Sunteți pe pagina 1din 25

ELEC 353 1

Semiconductors 2007-01-31 Neil R Thomson & John L Bhr


SEMICONDUCTORS

Semiconductor materials, particularly silicon, form the basic building material of modern
solid-state devices such as diodes and transistors in their various forms. These devices, in turn, are
the building blocks from which integrated circuits, computers and most other electronic equipment
are built.

Our aim will be to get a useful quantitative understanding of the theory of how some of the
more important of these semiconductor devices work; a mainly classical standpoint will frequently
suffice rather well, thus giving us the best value for effort.

Firstly, we will review the mechanism of electrical conduction.
CONDUCTION
Electric current is carried by moving charges, most commonly electrons. For conduction in
solids, the range of conductivity between the poorest and the best covers a factor of about 10
25
.
We can distinguish 3 classes: insulators, semiconductors and conductors.

1) Insulators have very nearly all their electrons tightly bound to their atoms in their
lattices with very, very few free to move for conduction. A good insulator might have one
charge carrier for each 10
20
or more atoms.

2) Semiconductors have conductivities intermediate between insulators and
conductors. The electrons in their outer shells are less tightly bound to their atoms than in
insulators but none-the-less much more tightly bound than in conductors such as metals. For
example, in pure gallium arsenide at room temperature there is approximately one free charge
carrier for every 10
16
atoms while in doped germanium, there might be one free carrier per
10
6
atoms. These carriers are free to move through their lattices under the influence of an
electric field and thus carry an electric current.

3) Conductors such as metals have the best conductivities because they have the
greatest number of charge carriers per atom. Typical metallic conductors such as silver have
about one free electron per atom which gives them a much higher conductivity than the
semiconductors.
Conduction Model
In a typical metal (or semiconductor), the atoms of the material are arranged in a fixed
regular array called the crystal lattice. Some or all of their electrons in their outer shells are free to
move around the whole of the solid forming a sea or gas of electrons. At room temperature these
electrons are in rapid random motion with thermal velocities of the order of 10
5
- 10
6
m/s rather
like the molecules in a gas such as the air around us. They are frequently colliding with the fixed
lattice 'centres'; their mean free path and the mean time between collisions, , being determined by
this thermal velocity.

When an electric field, E, is applied, each electron experiences a force, eE, and hence an
acceleration E
m
e
in the direction of the applied field. Thus in time, , it acquires a velocity E
m
e

in the direction of the applied field. Typically, the direction of this velocity will be randomized by
the collision which occurs with a lattice centre after seconds. The picture is thus one of the
electrons being repeatedly accelerated from an average velocity of zero in the direction of the
ELEC 353 2
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
electric field for a typical time, , and then having this (directed) part of their velocity, E
m
e
,
reduced to zero again by random collisions. Their average velocity will thus be given by
E
m
e
u
2
=

This can be written as E u = where is called the mobility and is a constant for the
particular material of which the solid is made. Thus the application of an electric field to a solid has
the net effect of causing the electrons to move with a constant velocity, u, known as the drift
velocity. Typically drift velocities are less than 1 mm/s.

Actually the atoms in their regular array on the lattice do not act as collision centres, at least not when the
temperature is absolute zero. The electrons' wave properties allow them to diffract around stationary atoms in a regular
array without collisional loss. However, at higher temperatures such as room temperature, the atoms are randomly
vibrating about their average position and it is this that provides the collisional centres. Also any impurities which
cause lattice irregularities act as collision centres too. The resistance of a typical metal thus falls (slowly) as the
temperature falls (because of the decreasing amplitude of vibration) eventually reaching a minimum value determined
by lattice imperfections.
Current Density
Consider a wire of crossectional area A square metres, which has n electrons per m
3
moving
with velocity u along it:

u /m

A/m
2

n /m
3
Wire
current

All electrons which have passed the arrowed point on the wire in the last second are shown as
dots; they occupy a volume uA m
3
and so there are nuA of them. The current is just the total charge
which has passed the arrowed point in one second and is thus neuA coulombs/seconds (or amps),
where e is the charge on the electron (1.6 x 10
-19
coulombs). The current per unit area or current
density, in amps/m
2
is thus given by
neu J = .
When we substitute u = E into this we have J = neE and since for a particular conducting
material n, e, and are all constant, we can write
E J =
where = ne is called the conductivity.
ATOMIC STRUCTURE IN SEMICONDUCTORS
Chemistry tells us that atoms like to have eight electrons in their outer shells. They are the
most stable in this configuration. The atoms of certain elements such as neon and argon already
have 8 electrons in their outer shells and are (chemically) very unreactive and very stable. The
atoms of most other elements try to achieve this stable configuration by sharing their outer shell
electrons with other atoms in crystal lattices or chemical compounds.

In the case of a typical metal such as sodium, there is one electron in the outer shell and 8 in
the next shell in. The atoms in a metallic sodium lattice each release this outer most electron to
wander freely about the lattice while the rest of each atom (which is an Na
+
ion) remains fixed in
the lattice. Each atom/ion on the lattice thus has the desired 8 electrons in what is now its outer
shell.
ELEC 353 3
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
The atoms of typical semiconducting materials such as silicon (Si) or germanium (Ge) have
four electrons in their outer shell. In a solid Si each atom shares its outermost shell electrons with
four neighbours. This sharing is called a covalent bond; these bonds hold the lattice together. In
this way each Si atom on the lattice gets a share in 8 electrons. The diagram below is a schematic
two-dimensional 'illustration' of this.


silicon
electron


In the diagram the big circles are the silicon atoms without their outer shell electrons since
these electrons (four per atom of Si) are shown as separate dots. These (outer shell) electrons are
fairly tightly bound to the atoms in the lattice; energy of about 1 eV is required to tear one of them
from its bond. This is, however, very much less than the corresponding energy for diamond which
is about 7 eV. This corresponds with the fact that diamond is an insulator and silicon is a
semiconductor.

The typical energy associated with random atomic motions at room temperature (say T = 300
K) is kT joules or kT 0.025 eV since k = 1.38 x 10
-23
J/K (Boltzmann's constant) and e = 1.6x10
-19

coulombs. Of course, at any given time there will be some silicon atoms which have more
vibrational energy than this and some which have less; our 0.025 eV is simply an estimate of the
mean energy. Indeed very occasionally some atom/bond on the lattice will have as much as 1 eV of
energy by random chance and that will allow one of its electrons to break free of the bond. This
electron can then move freely through the lattice and take part in electrical conduction. It leaves
behind it a hole which is a site with one electron missing and with a positive charge of +e since the
site was electrically neutral before the electron left.

