CHE
304/2013
2nd
Midterm
Solution


Analysis
of
impurity
isotops
in
rocks
can
yield
information
on
the
‘history’
of
those
rocks.
Garnets
are

not
only
expensive
gemstones,
but
a
useful
‘recorder’
of
conditions
such
as
pressure,
heat/cool

cycles,
and
environmental
chemistry.
However,
to
understand
the
information
embedded
in
these

rocks
required
better
understanding
of
the
processes
that
drive
impurities
into
the
rocks
in
the
first

place.


Van
Orman,
et
al
(2002)
measured
the
diffusion
coefficient
of
different
rare‐earth
molecules
in

garnet.
The
experiment
was
set
up
as
follows:
A
large
crystal
of
garnet
was
coated
by
a
layer
of
the

rare­earth
oxides.
The
samples
were
then
heated
to
the
test
measurement,
and
kept
in
it
for
a
period

of
time.
After
that
time,
the
concentration
profiles
were
measured
using
a
technique
called
Secondary

Ion
Mass
Spectrometry
(SIMS),
the
best
method
for
this
cases,
because
it
is
a
method
that
measures

1012­1016
atoms/cm3,
the
relevant
values
needed
in
these
systems.


The
density
of
garnet
is
4gr/cm3,
and
the
MW
is
400.



Recall
that
Avogadro’s
number
is
6.02*1023.


Mathematical
Information
you
MAY
need


F(s)=L{f(t)}
 f(t)
 F(s)=L{f(t)}
 f(t)



 
 
 



 
 




in
the
limit
of
x<<1,
 


1.2

0.8
erfc(x)

0.6

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3

x
1.



A.
Calculate
the
expected
profile
for
the
concentration
of
the
impurities
as
a
function
of
depth
(z)
and

time
(t).
Use
D
for
the
diffusion
coefficient,
and
Ci
for
the
concentration
in
the
garnet
at
the
interface

with
the
oxide
layer.
You
may
assume
that
initially
there
were
no
impurities
in
the
garnet.



 1

 Ci



Semi‐infinite
medium,
with
C(t=0)=C(z‐>∞)=0,
and
C(z=0)=Ci.
The
solution
is



C  z 
= erfc 

Ci  2 Dt 

B.
Justify
the
assumptions
made
to
allow
the
solution
you
found
in
(A).



Assumptions:


€
‐Only
the
impurity
is
diffusing.
This
is
correct‐
the
garnet
atoms
are
not
diffusing


‐D
uniform
and
constant


‐no
bulk
reaction


‐semi‐infinite
(the
thickness
is
much
larger
than
the
distance
for
diffusion‐see
question
3)


‐y
of
the
impurity
is
much
smaller
than
1.
The
units
of
Ci
are
1016
atoms/cm3
(this
is
the
higher
end
of

the
measured
values).
So
we
need
this
to
be
much
smaller
than
the
atoms
garnet/cm3.
The
molar

conc
of
garnet
is
4/400=0.01
moles/cm3.
Multiply
by
Av
number
gives
1021.
This
is
5
orders
of

magnitude
larger
than
the
impurities,
so
we
are
OK.


2.
The
data
in
the
plot
provided
was
measured
after
4
hrs
at
1,4500C.
The

concentrations
were

measured
in
arbitrary
units.




A) Fitting
this
data
with
a
diffusion
model
yielded
a
diffusion
coefficient
for
Ce
that
is
4*10­19m2/s,

and
for
Sm
it
was
3.6*10­19m2/s.
Is
this
reasonable,
or
was
there
a
calculation
error?
Explain.





It
seems
unreasonable:
How
can
CE
have
the
a
higher
diffusion
coefficient,
if
its
conc
in
the
garnet
is

practically
zero,
while
SM
is
not
zero
with
a
lower
D
value.
However:
D
defines
the
RATE
of
diffusion,

that
is,
it
sets
the
slope
at
z=0.

The
value
of
Ci
determines
how
much
(if
at
all)
will
get
into
the
garnet.



 2

 2

1.5
relative conc

The
value
of
Ci
for
CE
is
very
low,
much
lower
than
that
of
SM.
So,
it
is
possible
that
they
ahev
these

1 Ds‐
the
profiles
here
are
set
by
the
solubility
of
these
impurities
in
garnet.


By
the
way:

0.5

1.
Many
of
you
confuse
initial
conc
(which
means
at
t=0)
with
surface
conc,
which
is
at
z=0!

0
0 2000 4000 2.
If
you
though
that
the
‘relative
conc’
was
C/Ci,
then
it
had
to
be
1
at
z=0
for
ALL
the
impurities.

