Radiated-Power Distribution

in the Far Zone of a Moving System

Kirk T. McDonald
Joseph Henry Laboratories, Princeton University, Princeton, NJ 08544
(April 24, 1979; updated April 28, 2017)

1 Problem
Show that the angular distribution,
dP  d2 U 
= = f(cos θ , φ ), (1)
dΩ dt dΩ
of the power of electromagnetic radiation in the far zone of a system whose center of
mass/energy is instantaneously at rest in the (inertial)  frame has the form,
 
dPsource dU 1 cos θ − β
= = 4 f ,φ , (2)
dΩ dt dΩ γ (1 − β cos θ)3 1 − β cos θ

for radiation by the source in the (inertial) lab frame where the system has velocity v along
   
z ) axes of the spherical coordinate systems (r, θ, φ) and (r , θ , φ ), β = v/c
the polar (z,
and γ = 1/ 1 − β 2, with c being the speed of light in vacuum. Comment on the angular
distribution of radiation as detected by distant, fixed observers in the lab frame.
Comment also on the case of acoustic radiation.

2 Solution
2.1 Effect of Retardation in the Lab Frame
While the main theme of this problem is the transformation between frames of power emitted
by a source, we first take note of the distinct, but related issue that the power emitted by
a moving source into some solid angle, as measured in the lab frame, is different from that
received by a fixed observer (in the lab frame) who subtends the same solid angle.

As suggested by the sketch above, the energy dUsource emitted by a moving source at
angle θ with respect to its velocity v during time interval dt is compressed into the spatial

1
interval (c − v cos θ)dt along the direction of emission, where c is the speed of propagation
in the lab frame. This energy is later received by a fixed observer during the time interval
(1 − v cos θ/c)dt, such that the power received is related by,
dPobserver 1 d2 Usource 1 dPsource
= = . (3)
dΩ (1 − v cos θ/c) dt dΩ (1 − v cos θ/c) dΩ
The relation (3) can also be considered as a consequence of the fact that the energy
received by the observer (in solid angle dΩ) during time interval dt was emitted by the
source during the retarded time interval dt (into the same solid angle dΩ), where,1
   
 r(t )  r(t )  1 dr  v
t = t− , t=t + , dt = dt 1 + = dt 1 − cos θ , (4)
c c c dt c
and hence,
dPobserver d2 U 1 d2 U 1 dPsource
= = = , (5)
dΩ dt dΩ (1 − v cos θ/c) dt dΩ
 (1 − v cos θ/c) dΩ
as found previously.
Both t and t are measured in the lab frame, so the difference between Psource and Pobserver
(both in the lab frame) is not an effect of “relativity”, but can be attributed to retardation.
In the instantaneous rest frame of the source, where β  = 0, the power radiated by the
source in some solid angle is the same as that detected by a fixed observer with that solid
angle.

2.2 Electromagnetic Radiation

2.2.1 General Transformation of the Angular Distribution
The arguments here are based on the semiclassical insight that far from its source, the energy
and momentum densities u and p of electromagnetic radiation (in vacuum) are related by,
u = pc, (6)
as for a single photon. Hence, we can use a light-like 4-vector (du, c dp) to describe the
propagating energy and momentum of a small volume element (which moves at the speed of
light) in the far zone.
We take the z-axis to be along the direction of the velocity v of the source at some
(retarded) time in the lab frame, and let θ be the angle between the momentum vector dp
and the z-axis. Then, the energy and momentum densities in the instantaneous rest frame
(the  frame) of the (retarded) source are given by,
du = γ(du − v · dp) = γ du(1 − β cos θ), (7)
du
dp⊥ = dp⊥ = sin θ, (8)
c
du
dp = γ(dp − v du/c2 ) = γ (cos θ − β). (9)
c
1
The quantity 1 − v cos θ/c is sometimes called the Doppler factor. Its early history, with some initial
controversy over its interpretation can be traced in [1, 2, 3, 4, 5]. This factor was well discussed by Schott,
sec. 13 of [6], but apparently needed “rediscovering” in 1969, as on pp. 406-407 of [7]. The author learned
of this factor in graduate school (Caltech) in 1968, where it was presented as “well known”.