In pure silicon at room temperature at any time, only about one atom in about 3 x 10
12
has
released a free electron into the lattice in this way. Pure silicon at room temperature has about 1.5 x
10
16
free electrons/m
3
, 1.5 x 10
16
holes/m
3
, and about 5 x 10
28
atoms/m
3
. Because of the extreme
dilution, only very occasionally will an electron 'collide' with a hole and recombine. This is
balanced by the random thermal generation discussed in the last paragraph.
BAND STRUCTURE
Electrons in free atoms can occupy only certain discrete energy levels (s, p, d). When the free atoms are moved
close together to form a solid, the atoms are no longer independent of each other and it turns out that when we take this
factor into account we find that only certain energy bands are allowed. Electrons are forbidden to have energies
between these allowed bands.
ELEC 353 4
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr

free atom solid (lattice)
allowed energy level
allowed energy level
allowed energy level
separation between atoms
forbidden band
forbidden band
allowed band
allowed band
allowed band
Energy
Energy

Just as each energy level in the free atom is allowed to hold only a certain maximum number of electrons, each
energy band in the solid can accommodate a certain maximum number of electrons and no more (Pauli exclusion
principle). For example the s level in an atom can accommodate only 2 electrons; if N electrons have been brought
together to form the solid then the allowed energy band in the solid, which corresponds to this s level, will be able to
accommodate 2N electrons. In this way all the electrons accommodated in the N free atoms are also accommodated in
the solid with N atoms on the lattice.

Just as all the lower energy levels are filled in atoms at room temperature, the lower energy bands in a solid tend
to be filled too. In some solids, such as diamond, the lower energy bands are all completely filled and above these all
the bands are completely empty. Diamond is an insulator. When an electric field is applied the electrons cannot
accelerate because none of them have unfilled energy states immediately above their present states to move into (unless
the electric field is so huge as to accelerate them directly into the next band up, but this would be breakdown).

Metals such as sodium have a partly full band immediately above their lowest full bands. Recall that each
sodium has one electron in its outer (3s) shell so that, if there are N atoms in the solid, there will be N electrons in the
corresponding band which could hold 2N electrons. Thus we could expect this band to be only half full. (Actually this
band overlaps with higher bands resulting in a rather broad band which is less than half full.) This partly filled band is
called the conduction band because, when an electric field is applied, these electrons will readily move to slightly
higher energies within the conduction band and are thus able to conduct an electric current. The conductivity is very
good because the conduction band contains many electrons - about one per atom.

Semiconductors such as silicon have one very nearly full band
immediately above their lowest completely filled bands. This nearly full
band is called the valence band. The allowed band immediately above
this band is very nearly empty and is called the conduction band
because the electrons in it can move freely under the influence of an
electric field just as in a metal. The conductivity is much lower than in a
metal, however, because there are far fewer electrons in the conduction
band (only about one for every 3 x 10
12
atoms in pure silicon).

Conduction can also take place in the unfilled states or holes
near the top of the valence band. It turns out that these holes behave as
positive charges. The holes are separated from the electrons by the energy difference between the top of the valence
band and the bottom of the conduction band. This energy difference is called the band gap; it is about 0.7 eV for Ge,
1.1 eV for Si, and 7 eV for diamond. The probability of an electron filling (or not filling) a given energy state is
governed by the Fermi-Dirac statistical function (rather than Boltzmann) because of the Pauli exclusion principle.
HOLES AND ELECTRONS
The current in a semiconductor is thus carried by both holes and electrons,
J = (n
n
+ p
p
)eE = E
where n , p are the number densities of electrons and holes, and
n
,
p
are their respective
mobilities.


valence band
conduction band
Band gap
Energy
ELEC 353 5
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
In a pure semiconductor, n = p. At very low temperatures the valence band is completely full
and the conduction band is completely empty. As the temperature rises electrons are excited into
the conduction band leaving behind an equal number of holes in the valence band. The number of
holes and electrons rises rapidly as the temperature rises causing a rapid rise in conductivity (about
5% per degree C at room temperature for a typical case). Thermistors are temperature dependent
'resistors' using semiconductor material and are useful for temperature sensing and control.

Pure semiconductor is often referred to as intrinsic semiconductor and so the charge carrier
(electron or hole) number densities are often written as n
i
and p
i
. The charge neutrality condition in
the pure or intrinsic semiconductor can thus be written
i i
p n =
DOPING
Very tiny quantities of impurity in a semiconductor can increase the conductivity dramatically.
Atoms of a similar size but with 3 electrons (boron, aluminium) or 5 electrons (phosphorus, arsenic,
antimony) in their outer shells are particularly useful.

Such an atom with 5 electrons in its outer shell fits nicely into the lattice (in the place of say a
silicon atom) using four of its electrons for bonding and the fifth is then free to move through the
lattice as a charge carrier (in the conduction band). Only one such atom for every 10
9
Si atoms will
increase the number of free electrons and hence conductivity by a factor of ~10
3
or so since there
will now be about one free electron for every 10
9
Si atoms whereas in pure Si there was only about
one free electron for every 3 x 10
12
silicon atoms. A semiconductor doped in this way to have an
abundance of negative charge carriers (ie electrons) is known as n-type material.

Charge neutrality is, of course, maintained. When we add neutral phosphorus (P) atoms to
neutral Si the resulting material must be neutral. When each neutral P atom gives up its electron it
becomes a positive charge centre. This charge centre is, however, fixed to the lattice and can play
no part in conduction. It is fully bonded by its four remaining electrons together with a share in the
four electrons of each of its Si neighbours. It is just as tightly bonded to the lattice as any other Si
atom. The fixed positive charges of these atoms/ions of P balance the negative charges of the
mobile electrons which the P's released.


silicon
P+
free electron
bound electron
P+
phosphorus


In a similar way, suitable atoms with 3 electrons in their outer shells, such as boron (B), can
be added to pure Si. Each B atom fits neatly into the lattice in the place of a Si except that there is
one electron short.

ELEC 353 6
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr

silicon
bound electron
boron
B
B


A (bound) electron from a neighbouring Si atom can readily move in to complete the bond
around the boron. The boron site will now have a negative charge and the site from which the
electron came will be a positive mobile hole.


silicon
bound electron
boron
B-
+
+ hole
B-


This hole is now free to wander around the lattice. If an electric field is applied the hole will
drift in the direction of the field thus carrying an electric current. If one atom of boron is added for
every 10
9
Si atoms, this will increase the number of holes and hence the conductivity by a factor
~10
3
or so as compared to pure Si where there is only about one hole for every 3 x 10
12
Si atoms.
A semiconductor doped with an impurity such as boron that creates positive charge carriers (ie
holes) is known as p-type material.
HALL EFFECT
Experimental proof of the existence of positively charged holes as carriers of electric current
is found in the Hall Effect. The experiment is illustrated in the diagram below which shows two
(shaded) bars of semiconductor: the one on the left is p-type and the bar on the right is n-type. Both
bars are immersed in a uniform magnetic field, B, perpendicular to the page and directed into it,
and both bars are carrying a conventional current, I, flowing from left to right (as measured by
ammeters [not shown]).


-
- +
+
V
V
B
B
B
B
I I
p-type
n-type
u
u
u
u
u
u
F
F
F F
F
F


If the (conventional) current is being carried by holes, they will be moving to the right and
they will experience a magnetic force, F = e (u x B), which will be directed upwards as shown, and
will tend to make the upper face of the bar positive with respect to the corresponding lower face.

ELEC 353 7
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
On the other hand, if this same (conventional) current is being carried by electrons, they
will be moving to the left and they will experience a magnetic force, F = e (u x B), which will also
be directed upwards as shown because both charge and velocity are now negative, so that F has the
same direction as for positive holes. This will tend to make the upper face of the bar negative
with respect to the corresponding lower face.

