 S. O. Kasap, 1990 - 2002: v.1.

1)
Diffusion: Concepts and Solved Problems ( 1
An e-Booklet

DIFFUSION: CONCEPTS AND SOLVED PROBLEMS

Safa Kasap
Department of Electrical Engineering
University of Saskatchewan
Canada

1. Fick’s First Law

Fick’s first law relates the net diffusion flux of a given species of atoms to their concentration gradient.
Particle flux Γ is the number of particles crossing unit area per unit time. Thus if ∆N particles cross an
area A in time ∆t then, by definition, the particle flux is
∆N
Γ= (1)
A∆t
In describing flux, Γ, we must distinguish between species of particles. For example, the flux of
He atoms refers only the number of He atoms crossing unit area in unit time.
If Γ is the net particle flux at point x and C is the concentration at that point, then according to
Fick's first law, net particle flux, Γ, at a position, x, is proportional to the concentration gradient, dC/dx,
i.e. dC
Γ = −D (2)
dx
where the negative sign ensures that the flux direction is from high to low concentration regions
(negative gradient) and D is a proportionality constant called the diffusion coefficient. D depends on the
species of diffusing particles and the medium (e.g. the crystal structure) in which they are diffusing.
Equation (2), Fick's first law, represents the relationship between the net particle flux and the driving force
which is the concentration gradient. The diffusion coefficient D for atomic diffusions in solids is thermally
activated,

D = Do exp − d 
Q
(3)
 RT 
where Do is a constant, Qd is the activation energy for the diffusion process and R is the gas
constant.

Problem: Purification of hydrogen by diffusion

Palladium has the ability to absorb hydrogen gas 800-900 times its volume. The hydrogen molecules
dissociate on the Pd surface and become dissolved as H-atoms in the Pd structure taking up interstitial
sites. When there is pressurized hydrogen gas on one side A of a thin Pd sheet then, as a result of
diffusion, hydrogen gas appears on the other side B of the sheet as shown in Figure 2. If the pressures PA
and PB at A and B are kept constant with PA > PB then eventually steady state is reached in which there is a
constant hydrogen concentration gradient and there is a constant net flux of hydrogen atoms from A to B.
Hydrogen can be purified by diffusing the gas through a palladium sheet because the impurity gas
molecules would have a much smaller diffusion coefficient. In a certain purification process, the
concentration of hydrogen, during a steady state operation, at the high and low concentration sides of the
Pd sheet have been found to be 1 at.% and 0.2 at.%. How many kilograms of hydrogen can pass through
per day through a 5 mm thick Pd membrane having an area of 0.5 m2 at 500° C?
  S. O. Kasap, 1990 - 2002: v.1.1)
Diffusion: Concepts and Solved Problems ( 2
An e-Booklet

C(x) = Concentration of particles

= Net particle flux

dC
dx

x
A B
There is always a net flux,
diffusion, of particles from higher
to lower concentrations. The flux
depends on the concentration
gradient.

Figure 1
The diffusion coefficient D of hydrogen in Pd has an activation energy Qd of 24.6 kJ mol-1 and a
pre-exponential constant Do of 4.5 × 10-7 m2 s-1. The atomic mass and density of Pd are 106.5 g mol-1 and
12 g cm-3.
A B
H-gas Pd-Sheet H-gas

PA PB

C(x)

CA

CB
x
A B
d

A schematic illustration of hydrogen gas diffusion through a thin Pd sheet under

steady state conditions. There is a concentration gradient of H atoms from
surface A to surface B. The pressure PA is greater than the pressure PB .

Figure 2

Solution
The problem is schematically illustrated in Figure 2. The concentration of hydrogen at A and B is given as
an atomic percentage. To obtain the number of H-atoms per unit volume we first need to find the atomic
concentration in Pd. NA is Avogadro's number, Mat is the atomic mass and ρ is the density of Pd then the
number of Pd atoms per unit volume is
  S. O. Kasap, 1990 - 2002: v.1.1)
Diffusion: Concepts and Solved Problems ( 3
An e-Booklet

ρN A (12.0 × 10 kg m )(6.022 × 10 mol )

3 −3 23 −1

nPd = =
Mat (106.4 × 10 −3 kg mol −1 )
i.e. nPd = 6.8 × 1028 atoms m-3.
The concentrations of H-atoms at A and B are,
CA = (0.01)nPd = 0.01(6.8 × 1028 m-3) = 6.79 × 1026 H-atoms m-3.
and CB = (0.002)nPd = 0.002(6.8 × 1028 m-3) = 1.36 × 1026 H-atoms m-3.
As the pressures at A and B are constant we can take the concentrations at A and B to be constant
as well. Given d = 5 × 10-3 m, the concentration gradient of H-atoms through the Pd sheet is then
dC C − CB
=− A =−
(6.79 × 10 26
m −3 ) − (1.36 × 10 26 m −3 )
dx d (5 × 10 −3
m)

= −1.086 × 1029 m-4.