2
The angle θ of the vector dp with respect to the z/z  axis is given by,


dp dp cos θ − β cos θ − β
cos θ = = = = , (10)
p p 2 + p⊥ 2 (cos θ − β)2 + (1 − cos2 θ)/γ 2 1 − β cos θ

and hence,
d cos θ
d cos θ = . (11)
γ 2 (1 − β cos θ)2

Of course, the azimuthal angles φ and φ have the same values, so the transformation of
solid angle dΩ = d cos θ dφ is,
dΩ
dΩ = . (12)
γ 2 (1 − β cos θ)2
Also, a time interval dt for a fixed observer in the lab frame is related to the time interval
dt in the moving  frame by time dilation,

dt = γ dt. (13)

Combining eqs. (7) and (10)-(13), the angular distribution of the power radiated by the
moving source can be written as,
 
dPsource d2 u 1 d2 u 1 cos θ − β
= = 4 = 4 f , φ . (14)
dΩ dt dΩ γ (1 − β cos θ)3 dt dΩ γ (1 − β cos θ)3 1 − β cos θ
This result is based on the transformation of the energy and momentum far from the source,
and so corresponds to the angular distribution of power according to a distant observer.
According to the argument of sec. 2.1, the angular distribution of power as detected by a
fixed, distant observer is,
 
dPobserver 1 dPsource 1 cos θ − β
= = 4 f ,φ . (15)
dΩ 1 − β cos θ dΩ γ (1 − β cos θ)4 1 − β cos θ

For example, although a single electric charge cannot emit radiation with an isotropic
angular distribution, this is possible for an appropriate current distribution [8]. If that
current distribution had bulk velocity v, the angular distribution observed in the lab frame
would be,
dPobserver 1
= 4 , (16)
dΩ γ (1 − β cos θ)4

and the forward/backward ratio would be,

⎧
dPobserver(θ = 0)/dΩ (1 + β) ⎨ 1 + 8β (β  1),
4
= ≈ (17)
dPobserver(θ = π)/dΩ (1 − β)4 ⎩ 32γ 8 (γ , 1, β ≈ 1 − 1/2γ 2 ).

3
2.2.2 Radiation by a Single Accelerated Charge
If an electric charge e has acceleration a in its instantaneous rest frame, then its angular
distribution of radiated power in that frame is given by the differential version of the Larmor
formula (in Gaussian units),

dP  e2a2 sin2 α
= , (18)
dΩ 4πc3
where α is the angle between the acceleration vector and the direction to the observer in
the  frame.
To relate the lab-frame acceleration 3-vector a to the acceleration a in its instantaneous
rest frame we recall that a velocity 4-vector can be defined as,

vμ = γ(c, v) = cγ(1, β), (19)

and then the acceleration 4-vector is given by,

dvμ dvμ
aμ = =γ = cγ 2 (γ 2(β · β̇), β̇ + γ 2 β(β · β̇)), (20)
dτ dt
noting that,
dγ
= γ 3(β · β̇), (21)
dt

where β̇ = dβ/dt. The acceleration 4-vector in the instantaneous rest frame has components

c(0, β̇ ) = (0, a ), so the invariant square of the acceleration 4-vector is,
2
aμ aμ = −a2 = −c2 γ 6[β̇ − (β × β̇)2] = −γ 6 [a2 − (v/c × a)2 ]. (22)

Using eqs. (18) and (22) in (14), we the power radiated by the accelerated charge is,

dPsource γ 2e2 [a2 − (v/c × a)2 ] 2 

= sin α (θ, φ). (23)
dΩ 4πc3 (1 − β cos θ)3

The power radiated by an accelerated charge, as detected by a distant, fixed observer,

is often computed from the Poynting vector of the Lienard-Wiechert fields of an accelerated
charge, yielding,2

dPobserver r2 Efar
2
e2[n̂ × ((n̂ − β) × a)]2
= r2 Sfar = = , (24)
dΩ 4πc 4πc3(1 − β cos θ)6
dPsource dPobserver e2[n̂ × ((n̂ − β) × a)]2
= (1 − β cos θ) = , (25)
dΩ dΩ 4πc3 (1 − β cos θ)5

where n̂ is a unit vector along the direction from the (retarded) source to the observer in
the lab frame. The forms (23) and (25) are equivalent, despite their apparent differences.
2
See, for example, sec. 14.3 of [9].