Solid state physics theory, which takes into account the wave nature of the electron, 'Bragg
reflections' from the lattice, density of allowed energy states in the conduction and valence bands,
and Fermi-Dirac statistics etc., successfully predicts the (effective) positive nature of holes. Here
we will be content to rely mainly on the strong supporting evidence which the Hall Effect
experiments give.
GENERATION AND RECOMBINATION - EQUILIBRIUM
In a semiconductor, doped or pure, there are new hole-electron pairs being continually
generated by some random lattice point momentarily acquiring a large and sufficient amount of
energy to release an electron into the lattice (ie excite an electron into the conduction band). This
generation is balanced at equilibrium by the random recombination of electrons and holes.

The rate of recombination will be proportional to the concentration of electrons, since
doubling the number of electrons doubles the chances of an electron coming close enough to a hole
to recombine with it. Similarly the rate of recombination will be proportional to the concentration
of holes. The recombination rate, R, can thus be written
R = rnp

where r is the proportionality constant for the material's recombination rate.

At equilibrium, this recombination rate must equal the generation rate, g; ie
g = rnp

Now, in any of the silicon samples discussed so far, the vast majority of atoms on the lattice
are Si and so the thermal generation rate of hole-electron pairs will be independent of the doping.
The vast majority of the generations occurring at equilibrium will happening at Si atoms since
there are so many of them. In particular, the generation rate in the pure material will be the same as
the generation rate in doped material. We thus have
rn
i
p
i
= g = rnp .

From which we can conclude n
i
p
i
= np

and since n
i
= p
i
, we have
2 2
i i
p n np = =

This is a most useful relationship between the electron and hole concentrations in a doped
semiconductor and the electron or hole concentration in the pure semiconductor. In particular, it
shows us that there are holes in n-type semiconductor although their concentration is low especially
if the doping level, and hence electron concentration, is high. Similarly there are electrons in p-
type semiconductor although, again, their concentration is inversely proportional to the hole
concentration.

(Actually the dominant recombination mechanism for holes and electron is not normally direct recombination; most
recombination (and generation) takes place at a small number of impurity centres in the band gap but the equilibrium
result is the same as above. The above theory is adequate for most of the recombination situations which we will
encounter here but none-the-less does have its limitations. See e.g. van der Ziel's book for more details)

ELEC 353 8
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
A doping material with 5 electrons (such as phosphorus) is often known as a donor, since it
donates its extra electron for conduction. Similarly a doping material with 3 electrons (such as
boron) is known as an acceptor.

If a sample of n-type semiconductor has N
d
donor atoms per m
3
, then there will be N
d
fixed
positive charge centres per m
3
; there will also be a few holes, say p
n
per m
3
so that the total
positive charge per m
3
must be N
d
+p
n
. The only negative charges are the copious electrons, say n
n

per m
3
so that the charge neutrality requires

n d n
p N n + =

Practical doping usually involves adding far more donor atoms than there were electrons in
the pure material so that

d n
N n to a very good approximation.

Since np = n
i
2
, the concentration of holes in n-type semiconductor will be given by

d
i
n d
i
n
N
n
p N
n
p
2 2

+
=

Similarly in a p-type semiconductor, where the acceptor (impurity) concentration is N
a
, we
have
p a p
n N p + = ,
a p
N p , and
a
i
p a
i
p
N
n
n N
n
n
2 2

+
=
THE PN JUNCTION OR SEMICONDUCTOR DIODE
When a piece of p-type semiconductor joins with a piece of n-type to form a junction we find
we get a diode. Current flows much more readily (perhaps a million times more readily) from p to
n than vice-versa. We are now in a position to understand why this is so.

We will assume that the p-material has been doped
much more heavily than the n-type material. This is not
essential for our discussion but will allow us to focus our
attention rather more on the holes instead of having to spread
it evenly between the holes and the electrons.
Recall that p
p
n
p
= n
i
2
=p
n
n
n
so that the number densities
on each side of the junction will be rather like those in the log
plot shown.

The electrons and holes will be moving around in the
semiconductor with random thermal velocities rather like the
particles in the gas in a room. (Both the electrons in the conduction
band and the holes in the valence band are far from the Fermi level; also
the number of particles per m
3
is very low so most states will be unfilled.
Hence rather like ordinary Boltzmann gas.)

There will be far more holes moving towards the junction from the left than from the right,
because there are far more holes in the p-type material; thus for a short time after the junction is
formed holes will move from p to n. This will leave 'uncovered' (ie unneutralized) negative charge
centres (boron, say) just inside the p-material on the left hand side of the junction. Once the holes
cross the junction, they will combine with the electrons just inside the n-material which will leave

p n
log(number density)
p
p
n
p
n
n
p
n
Distance
(from junction)
ELEC 353 9
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
'uncovered' ( ie unneutralized) positive charge centres (phosphorus, say) on the n side of the
junction. This is illustrated in the top part of the diagram below.

This transfer of charge from p to n due to hole flow has made the n material positive and the
p-material negative. Before very much charge has flowed, a reverse electric field (ie an electric
field directed from n to p) will build up to just the right size to oppose the flow. Our junction in
equilibrium will have a pressure gradient force (due to the hole density gradient) pushing to the left
which will be exactly balanced by the electric field force to the right. (Note that the amount of charge that
has flowed across the junction to create this field is much too small to affect either the p or n concentrations.)

Exactly the same statements are true for the electrons; the same electric field will serve to
oppose their pressure and maintain them in equilibrium. However, we have assumed there are far
fewer of them so that we need hardly concern ourselves with them.
EQUILIBRIUM JUNCTION POTENTIAL
We can now calculate the potential difference across the junction (due to this electric field) in
terms of the hole and electron densities well inside the p- and n-materials. The top part of the
diagram below shows the junction and the 'uncovered' charges; the lower part of the diagram greatly
expands the transition region (the region where there is this electric field and in which the particle
densities are smoothly changing from their equilibrium values on the p-side to those on the n-side).
The structure of the transition region and its 'uncovered' charge will be dealt with in more detail
later. The transition region width is typically ~ 0.5 m. Note that the number densities are plotted
logarithmically so that p
p
>> n
n
>> p
n
>> n
p
probably by a factor of well over 100 in each case.

P+dP
p-type n-type
transition region (expanded)
log(number density)
p
p
p
n
n
p
n
n
-
-
-
+
+
+
distance from (metallurgical) junction x
E (elecric field)
dx
P


We will focus our attention on the thin slice of holes of width dx (and crossectional area, A,
perpendicular to the paper and parallel to the junction boundary) shown in the lower right part of
the diagram above. These holes are experiencing a pressure force to the right due to the hole
pressure, P + dP , on the left hand side of the slice being greater than the hole pressure, P, on the
right hand side; this pressure gradient is caused by there being more holes to the left than to the
right. The net pressure force to the right will thus be A (P + dP) - AP = AdP. The volume of the
slice is Adx, so that
the force per unit volume due to the pressure gradient
dx
dP
=
(note that pressure is being designated by upper case P whereas hole density is lower case p.)

