Homework Set #2:

1a. Consider simple, body-centered, and face-centered cubic Bravais lattices. How many
“nearest neighbor” lattice points are there for each lattice point in the three lattice types?
How many lattice points are in each unit cell? (Note: It is conventional for a lattice point
“shared” by more than one unit cell to be divided evenly between the unit cells.)
Diamond cubic (Si, Ge) and zincblende (GaAs) lattices are face centered cubic.
However, each atom is tetrahedrally bonded to four nearest neighbors. Explain how a
diamond cubic or zincblende lattice can also be face centered cubic.

For a simple cubic lattice, it is clear that the nearest neighbor distance is just the
lattice parameter, a. Therefore, for a simple cubic lattice there are six (6) nearest
neighbors for any given lattice point.

For a body centered cubic (BCC) lattice, the nearest neighbor distance is half of
the body diagonal distance, a 3 2 . Therefore, for a BCC lattice there are eight
(8) nearest neighbors for any given lattice point.

For a face centered cubic (FCC) lattice, the nearest neighbor distance is half of
the face diagonal distance, a 2 2 . Hence, there are three groups of four lattice
points lying in three perpendicular face planes, that also lie at this distance from
any given lattice point. Therefore, there are twelve (12) nearest neighbors for any
given lattice point.

One can further observe that the FCC lattice is more densely packed since each
unit cell contains an equivalent of four lattice points, compared to two for the
BCC lattice, and one for a simple cubic lattice.

Both the diamond cubic and zincblende structures have an FCC Bravais lattice
with each lattice point associated with two atoms, i.e., basis group, rather than
just one as in the simple case. Alternatively, both the diamond cubic and the
zincblende lattice can be regarded as two primitive interpenetrating FCC lattices
offset by a displacement of a/4, a/4, a/4. (Of course, the lattice parameter, a, is
the same for both primitive lattices.) This geometry requires each atom to be
bonded to four other atoms in a tetrahedral coordination (bond angles 109° 28').
In the diamond cubic structure both atoms associated with a lattice point are the
same, e.g., silicon. In the zincblende structure, they are different, e.g., gallium
and arsenic.

b. Consider the intersection of [111]-type planes within a cubic crystal. What type of
solid figure does the intersection of [111]-type planes make? (Hint: The intersecting
planes need not be restricted to a single unit cell.) What then, is the angle between
adjacent [111] planes? Similarly, what is the angle between a [111] plane and a [100]
plane. (Use elementary geometry to find the angles.)
If one considers a cubic unit cell, any face of the cell is a [100]-type plane since
the origin can be translated arbitrarily and the unit cell has rotational and
inversion symmetry. Moreover, there are eight archetypical [111]-type planes,
viz., 111  ,  11 1  ,  1 11  ,  1 1 1  ,  1 1 1  ,  1 1 1  ,  1 1 1  , and
 1 1 1  planes. In this case, a solid figure will be generated if one considers
eight unit cells stacked to form a cube of dimension 2a:

The geometry of the resulting figure is illustrated above. Clearly, the desired
solid figure turns out to be an octahedron with all edges having a length of a 2 ,
i.e., each edge is a face diagonal.

Alternatively, each unit cell also has four [111]-type planes that intersect any
three corners for which each pair can be connected by a diagonal running along
a cell face, viz., 111  ,  11 1  ,  1 11  , and  1 1 1  planes. Therefore, if
one inscribes these four [111]-type planes within a single cell one can
immediately construct the figure:

Accordingly, it is evident that within a cubic unit cell [111]-type planes form a
tetrahedron. Moreover, it is further clear that any face diagonal lies in a [110]-
type plane and is perpendicular to a second [110]-type plane. Using this
observation, one can construct an isosceles triangle having a face diagonal as its
base and having two equivalent sides that both simulataneously lie in [110] and
[111]-type planes. Clearly, the sides of the triangle run from the precise center of
one face to diagonal cell corners on the opposite face. This is indicated by the
shaded region in the preceding figure. Naturally, the angle, , is the desired
angle between adjacent [111]-type planes as well as, by definition, the vertex
angle of the isosceles triangle. Obviously, the height, i.e., altitude, of the triangle
is merely the lattice parameter, a. Furthermore, by definition, the altitude runs
from the base to the vertex of the isosceles triangle and exactly bisects it into two
identical right triangles. Consequently, each hypoteneuse of the resulting two
right triangles corresponds to one of the equivalent sides of the original isosceles
triangle. It is now a simple matter to determine the hypoteneuse length, Lside, by
means of the Pythagorean Theorem:

2
Lside  a 2   a 2  a
3
 2 2

Of course, it follows from the elementary definition of the cosine (as the ratio of
the “adjacent side” to the hypoteneuse of a right triangle) that:

a 2
cos   
2 L 3
side

Therefore, , is given by the formula:

2
  2 arccos  70.52
3

In addition, the angle between [111] and [100]-type planes can easily be
determined by consideration of the same figure. Clearly, the two right triangles
obtained by bisection of the original isoceles triangle are, in addition to the right
angle, characterized by the same two angles. One of these is, of course, /2. The
other one (denoted as ) is precisely the angle between [111] and [100]-type
planes. (Obviously, the two base angles of the original isosceles triangle are both
equal to .) Since,  is clearly the complement of /2, one finds that  is given by
the trivial formula:

 
   54.74
2 2

Alternatively, the same result follows from the elementary definition of the sine
(as the ratio of the “opposite side” to the hypoteneuse of a right triangle):
 a 2
sin   
Lside 3

Hence, one finds that:

2
  arcsin  54.74
3

The geometry is illustrated below:

O 

2a. Assuming that rapid stirring conditions prevail, a CZ grown arsenic doped silicon
crystal is required to have a resistivity of 2.0  cm at a point exactly halfway between the
top and bottom of the ingot. Further assuming an initial 100 kg charge of pure silicon and
neglecting any silicon added to the melt by the seed, what is the amount of 0.01  cm
arsenic doped polysilicon that must be added to the melt to obtain the desired result? For
this condition, what is the resistivity of the ingot one quarter and three quarters of the way
between top and bottom? Assume no silicon is wasted during initial growth or the final
“pull-out” of the ingot, that the arsenic distribution coefficient of 0.27, that all resistivities
are measured at 300K, and that electron mobility is 1350 cm2/volt-sec.

Arsenic is, of course, an n-type extrinsic dopant. Thus, the concentration

corresponding to a resistivity of 2.0  cm can be determined simply from electron
mobility:

1
N As  n 
qe

1
N As  19
 2.312(1015 ) cm 3
(1.602(10 ) C)(2.0  cm)(1350 cm /volt sec)
2

The dopant concentration as a function of growth is given by the formula:

 K 1
 W 
Cs  KCo 1  
 Wo 

Since the melt is well-stirred, one uses the ideal value of the distribution
coefficient, K, which for arsenic is 0.27. When the crystal is half grown W Wo
equals one half irrespective of the initial amount of silicon. Thus, one can solve
for the initial concentration, Co, as follows:

1 K
C  W 
Co  s 1  
K  Wo 

2.312(1015 ) cm 3  1 
0.73
Co     5.5163(1015 )cm 3
0.27 2

The concentration of arsenic in the doping charge is:

1
N As  19
 4.624(1017 ) cm 3
(1.602(10 ) C)(0.01  cm)(1350 cm /volt sec)
2

This is much more concentrated than is required. Therefore, if one notes that the
volume of silicon is just the mass divided by the density, and then if w is defined
as the mass of doped silicon to be added, it follows that:

w ( w  100 kg)
(4.624(1017 ) cm 3 )  5.5163(1015 ) cm 3
 

Of course, one assumes that the density of silicon is unchanged by doping.

Clearly, density “cancels out” and, thus, it follows that:

5.5163(1017 ) kg cm 3
w  1.142 kg
4.5028(1017 ) cm 3

Hence, 1.142 kg of arsenic doped silicon must be added to the 100 kg of pure
silicon to obtain the correct initial doping.

To calculate resisitivity at one quarter and three quarters down the ingot, one
combines the formula for resistivity with the CZ growth formula:

1
 K 1
 W 
qKCo 1   e
 Wo 
 Substituting numerical values gives:

qKCoe  (1.602(1019 ) C)(0.27)(5.5163(1015 ) cm 3 )(1350 cm 2 /volt sec)

qKCoe  1.206 1 cm 1

Therefore at one quarter down the ingot resistivity is: 2.689  cm

and at three quarters down the ingot, resistivity is: 1.206  cm
Naturally, the resistivity falls because the dopant concentration rises from top to
bottom of the ingot.

b. If instead of rapid stirring, programmed growth conditions are imposed such that
effective boundary layer thickness, , is 0.5 cm during growth of the crystal and dopant
diffusivity in the melt, D, is 8(104) cm2/sec, what freezing rate is necessary to achieve an
effective segregation coefficient of 0.99? Would this seem a reasonable pull rate for a
practical CZ process?