4
2.2.3 Acceleration Parallel to the Velocity
When the acceleration a is parallel to the velocity v, then a2 = γ 6 a2 and the polar angle θ
is the same as angle α. Then,
(cos θ − β)2 (1 − β 2 ) sin2 θ sin2 θ
sin2 α = sin2 θ = 1 − cos2 θ = 1 − = = ,(26)
(1 − β cos θ)2 (1 − β cos θ)2 γ 2(1 − β cos θ)2
and radiated power in the lab frame follows from eqs. (14) and (18), or from eqs. (23)-(24),
as,
dP,source e2a2 sin2 θ dP,observer e2a2 sin2 θ
= , = . (27)
dΩ 4πc3 (1 − β cos θ)5 dΩ 4πc3 (1 − β cos θ)6
The dipole radiation pattern is folded forwards, as shown on the left below, with maximum
intensity at angle θmax ≈ 1/2γ when γ  1.

The total power radiated by the accelerated charge is,

 
dP e2a2 1 1 − cos2 θ γ 6 e2 a2
P,source = dΩ = d cos θ = , (28)
dΩ 2c3 −1 (1 − β cos θ)5 3c3
after integrating by parts twice (or using 2.153 of [10]).

2.2.4 Acceleration Perpendicular to the Velocity

When the acceleration a is perpendicular to the velocity v, then a2 = γ 4 a2 according to
eq. (22). Taking the acceleration a to lie in the x-z plane, angle α is related to the polar
coordinates (θ , φ ) by,
(cos θ − β)2 + (1 − β 2 ) sin2 θ sin2 φ
sin2 α = cos2 θ + sin2 θ sin2 φ = (29)
(1 − β cos θ)2
The radiated power observed the lab frame follows from eqs. (14) and (18) as,
dP⊥,source e2 a2
= [(cos θ − β)2 + (1 − β 2 ) sin2 θ sin2 φ]
dΩ 4πc3(1 − β cos θ)5
e2 a2
= [(1 − β cos θ)2 − (1 − β 2) sin2 θ cos2 φ]
4πc3(1 − β cos θ)5
 
e2 a2 sin2 θ cos2 φ
= 1− 2 ,
4πc3(1 − β cos θ)3 γ (1 − β cos θ)2
 
dP⊥,observer e2 a2 sin2 θ cos2 φ
= 1− 2 . (30)
dΩ 4πc3(1 − β cos θ)4 γ (1 − β cos θ)2

5
The radiation pattern is peaked in the forward direction, as shown on the right above, with
the null in the pattern at θ ≈ 1/γ for γ  1.
The total power radiated by the accelerated charge is,
  1  1 
dP⊥ e2 a2 d cos θ 1 1 − cos2 θ
P⊥,source = dΩ = − 2 d cos θ
2c3 −1 (1 − β cos θ) 2γ −1 (1 − β cos θ)5
dΩ 3

γ 4 e2 a2
= , (31)
3c3
using 2.153 of [10] and the integral in eq. (28).

2.2.5 Invariance of the Total Power Radiated by the Source

The total power radiated by the accelerated charge can be found more quickly by expressing
the radiated energy in an invariant form.
We start in the instantaneous rest frame of the charge where the integral form of the
Larmor formula can be written as,
dU 
 2e2 a2 2e2aν aν
Psource== = − , (32)
dt 3c3 3c3
recalling eq. (22). The last form of eq. (22) is a Lorentz scalar, so we infer that the total
power Psource radiated by an accelerated charge is a Lorentz scalar, and in any (inertial)
frame we can write,
dU 2e2aν aν 2γ 6e2 [a2 − (v/c × a)2 ]
P = =− = = P . (33)
dt 3c3 3c3
The results (28) and (31) illustrate this general relation for the special cases that a  v and
a ⊥ v.
If the charge has rest mass m0 its 4-momentum is,
pμ = m0 vμ = (E/c, p) = (γm0 c, γm0 v), (34)
where E and p and its energy and momentum. When the charge is subject to an external
electromagnetic field with 4-tensor Fμν , the motion of the charge is described by,
dpμ
= m0 aμ = fμ = eFμν v ν = eγ(E · β, E + β × B), (35)
dτ
so the total radiated power (33) can be written as,
2e2 aν aν 2γ 2 e4[(E + β × B)2 − (E · β)2 ]
P =− = . (36)
3c3 3m20 c3
For relativistic motion (γ  1) in zero electric field with B ⊥ v ≈ cv̂, the circular trajectory
has radius R ≈ γm0 c2 /eB, and the radiated power is,
2γ 2 e4B 2 2γ 4 c e2
P ≈ ≈ . (37)
3m20 c3 3R2
This synchrotron radiation limits the performance of circular high-energy particle accelera-
tors.3
3
See, for example, [11, 12].

