Apeiron, Vol. 10, No. 4, October 2003

29

Relativistic Doppler Eﬀect and the Principle of Relativity
W. Engelhardta
Abstract The frequency shifts predicted by the ‘relativistic’ Doppler eﬀect are derived in the photon picture of light. It turns out that, in general, the results do not depend exclusively on the relative velocity between observer and light source. In this respect, the relativistic Doppler eﬀect is not distinct from the classical one, where the shifts are also diﬀerent depending on whether the source or the observer moves. The ‘relativistic’ formulae for these two cases have been conﬁrmed by experiment and are described in many textbooks. It was, however, not recognized that they are at variance with Einstein’s relativity principle extended to electromagnetic waves.
Keywords: Relativity, Doppler Eﬀect, Aberration aprivate address: Fasaneriestrasse 8, D-80636 Mu¨nchen,
Germany, wolfgangw.engelhardt@t-online.de postal address: Max-Planck-Institut fu¨r Plasmaphysik, D-85741 Garching, Wolfgang.Engelhardt@ipp.mpg.de
c 2003 C. Roy Keys Inc. – http://redshift.vif.com

Apeiron, Vol. 10, No. 4, October 2003

30

I Introduction

When a wave with angular frequency ω and phase velocity c propagates in a medium, an observer moving with velocity v parallel to c experiences a shifted frequency ω1 = ω (1 − v/c). When the source oscillating with frequency ω moves with velocity v away from the observer, who is at rest in the medium, he will also measure a shifted frequency ω2 = ω/ (1 + v/c) which is, however, diﬀerent from ω1. These frequency shifts are known as ‘classical Doppler eﬀect’ and may be observed in sound waves, for example. Because of the intervening medium, they do not merely depend on the relative velocity between observer and source, in contrast to other mechanical phenomena such as collision processes between material bodies which, in Newtonian mechanics, depend only on relative motion.
Electromagnetic waves share with sound the property that their propagation velocity is independent of the motion of the source, as Einstein postulated in 1905 [1]. He formulated in the same paper a second postulate, his ‘extended principle of relativity’, namely that all physical phenomena, including the electromagnetic ones, should only depend on relative velocities, as far as constant motion is concerned. From these postulates he deduced the Lorentz transformation, which led to the consequence that the velocity of light is constant in all inertial systems, regardless how they move relatively to each other.
It is obvious that Einstein’s postulates are not compatible with a medium (‘ether’) theory of light. If both postulates are realized in nature, the Doppler eﬀect of light must distinguish itself from that of sound, insofar as the frequency shift of light can only depend on the relative velocity between source and

c 2003 C. Roy Keys Inc. – http://redshift.vif.com

Apeiron, Vol. 10, No. 4, October 2003

31

observer. Amazingly, the new ‘relativistic’ Doppler formulae,

which are derived from the Lorentz transformation in textbooks,

do not have this property: As in the case of sound, there is still

a diﬀerence depending on whether the observer moves and the

source is at rest, or whether the source moves and the observer

is at rest. Even more surprising, the formulae applying to the

diﬀerent situations have been checked by experiment to some

extent and were found to be correct within experimental error.

As far as we can see, it was not recognized that these results

cast doubt on the validity of Einstein’s relativity principle with

regard to the physics of electromagnetic waves.

In this paper the formulae for the relativistic Doppler eﬀect

are derived without using the Lorentz transformation (Section

II). This can be done in the photon picture of light in combina-

tion with the relativistic modiﬁcation of Newtonian mechanics

applied to detector and source. We ﬁnd indeed diﬀerent fre-

quency shifts depending on whether the observer moves or the

source. In Section III these results are compared with those de-

duced by Einstein from the Lorentz transformation. It turns out

that there is practical agreement with the formulae derived in

Section II. In Section IV the conclusion drawn from the Lorentz

transformation that a wave front ‘appears’ tilted, if there is a

relative velocity between observer and source, is scrutinized and

found to be untenable. In Section V consequences of our analysis

for the principle of relativity and for the Lorentz transformation

are discussed.

c 2003 C. Roy Keys Inc. – http://redshift.vif.com

Apeiron, Vol. 10, No. 4, October 2003

32

II Doppler eﬀect derived in the photon picture

of light

Without using the Lorentz transformation, the Doppler effect can be explained in the photon picture which was established by the Planck-Einstein quantum theory of light. According to this model the energy of a photon is:

E= ω

(1)

and its momentum:

p = ω/c

(2)