We need the diffusion coefficient at T = 500 °C, Qd = 24.6 × 103 J mol-1 and Do = 4.5 × 10-7 m2 s-
1
. Then at T = 500 + 273 = 773 K,

D = Do exp − d 
Q
 RT 

i.e. D = ( 4.5 × 10 −7

m s ) exp −
2 −1 ( 24.6 × 10 3 J mol −1 ) 

 (8.3145 J mol K )(773 K ) 
−1 −1

or D = 9.80 × 10-9 m2 s-1.

Applying Fick’s first law,

= −(9.80 × 10 −9 m 2s −1 )( −1.086 × 10 29 m -4 )
dC
Γ = ±D
dx
i.e. Γ = 1.064 × 1021 H-atoms m-2 s-1.
The mass of one H-atom is
mH = Mat/NA = (1.008 × 10-3 g mol-1)/(6.022 × 1023 mol-1) = 1.67 × 10-27 kg.
Over a time interval ∆t = 1 day, ∆t = 24 × 60 × 60 seconds, and over an area A = 0.5 m 2, the total
mass Mday of hydrogen diffusing through is
Mday = (mH)(Number of H-atoms flowing in 24 hours) = mH(A∆tΓ)
= (1.67 × 10-27 kg)(0.5 m2)(24 × 60 × 60 s)(1.064 × 1021 m-2 s-1)
= 0.077 kg or 77 grams
It takes 13 days to purify 1 kg of H-gas.
  S. O. Kasap, 1990 - 2002: v.1.1)
Diffusion: Concepts and Solved Problems ( 4
An e-Booklet

Problem: Parabolic growth in diffusion limited oxidation

Consider the growth of an already present oxide (SiO2) layer on the surface of a Si crystal. The actual
oxidation reaction occurs at the Si-SiO2 interface. If the oxide is sufficiently thick then the oxidation
reaction that takes place at the Si-SiO2 interface will be limited by the rate at which oxidant molecules arrive
by diffusing through the oxide layer as depicted in Figure 3. Show that if x o is the initial thickness of the
oxide at time t = 0 then the thickness at time t is
x 2 = x o2 + Bt

where B is a thermally activated constant , that is B = Boexp(−Qd/RT).

O2
Gas
dx / dt =Growth rate
CO Oxide surface

Oxide x = Thickness
CI Interface The oxidation reaction occurs
at the Si-SiO2 interface. The
Si crystal oxidant first has to diffuse
through the oxide layer to the
interface.

Figure 3

Solution
For simplicity assume a unit surface area (A = 1). Since x is the thickness, the growth dx/dt is limited by
the flux Γ of O2 molecules arriving at the interface. Using Fick's first law,
dx dC
∝ Γ(O2 molecules) = − D
dt dx
dx C − CI
= KD O
dt x
where K is a proportionality constant (temperature independent), D is the diffusion coefficient of
the oxidant (O2) in SiO2, and CO and CΙ are the concentrations of the oxidant at the surface and at the
interface respectively. CO is determined by the partial pressure of O2 on the surface and CΙ is determined
by the concentration of O2 in the oxide (number of oxygen molecules in SiO2 per unit volume) since the
oxide formation is at the interface. We can view the oxidation at the interface as the immediate removal
and incorporation of the arriving O2 molecules into the formation of an oxide so that the O2 concentration
here, CΙ, cannot exceed that in the oxide.
To obtain the thickness x at time t we integrate the above equation,

xdx = KD(CO − CI )∫ dt
x t
∫xo 0

giving x 2 = x o2 + Bt

where B = KD(CO−CI). Since D = Doexp(−Ed/kT), B is thermally activated. Note that increasing the
pressure increases CO and thus B and hence the growth rate. The activation energy Ed depends on the
oxidant species diffusing through the oxide. For oxidation in an oxygen gas, diffusing oxidants are O2
and Ed is 1.24 eV atom-1. For oxidation in steam, the diffusing species are H2O and Ed is 0.79 eV atom-1.
Note that the oxide thickness, after a given oxidation time, increases with temperature and pressure.
  S. O. Kasap, 1990 - 2002: v.1.1)
Diffusion: Concepts and Solved Problems ( 5
An e-Booklet

2. Fick’s Second Law and Diffusion from an Unlimited Supply

Fick’s second law describes the kinetics of time-dependent diffusion in which the concentration C(x,t) at a
point x at time t changes, that is the diffusion process is not steady state. Fick's second law relates the
space and time derivatives of the concentration C(x,t), through a partial differential equation:
∂ 2 C( x, t ) ∂C( x, t )
D = (4)
∂x 2 ∂t
where ∂C /∂t represents differentiation with respect to t only, keeping x constant, ∂C /∂x represent
differentiation with respect to x only, keeping t constant. D is the diffusion coefficient of diffusing species
of atoms in the host material. This equation has to be solved for a given set of boundary conditions. The
solution depends intimately on the boundary conditions.
He gas Pd Crystal He gas Pd Crystal C(x, t) = He concentration in Pd