6
2.2.6 Radiated Energy-Momentum 4-Vector
In general, an accelerated charge radiates momentum as well as energy, so we seek a 4-vector
description that combines these effects.
For this, we first rewrite eq. (32) as,

dU  dU  2e2 a2 2e2 aν aν

= = = − , (38)
dt dτ 3c3 3c3
noting that in the instantaneous rest frame the proper time interval is simply dτ = dt.
Further, the total radiated momentum P is zero in this frame,
  
dP dP d2 p  1  du

 e2 a 2
= = dΩ = n̂   dΩ = n̂ sin2 α dΩ = 0. (39)
dt dτ dt dΩ c dt dΩ 4πc4
If we a 4-vector for increments of the radiated energy and momentum according to,

dUμ = (dU, c dP), (40)

Then eqs. (38)-(39) suggest that we can write,

dUμ 2e2 aν aν
=− bμ , (41)
dτ 3c3
where bμ is a 4-vector whose components in the  frame are (1, 0). The simplest choice is
that bμ = vμ /c = γ(1, v/c). Hence, it is plausible that,

dUμ 2e2aν aν
=− vμ . (42)
dτ 3c4
The momentum radiated in the lab frame is then given by,

dP 1 dP 2e2aν aν v 2γ 6 e2[a2 − (v/c × a)2 ] dU v

= =− 4
= 5
v= . (43)
dt γ dτ 3c c 3c dt c2

2.2.7 Angular Distribution of Radiated Momentum

If we denote by n̂ the unit vector from the (retarded) source position to the observer in the
lab frame, then the angular distribution of the momentum radiated by an accelerated charge
in the lab frame is related to the angular distribution of the radiated energy by,
dP 1 dU 1 dPsource
= n̂ = n̂ . (44)
dt dΩ c dt dΩ c dΩ
This can be evaluated using eqs. (23) or (25), etc.
As an example, consider a charge e in a uniform magnetic field, for which the motion
is helical in general. The acceleration a = −a ρ̂ is perpendicular to the instantaneous
velocity v of the charge, and is towards the axis of the helix (to which unit vector ρ̂ is
perpendicular). The acceleration a in the instantaneous rest frame is also perpendicular
to the axis, a = −γ 2 a ρ̂. Hence, the radiation pattern in the instantaneous rest frame has

7
the form sin2 ψ , where angle ψ is measured with respect to ρ̂. This radiation pattern (in
the  frame) is not symmetric about the direction of the velocity v of the charge in the lab
frame.4 According to eq. (2), the radiation pattern in the lab frame is also not symmetric
about the the direction of v. Yet, according to eq. (43), the total momentum radiated in
the lab frame is along v (as measured by observers for which the present time in the lab is
their retarded time).5

It is tempting to cast the expressions for the angular distributions of radiated energy and
momentum into a Lorentzian form. For this, we introduce a 4-vector nμ associated with the
unit vector n̂ by defining its components in the instantaneous rest frame to be,

nμ = (0, n̂ ). (46)

Then, the square of its invariant length is,

nμ nμ = −1. (47)

The components of nμ in the lab frame are,

nμ = (γ(n̂ · β), n̂ + (γ − 1)(n̂ · β)β) = (n0 , n). (48)