6000 8000 10000

Since
it
was
not,
it
means
that
it
is
some
conc
unit.
z(nm)

For
both
B
and
C:
The
important
parameter
is
z/√4Dt.
Increasing
t
or
D
will
have
a
similar

effect.
The
diffusion
coefficient­
even
in
solids­
INCRASES
with
T!


B) Sketch
(qualitatively)
the
conc
profile
expected
for
Dy
at
time
close
to
the
beginning
of
the

experiment,
and
after
a
day.



long time

short time



The
concentration
at
the
interface
is
the
SAME
(Ci
is
set
by
the
solubility
in
garnet,
and
does

not
change‐
you
made
that
assumption
in
(1)!).
As
time
gets
longer,
more
diffusion
into

deeper
into
the
garnet,
but
the
shape
stays
the
same.



C) Sketch
(qualitatively)
the
conc
profile
expected
for
Dy
after
4
hrs
at
a
lower
T
(~8000C)
and

higher
temperature
(~25000C).



higher T

lower T



At
the
same
time,
Ci
is
the
same.
However,
D
is
higher
at
T
higher,
so
more
diffusion.



3.
A
sample
of
garnet
with
a
thickness
of
10µm
(10,000nm)
with
an
oxide
layer
of
Yb
was
forgotten
in

the
oven,
at
T=20000K,
for
2
days
(48
hrs).
Testing
it
gave
the
following
concentration
profile:



 3

 2

1.5
relative conc

0.5

0
0 2000 4000 6000 8000 10000

z(nm) 






A) Explain,
in
1­2
sentences,
why
this
profile
is
different
from
the
ones
shown
in
question
2.


Aftre
this
period
of
time,
the
impurity
diffused
until
it
‘reached’
the
end
of
the
garnet.
So
now
it
is

NOT
a
semi‐infinite
medium
anymore.
With
time,
the
conc
at
z=10000,
the
‘end’
of
the
garnet,
will

increase
even
more.
At
t‐>
inifinity
it
will
be
a
straight
line
and
C=Ci
everywhere.


Also:
Many
of
you
got
confused
about
the
fact
that
I
mentioned
the
oxide
layer
here.
If
you’d
read
the

intro
on
page
1,
you’d
see
that
ALL
systems
have
an
oxide
layer.
That’s
the
‘reservoir’
that
provides

the
impurity
into
the
system!!






B) From
other
experiments,
the
diffusion
coefficient
for
Yb
was
found
to
be

D=2*10­8exp(­343/(RT))
m2/s,
where
R=8.3*10­3
and
T
is
in
0K.


For
the
solution
you
derived
in
(1)
to
apply,
estimate
the
minimum
thickness
that
a
sample
of
garnet
has

to
be,
for
an
experiment
at
20000K
and
24
hrs.
Explain
your
reasoning!


If
we
want
the
garnet
to
be
considered
‘semi
infinite’,
we
want
the
concentration
at
L,
the
thickness,

after
this
period
of
time,
to
be
practically
zero
(since
erfc
is
an
exp‐like
function,
it
is
never

 z  to be close to zero for z=L. From the plot,
ABSOLUTELY
zero).

That
means
we
want
 C = erfc 
Ci  2 Dt 
any value for L/(2√Dt) that is larger than 2 or 3 would be acceptable.

Common mistakes:
€
(1) we do NOT want C/Ci to be equal to 0.1. That was our indicator for the previous system that the sample
was NOT thick enough.

(2) We do not want C/Ci to be equal to 1. That is going to be ALWAYS at z=0, since it is our boundary
condition! – and if you got a value that is not zero, something is very wrong.

(3) It is true from dimensional analysis that we want L to be some function of √Dt. But what should be the
proportionality constant? Why 1, and not 0.1 or 100? 


4.
The
diffusion
coefficients
calculated
from
fitting
the
data
to
the
solution
in
question
(1)
gavea
large

distribution
(±40%
in
many
cases).
The
differences
seem
to
be
due
to
the
method
that
the
oxide
layer
at

the
surface
was
formed.
One
suggestion
is
that
the
concentration
at
the
garnet
interface
is
controlled
by

the
oxide
reaction
(which
produces
the
impurity
atoms).
As
a
result,
the
flux
at
the
interface
is
given
by

q/√t,
where
q
has
units
of
mole/(area*√time).
At
t=0
the
flux
is
practically
infinite,
because
of
the



 4

 large
gradient
at
the
interface.
When
t
goes
to
infinity,
the
flux
goes
to
zero
because
there
is
no
more

oxide
to
provide
diffusing
atoms.