For a body with ﬁnite mass the relativistically modiﬁed formulae of Newtonian mechanics apply:

E = m c2

(3)

p = mv

(4)

m=

m0

(5)

1 − |v |2 /c2

They are valid as long as the De Broglie wavelength of a moving body is small compared to its dimensions. The rest-mass m0 takes into account any internal energy which may be of kinetic or potential form.
Let us assume that a detector with mass m moves with velocity v in x-direction and absorbs a photon with total momentum p, as depicted in Figure 1. The momentum balance before and after absorption yields:

p cos α + m v = m vx

(6)

p sin α = m vy

(7)

c 2003 C. Roy Keys Inc. – http://redshift.vif.com

Apeiron, Vol. 10, No. 4, October 2003

33

y

m ©v

p

detector

α

x

Figure 1: Absorption of a photon in a moving detector

where m = m0/

1−

v

2 x

+

v

2 y

/c2 is the mass of the detec-

tor after absorption, and vx,y are its velocity components. The energy balance requires:

c p + m c2 = m c2

(8)

The rest-mass of the detector is increased when it absorbs a certain amount of energy W :

W

m0 = m0 + c2

(9)

With (1) this energy may be expressed by a frequency ω1 = W , and can be determined experimentally by measuring the excitation energy of the detector. Eliminating from (6) and (7) the velocity components after absorption one obtains together with (9) and (5):

m2=

W m0 + c2

21 + c2

p2 + m2 v2 + 2 p m v cos α

(10)

c 2003 C. Roy Keys Inc. – http://redshift.vif.com

Apeiron, Vol. 10, No. 4, October 2003

34

Inserting this into (8) yields:

1 − (v/c) cos α W = ω1 = p c 1 − v2/c2 (1 + W/ (2 m0 c2)) (11)

The very small ratio of photon energy to rest-energy of the detector may be neglected, and one obtains with (2) and the abbreviation β = v/c the ﬁnal result:

1 − β cos α

ω1 = ω 1 − β2

(12)

It is distinguished from the classical formula for sound by the de-

nominator which describes the ‘quadratic’ part of the Doppler ef-

fect for light and leads to an increase of the measured frequency.

The necessary energy is, of course, taken from the kinetic energy

of the detector.

Let us assume now that a source with mass m moves with

velocity v in x-direction and emits a photon, as depicted in Fig-

ure 2. The momentum balance before and after emission yields

in this case:

m v = p cos α + m vx

(13)

0 = p sin α + m vy

(14)

and the energy balance requires:

m c2 = c p + m c2

(15)

The excitation energy of the emitting atom may be expressed by a frequency: W = ω. After de-excitation the rest-mass of the source is decreased by the amount W/c2:

W

m0 = m0 − c2

(16)

c 2003 C. Roy Keys Inc. – http://redshift.vif.com

Apeiron, Vol. 10, No. 4, October 2003

35

y

p

α m ©v

source

x

Figure 2: Emission of a photon by a moving source

Eliminating from (13-16) the mass m and the velocity components of the source after emission one obtains with (5):

W pc =

1 − β2 (1 − W/ (2 m0 c2))

(17)

1 − β cos α

Ignoring again the ratio W/ (2 m0 c2) one ﬁnds with (1) and (2) the frequency of the emitted photon:

1 − β2

ω2 = ω 1 − β cos α

(18)

This is the classical Doppler formula for a moving source, multiplied with the ‘dilation’ factor 1 − v2/c2, which leads to a relative decrease of the emitted frequency.
If both source and detector move with the same velocity, the motion dependent factors in (12) and (18) cancel out so that just the frequency ω = W is measured. The cancellation occurs
c 2003 C. Roy Keys Inc. – http://redshift.vif.com

Apeiron, Vol. 10, No. 4, October 2003

36

also when detector and source move in opposite directions, but

α is 90 degrees. The null result of the experiment published in

[2], where source and detector are placed at the periphery of a

rotating disk, conﬁrms this expectation.