Recalling the Larmor formula eq. (18), we can write,

a2 sin2 α = a⊥ 2 = a2 (1 − cos2 α ) = a2 − (a · n̂ )2 , (49)

where a⊥ is the component of the acceleration in the instantaneous rest frame perpendicular
to the direction to the observer, n̂. That is,

a⊥ = a − (a · n̂ ) n̂ = a + aν nν n̂ , (50)

noting that aν nν = −a · n̂. This leads us to define the 4-vector,

a⊥,μ = aμ + aν nν nμ , (51)
4
This can be illustrated via consideration of the  frame that moves with velocity −v to the lab frame,
where v is the projection of v onto the axis of the helix. In the  frame, the motion of the charge in is
a circle, and the radiation patterm (see sec. 2.2.4) is symmetric about the plane perpendicular to ρ̂ that
contains the charge.
In particular, we consider a photon that is emitted in the forward direction in the  frame, i.e., in the
direction of v . We can say that this photon is a particle with velocity u = c v̂ = c v̂⊥ .
A Lorentz transformation from the  frame back to the lab frame takes v to v and u to u. The velocity
transfrom of u is such that the angle θ u of the velocity vector with repect to the axis of the helix is,
 
1 − v2 /c2 u⊥ 1 − v2 /c2 c
tan θu = = , (45)
u + v v


while, noting that v⊥ = v⊥ / 1 − v2 /c2 , the transform for angle θ v yields tan θv = v⊥ /v . In general,

v⊥ = 1 − v2 /c2 c, so a photon radiated in the forward direction in the  frame is not in the forward
direction in the lab frame.
5
This result is implied just before eq. (5) of [13], where is it noted that the radiation of energy/momentum
by a charge in helical motion does not change the direction of the charge, while reducing its speed slightly.

8
whose invariant length squared is,

a⊥,μ aμ⊥ = aμ aμ + (aν nν )2 nμ nμ + 2(aν nν )2 = −a2 + (aν nν )2 = −a2 + (a · n̂ )2 = −a⊥ 2 .(52)

With this notation (inspired by sec. 5.1 of [14]), we can write eq. (18) as,

dP  dU  e2a2 sin2 α e2a⊥ 2 e2 a⊥,ν aν⊥

= = = = − , (53)
dΩ dt dΩ 4πc3 4πc3 4πc3

The angular distribution of radiated momentum in the frame can then be written as,

dP 1 dU   e2a⊥,ν aν⊥ 

= n̂ = − n̂ , (54)
dt dΩ c dt dΩ 4πc4
It is suggestive to combine eqs. (53)-(54) into a quantity which has the appearance of a
4-vector,
 
dU  dP e2a⊥,ν aν⊥  e2a⊥,ν aν⊥  vμ  ? dU
μ
 
, c  
= − 3
(1, n̂ ) = − 3
+ nμ = . (55)
dt dΩ dt dΩ 4πc 4πc c dτ dΩ

However, this cannot be so, as can be seen in various ways. While dUμ /dτ is a 4-vector, dΩ
is not a Lorentz invariant as it transforms according to eq. (12). Also, as discussed at the
beginning of this section, dU/dt dΩ and c dP/dt dΩ are in the ratio,

n γ[n̂ + (γ − 1)(n̂ · β)β]
1 : n̂ = =  , (56)
n γ 2 + (γ 4 − 2γ 2 + 2γ − 1)(n̂ · β)2

whereas in the lab frame v0 /c + n0 = γ(1 + n̂ · β), while the 3-vector part of vμ /c + nμ is,

γβ + n̂ + (γ − 1)(n̂ · β)β. (57)

Hence, this author is skeptical that eq. (5-16) of [14], i.e., eq. (55) above, holds in the
lab frame, although it is correct in the instantaneous rest frame.

2.3 Acoustic Radiation

The speed c of electromagnetic radiation is the same (in vacuum) for any inertial observer,
whereas the speed of acoustic radiation is different for observers with different velocities. As
such, no general transformations of the angular distribution of acoustic radiation between
observers with different velocities can be given for acoustic radiation.
As an example of the greater complexity of acoustic radiation compared to electromag-
netic radiation, [15] quotes three different results by three different authors for the for-
ward/backward ratio for (monopole) sound intensity for source and observer at rest, but for
air moving along the direction between them. Lighthill [16], p. 476, makes the claim that the
forward backward ratio is unity, independent of the speed of the air. This seems impossible
for air speeds greater than the speed of sound, but it could be true for subsonic airspeeds.
For electromagnetic radiation, if the source and observer are at rest in a common inertial
frame, the angular distribution of radiation in that frame is independent of the velocity of

9
that frame, as there is no absolute rest frame for electromagnetic waves (no electromagnetic
æther). Hence, it is “obvious” that the forward/backward ratio for an isotropic electromag-
netic radiator is unity, no matter what is the velocity of the frame of the source and observer.
In contrast, the angular distribution (in the rest frame of the source) of sound waves emitted
by a source depends, in general, on the velocity of the air in that frame; the general sense
is that the angular distribution is “blown downwind”. So, it is less obvious that Lighthill’s
claim is true for sound waves than for electromagnetic waves.

