Section 4.2: Einstein Quotes

4.2.1 Volume 2: The Swiss Years: Writings 1900-1909, Doc 47: On the relativity principle and the conclusions drawn from it, 1907
Newton’s equations of motion retain their form when one transforms to a new system of coordinates that is in uniform translational motion relative to the system used originally. One could not doubt that the laws of nature are the same, without regard to which of the coordinate systems they are referred. This independence from the state of motion of the system of coordinates used we call “the principle of relativity”. 

After the acceptance of that [Lorentz] theory one had to expect that one would succeed in demonstrating an effect of the terrestrial motion relative to the luminiferous ether on optical phenomena. Lorentz proved that in optical experiments an effect of that relative motion on the ray path is not to be expected, as long as the calculation is limited to terms in which the ratio v/c, the relative velocity to the velocity of light in a vacuum, appears in the first power. But the negative result of Michelson and Morley’s experiment showed that an effect of the second order, (proportional to v²/c²), was not present either.

This contradiction was formally removed by the postulate of Lorentz and FitzGerald, according to which moving bodies experience a certain contraction in the direction of their motion. However, this ad hoc postulate seemed to be only an artificial means of saving the theory. Michelson and Morley’s experiment had actually shown that phenomena agree with the principle of relativity.

It turned out that a sufficiently sharpened conception of time was all that was needed to overcome the difficulty. One had only to realize that an auxiliary quantity introduced by Lorentz and named by him “local time” could be defined as “time” in general. If one adheres to this definition of time, the basic equations of Lorentz’s theory correspond to the principle of relativity, and the above [Galilean] transformation equations are replaced by ones that correspond to the new conception of time. [Then] Lorentz’s and FitzGerald’s hypothesis [of length contraction] appears as a compelling consequence of the theory.

4.2.2 Volume 3: The Swiss Years: Writings 1909-1911, Doc 2: The principle of relativity and its consequences in modern physics, 1910

Part 2: The optics of moving bodies and the ether
Assuming that ether is completely immobile, Lorentz conceived in 1895 a very satisfactory theory of electromagnetic phenomena, a theory which not only permitted a quantitative prediction of Fizeau’s experiment, but also provided a simple explanation of almost all the experiments that one can imagine in this sphere.

Part 3: Experiments and consequences not reconcilable with the theory
From Fizeau’s experiment, one had to conclude that the ether is not carried along completely by matter in motion, but that, instead, there occurs a relative displacement. The earth being a body that rotates around its axis, and revolves around the sun, with velocities that change their directions, one was bound to believe that, in our laboratories, the ether would take a slight part in the motion of the earth.

From this, it would seem to follow that the relative velocity of the ether, with respect to our equipment, must vary with time, and that one therefore should expect that an apparent spatial anisotropy be observed in optical phenomena, ie these phenomena should depend on the orientation of the equipment. Thus, in vacuum, or in the atmosphere, light should propagate faster in the direction of the earth’s motion than in the opposite direction. 

Experimental verification of this consequence of the theory was unthinkable, because the order of magnitude of the term considered is that of the ratio of the velocity of the earth to the velocity of light, and one could not hope to attain such precision in the direct determination of the velocity of light. Also, and this is a most important point, all terrestrial methods for measuring the velocity of light employ light rays that travel along a closed (back and forth) rather than a simple path, this due to the fact that the times of departure and arrival of the rays must be determined with the help of one and the same device.

The negative result of one of these experiments presented a real headache for the theoreticians, and as a result the foundation of Lorentz’s theory seemed shaky. Lorentz and FitzGerald resorted to a strange hypothesis: they assumed that each body in motion with respect to the ether contracts in the direction of motion by a fraction equal to ½(v/c)², or if only terms of second order are considered, that the length of the body is diminished in that direction in the ratio: 1 to √(1-v²/c²).

Part 4: The principle of relativity and the ether
What is the source of the difficulties we have just seen? Lorentz’s theory contradicts the purely mechanical models to which physicists hoped to reduce all phenomena. For while mechanics admits no absolute motion, but only the motions of bodies relative to each other, there is a particular state in Lorentz’s theory that corresponds physically to the state of absolute rest: that is the state of a body which is not in motion with respect to the ether.

If the equations of Newtonian mechanics of a coordinate system that is not undergoing accelerated motion are referred, by means of the relations t’ = t, x’ = x -vt, y’ = y, z’ = z, to a new coordinate system which is in uniform translational motion with respect to the first, one obtains equations in t’, x’, y’, z’ that are identical to the original equations in t, x, y, z. In other words, the Newtonian laws of motion transform to laws of the same form when one passes from one coordinate system to another one that is in uniform translational motion with respect to the first. 