The relativistic Doppler formulae contain - in contrast to the

classical formulae - the dilation factor. From the above deriva-

tion it is evident that the dilation factor originates from the fact

that inelastic scattering processes were considered. If the pho-

tons are neither emitted nor absorbed, but just elastically scat-

tered at moving mirrors, the rest-mass of the mirrors remains

unchanged and the dilation factor does not play a role. This

may be shown explicitly by replacing the absorbing detector in

Figure 1 by a mirror which reﬂects the impinging photon into

the y-direction. Momentum and energy balance become now at

constant rest-mass:

m v + p1 cos α = m vx , p1 sin α = m vy + p2

p1 − p2 = (m − m) c

(19)

Eliminating the velocity components after reﬂection yields with (5) the relationship:

m 2 − m2 c2 = 2 m v p1 cos α + p21 + p22 − 2 p1 p2 sin α (20)

Together with the energy balance in (19) one obtains the result:

p2

=

p1

−

2mv

p1

cos α + p21 + p22 − (m + m) c

2 p1

p2

sin α

= p1 (1 − β cos α) , for m0 → ∞

(21)

For an inﬁnitely large rest-mass of the mirror, this is the classical Doppler formula without the dilation factor, in contrast to equation (12).

c 2003 C. Roy Keys Inc. – http://redshift.vif.com

Apeiron, Vol. 10, No. 4, October 2003

37

Just as in the classical description, the frequencies ω1 and ω2 in the relativistic formulae (12) and (18) are not equal at α = 0,

if one identiﬁes in (18) β, the velocity of the source, with −β in

(12), the velocity of the detector. This means that the relativis-

tic Doppler eﬀect does not depend exclusively on the relative

velocity between observer and source, which clearly contradicts

Einstein’s extended relativity principle. On the other hand, for-

mula (12) has been conﬁrmed in a Mo¨ssbauer experiment with

a moving detector [3], and formula (18) was early veriﬁed by

measuring the emission from moving atoms [4]. The outcome

of these experiments and the null result reported in [2] refute

Einstein’s relativity postulate with respect to its application on

electromagnetic waves. This was apparently not recognized in

the literature, but it was erroneously claimed that the exper-

iments conﬁrm the extended relativity principle. The reason

for this misinterpretation must be sought in the fact that the

relativistic Doppler formulae have also been derived from the

Lorentz transformation which rests on Einstein’s relativity prin-

ciple. The connection between Lorentz transformation and the

relativistic Doppler eﬀect will be discussed in the next Section.

III Doppler eﬀect and Lorentz transformation
Einstein derived on page 911 of his paper [1] from the Lorentz transformation a new formula for the Doppler eﬀect which applies to a moving observer. His deduction may be delineated as follows. Let us assume that there is a coordinate system (x, y) and a second system (x , y ) which coincides with the unprimed system at t = 0, but moves with velocity v in positive x-direction. The Lorentz transformation and its inverse are then

c 2003 C. Roy Keys Inc. – http://redshift.vif.com

Apeiron, Vol. 10, No. 4, October 2003

38

given by the expressions:

t − v x/c2

x−vt

t=

, x=

, y =y

1 − v2/c2

1 − v2/c2

t + v x /c2

x +vt

t=

, x=

, y=y

(22)

1 − v2/c2

1 − v2/c2

A plane light wave travelling in vacuo in the direction of the wave vector k with phase velocity ω/k = c in the unprimed system is described by:

cos (ω t − k x cos α − k y sin α)

(23)

In the primed system it is also a plane wave described by:

cos (ω t − k x cos α − k y sin α )

(24)

When the amplitude of the wave vanishes at a certain spacetime point, it must vanish in both systems which requires that the phase in (23) at (x, y, t) must be equal to the phase in (24) at (x , y , t ):

ω t − k x cos α − k y sin α = ω t − k x cos α − k y sin α (25)

The propagation velocity is the same in both systems:

ω/k = ω /k = c

(26)

which follows from diﬀerentiation of the phases in (25) with respect to time in the two systems. Inserting the second line of (22) into (23) one obtains:

cos t ω − k v cos α − x k cos α − ω v/c2 − y k sin α (27)

1 − v2/c2

1 − v2/c2

c 2003 C. Roy Keys Inc. – http://redshift.vif.com

Apeiron, Vol. 10, No. 4, October 2003

39

Since, in view of (25), the expression (27) should be equal to (24)

for all values of (x , y , t ), we ﬁnd with (26) the identiﬁcations:

ω (1 − β cos α)

ω=

(28)

1 − β2

k (cos α − β)

k cos α =

, k sin α = k sin α (29)

1 − β2

These transformation formulae are independent of where the source or where the detector is located. If one assumes with Einstein that the detector is at rest in the moving primed system, the measured frequency ω1 is according to (28):