Cs
t = t2

t=0 t = t1

Co x
t=0 t = t1 x=0

Diffusion with an unlimited supply of diffusing species. The surface

concentration is constant.
Figure 4
Consider the diffusion of a given species of atoms (such He) into a given host crystal (such as Pd)
as shown in Figure 4. Suppose that the surface concentration Cs of diffusing atoms is constant, there is a
small concentration C o in the bulk (called the background concentration) and the crystal is infinitely long.
Then these conditions are
C(0, t ) = Cs at the surface at any time 

C( x, 0) = Co at t = 0 anywhere in the bulk  (5)
C(∞, t ) = Co at any time at far end 
These boundary condition define diffusion from an unlimited supply (or constant surface source).
Given the boundary conditions in Eq. (5), Fick's second law in Eq. (4) can be solved to obtain the
concentration C(x,t) at x at time t,
C( x, t ) − Co  x 
= 1 − erf   (6)
Cs − Co  2 Dt 
where erf(z), called the error function of z, is a function in which z is the argument and is tabulated
like many other functions such as log(z), sin(z) etc. It is clear that in this particular diffusion case, z =
x/[2√(Dt)]. The error function is defined in terms of an integration as
2 z − y2
erf ( z ) = ∫e
π 0
dy (7)

Note: 1−erf(z) = erfc(z) is called the complementary error function.

  S. O. Kasap, 1990 - 2002: v.1.1)
Diffusion: Concepts and Solved Problems ( 6
An e-Booklet

Problem: Carburization of steel

The surfaces of steels for certain applications are hardened by diffusing carbon atoms into the steel at
elevated temperatures. Typically a steel component is exposed to methane gas at high temperatures. The
methane molecules dissociate at the surface and the carbon diffuses into the steel. Suppose that 0.15 wt.%
C steel (iron with 0.15 wt.% C) is carburized at 700° C for 2 hours during which the surface concentration
of carbon is 1.5 wt.%. What is the carbon concentration at 1 mm beneath the surface?
What is the time required to achieve the same concentration by carburizing at 600° C? The
diffusion coefficient of carbon in BCC iron has Qd = 80.3 kJ mol-1 and Do = 6.2 × 10-7 m2 s-1.

Solution
Since the surface concentration is maintained constant, the diffusion process obeys Eq. (6). At the
surface, Cs = 1.5 wt.%. At time t = 0, the concentration of carbon in the bulk is Co = 0.15 wt.%. At T =
700 + 273 = 973K we have

 Q  
D = Do exp − d = 6.2 × 10−7 exp −
(80300 J mol−1) 

 (8.315 J mol K )(973K) 
 RT  −1 −1

= 3.03 × 10-11 m2 s-1

At x = 10-3 m (1mm), at time t = 2 × 60 × 60 seconds (2 hrs) the argument of erf(z) in Eq. (6) is

z=
x
=
(1 × 10 −3
m)
= 1.070
2 Dt 2 (3.03 × 10 −11
m 2 s-1 )(2 × 60 × 60 s)
Thus, Eq. (6) for C(x,t) = C(1 mm, 2 hrs) is,
C(1 mm, 2 hrs) − 0.15
= 1 − erf (1.070)
1.5 − 0.15
To find C at x = 1 mm and t = 2 hrs we therefore need erf(z) = erf(1.070). We can carry out a
numerical integration of Eq. (7), use a mathematical software package (or a calculator that has the error
function), or use the error function tables with appropriate interpolation. The simplest is the numerical
integration,
2 1.070
erf ( z ) = erf (1.070) = ∫ e − y dy = 0.8698
2

π 0

C(1 mm, 2 hrs) − 0.15

Then, = 1 − erf (1.070) = 1 − 0.8698
1.5 − 0.15
or C(1 mm, 2 hrs) = 0.326 wt.%
At the lower temperature of 600° C the diffusion coefficient will be smaller. At T′= 600 + 273 =
873K we have,

′
 Qd 
D = Do exp − = . ×
6 2 10 −7

exp −
(80300 J mol−1 ) 

 (8.315 J mol K )(873 K) 
 RT ′  -1 −1

= 9.73 × 10-12 m2 s-1.