This pressure force is balanced at equilibrium by the electric field force per unit volume, eEp.
Hence
dx
dP
eEp = , in our equilibrium junction.
ELEC 353 10
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
But recall the pressure due to an ideal gas is given by nkT P = where n is the number of
particles per unit volume (the number density). We've been using n for the electron number density
and p for the hole number density; so our hole pressure will be given by pkT P = .

(You might like to recall the familiar ideal gas law PV = nRT but remember n is the number of moles in this
formula. This gas law can also be written PV = NkT where N is the number of particles in volume, V, of the gas. If
we divide this last equation through by V we get P =
N
V
kT or P = nkT where n is the number of particles per unit
volume. We will not use V for volume except in this paragraph; V will normally be voltage or potential difference.)

Hence, at equilibrium, we have
dx
dp
kT
dx
dP
eEp = =

so that
p
dp
dx
kT
eE
=

We need only integrate this equation across the junction because the potential across the
junction, V
jo
, is just the integral of Edx across the junction.
Hence
n
p
j
p
p
V
kT
e
ln
0
= (ln log
e
)
Thus for the unbiased pn junction
n
p
jo
p
p
e
kT
V ln = .
At room temperature, approximately,
|
|

\
|
=
n
p
jo
p
p
V ln 25 millivolts; for most diodes V
jo
is a few tenths
of a volt.
Contact Potential
It might be tempting to think that if we put wires into the n- and p-type materials that the
junction potential difference would make a current flow in an external circuit; all we would have to
do would be to have a warm reservoir maintain the (room) temperature of the junction by supplying
the heat required by conservation of energy. This would, of course, violate the second law of
thermodynamics; we would be taking heat from a warm reservoir, converting it entirely into work
and having no other effect. We would be getting order (electrical work) from chaos (random
thermal motion).

In practice there will be contact potential differences set up both where the wire is in contact
with the p-material and where the other wire is in contact with the n-material. The sum of these
three potentials (pn, p/wire, n/wire) will be zero and no current will flow.
DIODE ACTION
We can now give a (qualitative) description of why the diode conducts current very much
more readily in one direction than the other.
Suppose we apply a 'forward' external voltage, V, between
the p- and n-materials as shown. This will decrease the
equilibrium reverse potential (calculated above) which was
holding back the large number of holes in the p-material. Some
of the numerous holes in the p-material will now be able to flow into the n-material giving a current
flow anti-clockwise. Electrons in the n-material will no longer be fully held back by the junction

p n
+
_
V
ELEC 353 11
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
potential and will be able to flow into the p-material also creating a conventional current from left
to right. The diode thus conducts well in the forward direction.

If the battery is reversed, however, this will increase the
reverse potential across the junction and tend to stop the
majority carriers (holes in the p-material and electrons in the
n-material) from crossing the junction.

The battery is now trying to drive current from the n-material to the p-material but there are
very few holes in the n-material so very little current can be carried by holes. The only other way a
current can flow from n to p is for electrons to flow from p to n but there are very few electrons
indeed in the highly doped p-material. The reverse current is thus limited by the concentration of
the minority carriers (electrons in the p-material and holes in the n-material) and is consequently
very small, very much smaller than the forward current.
DIODE EQUATION
When the voltage of the external battery across a diode is increased from zero to V in the
forward direction, we can expect the potential difference across the pn junction to decrease from
V
jo
to V
jo
- V. This allows holes to flow across the transition region into the n-material as shown in
the diagram below. The hole concentration in the n-material just beyond the transition region will
rise from p
n
(the unperturbed value) to p as shown.


V
+ -
p n
log p
transition
region
x
x = 0
p
p

p
n

p
p'
p'


The hole number density will decrease as we go further into the n-material because of
recombination of our injected holes with the numerous ambient electrons. This hole density
gradient (falling from p' to p
n
) will drive the hole current in the n-region where there is virtually no
electric field because of the high conductivity; virtually all of the applied voltage is dropped across
the junction.
The voltage drop across the unbiased junction was
n
p
j
p
p
e
kT
V ln
0
= in terms of the hole
number densities on either side of it. By exactly the same reasoning the voltage drop across our
forward biased junction (ie across the transition region) must be similarly related to the (new) hole
number densities on either side of the transition region
ie
'
ln
0
p
p
e
kT
V V
p
j
=

so that the new hole density, p', just inside the n-region, can be obtained as
p n
+
_
V
ELEC 353 12
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
|

\
|
|
|

\
|
=
|
|

\
|
=
kT
eV
kT
eV
p
kT
V V e
p p
j
p
j
p
exp exp
) (
exp '
0
0


|

\
|
=
kT
eV
p p
n
exp ' since p' = p
n
when V = 0.

As we shall show a little later on, the excess hole density falls exponentially with distance in
the n-region (hence the straight line on the log plot in the diagram in dropping from p' to p
n
);
ie ( )
|
|

\
|
=
|
|

\
|
= =
p p
n n
L
x
p
L
x
p p p p p exp ' exp ' where L
p
is the distance constant of the
exponential hole decay known as the hole diffusion length, and the other symbols are defined by
the equations and the diagram above.

The hole current in the n-region will be proportional to the hole density gradient (twice the
gradient, twice the current). Just inside the n-region the hole current is the whole current; only
further in to the n-region is part of the current carried by the electrons. (Deep into the n-region all
the current is carried by electrons.) The total diode current is thus proportional to the hole gradient
just inside the n-region. But the slope of any exponential function (like the excess hole density) is
proportional to the size of the function itself (ie the excess hole density).
ie for the diode current I p'
but
(

\
|
=
|

\
|
= = 1 exp exp ' '
kT
eV
p p
kT
eV
p p p p
n n n n

Hence
(

\
|
1 exp
kT
eV
p I
n

or
(

\
|
= 1 exp
kT
eV
I I
s

which gives the relationship between the voltage across a diode and the current through it; it is
known as the diode equation.

In the diode equation, I
s
appears in the above derivation as the proportionality constant. It is
also the maximum reverse current, or reverse saturation current, since, when V is negative by
more than a few multiples of
e
kT
= 0.025 volts, I -I
s
.

At room temperature, approximately I = I
s
[exp(40V) - 1]

The above diode equation works well for germanium under most conditions. For silicon, the theory is correct
too, but additional processes are significant at high and low currents. At very low forward currents, recombination
within the transition region may be the dominant factor while at very high currents the hole injection into the n-region
may result in hole concentrations there which are comparable with or greater than the electron concentration. This
effect, together with the finite conductivity of the n-region, may result in a significant part of the applied potential drop
occurring in the n-region rather than at the junction itself with high currents. Even at high currents, the equation still
predicts the junction potential itself quite well. In particular, the equation will give a reasonable value for the slope (ie
the dynamic resistance) of a silicon diode at common moderate currents.
ELEC 353 13
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
DIFFUSION CURRENT
The (forward) hole current in the n-region of our diode is driven by its density or pressure
gradient. Such currents are called diffusion currents to distinguish them from currents driven by
electric fields which are called drift currents.