ω1

=

ωS

(1

−β 1−

cos α) β2

(30)

in agreement with (12) and the formula given by Einstein in [1]. It can also be found in textbooks, for example [5, 6]. The angle α is, in principle, a function of time, if the source is located at a ﬁnite distance R from the detector:

vt

α (t) = α + ∆α (t) α + sin α

(31)

R

Choosing the distance suﬃciently large, however, the temporal change can be made arbitrarily small for ﬁnite observation times.
If one assumes that the source is at rest in the moving primed system, it has the velocity −v with respect to the detector. Equation (28) yields in this case:

ωS

=

ω2

(1

+β 1−

cos α) β2

(32)

c 2003 C. Roy Keys Inc. – http://redshift.vif.com

Apeiron, Vol. 10, No. 4, October 2003

40

Resolved with respect to the detector frequency ω2 one obtains:

ω2

=

ωS 1+

1 − β2 β cos α

(33)

in agreement with (18) for v → −v. Again, the angle α changes slowly in time as in (31), but for R → ∞ it stays constant. Formula (33), which was ﬁrst conﬁrmed experimentally by Ives and Stilwell [4], can be found in all textbooks dealing with the subject.
As demonstrated above, both formulae of Section II can also be derived from the Lorentz transformation, but they are not equal, since:

ω1 − ω2 =

ωS β2 sin2 α 1 − β2 (1 + β cos α)

(34)

With the exception of the case α = 0, which, exclusively, is considered by Feynman [5], the frequencies (30) and (33) have the same property as the classical ones: They do not depend on the relative velocity alone, in contradiction to the relativity principle.
It should be noted that the derivation of (33) did not make any use of equations (29), which predict a change of the angle α when the source starts to move. From equation (31) we rather concluded that α is neither inﬂuenced by the velocity of the observer, nor by the velocity of an inﬁnitely distant source. This was in perfect agreement with the postulate that the propagation of the wave is independent of the motion of the source. The apparent inconsistency between (31) and (29) requires to take a closer look at the phenomenon of aberration, which will be discussed in the next Section.

c 2003 C. Roy Keys Inc. – http://redshift.vif.com

Apeiron, Vol. 10, No. 4, October 2003

41

IV Aberration

Einstein has used the Lorentz transformation to give a new interpretation of the astronomically observed phenomenon of aberration. From the two equations (29) one can eliminate k and obtain the transformation law for the angle between the velocity and the wave vector:

cos α

=

cos α − β 1 − β cos α

(35)

This equation can be found on page 912 of Einstein’s paper [1].

The interpretation of (35) is that a wave travelling at the angle

α in the unprimed system with respect to the x-axis will ‘appear’

to travel at the angle α with respect to the x -axis in the moving

primed system where the observer is at rest. For small β one

obtains:

α−α

= αA

=

β sin α 1 − β cos α

(36)

Since the earth changes annually its velocity with respect to the ﬁxed stars, one should observe an aberration angle αA given by (36) in the focal plane of a telescope. This is indeed the case and was already discovered in the early eighteenth century by Bradley [7]. He gave an explanation on the basis of the ether theory and the ﬁnite velocity of light. Aberration is a ﬁrst order eﬀect and follows also from the Galilei transformation. Hence, the explanation did not have to wait for the construction of the Lorentz transformation.
Similarly as in (28), it is not speciﬁed in (35) in which system the source or the observer are located. If we assume now that the observer is at rest and the source moves, we have to interchange

c 2003 C. Roy Keys Inc. – http://redshift.vif.com

Apeiron, Vol. 10, No. 4, October 2003

42

the primes in (35) and put v → −v as before. The result is:

cos α + β

cos α =

(37)

1 + β cos α

which is not distinguished from (35). This means that aberration is a symmetric phenomenon according to the Lorentz transformation which depends only on the relative velocity between source and observer.
In Figure 3 four cases are selected with diﬀerent velocities of a telescope and a star. In the upper row the predictions of the Lorentz transformation concerning the occurrence of aberration are sketched on the basis of equations (35) and (37). In case I a telescope is at rest and a star - also at rest - is located on its optical axis at inﬁnite distance. The plane wave fronts of the light entering the telescope are imaged by the lens to a diﬀraction spot which lies in the focal plane on the optical axis. When the telescope moves parallel to the wave fronts (case II) equation (35) predicts a tilting of the wave fronts so that the image of the star is shifted to the left of the optical axis. In case III the star moves with the same velocity as the telescope to the right. This case is equivalent to case I, as the relative velocity vanishes. Equations (35) and (37) predict no aberration of the image in this case, since α = α = 90 degrees. In case IV the telescope is at rest and the star moves to the left. The relative velocity is not distinguished from that of case II and equation (37) predicts, consequently, the same aberration of the image as in case II.
In the lower row of Figure 3 the aberration shifts are sketched which can be inferred from astronomical observations. As the