This is the property we express when we say that the principle of relativity is satisfied in classical mechanics. More generally, we will state the principle of relativity in the following way: The laws governing natural phenomena are independent of the state of motion of the coordinate system with respect to which the phenomena are observed, provided that this system is not in accelerated motion.

Part 5: On the arbitrary hypotheses contained implicitly in the customary notions of time and space
The question arises: Is it really impossible to reconcile the essential foundations of Lorentz’s theory with the principle of relativity? If we wish to attempt such a reconciliation, the first step we must take is to give up the ether. For, on the one hand, we have been obliged to admit that the ether is stationary, whereas, on the other hand, the principle of relativity demands that the laws of natural phenomena referred to a uniformly moving coordinate system S’, be identical with the laws of these same phenomena referred to a system S, at rest with respect to the ether. But there is no reason to assume the immobility of ether any more with respect to the system S’, than with respect to the system S; these two systems cannot be distinguished from each other, and it is therefore improper to make one of them play a special role by saying that it is at rest with respect to the ether. It follows that the only way to arrive at a satisfactory theory is to give up the notion of a medium filling all of space.

To go a step further, we must reconcile the principle of relativity with an essential consequence of Lorentz’s theory, because giving up this consequence would amount to giving up the most fundamental formal properties of the theory. And here is the consequence in question: A ray of light in vacuum always propagates with the same velocity c, which velocity is independent of the motion of the body that emits the ray.

In Lorentz’s theory this principle holds only for a system in a special state of motion: In effect, the system must be at rest relative to the ether. If we want to preserve the principle of relativity, we must assume that the principle of the constancy of the velocity of light holds for any arbitrary system not in accelerated motion. At first glance this seems impossible. For let us consider a light ray that propagates with velocity c with respect to the system S, and suppose that we seek to determine the velocity of propagation relative to a system S’, that is in uniform translational motion with respect to the first system. Applying the rule of addition of velocities, we will find a velocity different from c; in other words, the principle of the constancy of the velocity of light that is valid with respect to S is not valid with respect to S’. 

So that the theory based on these two principles should not lead to contradictory results, one must renounce the customary rule of addition of velocities. Well founded as this rule may seem to be at first glance, it conceals no less than two arbitrary hypotheses, which hold sway over all of kinematics. 

The first hypothesis we wish to discuss concerns the physical notion of time measurement. The meaning of this definition is perfectly clear as long as the clock is sufficiently close to the place at which the event occurs, so that the clock and the event can be observed simultaneously. If, on the contrary, the event is taking place in some corner far away from the clock, then it will no longer be possible to establish immediately a correspondence between the different phases of the event, and the different positions of the clock’s hands.

To determine the time at each point in space, we can imagine it populated with a very great number of clocks of identical construction. Suppose, for example, that we are studying the motion of a material point whose trajectory passes through the points A, B C. Let us consider the points A, B, C, each of which is furnished with a clock, and is referred to a system in non-accelerated motion with the aid of time-independent coordinates. We will now be able to know the time at any of these locations. 

If we choose a sufficiently large number of clocks, so that we can ascribe to each of them a sufficiently small domain, we will be able to fix any instant whatsoever, at any location in space, to any degree of accuracy desired. But we cannot obtain in this manner a definition of time useful to a physicist, because we did not say what the position of the clock hands should be at a given instant of time at different spatial points. We forgot to synchronize our clocks.

To get a complete physical definition of time, we have to say in what manner all of the clocks have been set at the start. We proceed as follows: First, we furnish ourselves with a means of sending signals, be it from A to B, or from B to A. This means should be such that we have no reason whatsoever to believe that the phenomena of signal transmission in the direction AB will differ in any way whatsoever from the signal transmission in the direction BA.

In that case there is, obviously, only one way of regulating the clock at B against the clock at A, in such a manner that the signal travelling from A to B would take the same amount of time, measured with the clocks described above, as the signal travelling from B to A. That is: The reading of the clock at B at the moment the signal AB arrives at B (ie t(b)) minus the reading of the clock at A at the moment the signal AB leaves A (ie t(a)) equals the reading of the clock at A at the moment the signal AB arrives at A (ie t‘(a)) minus the reading of the clock at B at the moment the signal BA leaves B (ie t‘(b)), ie t(b) – t(a) = t’(a) – t’(b)

For these signals we can use, for example, sound waves that propagate between A and B, through a medium that is at rest with respect to these points. We can use light rays propagating through the vacuum, or through a homogeneous medium at rest with respect to A and B. It does not make any difference whether we choose this or that kind of signals. Still, of all the signals that can be used, we are going to prefer those that make use of light rays propagating in the vacuum, because the synchronization requires that the path out, and the path back, be equivalent, and in our case this equivalence is satisfied, by definition, by virtue of the principle of the constancy of the velocity of light, as light in the vacuum always propagates with the velocity c.