Recall, that if a current is driven by an electric field, the balance between the collisional
friction and the driving force results in a constant velocity given by u = E, and if we multiply
through by pe/ we get

peu
= peE. The right hand side of this is readily identified as the driving
force per m
3
, so that the left hand side can then be identified thus:

peu
=
3
m per drag frictional l collisiona the

This must be true whether the driving force is an electric field or a pressure/density gradient.
The current is determined by the balance between the driving force and the collisional friction of the
flow of current. When determining the voltage across the unbiased junction, we showed that the
magnitude of the force per m
3
due to a pressure/density gradient was
dx
dp
kT
dx
dP
= , hence our
diffusion current force balance for the injected holes in the n-region side of the junction is

peu
dx
dp
kT =

which after multiplying by (and recalling that J = peu) gives
J peu
dx
dp
kT = = for the diffusion current.

Differentiating the first part of this gives
( )
dx
pu d
e
dx
p d
kT =
2
2

Now pu is the hole flow rate (number m
-2
s
-1
) so that
dx
pu d ) (
is the loss rate or
recombination rate (m
-3
s
-1
) which we already know is rn
n
(p
n
+ p) - g = rn
n
p. We will choose to
write
p
n
p
p rn

= for the recombination rate (m


-3
s
-1
) where
p
, so defined, is called the carrier
lifetime of the holes in the n-type material. We thus have
p
p
e
dx
p d
kT


=
2
2

but, since p = p
n
+ p and p
n
is a constant independent of x, we can write
p
p
e
dx
p d
kT

2
2

The solution of this is
|
|

\
|

=
p
L
x
p p exp '
where
p p
e
kT
L

= as can be readily verified by substitution


and checking the boundary conditions that p =p ' at x = 0, and p = 0 at x = . Thus we see that
ELEC 353 14
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
the hole density (or minority carrier density) does indeed decay exponentially with distance in the
n-region. L
p
is called the diffusion length for holes in n-material.
REVERSE SATURATION CURRENT
We can now calculate the reverse saturation current (ie the small current, I
s
, which flows
when we apply a reverse voltage of many times 0.025 V (kT/e) to our diode. We will continue
assuming that the p-material is much more heavily doped than the n-material, so that the reverse
current will be carried mainly by holes from the n-material since there will be far fewer electrons in
the p-material (the only other possible carrier source for reverse current).

In the n-material just outside the transition region, the hole concentration will be zero because
any holes on the transition region boundary will be rapidly removed by the strong electric field
directed towards the p-material. Deep in the n-material, far from the transition region, the hole
concentration will be p
n
, the equilibrium value. As we approach the transition region from deep in
the n-material we find the hole concentration, p, as shown in the diagram decreases steadily from p
n

to zero at the transition. This gradient drives the hole diffusion current which constitutes the small
reverse current.


V
+ -
p n
log p
transition
region
x x = 0
p
p

p
n

p
n
n


The hole force balance at a point in the n-region such as that marked with the p will be
pressure gradient force = collisional friction force
ie

peu
dx
dp
kT =

which is just the same equation as for the forward (minority) hole current in the n-region except that
the number density gradient and the velocity (and hence current) are reversed left for right. (Also
the hole concentration in the current flow is very much smaller because it is having to be supplied
from the n-region and not by injection from the p-region.) Thus after multiplying by and
differentiating we again get
( )
dx
pu d
e
dx
p d
kT =
2
2

Deep in the n-material the reverse current is being carried by electrons. As we approach the
transition region, more and more of this reverse current is carried by holes until, at the transition
region boundary, all the reverse current is being carried by holes. Near the transition region the
hole concentration is less than the equilibrium value and so there will be a net generation of holes
which can be expressed as
( )
p
n n n n n n n
p
p rn p p rn p rn p rn p rn g
dx
pu d

= = = = =
) (
where p = p
n
- p here.
ELEC 353 15
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
Hence just as in the forward bias case we can write
p
p
e
dx
p d
kT

2
2

The solution of this is
(
(

|
|

\
|

=
p
n
L
x
p p exp 1

where L
p
=
p
e
kT

as can be readily verified by substitution and checking the boundary


conditions that p = 0 at x = 0, and p = p
n
at x = .

The hole current is now obtained from the force balance equation we started with,

peu
dx
dp
kT =
and the usual current density relation J
s
= peu,
which gives
p
n
s
L
p
kT
dx
p d
kT
dx
dp
kT J

= =
at x = 0
and substituting for L
p
gives finally for the reverse saturation current

p
n s
ekT
p J

=

The reverse saturation current of a pn diode can thus be expected to be given by I
s
= AJ
s
, where A is the
crossectional area of the diode. If this value of I
s
is used in the diode equation we get quite good agreement for
germanium pn junctions over the full current range.

In the case of silicon pn junctions, we get reasonable predictions for the forward current only when the forward
current is neither very high nor very low. When the external bias voltage is such that the diode current is very low, we
find that there is a component of current in addition to that described in the theory above. This is because
recombination-generation centres within the transition region itself contribute a current which, though small compared
to normal forward currents, is none-the-less much larger than the I
s
described above. Thus the reverse current flowing
in most silicon diodes, when a few volts of reverse bias is applied, is normally in the range 1 pA - 1 nA which is several
orders of magnitude greater than the I
s
described above.
TRANSITION REGION
We will now examine the transition region of our pn junction in more detail and calculate its
width and capacitance as the (reverse) bias is varied. This fundamental capacitance limits the high
frequency response of many electronic circuits. The capacitance can also be useful because it turns
out to be voltage dependent which means it can be used in such applications as automatic frequency
control, for example the locking of the frequency of one crystal oscillator to another or to an atomic
standard.

Both p- and n-materials are neutral to start with; in the p-material the positive charges of the
mobile holes are balanced by the fixed negative charges at the lattice points occupied by the
acceptor atoms (e.g. boron), while in the n-material the negative charges of the mobile electrons are
balanced by the fixed positive charges at the lattice points occupied by the donor atoms (e.g.
phosphorus).

ELEC 353 16
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
In reverse bias the externally applied voltage, V,
reinforces the built-in reverse potential, V
jo
, and the
corresponding electric field repels both the mobile holes
in the p-material and the mobile electrons in the n-
material (ie the majority carriers) away from the
junction. This means the fixed charges near the
junction are no longer balanced by the mobile charges.
These unbalanced or uncovered charges are the
physical source of the electric field in the transition
region, as shown in the diagram below.

The transition region thus contains no mobile
charges; any that enter will be quickly swept away by
the electric field. However, as explained, the region
contains fixed charge uniformly distributed throughout
its space or volume (ie the fixed charge is not just on
the surface). For this reason the transition region is
commonly known as the space charge limited region
or SCL. This region is also known as the depletion
region since it is depleted of mobile charge carriers. In
the p-side of the SCL there will be N
a
fixed charges per
m
3
if the p-material was made with N
a
acceptor atoms
per m
3
, while on the n-side of the SCL there will be N
d

fixed charges per m
3
assuming the n-material was
doped with N
d
donors per m
3
. We have continued our
assumption that the p-material is much more heavily doped than the n-material so that N
a
N
d
.