c 2003 C. Roy Keys Inc. – http://redshift.vif.com

Apeiron, Vol. 10, No. 4, October 2003
v star

43 v

telescope relativity theory I

v

v

II

III

IV

v

v

v

v

experiment I

II

III

IV

Figure 3: Aberration shift as predicted by the Lorentz transformation (upper row). Aberration shift as inferred from astronomical observations (lower row)

c 2003 C. Roy Keys Inc. – http://redshift.vif.com

Apeiron, Vol. 10, No. 4, October 2003

44

earth reverses its velocity within half a year, one observes aber-

ration for all ﬁxed stars by the same amount in agreement with

equation (35). Case II applies and the aberration angle is given

by the ratio of the earth’s velocity to the velocity of light accord-

ing to equation (36). This is also in agreement with Bradley’s

ether theory. If the motion of the stars would not change dur-

ing terrestrial observation periods, the relevant relative velocity

would be exclusively determined by the annual changes of the

earth’s velocity. Any motion of the stars, as in case III or IV,

would have no inﬂuence on the actual observation, since the true

position of the stars is unknown.

There are, however, double star systems [8] the components

of which change their velocity on a time scale ranging from days

to years. Spectroscopic binaries, in particular, which are too

close to be resolved by a telescope, have velocities exceeding

the earth’s velocity round the sun. They revolve around their

common center of gravity within days, a period during which

the motion of the earth is practically constant. In this case

equation (37) predicts an aberration shift which should have a

magnitude given by the ratio of the star’s velocity to the velocity

of light. The components of the binary system should be easily

separable, when their changing velocities are comparable to the

earth’s velocity round the sun. This is, however, not observed.

The binary components remain unresolved which means that

their velocity has no inﬂuence on aberration, as depicted in case

IV in the lower row, in agreement with equation (31). Rotating

binary systems follow the same pattern as all ﬁxed stars and

are observed within a period of a year under the same universal

aberration angle, i.e. their apparent position changes with an

c 2003 C. Roy Keys Inc. – http://redshift.vif.com

Apeiron, Vol. 10, No. 4, October 2003

45

annual period common to all distant stars, as depicted in case

III, lower row.

The astronomical observations on double star systems favor

apparently equation (31) rather than equation (37), as (31) pre-

dicts no detectable tilting of the wave front when the star moves,

provided it is suﬃciently far away. If a star would rotate on an

orbit which has the same diameter as the earth’s orbit round the

sun, it would appear under the same parallax angle under which

a star at rest is normally observed from the orbiting earth. It is

well known that the parallax angle even of the closest stars is an

order of magnitude smaller than the universal aberration angle

due to the earth’s velocity. The parallax angle of spectroscopic

double star systems is so small that it cannot be resolved by a

telescope. Their large velocity does not change this in any way.

The correct prediction of the terrestrially observed aberra-

tion angle by equation (36) must be considered as fortuitous.

The Lorentz transformation fails completely to predict correctly

the observations made on double star systems, when equation

(37) is applied. It is, of course, a contradiction to say on the

one-hand-side that two coordinate systems move such that the

y - and the y-axis remain parallel during the motion, as follows

from the Lorentz transformation (22), and say on the other hand

, one and the same wave front spans an angle α with the y-axis,

but a diﬀerent angle α with the y -axis. The usual terminolog-

ical escape from the discrepancy is to say that the wave fronts

appear tilted rather than they are tilted. This wording only

covers the problem, but it does not solve it.