We have to synchronize our clocks in such a way that the time spent by a signal travelling from A to B, be equal to the time spent by an identical signal travelling from B to A. Now we possess a well-defined method by which to synchronize two clocks, with respect to each other. Once the synchronization has been done, we will say that the two clocks are in phase. The totality of the readings of all of these clocks in phase with one another, is what we will call the physical time.

By an elementary event, we will understand an event that is concentrated in one point, and is of infinitely short duration. By the time coordinate of an elementary event, we understand the indication, at the instant of the event’s occurrence, of a clock that is situated infinitely close to the point where the event takes place. An elementary event is thus defined by four coordinates, the time coordinate, and the three coordinates that define the spatial position. Thanks to this, we can give a perfectly defined meaning to the concepts of simultaneity, and non-simultaneity, of two events occurring at locations removed from one another.

In order to define the physical time with respect to a coordinate system, we used a group of clocks in a state of rest relative to that system. According to this definition, the time readings, or the establishment of the simultaneity of two events, have meaning only if the motion of the group of clocks, or that of the coordinate system, is known.

Consider two non-accelerated coordinate systems, S and S’, in uniform translational motion with respect to one another. Suppose that each of these systems is provided with a group of clocks attached to it, and that all clocks belonging to the same system are in phase. Under these conditions, the readings of the group attached to S will define the physical time with respect to S; the readings of the group attached to S’ define the physical time with respect to S’. Each elementary event will have a time coordinate t with respect to S, and a time coordinate t’ with respect to S’. 

But we have no right to assume, a priori, that the clocks of the two groups can be set in a manner that the two time coordinates of the elementary event would be the same, or in other words, in a way that t would be equal to t’. To assume this would mean to introduce an arbitrary hypothesis which has been introduced into kinematics up to the present time.

The second arbitrary hypothesis introduced in kinematics concerns the configuration of a body in motion. Consider a bar AB moving in the direction of its axis with velocity v, with respect to a coordinate system S not in accelerated motion. What should we understand by the “length of the bar”? One is at first inclined to believe that this concept does not require any special definition. However, we will immediately see that nothing of the sort is true, if we consider the following two methods of determining the length of the bar:

  1. One accelerates the motion of an observer furnished with a measuring rod until he attains the velocity v, ie until he is at relative rest with respect to the bar. The observer then measures the length AB by successively applying the measuring rod along the bar. 
  2. Using a group of clocks in phase with each other, and at rest with respect to the system S, one determines the two points P₁ and P₂ of S where one finds the two ends of the bar, at the instant t; after that, one determines the length of the straight line connecting the two points P₁ and P₂ by successively applying the measuring rod along the line P₁P₂, which is assumed to be a material line. 

As one can see, it is with some justification that the results obtained in the first, and in the second, case are designated as the “length of the bar.” But in no way does this mean, a priori, that these two operations must necessarily lead to the same numerical value for the length of the bar. All that one can deduct from the principle of relativity, and this is easy to demonstrate, is that the two methods lead to the same numerical value for the length, only when the bar AB is at rest relative to the system S. But in no way is it possible to assert that the second method yields a value for the length, independently of the velocity v of the bar.

More generally, if the configuration of a body in uniform translational motion, with respect to S, is determined by ordinary geometric methods, by means of measuring rods, or other solid bodies, moving in exactly the same way, the measurements turn out to be independent of the velocity v of the translation: these results give us what we will call the geometric configuration of the body. By contrast, if one marks in the system S the positions of various points of the body at a given instant, and determines the configuration formed by these points by geometric measurements using measuring rods at rest with respect to S, one obtains as a result what we will call the kinematic configuration of the body, with respect to S.  The second hypothesis used unconsciously in kinematics can thus be expressed as follows: The kinematic configuration and the geometric configuration are identical. 

Part 6: The new transformation equations (the Lorentz transformation) and their physical meaning
The rule of the parallelogram of velocities, which made one think that Lorentz’s theory cannot be reconciled with the theory of relativity, is based on unacceptable hypotheses which lead to the following transformation equations: t’ = t, x’ = x – vt, y’ = y, z’ = z. The first of these equations, expresses an ill-founded hypothesis about the time coordinates of an elementary event taken with respect to two systems, S and S’, that are in uniform translational motion with respect to each other. The other three equations, express the hypothesis that the kinematic configuration of the system S’, with respect to the system S, is identical with the geometric configuration of the system S’. 