Each electric field line starts on a fixed positive (uncovered) charge in the n-material and ends
on a fixed negative charge in the p-material. Hence the electric field strength is greatest at the
metallurgical junction and falls to zero at the edges of the transition region where the high
conductivity n- or p-regions begin. The accompanying graph shows the electric field, E, as negative
since it is directed in the -x direction. In the potential graph, we have arbitrarily taken the p-
material to be at the zero of potential. Note that most of the potential difference across the junction
occurs in the n-material because the distances are greater (potential difference =

Edx ).
Width of the transition region:
The quantity of uncovered charge on each side of the metallurgical junction must be the same
(each field line begins on a plus and ends on a minus) so that
N
a
W
p
= N
d
W
n


where W
p
and W
n
are the widths of the SCL (transition region) in the p- and n-type materials
respectively.
Poisson's equation gives us


= V
2

or, in our one dimensional case, SCL the of side - n in the
2
2

d x
eN
dx
V d
=

Since -

= Edx V
x
, this becomes

d
eN
dx
dE
=

+
-
V
+
-
+
+
+
-
-
p n
x
E
E
max

x
V
i
Potential
-
-
-
-
-
+
+
+
+
+ -
E
+ + + + + + + +
-
-
-
-
-
-
-
-
N
x
N
a
W
n
W
p
N
d
ELEC 353 17
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
This last result can also be readily obtained by
applying Gauss' Theorem to the 'pillbox' in the diagram:

EdA = charge enclosed =eN


d
x volume = eN
d
Adx
(A is the junction crossection area)

Hence AdE = eN
d
Adx from which the result above
follows.
Integrating

d
eN
dx
dE
= from x to W
n
and taking E = 0 at x W
n
(since all the electric field
lines have terminated by then) gives

( )
n
d
W x
eN
E =

in the n-side of the SCL (ie W


n
x 0).
Note that E is negative since it is directed along the x axis.

We will neglect the small potential drop on the p-side of the SCL (it is small because the p-
material being much more highly doped than the n-material means W
p
W
n
) and take V = 0 at x =
0, ie at the metallurgical junction. We thus have for the potential at x

( )
n
d
x
xW x
eN
Edx V 2
2
2
= =




but we want the potential across the junction, V
j
, (relative to V = 0 at x = 0), ie we want the potential
at x = W
n
which is given by
2
2
W
eN
V
d
j

=

where we have put W
n
= W for the width of the SCL (or transition region) since this width is
mainly in the n-side. Note that V
j
= V
jo
- V, where
n
p
j
p
p
e
kT
V ln
0
= is the unbiased or built-in
junction potential and V is the reverse voltage applied externally between the diode's inputs. (Thus
-V in the equation is positive for a reverse bias and so a reverse bias increases the magnitude of the
potential across the junction).
Capacitance of the transition region:
The (uncovered) charge stored in the junction when there is a potential V
j
across it is given by
q = eN
d
AW
n
= eN
d
AW

The dynamic (or small signal) capacitance of the junction is defined as

j j
T
dV
dW
dW
dq
dV
dq
dV
dq
C = = =

hence
1
|
|

\
|
= =
dW
dV
A eN
dV
dW
A eN C
j
d
j
d T

which, after differentiating our formula for V
j
above gives for the transition capacitance,
W
A
C
T

=


-
+
-
-
-
-
+
+
+
+ -
E
Gaussian pillbox
p n
SCL
+
ELEC 353 18
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
This is the same formula as for a parallel plate capacitor because, if we alter the junction
potential by a small amount, all the change in charge occurs on the edges of the depletion region
which are separated by a distance W. For silicon 12

(ie silicon's relative permitivity is


about 12).

Since C
T
W
-1
and V
j
W
2
we have

j
T
V
C
1
for an abrupt pn junction.

For a linearly graded junction (where the doping varies linearly across the junction instead of
abruptly), it can be shown that
3
1
j
T
V
C .
BREAKDOWN - AVALANCHE AND ZENER DIODES
From the linear graph of E versus x for the abrupt junction given above, it is readily seen that
W E V
j max
2
1
=
Now from W
eN
V
W W
eN
V
d
j
d
j
for substitute can that we so
2
have we
2

2

= = obtaining

j d
V eN
E
2
max
=
If this maximum electric field exceeds about 30 volts/m (3 x 10
7
volts/m), the electrons (or
holes) in the (normally small) reverse current flow in the depletion region are accelerated between
each lattice collision up to an energy which is sufficient to knock another electron from its bond (ie
create a new electron-hole pair) at the next collision. These 2 electrons are similarly accelerated
and produce 2 more or a total of 4 after two mean free paths. These 4 give rise to 8 and so on. This
is the avalanche effect. Many new electron-hole pairs are thus generated in the depletion region;
the new holes drift rapidly to the p-region and the new electrons drift rapidly to the n-region under
the influence of this large electric field. Thus a substantial reverse current can be expected when V
j

is large enough for E
max
to exceed 30 V/m.

The above formula for E
max
shows that the manufacturer can arrange at what junction
potential the avalanche effect will set in just by suitably arranging the doping level, ie by adjusting
N
d
. A diode intended for rectifying large AC voltages (e.g. the mains, 230 V rms, 325 V peak)
needs a large PIV (peak inverse voltage) rating. Such a diode needs a low value of N
d
so that even
at a reverse voltage of 325 volts (much more to be safe!) E
max
is still less than 30 V/m.

By having a sufficiently high doping it is possible to have the avalanche effect set in with
reverse voltages as low as about 5 - 7 volts. When the doping is made even higher, the width of the
depletion region becomes too narrow for there to be enough 'stages' (collisions) in the avalanche to
give significant extra current. When the reverse voltage (or doping) is increased until E
max
is about
100 volts/m (10
8
V/m) the valence electrons in the depletion region start to be torn straight from
their bonds - not in the classical sense (the field is not nearly strong enough for that) but in the
quantum mechanical sense that there becomes a significant probability of them escaping or
tunneling out of the potential well of their bonds. This is the Zener effect. As one would expect
at such high fields it is normally amplified by the avalanche effect.

ELEC 353 19
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
In some diodes and circuits this breakdown effect is put to advantage. Manufacturers sell
'zener diodes' with specified (reverse) breakdown voltages most commonly between about 5 V and
30 V and less commonly from about 3 V to 200 V. Such diodes with 'zener voltages' (or
breakdown voltages) greater than about 7 V exhibit the avalanche effect rather than the zener effect.
The zener effect only begins to dominate for diodes with breakdown voltages below about 5 V.
This is, however, not of great consequence to the user.

The 'breakdown' does not usually destroy the diode; provided the power dissipation is not so
high as to overheat the junction, the diode will not be damaged. For example, if the manufacturer
sells you a 20 V, 1 watt zener diode the maximum reverse current you safely can allow to go
through it will be
V 20
watt 1
= 50 mA. If it were a 20 V, 0.1 W zener then the maximum reverse current
which would not spoil the diode would be 5 mA.