We need the time t′ that it takes for the concentration at x = 1 mm to reach the same value of 0.326
wt.%. The concentrations C(x,t), Cs and Co are therefore the same, 0.326 wt.%, 1.5 wt.% and 0.15
  S. O. Kasap, 1990 - 2002: v.1.1)
Diffusion: Concepts and Solved Problems ( 7
An e-Booklet

wt.% respectively. The left hand side of Eq. (6) thus remains the same which means that we must have
the same z at time t′.
x x
z= =
2 Dt 2 D′t′
or D′ t′ = D t

Dt ( 3.03 × 10 m s )(2 hrs)

−11 2 −1

giving t′ = = = 6.2 hrs

D′ (9.73 × 10-12 m2 s−1)

Problem: Diffusion doping in semiconductor technology

In electronic device fabrication, semiconductors are doped by diffusion processes. A Si crystal is to be
doped with boron (B) atoms by exposing the Si crystal surface to a B gas at high temperatures thereby
allowing the B atoms to diffuse into the bulk. At the crystal surface the boron concentration Cs is constant
and is determined by the solubility limit of B in Si at the diffusion temperature. At 1000 °C the solubility
limit of B in Si is about 3.5 × 1020 cm-3.
Consider a wafer of Si crystal which has a uniform phosphorus (P) doping of 1 × 1016 cm-3 in the
bulk. The crystal surface is then exposed to a high concentration of B atoms which diffuse into the
crystal. When at a certain point in the crystal, at x = x j from the surface, the B concentration, C(x j,t), is
equal to the P concentration, then that point x j becomes the junction of the pn junction device (diode).
Suppose that the above Si wafer has no background B doping in the bulk and is exposed to a B gas at
1000 °C. The pn junction is required to be at a distance 0.5 µm from the surface. How long should the
diffusion process be allowed to take place? Where will be the junction if the diffusion goes on for 2hrs?
Note: The diffusion coefficient of B in Si is 1.5 × 10-14 cm2 s-1 at 1000° C
C(x, t)

Cs B concentration profile at t = tj

P concentration
CP

Co x
xj

Diffusion of B atoms into a P doped Si

crystal
Figure 5

Solution
Since the surface concentration is maintained constant, the diffusion process obeys Eq. (6). At the
surface, Cs = 3.5 × 1020 cm-3 and there is no background B concentration in the crystal at t = 0 so Co = 0.
At the junction, x = xj = 0.5 µm, Eq. (6) has C(x,t) = CP (P concentration = 1 × 1016 cm-3)

( )
1 − erf z j =
( )
C x j , t − Co
=
1 × 1016 cm -3 − 0
= 2.86 × 10 −5
Cs − Co 3.5 × 10 20 cm -3 − 0
  S. O. Kasap, 1990 - 2002: v.1.1)
Diffusion: Concepts and Solved Problems ( 8
An e-Booklet

where zj = xj/[2√(Dt). We need the zj value for erfc(zj) = 1− erf(zj) = 2.86 × 10-5. From erfc(z)
vs. z graphs or from erfc(z) tables, zj ≈ 2.9. We note that it is more convenient to work with the
complementary error function than the error function itself when erf(z) is close to 1; here erf(zj) = 1 − 2.86
× 10-5 = 0.9999714. Thus, 2from z = x/[2√(Dt)],
1  xj 
2
1  0.5 × 10 −6 m 
t=  =
D  2 z j  1.5 × 10 −14 × 10 −4 m 2 s-1  2(2.9) 
i.e. t = 4954 s or 1.37 hrs.
If the diffusion continues for a time t′ = 2 hrs, then the junction is at x j′. Since the junction is
defined as the location where C(x,t) = CP, the term erfc(zj) = 1− erf(zj) = 2.86 × 10-5 remains the same and
hence zj′ = 2.9.

x ′j = 2 z′j Dt′ = 2(2.9) (1.5 × 10−14 × 10−4 m2 s-1 )(2 × 60 × 60 s)

i.e. xj′ = 6.03 × 10-7 m or 0.60 µm.

3. Fick’s Second Law and Gaussian Diffusion

Consider diffusion involving a constant number of diffusing species of atoms all initially concentrated in a
very thin surface region of the crystal as depicted in Figure 6. This type of diffusion is often called tracer
diffusion because we can follow the time evolution of the concentration profile of diffusing atoms if the
diffusing species are radioactive. The boundary conditions in this case are such that the total number of
diffusing atoms per unit area remains constant and equal to N o . Suppose that initially, at t = 0, all the
diffusing atoms are contained within a very small thickness xo. Then the concentration at the surface Co =
N o /x o . Assuming that the crystal is very long then the concentration profile at any point x at any time t
obeys a Gaussian function of the form
No  x2 
C( x , t ) = exp −  ; x > xo (8)
πDt  4 Dt 
C(x, t)

Cs
Crystal Crystal
t=0
t = t1

t = t2
t=∞
x
xo
t=0 t = t1 x = 0 x1

Diffusion with a limited supply of diffusing species. The total number of

diffusing particles is constant.

