........................
............ ...·············· ··········
............................................................
.............................................................................................................................
11111iirniii•i1rnrn•irnrni•rni1rn•1rnrni•irnrn1•rn1
.............................................................................................................................
~;g~ ~ ~ ~ ~ ~ ~ !;~;;;;
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~ ~;;;;;;; ;;;1 µm ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~
11111111111111111111111111111rn111rnrn11111111rn1111rnrn111rn1
·····························································································································
.............................................................................................................................
.............................................................................................................................
...............................................................................................................................
1111111111111•11rnrn•1rnrni•rnrnii•11111111•1rnrni•rni
.............................................................................................................................
.............................................................................................................................
...............................................................................................................................
>-Figure P2.17
'A reactive-ion etch is used for the trench etch into the silicon substrate, which is optimized to yield vertical
walls in silicon with negligible mask undercutting.
~ polysilicon mask .........................

·························
Ed (clear field) .........................
..........................
..........................
.........................
·························
contact mask
II (dark field)
metall mask
~ (clear field)
via mask
(dark field)
metal2 mask
(clear field)
A
> Figure P2. 18
(b) Accurately sketch the cross section B-B, assuming that the deposited
P2.19 Using the double metal process from P2.18, consider the layout in Fig. P2.19.
(a) Accurately sketch the cross section A-A, assuming that the deposited
(b) Accurately sketch the cross section B-B, assuming that the deposited
CVD films conformally coat the surface. Note that the rough topogra-
phy will make lithography difficult. The geometric design rules can pre-
vent this situation by prohibiting the stacking of metall and metal2.
P2.20 The mask layout in Fig. 2.20 uses the simple MOSFET process in Fig.
2.19(b).
(a) Sketch the cross section along A-A.
F9 polysilicon mask
EiJ (clear field)
contact mask
II (dark field)
~ metall mask
~ (clear field)
via mask
(dark field)
meta12 mask
(clear field)
> Figure P2.19

(b) Sketch the cross section along B-B.
(c) Sketch the cross section along C-C.
P2.21 The resistor whose layout is shown in Fig. P2.21 is a test structure on an inte-
grated circuit.
The measured resistance is R = 22.3 kQ. What is the sheet resistance of the
ion-implanted layer in ohms per square? You can assume that the contact
regions each contribute 0.65 squares.
oxide mask
D (dark field)
polysilicon mask
[Jill
. (clear field)
•
~
contact mask
(dark field)
metal mask
(clear field)
c
> Figure P2.20
--------L=150µm ____ ..,
> Figure P2.21

P2.22 The two ion-implanted resistors in Fig. P2.22 have a different style of contact
region compared to that shown in Fig. 2.22. The longer one has a measured resistance
of R 1 = 395 n, with the shorter one measuring R 2 = 270 Q.
From these measurements and the dimensions, what is the sheet resistance of this
layer and the effective number of squares for this type of contact region?
P2.23 Given a p-type ion-implanted resistor with an average acceptor concentra-
tion of Na1 = 1016 cm- 3 over the depth O < x < 0.5 µm and Na2 = 7.5 x 1015 cm-3 over the
depth 0.5 µm < x < 1.25 µm. Find the sheet resistance in Q/D.
P2.24 Given an n-type ion-implanted layer with thickness t = 1 µm and average
doping concentration Nd= 1017 cm-3 .
(a) What is the sheet resistance?
(b) In order to meet the target sheet resistance of 750 Q/D, we can perform
a second implant anneal to add additional dopants to a layer of thick-
ness 0.25 µm. Neglecting the change in the overall layer thickness after
annealing, what dose and dopant type (nor p) are needed?
P2.25 The normalized uncertainty in the average acceptor concentration in an
implanted layer is EN = 0.025 . The uncertainties in the mobility and in the layer
thickness ( t = 1.4 µm ) are negligible for this problem. Finally, the dimensions of the
resistor pattern have an uncertainty of 8 = 400 A.
17 3
(a) For a 120 square resistor with Na = 10 cm- and W = 2.5 µm, what is
the average resistance and its uncertainty?
(b) If we increase the doping by a factor of two, what is the width necessary
in order to keep the average resistance the same? What is the new
uncertainty?
P2.26 It is an attractive idea, at first glance, to compensate a silicon region so that
Nd= Na and n 0 = p 0 = n;. Assume that the wafer is doped with acceptors at precisely
Na= 1016 cm-3 and that we can introduce donors uniformly into a 1 µm-thick surface
region with a normalized uncertainty of 0.005.
(a) If we wish to guarantee that the surface region stays p-type, what donor
concentration should we add? Note that the normalized uncertainty is
0.005.
W=4µm
L=60µm
~...-----: ~
W--+---•l
L = 40 µm -----
> Figure P2.22

(b) What is the range of sheet resistances for the surface region, given the
doping in part (a)?
DESIGN PROBLEMS
02.1 There are often geometric constraints in implementing resistors on an inte-

grated circuit. Figure 02.1 defines the outline of the area in which the 160 kn resistor
must fit, given the following geometric design rules. The locations for the "dogbone"
style contacts are set by the other circuit elements are cannot be moved. The oxide
mask pattern can abut the edges of the outline in Fig. 02.1.
Geometric design rules:
minimum width of oxide mask pattern = minimum resistor width=
W=2µm
minimum spacing between oxide patterns = Wg= 2.5 µm
Resistor layer characteristics:
17 3
doping concentration Nd= Nd(I±£ ) = 10 cm- (1±0.025)
1
thickness t = t ( 1 ± £ 1) = 0.6 µm (1 ± 0.02)
variation in resistor lateral dimension: o = 650 A.
(a) Lay out the 160 kn resistor to fit inside the outline in Fig. 02.1.
(b) What is the uncertainty in the resistance?
(c) Modify your layout in (a) to minimize the uncertainty, while keeping
within the boundaries of the outline. By what factor can you reduce the
uncertainty?
02.2 In some cases, the IC designer aims for the maximum resistance possible in
a particular area. In this problem, you will lay out a polysilicon resistor in an area of
100 µm x 120 µm. The contacts should use the "dogbone" style in Fig. 2.22(a) and are
,_______ 45 µm -----~ ...,.