c 2003 C. Roy Keys Inc. – http://redshift.vif.com

Apeiron, Vol. 10, No. 4, October 2003

46

V Discussion and conclusion

The Lorentz transformation may be seen as a formalism of merely kinematic nature to ensure that the velocity of light is the same in all inertial systems, independent of their motion. It ignores completely the dynamics involved in the interaction of light with matter. Hence, it would be surprising, if it could predict the exact formulae of the relativistic Doppler eﬀect, which were rigorously derived in Section II on the basis of the dynamic formulae (1-5). The ratio of photon energy to rest-energy in (11) and (17), although negligible in practice, did not appear in the derivations of Section III. Under this aspect the agreements with the approximate formulae in Section II must be considered as more or less fortuitous.
Surprisingly, it was not recognized in the literature that the correct Doppler formulae of Section II, which appear to be conﬁrmed by the Lorentz transformation, are at variance with its underlying idea: Extension of the principle of relativity from mechanics to light. Given now the fact that the relativistic Doppler formulae are conﬁrmed by experiment, Einstein’s extended relativity principle must be rejected. His conclusion drawn from his two postulates, namely that the velocity of light is constant in all inertial systems, appears untenable. Only Einstein’s postulate that the velocity of light is independent of the motion of the source, which holds also for a medium theory, has been conﬁrmed by experiment. Observations on the cosmic background radiation, as measured with earthbound detectors, show that there is a unique reference system in which the background radiation is isotropic and electromagnetic waves propagate with velocity c in all directions. The motion of the earth relative

c 2003 C. Roy Keys Inc. – http://redshift.vif.com

Apeiron, Vol. 10, No. 4, October 2003

47

to this system can clearly be determined by using the Doppler eﬀect.
The Lorentz transformation predicts an apparent tilting of the angle spanned by the wave vector and the relative velocity between observer and source. This claim is refuted by astronomical observations on moving light sources, as pointed out in Section IV. Neither the velocity of an inﬁnitely distant source nor the velocity of an observer inﬂuences the direction of the wave vector. This is in agreement with the postulate that the propagation of light in free space is independent of the motion of the source. As aberration is of ﬁrst order in β, the observed asymmetry of the eﬀect proves even more convincingly than the comparison of the diﬀerent Doppler formulae that the Lorentz transformation contradicts experimental facts. The terrestrial observation of aberration, which depends only on the velocity of the earth, has nothing to do with a hypothetical tilting of the wave fronts, but can be explained satisfactorily by the ﬁnite travel time of light between the imaging element and the focal plane of a telescope.
Apart from the relativistic Doppler eﬀect, the null result of the Michelson-Morley experiment has been taken as experimental proof for the validity of the Lorentz transformation. In recent work by the present author [9] it has been shown in detail that the outcome of the Michelson-Morley experiment proves the opposite: The Lorentz-Fitzgerald contraction, which is derived from the Lorentz transformation, cannot be real. If this were the case, Michelson should have measured a fringe shift. The misinterpretation of this experiment in the literature is due to a wrong assumption concerning the reﬂection angle of photons at a moving mirror. Adopting the correct physical law of

c 2003 C. Roy Keys Inc. – http://redshift.vif.com

Apeiron, Vol. 10, No. 4, October 2003

48

reﬂection, a fringe-shift is not expected.

Taking the experimental evidence together, suﬃcient proof

has been presented that the Lorentz transformation does not

represent a physical law which is realized in nature. Its under-

lying principle, that optical phenomena depend only on relative

velocities, cannot be maintained.

Acknowledgments
The author is indebted to Dr. O. Kardaun at Max-PlanckInstitut fu¨r Plasmaphysik, Garching, for carefully reading the manuscript and making useful suggestions for the ﬁnal presentation of the material.

c 2003 C. Roy Keys Inc. – http://redshift.vif.com

Apeiron, Vol. 10, No. 4, October 2003

49

References

[1] A. Einstein, Annalen der Physik 17 (1905) 891.
[2] D. C. Champeney, G. R. Isaak, A. M. Khan, Phys. Letters 7 (1963) 241.
[3] H. J. Hay, J. P. Schiﬀer, T. E. Cranshaw, P. A. Egelstaﬀ, Phys. Rev. Letters 4 (1960) 165. See also T. E. Cranshaw in Proceedings of the International School of Physics, Varenna, Course XX, 1961, Academic Press, New York (1962), page 208.
[4] H. E. Ives, G. R. Stilwell, J. Opt. Soc. Am. 28 (1938) 215.
[5] R. P. Feynman, R. B. Leighton, M. Sands, The Feynman Lectures on Physics, Vol. I, chapter 34-4, Addison-Wesley Publishing Company, Reading, Massachusetts (1963).
[6] J. D. Jackson, Classical Electrodynamics, Second Edition, John Wiley & Sons, New York (1975), Sect. 11.3.
[7] J. Bradley, Phil Trans. 35 (1728) 637.
[8] E. Eisner, Am. J. Phys. 35 (1967) 817.
[9] W. Engelhardt, Interpretation of the Michelson-Morley experiment in the photon picture of light, manuscript available on request.

c 2003 C. Roy Keys Inc. – http://redshift.vif.com