The effective distribution coefficient is related to the thermodynamic distribution

coefficient by the formula:

K
Ke   R
K  (1  K )e D

Inverting this expression one finds that:

 1  K e  K   R D
  e
 K e  1  K 

From this, one constructs an expression for the freezing rate as follows:

D   1  K  K e  
R ln    
   K  1  K e  

Substituting numerical quantities gives:

(8(104 ) cm 2 /s)   1  0.27  0.99   3

R ln       8.944(10 ) cm/s
0.5 cm   0.27  1  0.99  

This rate implies that 32.1 cm of ingot are grown per hour. Thus a 2 meter ingot
would require 6.2 hours for growth. In practice, this is probably too fast for good
control of temperature and heat fluxes; however, it is not unreasonably fast.
 Accordingly, one can trade-off pull rate and rotation to optimize crystal quality.
Practical values result in growth times on the order of a day or a day and a half.

3. The lattice parameter of crystalline silicon is known with very high precision to have a
value of 0.543102064 nm. Assuming that energies of formation of a single vacancy and a
single self-interstitial (Ev and Ei) are 2.3 and 1.1 eV, respectively (values taken from
Handbook of Photovoltaic Science and Engineering by Antonio Luque and Steven
Hegedus), estimate equilibrium concentrations of vacancies and self-interstitials within a
pure silicon crystal at 27C and 1327C. Also, estimate the equilibrium constant at these
two temperatures. (Boltzmann’s constant  8.61735(105) eV/K; Avogadro’s
number  6.023(10 ) mole1)
23

As mentioned in the notes, N can be considered as atomic number density rather

than an absolute number of atoms. Accordingly, one can easily determine the
atomic number density in a silicon crystal by recalling that each unit cell of a
perfect crystal “contains” eight silicon atoms, hence:

8 atoms
N 7 3
 4.994(1022 ) atoms/cm 3
(0.5431(10 ) cm)

Although vacancy and self-interstitial “concentrations” can be ideally

constructed by considering “reactions” in which a defected crystal is formed
from a perfect crystal by moving atoms from the interior to the surface or vice-
versa, in reality within a bulk crystal they can be regarded as satisfying a mass
action equilibrium of the usual form:

K eq  [V ][ I ]

Here, [V] and [I] are defined as vacancy and self-interstitial concentrations,
respectively. Accordingly, an expression for the equilibrium constant can be
constructed as follows:

5 2  E  Ei 
K eq  N exp   v 
8  kT 

Upon substitution of the previous value for N, one finds that as a function of
temperature, the equilibrium constant has the form:

 39455.3K 
K eq  (1.55873(1045 ) cm 6 ) exp   
 T 

Thus:
at 27C(300K), the equilibrium constant is 1.18959(1012) cm6
 at 1327C(1600K), the equilibrium constant is 3.04273(1034) cm6

In analogy to mobile carriers, one expects that vacancy and self-interstitial

concentrations strictly generated thermally must be equal; hence:
at 27C(300K), [V] equals [I] equals 1.0906(106) cm3
at 1327C(1600K), [V] equals [I] equals 1.7443(1017) cm3

Clearly, at 300K equilibrium concentrations of vacancies and self-interstitials

are ridiculously low. Nevertheless,, significant concentrations of point defects
may exist at room temperature because they can be formed by athermal processes
such as damage or stress and, once formed, can be expected to recombine very
slowly. In contrast, at 1600K equilibrium concentrations are much larger and,
moreover, significant in comparison to atomic density. Indeed, a typical
published value for the melting point of silicon is 1685K and, of course, in the
neighborhood of this temperature one expects vacancy and self-interstitial
concentrations to approach the atomic concentration since melting completely
disrupts the crystal lattice.

In addition, it should be mentioned that vacancies and interstitials have a

fundamental role in impurity diffusion in silicon. These processes are generally
carried out at temperatures in the range of 1100-1300K. Accordingly, although
not large, vacancy and interstitial concentrations can be expected to be
significant.

4. Ignore temporarily the effect of lattice strain on the free energy of an oxygen
precipitate in a silicon crystal. At 1100C, the energy of formation of SiO2 is 900 kJ/mole
and the free energy of formation is 644 kJ/mole. Suppose that the surface energy of an
Si/SiO2 interface is 25000 erg/sq. cm. What is the size of a critical nucleus (i.e., what is
the minimum radius of a stable oxygen precipitate?) What is the effect of the omitted
strain term on the critical nucleus size? Assuming that strain energy is simply
proportional to the volume of the precipitate, what is the critical value of strain energy per
unit volume above which oxygen precipitates cannot be formed? (Note the unit
conversions: 1 J  1(107) ergs)

By definition, the free energy of formation of SiO2 is given by:

ASiO2  ESiO2  TS SiO2

Therefore, the free energy of formation of a precipitate containing N SiO2

“molecules” is:

A  NESiO2  NTSSiO2  A  gV
Here, A is the surface area and V is the volume of a single precipitate. For the
moment, one takes g equal to zero. It follows that N is given by the formula:

 No 
N  V  

 M SiO2 

where N o is Avogadro's number, MSiO2 is the formula weight of SiO2, and  is

mass density of SiO2. Thus,

 No 
A  V  (ESiO  TSSiO )  A  gV
 M SiO  2 2
 2 

 No 
A  V  ASiO  A  gV

 M SiO2 
2

Now, as is usual, one considers the precipitate to be spherical. The volume of a

sphere of radius, r, is 4r 3/3, and the surface area is 4r 2, thus,

A r 3  N o 
  ASiO2  g   r 2
4 3  M SiO2 

The critical nucleus size is determined by the maximum of the free energy as a
function of radius. Thus, one determines the criticality condition as follows:

1   N o 
0 A  rcrit
2
 A  g   2rcrit 
4 r M SiO 
 SiO2 
2

 N o 
0  rcrit  ASiO2  g   2
 M SiO2 

 2
rcrit 
 N o 
 A  g 
M SiO 
 SiO2 
2

One can now substitute the numerical quantities into this expression. At this
point it is clear that the heat of formation is not needed since the free energy of
formation is given explicitly. First, however, convert free energy to cgs units:
 (644 kJ/mole )(1000 J/kJ)(107 erg/J )
ASiO2 
6.023(1023 ) mole 1

ASiO2  1.069(1011) erg

Clearly, the free energy of formation must be formally negative since SiO2 forms
spontaneously. Thus, if strain energy is ignored, the critical radius is given by the
expression:

 2(25000 erg/cm 2 )
rcrit   2.056 nm
 (2.27 g/cm 3 )(6.023(1023 ) mole 1 )(1.069(10 11) erg ) 
 
 60.1 g/mole 

If one considers the functional behavior of the various components of the free
energy, it is clear that for r equal to zero, the surface term invariably dominates
so that an SiO2 precipitate is unstable, i.e., the free energy increases with
increasing radius. At the critical radius, the free energy of formation term begins
to dominate which causes the SiO2 precipitate to become stable, i.e., the free
energy decreases with increasing radius. Accordingly, precipitate growth is
thermodynamically favored. The result obtained above indicates that the critical
radius is fairly small and that only a small cluster of SiO4 tetrahedra would be
sufficient to nucleate stable precipitates at this temperature. Therefore, if this
result is valid, then oxygen precipitates can be expected to be prevalent in a
silicon crystal even at a high temperature such as 1100C. It is well-known that
this is not the case. Why is this result wrong?

Obviously, the effect of strain energy has not been included. Addition of a
positive strain energy term serves to make the critical radius larger since it offsets
the negative free energy of formation. Clearly, the strain energy must make a
significant contribution to the overall free energy of an SiO2 precipitate. Indeed,
if g were to exactly balance the free energy of formation, the denominator of the
preceding formula would vanish implying that critical radius becomes infinite and
indicating that SiO2 precipitates never become stable. Of course, if g dominates,
the calculated radius becomes negative which is an unphysical result and also
indicates that precipitates are exceedingly unstable.

Consequently, it is clear from preceding observations that oxygen precipitates can

never be stable if the strain energy entirely compensates the free energy of
formation. This will occur for the condition:

N o
g ASiO2
M SiO2
 (2.27 g/cm 3 )(6.023(1023 ) mole 1 )(1.069(1011) erg )
g  2.434(1011) erg/cm 3
60.1 g/mole

5. Consider an ingot of crystalline of silicon uniformly contaminated with antimony (Sb)

which has a distribution equilibrium coefficient, K, of 0.020. (Assume, K and Ke are the
same.) What is the lowest concentration of Sb that can be obtained by a single pass zone
refining process? Where in the crystal does this occur? What is the Sb concentration at a
point halfway along the ingot?

The initial unrefined antimony concentration, although uniform, was not specified
numerically, but it can be represented as a variable, Co. The equation relating
concentration in the solid to the initial concentration for a single pass zone
refining process was given in the notes and is as follows:

Cs ( x)  Co 1  (1  K )e L 
 Kx

 

Obviously, the lowest concentration occurs at the very beginning of refining and
is merely the product of K and Co. Hence, for the present case this is just
0.02Co. Concomitantly, at a poiny halfway along the ingot, x/L is 0.5.
Therefore, upon substitution into the preceding formula, one finds that:

Cs (0.5)  Co (1  (1  0.02)e0.020.5 )  (0.0297511)Co

Clearly, the concentration is almost 50% larger halfway along the ingot. Even
so, it is still quite small compared to Co , thus, for contaminants with small
distribution coefficents, zone refining is quite effective.