Figure 6
The concentration profiles C(x,t) vs. x at times t = t1 and t = t2 in Figure 6 follow Eq. (8). If we
were to substitute Eq. (8) into Fick's second law we would find that it satisfies the law; it is a solution of
Eq. (4). This type of diffusion is called Gaussian diffusion. It arises whenever there is a diffusional
spread of a constant number of particles which are initially contained in a thin sheet. The number of
diffusing species of particles remains constant. The area under the C(x,t) vs. x profiles in Figure 6
  S. O. Kasap, 1990 - 2002: v.1.1)
Diffusion: Concepts and Solved Problems ( 9
An e-Booklet

therefore represent the number of diffusing particles per unit area or No. For an infinitely long sample, the
concentration profile across the specimen after infinite time becomes very small or C(x,t)→ 0 as t → ∞,
though the area under the curve is always N o . In practice, the sample length is finite and the eventual
concentration after infinite time is not zero but Co where CoL = C sxo and L is the sample length as shown
in Figure 6.

Problem: Diffusion doping in semiconductor technology

The solubility limit of boron atoms on the surface of the silicon crystal is of the order of 4 × 1020 cm-3
around 1000 °C. Suppose that at t = 0 at 1100 °C, the surface of a Si crystal has been suddenly saturated
with B atoms and then the boron supply is turned off. The boron atoms with time will diffuse into the
bulk of crystal. If we assume initially all the B atoms were within a lattice dimension a (0.543nm)what
will be concentration of B at 1 µm (micron) from the surface after 1 hour at 1100 °C? How many B atoms
are there in parts per billion (ppb)? What would have been the concentration had the temperature been
1000 °C (9% lower temperature)?
The diffusion coefficient of B in Si has Do = 0.76 cm2 s-1, and Ed = 3.46 eV/atom.

Solution
We are given the surface concentration of B as Cs = 1 × 1020 cm-3. Within a lattice parameter a there are N o
number of B atoms per cm2, where
No = a Cs = (0.543 × 10-7 cm)(4 × 1020 cm-3) = 2.17 × 1013 cm-2.
The diffusion coefficient D at 1100 °C or 1373K is
 Ed   (3.46 eV)(1.602 × 10 −19 J/eV) 
 kT  (
D = Do exp − = 0.76 cm s ) exp −
2 −1

 (1.3807 × 10 J K -1 )(1373 K ) 
−23

= 1.52 × 10-13 cm2 s-1.

Distance of interest x = 1 µm or 10-4 cm and time of interest t = 1 × 60 × 60 seconds. Then,

(10 cm)−4 2
x2
= = 4.57
4 Dt 4(1.52 × 10 −13 cm 2 s-1 )(1 × 60 × 60 s)

and the concentration at x = 1 µm after 1 hr can be found from

No  x2 
C( x, t ) = exp − 
πDt  4 Dt 
2.17 × 1013 cm −2
so that C(1 µm,1 hr ) = exp( −4.57)
π (1.52 × 10 −13 cm 2 s −1 )(1 × 60 × 60 s)

= 5.43 × 1015 cm-3.

The concentration of Si atoms in the crystal is
ρN A (2.33 × 10 kg m )(6.022 × 10 mol )
3 −3 23 −1

nSi = = = 5 × 10 28 m −3
Mat (28.09 × 10 kg mol )
−3 −1

Relative B concentration is (5.43 × 1015 cm-3 )/(5 × 1022 cm-3) or 1.086 × 10-7 or 108.6 ppb (parts
per billion).
  S. O. Kasap, 1990 - 2002: v.1.1)
Diffusion: Concepts and Solved Problems ( 10
An e-Booklet

We can repeat the calculation for diffusion at 1000 °C. The new diffusion coefficient at T′ = 1000
+ 273 = 1273K becomes, D′ = 1.525 × 10-14 cm2 s-1. Then

(10−4 cm)
2
x2
= = 45.54
4 D′t 4 (1.525 × 10−14 cm2 s-1 )(1 × 60 × 60 s)

2.17 × 1013
which means C(1µm,1hr ) = exp( −45.54) ≈ 2.76 × 10 −2
π (1.525 × 10 −14
)(1 × 60 × 60)
It is apparent that when the temperature is decreased by 9%, the diffusion becomes almost
insignificant over this time scale. Indeed, for the concentration at x = 1µm to reach 5.43 × 1015 cm-3 at
1000 °C we have to hold the crystal at this temperature for ~10 hours. This can be shown quite simply by
finding the new time t′ which makes the concentration at x = 1µm at 1000 °C equal to 5.43 × 1015 cm-3,
i.e.
No  x2 
C ( x, t′) = exp−  = 5.43 × 1015 cm−3
πD′t′  4 D′t′ 

The above can be solved by plotting C(x,t′) against t′ and finding that value of t′ where C(x,t′) =
5.43 × 1015 cm-3. This is left as an exercise for the student. (Answer 10 hours.) Clearly the importance of
temperature cannot be overemphasized.