In contrast, a rectifying diode with a reverse breakdown voltage of 320 V rated at 1 watt could
handle only 1/320 amp or about 3 mA of reverse current before overheating and destroying itself.
As the current is most unlikely to be limited to only 3 mA in a circuit of this voltage, breakdown
may well mean destruction.
Voltage Regulator
The diagram below shows the circuit of a simple zener diode voltage regulator (note the
zener diode circuit symbol), a graph of the (current/voltage) characteristics of the zener diode, and a
graph of the voltage, V
in
(from a simple rectifier circuit) which is to be the source of our regulated
output, V
o
.

V
0
V
in

-12
-8 -4 0
-10
-15
-20
v (volts)
i (mA)
1 k
Zener
diode
Slope=50
time
V
in

20
10
-16

We will begin by assuming there is a reverse current flowing through the zener diode (ie it is
doing something useful) and check this assumption later. From the zener characteristics we then
see that the voltage across the zener diode is about 12 volts (1 volt or so). ie V
o
12 V. From the
input voltage graph we see V
in
is varying between about 18 V and 22 V (ie a range of 4 V). We
wish to find by how much V
o
will vary. (Hopefully its range will be much less than 4 V.)

When V
in
= 22 V, the voltage across the 1 k resistor will be V
in
- V
o
22 - 12 = 10 V.
Hence the current through the 1 k resistor will be
k 1
V 10
= 10 mA. This will be the current through
the zener diode too - there is nowhere else for the current to go. Checking with the zener
characteristics, we see that with 10 mA through it, the zener will have very nearly 12 V across it;
this justifies our original assumption.

ELEC 353 20
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
When V
in
= 18 V, the voltage across the 1 k resistor will be V
in
- V
o
18 - 12 = 6 V.
Hence the current through the 1 k resistor will be
k 1
V 6
= 6 mA. Again, this will be the current
through the zener too and, again, the zener characteristics show us that the voltage across the zener
is very nearly 12 V.

Actually, in this second case with only 6 mA through the zener there will be slightly less
voltage across the zener than in the first case where there was 10 mA through it. Since the current
through the zener has changed by 10 - 6 = 4 mA, we can use the slope of the zener characteristic
(50 = 50 volts/amp = 50 mV/mA) to find the difference in the voltage across the zener in the two
cases (V
in
= 22 V and V
in
= 18 V):

change in zener voltage = i .slope = i.(dynamic resistance)

= (4 mA) x (50 )
= 0.2 V = V
o
since the voltage across the zener is just the output voltage.

Thus our simple voltage regulator has taken a raw input voltage with a range of 4 V (18 - 22 V) and
produced a regulated output with a range of 0.2 V; this is an improvement of a factor of 20
2 . 0
4
= .
THE (BIPOLAR JUNCTION) TRANSISTOR (BJT)
The top panel of the diagram below shows a pn junction (with the p-side much more highly
doped than the n-side) in forward bias. In particular, it shows the excess hole concentration in the
n-region decaying exponentially with distance,
V
+ -
p n
log p
transition
region
x x = 0
p
p

p
n

p
p
DIODE
p p n
x
p
p
e

p
b

emitter base collector
x = W
b

V
be

V
ce

+ + - -
TRANSISTOR
n
b

log p
p
c

x = 0

|
|

\
|

=
p
L
x
p p exp ' , where the diffusion length, L
p
, is a few tens of m. The width of the
transition region is a few tenths of a m.

As shown in the bottom panel, a transistor is made by having the n-region width, W
b
, much
less than L
p
(typically, 0.5 m W
b
3 m) and having a negatively biased p-region joined to the
ELEC 353 21
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
end of the shortened n-region to collect the holes. Such a transistor is called a (pnp) bipolar
junction transistor (BPJT) and looks like two back-to-back pn junctions with the n-region being
narrow and much more lightly doped than the left-hand p-region.

The p-region on the left, from which the hole current originates, is called the emitter while
the (negatively biased) p-region on the right, which 'collects' the holes is called the collector. The
central n-region is called the base.
Amplifier action:
Because W
b
L
p
, most of the holes succeed in crossing the n-type base without suffering
recombination with the numerous electrons there. In a typical transistor, like this one, only about
one hole in 100 will suffer recombination in the base. For every 100 holes which leave the emitter,
one will recombine with an electron in the base and 99 will make it through to the collector
(strongly encouraged by the negative bias, V
bc
).

Each time a hole recombines with an electron in the base, that electron has to be replaced by
an electron flowing into the base terminal. If it were not, then a positive potential would very
quickly build up on the base repelling further holes and the hole current would stop. Thus this
small base current (electrons in conventional current out) is controlling a (conventional)
current 100 times as large flowing into the emitter lead and out the collector lead (driven by the
external battery[s]). Thus the transistor is an amplifier.

For a given transistor, the ratio of collector current, I
c
, to base current, I
b
, is nearly constant:
ie =
b
c
I
I

where the constant, , is the (common emitter) current gain of the transistor. Thus is an
important figure of merit for a transistor. Most (discrete) transistors have current gains in the range
50 - 300, with 100 being common.

Note that current continuity requires that


b c b e
I I I I ) 1 ( + = + =

where I
e
is the current flowing into the emitter while I
b
and I
c
are the currents flowing out of the
base and collector, respectively. Since, typically,
100
c
b
I
I , we have to a good approximation

b c e
I I I = for many purposes.
Base Diffusion Current :
The (hole) current in the base is a diffusion current (ie it is driven by the density gradient)
and, as we saw when we discussed the diffusion current in the diode, it is given by
dx
dp
kT J = .
Since the hole current through the base is very nearly constant (99% of the current makes it all
the way from the emitter to the collector), we can see that
dx
dp
must be constant across the base and
this is why the hole density versus distance in the transistor diagram above is shown as a straight
line. Note that the 'y-axis' of this diagram is shown as a log scale for the high (hole) densities (to
accommodate them) but as a linear scale for the low (hole) densities to emphasize this linear hole
gradient in the base. The hole density on the collector side of the base is nearly zero (or at least
small compared with the other hole concentrations in the transistor) because the strong electric field
ELEC 353 22
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
in the reverse biased base-collector junction pulls any such holes rapidly into the collector. Thus, if
p
b
' is the hole density on the emitter side of the base we have
b
b
e
W
p
kT
dx
dp
kT J
'
= =

or
b
b
e
W
p
kTA I
'
=
where J
e
and I
e
are respectively the current density and current, flowing out of the emitter (into the
base) and A is the crossectional area of the transistor.
Bias Currents:
The transistor is normally operated with substantial steady forward currents, I
e
, I
b
, and I
c
,
flowing as described above (where, of course, current continuity requires I
e
= I
b
+ I
c
). In particular,
the transistor works usefully as an amplifier only when the currents are flowing in the
directions shown above. At first sight, it might thus seem that the transistor would not be suitable
for amplifying AC currents (ie sinusoidal signals) because these must travel equally in both
directions.

However, this apparent difficulty is readily solved by arranging to have steady DC currents
I
e
, I
b
, and I
c
which are much larger than the AC signals which we wish to amplify and allowing
these small AC currents to superpose on top of these steady DC currents. As they oscillate, these
small AC currents then add to, and subtract from, the much larger DC currents so that the total
currents are always going through the transistor in the right direction for amplification.