(Note,
the
1/√t
indicates
that
impurity
atom
transport
in
the
oxide
is
dominated
by
diffusion
as
well).


A. Calculate
the
concentration
profile
for
a
system
where
initially
the
concentration
of
impurity
is

zero,
and
at
the
interface
the
FLUX
is
given
by
q/√t.


We
have
a
FLUX
BC,
which
means
that
–D∂C/∂z
at
z=0
is
equal
to
q/√t.
This
does
NOT
mean
the

concentration
at
z=0
is
equal
to
q/√t!



To
solve
this
problem,
follow
HW
6
question
2
exactly.
The
only
difference
is
that,
when
solving

for
θ,
we
do
not
use
θ
(z=0)=L(δ),

but
the
condition
–D∂θ/∂z
at
z=0
is
equal
to
L(q/√t).
That

transform
was
given
to
you
in
the
1st
page
tables.



Note:
You
HAVE
to
transform
the
BC
from
t
to
s.
You
CANNOT
leave
it
as
–D∂θ/∂z=q/√t!!


Mathematically,
it’s
like
saying
that
if
x=3,
then
ex
is
also
equal
to
3.
We
do
some
mathematical

action
on
C
to
get
it
to
be
theta,
so
you
HAVE
to
apply
that
to
the
value
of
the
BC
at
that
point.
The

Laplace
operation
is
integrating
over
t
from
t=0
to
t=infinity,
so
there
is
no
t
anywhere
after
you

transformed
it.


B) The
data
seemed
to
fit
the
function
derived
in
question
(1).
Explain
why
this
does
not
contradict
your

new
calculation.,
that
is,
WHT
your
new
function
would
fit
the
data
as
well.




The
word
DATA
means
measurements
that
someone
spent
a
lot
of
time
in
the
lab
measuring.
It
is

FACTS.

You
can
see
in
question
1
that
there
are
DATA
POINTS‐
it’s
not
some
line
that
someone

calculated
from
a
mathematical
function.

The
question
here
was
:
These
measurements
are
fit
well
by
the
erfc
that
was
derived
for
a
semi‐
infinite
slab
with
a
constant
concentration
Ci
at
z=0.
Does
that
mean
that
whatever
you
derived
in
(A)

here
is
incorrect?

If
what
you
derived
in
(A)
was
an
erfc
(with
a
different
constant
multiplying
it),
then
the
answer
is

that
there
is
no
problem.
The
measurements
are
taken
at
some
t
and
show
the
z
dependence,
so
we

cannot
tell
what
the
pre‐factor
is.

If
what
you
derived
in
(A)
was
not
an
erfc,
then
you
need
to
see:
Is
your
function
similar
in
form
to

the
erfc?
(If
you
had
miraculously
and
erroneously
arrived
at
the
thin
film
function
that
is

appropriate
for
a
film
with
a
constant
concentration
at
the
boundary)
then
the
answer
is
still
yes
IF

the
time
was
short
enough
so
you
do
not
see
the
effects
of
the
finite
thickness.
Otherwise,
you
will
get

profiles
like
question
3.A
or
ones
with
even
higher
values
at
the
end
of
the
film,
and
there
will
be
a

problem.

If
you
got
some
other
type
of
function,
you
need
to
see
if‐
for
a
given
t,
it
has
a
similar
dependence
on

z
like
the
erfc.
If
not,
then
this
function
does
NOT
fit
the
data,
and
there
is
an
issue
with
the
math‐

NOT
the
measurements.



Saying
that
the
concentration
will
decrease
or
increase
with
t,
D
or
z
is
NOT
relevant.
The
function
1‐
z/L
decreases
with
z,
but
it
does
NOT
look
like
the
data
in
question
1.


Also‐
even
if
you
made
some
math
error
and
arrived
at
a
function
that
has,
for
example,
exp(
z+D),
at

least
say
that
you
know
that
(1)
you
cannot
add
length
to
l2/t,
and
(2)
that
you
cannot
take
the
exp
of

something
with
dimensions!!



 5