Problem: “Gaussian wind”

Consider the Gaussian diffusion of constant number N o of particles from the surface of a crystal into the
bulk. Show that the particle concentration C(x1, t) and flux Γ(x1, t) at a location reach their maxima at time
tc and tf given by
x2 x2
tc = 1 and tf = 1
2D 6D
so that tc/tf = 3, that is, the flux reaches a maximum first and then, three times later, the
concentration reaches a maximum. Sketch schematically the concentration and flux as a function of time at
x = x 1.
Solution
The concentration C(x,t) at position x and t is given by
No − 12  x2 
C( x , t ) = t exp − 
πD  4 Dt 

The flux Γ at position x and time t can be found from the concentration gradient because Γ = −
D(dC/dx). To find the gradient of C(x,t) at a given instant, we differentiate C(x,t) with respect to x but
keeping t constant. This is called partial differentiation of C(x,t) and is written as∂C/∂x,
∂C( x, t )  2 x  No − 12  x2 
Γ ( x, t ) = − D = t exp − 
∂x  4 Dt  πD  4 Dt 
No  x2 
Γ ( x, t ) =
− 23
i.e. xt exp − 
2 πD  4 Dt 
  S. O. Kasap, 1990 - 2002: v.1.1)
Diffusion: Concepts and Solved Problems ( 11
An e-Booklet

To find the maximum in C(x,t) at x = x1, occurring over time at t = tc, we use
 ∂C( x1 , t ) 
 ∂t  = 0
 t = tc

 No  1 − 23    x12   No − 12   x12 −2   x12  

− t
 πD  2 c   exp  −  + t  t  exp −  = 0
 4 Dtc   πD   4 D c   4 Dtc  
c

Canceling out terms and simplifying, we find,
1  x 
2
1 3
− t c − 2 + t c − 2  1 t c −2  = 0
2  4D 

x12
i.e. tc =
2D
To find the maximum in Γ(x,t) at x = x1, occurring over time at t = tf, we use
 ∂Γ( x1 , t ) 
 ∂t  =0
 t = t f

 No  3 5   x    No  2
3  x   x2  
x1   − t f − 2  exp − 1   + 
2
 x1t f − 2   1 t −f 2  exp − 1   = 0
 2 πD   2    4 Dt f    2 πD   4 D   4 Dt f  
  
canceling out terms and simplifying we find
x12
tf =
6D
Thus, the ratio tc/tf = 3 and the concentration reaches its peak after the flux. The expressions for
C(x1,t) and Γ(x1,t) are schematically shown in Figure 7 (it is left as an exercise for the student to verify the
sketches)
C(x1, t) (x1, t)

t t
tc tf

Concentration, C(x1, t) and particle flux (x1, t) at position x1 as a function of time.

Figure 7
  S. O. Kasap, 1990 - 2002: v.1.1)
Diffusion: Concepts and Solved Problems ( 12
An e-Booklet

NOTATION
A area (m2) L sample length (m)
a lattice parameter M at atomic mass
C concentration of particles m mass (kg)
D diffusion coefficient (m2 s-1) NA Avogadro’s number (6.022 × 1023 mol-1)
Do pre-exponential constant in diffusion equation (D No total number of diffusing particles per unit area
= Doexp(-Qd/RT)) P pressure (Pa)
∆ change Qd activation energy for diffusion process (J mol-1)
∆N number of particles R gas constant (8.31451 J K-1 mole-1)
∆t time period ρ density (kg m-3)
d thickness of the material that is being diffused T temperature (K)
through
t time
δ a small change, an incremental change
x a point; oxide thickness
Γ Particle flux (particles m-2 s-1)