___ 30 µm ---1
I
,,-----,
•
, -110
µm
I
I I
.----"T""" ___ .._ _ __.,,I -
I
/
\
'
. '
'\
I
'
cont:::-::~d~~---,,,
25 µm
j
i
15 µm
1
'----,-''
\I
I
contact
\ window
\
...............detail
',,
.............. .1 2µmx2µm
contact, spaced
2µmfrom
1~
comer of
outline
> Figure 02. 1

to be located as shown on Fig. D2.2. The polysilicon resistor can abut the edges of the
outline. The geometric design rules and parameters for the polysilicon film (lot-to-lot
uncertainties) are:
Geometric design rules:
minimum polysilicon line width W = 1 µm
minimum spacing between polysilicon lines = Wg= 1 µm
Polysilicon film parameters:
resistivity p = p ( l±EP) = (lmQ - cm) (1 ± 0.05)
thickness t = t(l ±E 1) = 0.5 µm (1 ±0.05)
variation in polysilicon line: o= 500 A.
(a) Design the polysilicon resistor layout for the maximum number of
squares in this area, given the design rules and fixed location of the con-
tacts.
(b) What is the maximum resistance and its variation?
I
120µm
50 µ m
lOOµm
contact
windows,,- - -
,'
I
I
..:-1
\ , .....__
--~,
detail
2
'
''
-lµm
50 µ
m ',, '
1 ~ 2
-lµm
~ Figure D2.2

Chapter 2

Uploaded by

Copyright:

Available Formats

Chapter 2

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Chapter 2

Uploaded by

Copyright:

Available Formats

•

Microelectronic engineers need a working knowledge of the physi-

+ The bond model for silicon-the electron and hole as mobile

:,...2.1 PURE SEMICONDUCTORS

To begin our discussion of semiconductor physics, we consider the

3sp tetrahedral bond

2.1.1 Electrons, Holes, and the Bond Model

4 valence electrons (-4 q),

~ border of bulk silicon region

two electrons in bond

incomplete bond (mobile hole)

Given the particular environmental conditions that determine G 1h(T) and G0 p,

Gth (T) + Gop

An idealized, special case of the steady-state condition called thermal equilib-

a;h(T) + G :P(T) (2.6)

where Ke/T) is the equilibrium constant.

I n; = 1010 cm-3at 300 K (2.10) intrinsic concentration

... 2.3 DOPING

border of bulk silicon region / immobile ionized donor

Silicon doped with Nd » n; is called extrinsic. The electron concentration n 0 in

do~~r since the atomic density of silicon is Ns; = 5 x 1022 cm- 3 .

1014 1014 4 . 1020 14

The electron concentration can be eliminated by substituting n = n/ I p

immobile negatively ionized acceptor

Assuming Nd - Na » n;, we find that

I no=~-~ (~ >~) (2.26)

Electron and Hole Concentrations

The electron concentration follows from the mass-action law

:,...·~o CARRIER TRANSPORT

'A,=v 1h 'tc = 10 nm= 0.01 µm (2.29)

2.4.1 Drift Velocity

llx = (xf, I -x;) + (xf,2 -x;) + (xf,3-x;) "'0

(a) Thermal Equilibrium, E = 0 (b) Electric Field E > 0

* x; = initial position * x1; n = final position of electron n after 7 collisions

where µn is the electron mobility with units of cm 2 /(Vs).

(2.35) hole .drift v:e,lpcify

reference plane reference plane

> EXAMPLE 2.3 Drift Current Density

-140 -100 -60

2.4.3 Diffusion Current Density

The proportionality of the diffusion current density to the concentration gradient

diff [1n(x = xr - A)AA -1n(x = xr +A)AA1 electron·aiffus,~n.

J!,iff (x = xr) (2.49)

..--t---Jn diff (negative)

~ Figure 2.13 One-dimensional picture for quantifying the electron diffusion

where k = 1.38 x 10-23 J /K is Boltzmann's Constant. This important result is called

The thermal voltage is defined as the quantity kT/q:

vth -- -kTq (2.53)

2.4.4 Total Current Density

n(x=O)= n(x= 1 µm)

~ Figure Ex2.4 Electron concentration through a slab of p-type silicon.

,> 2.5 SILICON INTEGRATED CIRCUIT TECHNOLOGY

2.5.1 Photolithography and Pattern Transfer

ultraviolet light illumination

Process and Cross Sections Oxide Mask

j 1. Grow 500 nm of Si0 2 j ~ thermal Si0

j 4. Etch Si0 2 in fluorine plasma I j s. Strip photoresist in 0 2 plasmaj

11.Implant phosphorus ions j I3. Phosphorus profile after annealing I

2.5.3 Deposited Conducting and Insulating Films

11.Grow 500 nm of thermal oxide and pattern using oxide mask I