References
[1] A.H. Bucherer, Mathematische Einführung in die Elektronentheorie, (Teubner, Leipzig,
1904), http://kirkmcd.princeton.edu/examples/EM/bucherer_electrons_04.pdf

[2] A.H. Bucherer, Introduction Mathématique a la Théorie des Électrons, Arch. Sci. Phys.
Nat. 22, 105 (1906), http://kirkmcd.princeton.edu/examples/EM/bucherer_aspn_22_105_06.pdf

[3] L. de la Rive, Sur l’introduction du facteur de Doppler dans la solution des équations
de la théorie des électrons, Arch. Sci. Phys. Nat. 22, 209, 321 (1906),
http://kirkmcd.princeton.edu/examples/EM/delarive_aspn_22_209_07.pdf

[4] A.H. Bucherer, Les Potentiels Retardes Transformés, Arch. Sci. Phys. Nat. 23, 129
(1906), http://kirkmcd.princeton.edu/examples/EM/bucherer_aspn_23_129_07.pdf

[5] L. de la Rive, Sur l’introduction du facteur de Doppler dans la solution des équations
de la théorie des électrons, Arch. Sci. Phys. Nat. 23, 433 (1907),
http://kirkmcd.princeton.edu/examples/EM/delarive_aspn_23_433_07.pdf

[6] G.A. Schott, Electromagnetic Radiation and the Mechanical Reactions Arising from it
(Cambridge U. Press, 1912),
http://kirkmcd.princeton.edu/examples/EM/schott_radiation_12.pdf

[7] V.L. Ginzburg and S.I. Strovatskii, Developments in the Theory of Synchrotron Radi-
ation and its Reabsorption, Ann. Rev. Astron. Astrop. 7, 375 (1969),
http://kirkmcd.princeton.edu/examples/EM/ginzburg_araa_7_375_69.pdf

[8] H. Matzner and K.T. McDonald, Isotropic Radiators (Apr. 8, 2003),

http://kirkmcd.princeton.edu/examples/isorad.pdf

[9] J.D. Jackson, Classical Electrodynamics, 2nd ed. (Wiley, 1975),

http://kirkmcd.princeton.edu/examples/EM/jackson_ce2_75.pdf

[10] I.S. Gradshteyn and I.M. Ryzhik, Table of Integrals, Series, and Products, 5th ed.
(Academic Press, 1994), http://kirkmcd.princeton.edu/examples/mechanics/gradshteyn_80.pdf

[11] K.T. McDonald and C.G. Tully, Maximum Energy of Circular Colliders (Dec. 10, 2001),
http://kirkmcd.princeton.edu/examples/lep.pdf

10
[12] K.T. McDonald, The Maximal Energy Attainable in a Betatron (Nov. 19, 2000),
http://kirkmcd.princeton.edu/examples/betatron.pdf

[13] N.S. Kardashev, Nonstationarity of Spectra of Young Sources of Nonthermal Radio

Emission, Sov. Astron. 6, 317 (1962),
http://kirkmcd.princeton.edu/examples/astro/kardashev_sa_6_317_62.pdf

[14] F. Rohrlich, Classical Charged Particles (Addison-Wesley, 1965; World Scientific, 2007).

[15] C.H.E. Warren, A Note on Moving Multipole Sources of Sound, J. Sound Vib. 44, 3
(1976), http://kirkmcd.princeton.edu/examples/fluids/warren_jsv_44_3_76.pdf

[16] J. Lighthill, The Propagation of Sound through Moving Fluids, J. Sound Vib. 24, 471
(1972), http://kirkmcd.princeton.edu/examples/fluids/lighthill_jsv_24_471_72.pdf

11