USEFUL DEFINITIONS
Activation energy is the potential energy barrier that prevents a system from changing from state to another. For
example, if two atoms A and B get together to form a product AB, the activation energy is the potential energy
barrier against the formation of this product. It is the minimum energy which the reactant atom or molecule must
have to be able to reach the activated state and hence form the product. The probability that a system has an energy
equal to the activation energy is proportional to the Boltzmann factor: exp(-EA/kT), where E A is the activation
energy, k is the Boltzmann constant and T is the temperature (Kelvins).
Diffusion is a random process by which particles move from high concentration regions to low concentration regions.
There is a flow of particles of a given species from high to low concentration regions by virtue of their random
motions. Diffusion is a process by which atoms migrate by virtue of their random thermal motions.
Diffusion coefficient is a measure of the rate at which atoms diffuse. It depends on the nature of the diffusion process
and typically it is temperature dependent. It is defined as the magnitude of diffusion flux per unit concentration
gradient.
Doping is the process in which dopants are incorporated into bulk crystals or thin layers of semiconductors by using a
variety of processes. These include doping during bulk material or layer fabrication (in-situ doping), and also use of
diffusion and ion implantation. These approaches all are designed to permit the controlled introduction of precise
amounts of an impurity to create n- or p-doped material. Using masking, implantation and diffusion processes
permits the doping of only selected regions of material.
Error function is a standard mathematical function for which there are extensive tables available, which is usually written
z
2 − y 2 dy . The error function has the following properties: erf(z) = erf(-
as erf(z) and defined as follows: erf( z ) = ∫e
π 0
z), erf(0) = 0 and erf() = 1. This function, and the so-called complementary error function erfc(z) = 1 - erf(z), are used
extensively in simplifying analytical solutions for heat and mass diffusion in solids.
Flux is a term used to describe the rate of flow through a unit area. If ∆N is the number of particles flowing through an area
A in time ∆t, then particle flux, Γ, is defined by Γ = ∆N/(A∆t). If an amount of radiation energy ∆E flows through
an area A in time ∆t, energy flux is ΓE = ∆E/(A∆t) which defines the intensity of an electromagnetic wave.
Gaussian diffusion is a type of diffusion which arises whenever there is a diffusional spread of a constant number of
particles which are initially contained in a thin sheet. The number of diffusing species or particles remains constant.
Mole of a substance is that amount of the substance which contains N A number of atoms (or molecules) where N A is
Avogadro's number (6.022 × 1023). One mole of a substance has a mass as much as its atomic (molecular) mass in
  S. O. Kasap, 1990 - 2002: v.1.1)
Diffusion: Concepts and Solved Problems ( 13
An e-Booklet

grams. For example, 1 mole of copper contains 6.022 × 1023 number of copper atoms and has a mass of 63.55
grams.
Oxidation is the term used to refer to the production of silicon dioxide on silicon. Typically, silicon is exposed to either
water vapor or oxygen at high temperatures to controllably produce a layer of oxidized silicon, or silicon dioxide.
Tracer diffusion is diffusion involving a constant number of diffusing species of atoms, all initially concentrated in a very
thin surface region of the crystal. The name comes from the fact that the time evolution of the concentration profile
of diffusing atoms can be followed, if the diffusing species are radioactive.

QUESTIONS AND PROBLEMS

1. Bipolar Junction Transistor The bipolar junction transistor operation is based on the
diffusion of charge carriers in the base region of the device. Consider an npn transistor as shown
in Figure 8 (a). The emitter is heavily doped and has plenty of electrons. When a forward bias
voltage is applied across the emitter-base junction as in Figure 8 (b), the emitter injects electrons
into the base. The injected electron concentration n(0) at the edge of the base (near the emitter-base
junction) is determined by the voltage VEB and is given by the law of the junction:
n(0) = noexp(eVEB/kT)
where no is the equilibrium electron concentration in the base k is the Boltzmann constant and T is
the temperature (K).
The reverse bias voltage VCB determines the electron concentration at the edge of the base at
the base-emitter junction at x = W B . With a few volts applied across the base-emitter, n(WB) = 0.
Under steady-state conditions (dc voltages and currents), the electron concentration profile
decreases linearly from n(0) at x = 0 to 0 at x = W B across the base width W B . Electrons diffuse
across the base to reach the collector. There is no electric field in the base to drift the electrons.
They are driven by Fick’s first law (concentration gradient). Use the definitions of the current
density J and particle flux Γ, to relate J to Γ and hence show that the current density due to the
diffusion of electrons through the base is given by

exp EB 
eDe no eV
J=
WB  kT 

E B C
n(x)
WB
IE n(0) IC
n(x) E
n+ p n

Emitter Base Collector x

IB
VEB VCB

(a) A schematic illustration of the npn bipolar transistor with three differently doped regions. (b) The npn
bipolar operated under normal and active conditions: The CB configuration with input and output circuits
identified.

Figure 8
  S. O. Kasap, 1990 - 2002: v.1.1)
Diffusion: Concepts and Solved Problems ( 14
An e-Booklet

These electrons diffuse and reach the base-collector junction. There is an electric field in this
junction region that sweeps the electrons into the collector. When the electrons reach the collector
they become collected by the battery and constitute the collector current I C . The transistor action is
the fact that the collector current IC in the output circuit is controlled by the emitter-base voltage VEB
in the input circuit.
The reason for the law of the junction is that electrons in the emitter and base have different
potential energies: PE = Charge × Voltage. The voltage VEB thus modulates the electron PE across
the emitter-base junction and this change in the PE, eVEB, controls the amount of injected electrons
into the base via the Boltzman factor, exp(eVEB/kT).
A particular npn transistor has the following properties:
Cross sectional area = 1 mm2
Base width = 4 µm,
Equilibrium electron concentration in the base, no = 1 × 105 cm-3
Electron diffusion coefficient in Si = 37 cm2 s-1.
Emitter-base forward bias, VEB = 0.6 V.
Calculate the collector current at room temperature.