In this way both the large DC component and the small AC component of the current in the
collector are respectively times greater than the corresponding components of the current in the
base.

These steady DC currents (on top of which the small AC signals are superimposed) are called
bias currents. A very common situation would be to arrange to have a collector bias current of
about 1 mA; the corresponding bias current in the base would then be about 0.01 mA = 10 A (for
100). Since I
e
= I
b
+ I
c
, I
e
= 1.01 1.0 mA. This is illustrated in the right hand panel of the
diagram below where the small AC signal is shown as a sine wave with (peak) amplitude of i
b
= 1
A in the base which corresponds to i
c
= 100 A = 0.1 mA in the collector. Note that all the
currents (AC, DC or total) in the base are 'mirrored' at an amplitude 100 (or ) times greater in the
collector. Note also the use of small i for the small AC signal currents and the use of big I for the
(big) DC bias currents.
0.60 0.65 0.70
1.0
2.0
I
e
(mA)
V
be
(V) time (s)
1.0
0.5
10
5
0
I
c
(mA) I
b
(A)
V
be
(V)
0.67
0.65


This bias current of about 1 mA flowing across the emitter-base junction is found to
correspond to a voltage drop of about 0.7 volt between the emitter and base terminals for a silicon
ELEC 353 23
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
transistor as shown in the left hand panel of the diagram above. Indeed, since this forward junction
current, I
e
, can be expected to be proportional to
|

\
|
kT
eV
be
exp as for the diode, and volts 025 . 0 =
e
kT
,
a change of emitter-base voltage by just 0.1 V will change I
e
(and hence I
b
) by a factor of (2.718)
4

60. This covers the normal range of bias currents used in most circuits and thus justifies the usual
assumption that a normally biased, working transistor has V
be
= 0.7 volt (0.1 V).
Dynamic Input Impedance:
As can be seen from the diagram above, the small AC oscillations in the base, emitter, and
collector currents are accompanied by small oscillations in the emitter-base voltage, V
be
.
Alternatively we might prefer to say that the small oscillation in V
be
(ie an AC input voltage of dV
be

= v
be
) is causing the small AC current in the base which, in turn, causes the small (though larger)
AC currents in the collector and emitter.

The ratio
e b
b
be
b
be
r
i
v
dI
dV
'
= = is called the (common emitter) dynamic input impedance of the
transistor. (The words 'common emitter' are used only to avoid possible confusion with some other [old] usages
which will not concern us here.) We can now calculate this dynamic input impedance.

Our discussion of the base diffusion current in the transistor showed that the emitter current
was proportional to the hole density gradient in the base, in particular that I
e
p
b
'. Our earlier
discussion of the pn junction under forward bias, when applied to our emitter-base junction, gives
|

\
|
=
kT
eV
p p
be
n b
exp ' .
Thus we can write |

\
|

kT
eV
I
be
e
exp
or
|

\
|
=
kT
eV
c I
be
e
exp where c is the proportionality constant.
Differentiating, we get


kT
eI
kT
eV
c
kT
e
dV
dI
e be
be
e
= |

\
|
= exp .

Hence, inverting,
e e
be
eI
kT
dI
dV
=
Recalling I
e
I
b
, gives
e
e b
b
be
eI
kT
r
dI
dV
= =
'
for the dynamic impedance of the emitter-base
junction. This is a resistance of about 5 k for a typical bias current of I
e
= 1 mA and a typical
good current gain of = 200. Note that this dynamic input resistance (the input resistance for small
AC signals) is inversely proportional to the (large) DC bias current.

In addition to this junction resistance, r
b'e
, between the emitter and base terminals there is
also the ohmic resistance of the (lightly doped) base region, r
bb'
, so that the total input dynamic
resistance between the base and emitter leads is given by

' ' bb e b be
r r r + = .
Typically r
bb'
is 100 - 200 so that, r
be
r
b'e
except at high currents or high frequencies.
ELEC 353 24
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
Diffusion Capacitance:
The holes in the base which constitute the diffusion current must be considered as stored
charge which must be increased or decreased when the current is altered and hence when the
emitter-base potential, V
be
, is altered. This constitutes a capacitance called the diffusion
capacitance which is one of the important factors which limit the high frequency response of the
(BPJT) transistor. (The negative charge on the 'capacitor' which balances the holes comes from a
slight [percentage] increase in the high electron concentration in the base.)

The charge stored in the base is given by

q
b
= e .
1
2
p
b
'. AW
b

which, after using the expression derived for the base diffusion current above, gives
kT
I W
e q
e b
b
2
2
1
=
so that the diffusion capacitance is given by
e b
b
be
e b
be
e
e
b
be
b
D
r kT
eW
dV
dI
kT
eW
dV
dI
dI
dq
dV
dq
C
'
2 2
2
1
2
1
= = = =
where r
b'e
is the dynamic resistance of the emitter-base (diode) junction, as above. It is worthwhile
to rewrite the equation as
kT
eW
C r
b
D e b
2
'
2
1
=
because this RC time constant, as we shall shortly see, is an important time constant in determining
the high frequency response of the transistor.

The above analysis assumes the whole of the charge across the full width of the base will change in response to a
change in V
be
but if this input is sinusoidal it will take time for the change to travel across the base. Millman and
Halkias (see reference list below) give this effect reducing C
D
by a factor of 2/3.
Two types of BPJT's:
In our description of the internal workings of the (bipolar junction transistor), we have
concentrated on the pnp type. The other type is npn and is at least as common and works in the
same way except that in an npn transistor electrons play the role that holes play in a pnp transistor.
Structure of BPJT's:
Transistors are not actually made by joining long thin bars of p- and n-type material as might
seem the case from the foregoing schematic diagrams. They are actually built up from a slab or
substrate of (doped) crystalline silicon by deposition from suitable gases (SiCl
4
, PH
3
, POCl
3
, B
2
H
6
)
in suitable concentrations at suitable temperatures. Diffusion is also used to implant suitable
doping into appropriate layers. Masking with photoresist and SiO
2
is needed to define the lateral
position of the layers.

The process of building up a nice regular lattice
plane by plane in this way is called the planar epitaxial
process. A description can be found in Millman &
Halkias (1972). The diagram below shows the structure
of a discrete npn (bipolar junction) transistor.

emitter
wire
base
wire
collector
wire
n
n p
ELEC 353 25
Semiconductors 2007-01-31 Neil R Thomson & John L Bhr
REFERENCES

Millman J and A Grabel
MicroElectronics, McGraw-Hill, 1987

Eisberg, R and R Resnick
Quantum Physics of Atoms Molecules, Solids, Nuclei, and Particles, Wiley, 1985.

Frederiksen, T M
Intuitive IC Electronics' , (National's Semiconductor Technology Series), McGraw-Hill, 1982.

Millman, J and C C Halkias
Integrated Electronics: Analogue and Digital Circuits and Systems, McGraw-Hill, 1972.

Sze, S M
Physics of Semiconductor Devices, Wiley, 1981.

van der Ziel, A
Solid State Physical Electronics, Prentice-Hall, 1968.

S-ar putea să vă placă și