Astrophysics 1

Tutorial 10 - Bremsstrahlung radiation

We are going to calculate the emission coefficient jν of bremsstrahlung radiation (following
Rybicki & Lightman “Radiative processes in astrophysics” section 5.1): We start from the
Larmor formula for the total power radiated by a charge e accelerating with acceleration a (t):
 2
 
2 e2 2 d̈ (t)
P = |a (t)| ≡
2
,
3 c3 3 c3
and in frequency space
8πω 4
Pω = 3
|dω (ω)|2 ,
3c
where d (t) is the time dependent dipole moment, eR, and R is the displacement vector of
the charge.
Let us consider an electron with charge −e, passing rapidly by an ion with charge +Ze,
with an impact parameter b. The electron is moving rapidly enough so that its trajectory is
almost a straight line with constant velocity (even though the acceleration isn’t zero, in our
approximation it will affect the radiation, not the motion). This approximation, called the
small angle approximation, simplifies the computation even though it isn’t necessary.
The second time derivative of d in our case is

d̈ = −ev̇, (1)

where v is the velocity of the electron. The Fourier transform of the former equation is
Z ∞
e
−ω dω = −
2
v̇ (t) eiωt dt. (2)
2π −∞

The interaction is significant during a time interval τ ∼ b/v. When τ ω ≫ 1, the integral
oscillates rapidly, and v̇ is small, so the result is negligible. On the other hand when ωτ ≪ 1

1
the exponent can be approximated by 1. So we can write
(
e
|∆v| , ωτ ≪ 1
dω ≃ 2πω 2
, (3)
0 ωτ ≫ 1

|∆v| is the change in the electron velocity during the collision. We can now write the Larmor
formula in the frequency domain
( 2
dE 8π 4 2e
|∆v|2 , ωτ ≪ 1
Pω = = 3 ω |dω | ≃ 2 3πc3
. (4)
dω 3c 0 ωτ ≫ 1

To estimate |∆v| we will integrate the acceleration component normal to the path, which is
the dominant part of it (for t → 0)
 
Ze2 vt b
a=− √ ,√ , (5)
m (b2 + v 2 t2 ) b2 + v 2 t2 b2 + v 2 t2
Z
Ze2 ∞ bdt 2Ze2
|∆v| ≃ = . (6)
m −∞ (b2 + v 2 t2 )3/2 mbv
Another way to estimate ∆v, is to assume that the interaction takes place only very near to the
closest approach point in a time interval of 2τ , and there it generates a constant acceleration,
which is the maximal acceleration,
Ze2 b 2Ze2
|∆v| ∼ amax 2τ ∼ 2 ∼ . (7)
mb2 v mbv
Inserting |∆v| into eq(4) results with
(
8Z 2 e6
, ωτ ≪ 1
Pω ∼ 3πc3 m2 b2 v 2
. (8)
0 ωτ ≫ 1
The total amount of electrons incident on one ion from impact parameter b is their flux
×surface area around the ion ∼ ne v2πbdb, and per unit volume we have ni ions so the to-
tal emission per unit volume unit time and unit frequency is
Z Z  
dE 16ne ni Z 2 e6 bmax db 16e6 bmax
∼ 2πne ni v Pω bdb ∼ 2
= 3 2 ne ni Z ln . (9)
dωdV dt 3c3 m2 v bmin b 3c m v bmin
The integration limits can be determined as follows: given v and ω the upper bound is the
impact factor at which the approximation
b
ωτ = ω ≪ 1, (10)
v
2
is inapplicable, so we will set it to bmax ∼ v/ω. The lower bound can be determined in two
ways: one is the impact factor in which the constant velocity-small angle scattering approxi-
mation is no longer works, this occurs when the kinetic energy is comparable to the maximal
potential energy

1 2 Ze2 2Ze2
mv ∼ =⇒ bmin ∼ . (11)
2 b mv 2
The second one is from QM’s uncertainty principle,∆x∆p ≳ h. Taking ∆x ∼ b, ∆p ∼ mv
gives
h
bmin ∼ . (12)
mv
We need to use the bigger lower bound. Anyway we see that they enter only as a logarithmic
factor, usually expressed via correction factor called the Gaunt factor (here the f −f subscript
refer to “free-free” emission, another term for bremsstrahlung)

dE 16πe6
∼ √ ne ni Z 2 gf−f (v, ω) , (13)
dωdV dt 3
3 3c m v2

√  
3 bmax
gf−f (v, ω) ≡ ln . (14)
π bmin
Now we need to know the distribution of electron velocities in order to average over them. If
the distribution is thermal, then it is the Maxwell-Boltzmann velocity distribution: the prob-
ability to find an electron with speed between v and v + dv is given by
mv 2
r  
v 2 e− 2kT dv 2 m 3/2 2 − mv2
Prob (v, v + dv) = R ∞ mv 2
= v e 2kT dv. (15)
v 2 e− 2kT π kT
0

Averaging over this distribution gives

  r   Z ∞ 2
dE 2π m 3/2 16e6 v − mv2
= jω ≃ n e ni Z 2
e 2kT gf−f (v, ω) dv, (16)
dωdV dt 3 kT 3c3 m2 vmin v

where vmin is determined by the condition that given ω, 21 mvmin

2
= ℏω. This condition assures
that a photon with energy ℏω could be produced in this interaction. So vmin = (2ℏω/m)1/2 .
r  
2π m 1/2 16e6 ℏω
jω ≃ 3 2
ne ni Z 2 e− kT ⟨gf−f (ω)⟩
3 kT 3c m

3
Now ⟨gf−f ⟩ is the Gaunt factor averaged over velocities. Translating into ν = ω/2π:
 3/2 
2π m 1/2 16e6
ne ni Z 2 e− kT ⟨gf−f (ν)⟩ .
hν
jν ≃ 3 2
3 kT cm

Now we can integrate over frequencies ν:

   3/2  Z
dE 2π m 1/2 16e6 ∞
e− kT ⟨gf−f (ν)⟩ dν,
hν
j= ≃ ne ni Z 2
dV dt 3 kT c3 m2 0

   3/2  
dE 2π m 1/2 16e6 kT

≃ n e n i Z 2
g f−f .
dV dt 3 kT c3 m2 h


Here g f−f is averaged over frequencies too, and is of order unity so I omitted it.
We can now insert the natural constants:
     1/2  2 1/2 2
dE 8π e4 2π mc e

≃ 4 ckT n e ni Z 2
g f−f ,
dV dt 3 m2 c4 3 kT hc
"   #
2π
1/2

 1 2
= 4 g f−f cσT ne ni Z αE kT.
3 βTh
This expression has the right form: cne is a flux of electrons, that when multiplied by Z 2 αE σT ,
the effective interaction cross section, gives the number of interacting electron per unit time.
ni are the number of scatterers per unit volume, and kT is the average energy released in each
interaction. To get the numerical factors and 1/βTh we had to do all this hard work.