2. Semiconductor fabrication Consider a wafer of Si crystal which has a uniform boron (B)
doping of 1 × 1017 cm-3 in the bulk. Suppose that the Si wafer is exposed to a phosphorus (P) gas
at 1250 °C for 10 minutes. During the diffusion process, the surface P concentration remains
saturates at about 1 × 1021 cm-3. Where is the junction from the surface? How long does it take to
have the same junction depth from the surface if the diffusion temperature is 1100 °C? State the
assumptions used in your calculations. The diffusion coefficient of P in Si has Do = 3.85 cm2 s -1,
and Ed = 3.66 eV/atom.
Note: B doping makes Si, p-type and P-doping makes it n-type. The junction is where B and P
concentrations are equal.
(Ans: 2.33 µm; 3.5 hrs)

3. Semiconductor fabrication Consider a wafer of Si crystal which has a uniform boron (B)
doping of 1 × 1016 cm-3 in the bulk. The surface region of the Si crystal has been predoped with P
so that the P concentration is approximately 1 × 1021 cm-3 over a small distance 0.01 µm. The Si
wafer is then placed in a furnace at 1100 °C for about 45 minutes. Where is the junction from the
surface? How long does it take to have the same junction depth from the surface if the diffusion
temperature is 1000 °C? State the assumptions used in your calculations. The diffusion coefficient
of P in Si has Do = 3.85 cm2 s-1, and Ed = 3.66 eV/atom.
Note: B doping makes Si, p-type and P-doping makes it n-type. The junction is where B and P
concentrations are equal.
(Ans: At 1100 °C, z = 7.97; xjunction = 1.1 µm; at 1000 °C, diffusion time ≈ 8.5 hours)

4. Nitriding of steels Steels may be strengthened by nitriding, which involves diffusion of

nitrogen into the surface at high temperatures. Suppose that the nitriding atmosphere produces 0.1
  S. O. Kasap, 1990 - 2002: v.1.1)
Diffusion: Concepts and Solved Problems ( 15
An e-Booklet

wt.% N at the surface of a steel initially containing 0.001 wt.%. Determine the nitrogen
concentration as weight percentage at a distance 1 mm beneath the surface after 5 hours of
exposure at 700 °C. Note: For N diffusion in BCC Fe, Qd = 76.57 kJ mol-1 and Do = 4.7 × 10-7
m2 s-1.
(Ans: 0.039 wt.%)
If the nitriding above is done at 650 °C, what is the time required to achieve the same nitrogen
concentration at the same depth?
(Ans: 8.34 hrs = 8h 20m 41s)
  S. O. Kasap, 1990 - 2002: v.1.1)
Diffusion: Concepts and Solved Problems ( 16
An e-Booklet

Table 1
Error and complementary error function tables (from Matlab).
z erf(z) erfc(z) z erf(z) erfc(z)
0 0 1 2 0.995322 0.00468
0.01 0.0113 0.988716 2.25 0.998537 0.00146
0.05 0.0564 0.943628 2.5 0.999593 0.000407
0.1 0.1125 0.887537 2.75 0.999899 378 0.000101
0.25 0.2763 0.723673 3 0.999977909 0.0000221
0.5 0.5205 0.4795 3.25 0.999995697 0.0000043
0.75 0.7112 0.288844 3.5 0.999999256901 0.000000743
1 0.8427 0.157299 3.75 0.999999886272 0.000000114
1.25 0.9229 0.077099 4 0.999999984582 0.0000000154
1.5 0.9661 0.033894 4.25 0.999999998149 0.00000000185
1.75 0.98667 0.00468 4.5 0.999999999803 0.000000000197

1x100
erf(z)

1x10-1

1x10-2

1x10-3
erfc(z)
1x10-4

1x10-5

1x10-6

1x10-7
0 0.5 1 1.5 2 2.5 3 3.5 4
z
Error and complementary error functions vs. their arguments

Figure 9
  S. O. Kasap, 1990 - 2002: v.1.1)
Diffusion: Concepts and Solved Problems ( 17
An e-Booklet

All material in this publication is copyrighted.

 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or
transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise,
without the prior written permission of the author.
Permission is granted to individuals for downloading this document from the author’s website or his CD-ROM
for self-study only. Permission is hereby granted to instructors to use this publication as a class-handout if the
author’s McGraw-Hill textbook Principles of Electronic Materials and Devices, Third Edition, has been adopted
as a requisite course textbook. The permission is valid only while the book remains in adoption.

SPECIAL CUSTOM PUBLISHED e-BOOKLET

 S.O. Kasap, 1990-2001
The author reserves all rights
Last Updated: 6 January 2002 (v.1.1)
First published in Web-Materials
(Established 1996)
http://Materials.Usask.Ca