Doppler Broadening

Hywel Owen
University of Manchester
hywel.owen@manchester.ac.uk
March 3, 2012
1 Wavelength and Temperature
We start by considering an individual molecule in motion toward an observer
with velocity v
x
; the velocities transverse to the observation direction we
denote v
y
and v
z
. If the molecule emits some frequency
0
in its frame of
reference, then the observer sees a frequency
=
0
c
c + v
x
. (1)
In terms of wavelength, we can write this as
v
x
c
=

0

0
=

0
, (2)
where
=
0
(3)
is the wavelength shift for an individual molecule. Squaring, we get

2
=

2
0
v
2
x
c
2
. (4)
Over many molecules, the variance of emitted wavelengths is therefore

2
=

2
0
c
2

v
2
x
. (5)
1
We have implicitly assumed here that the individual molecular velocities
are normally-distributed, which well show is true in the next section. We
know from equipartition theory that
1
2
m

v
2
x
=
1
2
kT. (6)
Therefore

0
=

kT
mc
2
. (7)
Note that everything still works for absorption rather than emission (think
of the shift of the wavelength being absorbed as seen by the moving molecule).
We also know that

0
=

k
k
0
=

0
, (8)
so that a relative linewidth in k is equivalent to a relative linewidth in .
2 The Shape of the Peak
Here we will show that the expected shape of the absorption peak is actually
Gaussian, by considering the velocity distribution. Recall that an ideal gas
in equilibrium at a particular temperature has a Boltzmann distribution in
energy, such that the probability of a molecule having a particular energy E
is
P[E]dE = Ae
E/kT
dE, (9)
where A is some constant will nd later on. Using E = p
2
/2m we can
re-write this as
P[E]dE = Ae
E/kT
= Ae
p
2
/2mkT
dE. (10)
But p
2
= p
2
x
+ p
2
y
+ p
2
z
, so we can write the distribution of p
x
values as
P[p
x
]dp
x
= Ae
(p
2
x
+p
2
y
+p
2
z
)/2mkT
dp
x
= A

e
(p
2
x
)/2mkT
dp
x
, (11)
where we have collected the p
y
and p
z
terms into the constant as they are
in equilibrium. So we can now write the velocity distribution simply as
2
P[v
x
]dv
x
= A

e
(mvx)
2
/2mkT
dv
x
= A

e
mv
2
x
/2kT
dv
x
(12)
We can normalise by observing that molecules must have some velocity
between and +, therefore

+
vx=
A

e
mv
2
x
/2kT
dv
x
. (13)
We use the standard result

e
x
2
dx =

. (14)
to nd
A

m
2kT
. (15)
Therefore the velocity distribution is
P[v
x
]dv
x
=

m
2kT
e
mv
2
x
/2kT
dv
x
, (16)
which we observe has a Gaussian shape. We also observe that it is sym-
metrical around v
x
= 0, and by comparing with the standard Gaussian
y = e
x
2
/2
2
(17)
we see that the expected width in velocities is just

2
v
=

v
2
x
=
kT
m
. (18)
Dividing through by c
2
we obtain

v
2
x
c
2
=
kT
mc
2
(19)
as desired.
3
3 Interpretation of the Dierential Peak
In the previous sections we have shown that the intensity of the emitted
wavelengths should have a one-to-one correspondence with the individual
molecular velocities. We also showed that the velocity distribution is Gaus-
sian, and therefore that the shape of the absorption peak is also Gaussian.
However, we are not looking at the absorption peak directly, but actually at
its derivative. So how do we determine the width of the absorption peak
by looking at its derivative?
Lets look at the absorption in the wavenumber domain k (recalling from
above that k/k = /). The true absorption peak has some variation of
intensity with k which we expect to be Gaussian:
I(k) = I
0
e

k
2
2
2
k
(20)
The observed data T(k) is the derivative of this:
T(k) =
dI(k)
dk
=
k

I
0
e

k
2
2
2
k
. (21)
We wish to look at the minimum and maximum of T(k) and from this
determine what
k
is. These values of k are given by
dT(k)
dk
= 0, (22)
therefore
dT(k)
dk
=
d
2
I(k)
dk
2
= 0. (23)
Taking the 2nd derivative of I(k) and collecting terms we have
dT(k)
dk
=
d
2
I(k)
dk
2
= I
0
(
k
2

4

1

2
)e

k
2
2
2
k
= 0. (24)
This has the trivial solutions
k = , k = +, (25)
but more interestingly our minima and maxima are located at
k = , k = + (26)
4
Therefore the distance between the minimum and maximum points T(k)
is 2.
5