If one abandons the ordinary kinematics and builds a new kinematics based on the new foundations, one arrives at transformation equations different from those given above. And now, we are going to show that based on:

1. The principle of relativity
2. The principle of the constancy of the velocity of light

we arrive at transformation equations that allow us to see that Lorentz’s theory is compatible with the principle of relativity. The theory based on these principles we shall call the theory of relativity.

[S and S’] each possess a group of clocks that run-in synchrony when the two systems are at relative rest with respect to each other. Thus [in that circumstance] the velocity of light in a vacuum must be expressed by the same number in the two systems. Let t, x, y, z be the coordinates of an elementary event with respect to S, and t’, x’, y’, z’ the coordinates of the same event with respect to S’. We seek to find the relations that link these two groups.

These relations must be linear, because of the homogeneity of time and space, and time t is therefore linked with time t’ by a formula of the form: t’ = At + Bx + Cy + Dz. We will count the time from the instant when the origins of the two systems coincide, ie when:

x’ = 0 and x – vt = 0,  y’ = 0 and y = 0,  z’ = 0 and z = 0, are equivalent

ie: x, y, z,  x’, y’, z’  are linked in the form: x’ = E(x – vt),  y’ =Fy,  z’ = Gz. 

[Footnote: it is easy to understand what we mean by the homogeneity of time and space, or why we assumed a priori that the transformation equations must be linear. For if a rate of a clock at rest with respect to S’, is observed from S, this rate does not have to depend on the location of the clock in S’, nor on the value of the time of S’ in the vicinity of the clock. An analogous remark applies to the orientation and length of a bar linked with S’, and observed from S. Only when the transformation equations are linear are these conditions satisfied].

To determine the constants A, B, C, D, E, F, G, we assert that, according to the principle of the constancy of the velocity of light, it has the same value, c, with respect to the two systems, ie the two equations: x² + y² + z² = c²t² &  x’² + y’² + z’² = c²t’² are equivalent. Replacing in the second of these equations t’, x’, y’, z’ by their values obtained from above [the two equations involving A – G], and equating it with the first of these equations, the transformation equations sought are of the form:

t’ = ϕ(v)β[t – (v/c²)x], x’ = ϕ(v)β(x – vt), y’ = ϕ(v)y, z’ = ϕ(v)z

where β = 1/√(1-v²/c²), and ϕ(v) is a function of v that is to be determined.

We can find ϕ(v) by introducing a third coordinate system S”, which is equivalent to the first two systems, is moving relatively to S’ with a uniform velocity -v, and is oriented with respect to S’, as S’ is oriented with respect to S. Applying the above set of equations twice, we obtain:

t” = ϕ(v)ϕ(-v)t, x” = ϕ(v)ϕ(-v)x, y” = ϕ(v)ϕ(-v)y, z” = ϕ(v)ϕ(-v)z

Since the origins of S and S’ are permanently coincident, the axes have the same orientation, and the systems are equivalent, ϕ(v)ϕ(-v) = 1. Moreover, as the relation between y and y’, (as also that between z and z’), does not depend on the sign of v, we have: ϕ(v) = ϕ(-v). It follows that ϕ(v) = 1, as ϕ(v) = -1 is inappropriate. The transformation equations are:

t’ = β[t – (v/c²)x], x’ = β(x – vt), y’ = y, z’ = z, or

t’ = [(t-(v/c²)x]/√(1 – v²/c²)], x’ = (x -vt)/√(1 – v²/c²), y’ = y, z’ = z,

where β = 1/√(1-v²/c²)

We call them the Lorentz transformations. 

Part 7: Physical interpretations of the transformation equations

1 Consider a body attached to S’. Let x’₁, y’₁, z’₁, and x’₂, y’₂, z’₂, be co-ordinates of two points of the body.  At any instant t of the system S, we have the following relations between these two co-ordinates: x₂- x₁= [(√(1 – v²/c²)(x’₂ – x’₁)],  y₂ – y₁ = y’₂ – y’₁,  z₂ – z₁ = z’₂ – z’₁.

The kinematic configuration of a body in uniform translational motion with respect to a co-ordinate system depends on the velocity v of the translation. 

Furthermore, the kinematic configuration differs from the geometric configuration solely by a contraction in the direction of the motion, a contraction which is in the ratio 1:√(1-v²/c²). A relative motion of two reference systems with a velocity v that exceeds the velocity of light in vacuum is incompatible with the principle here assumed. One recognizes in this the hypothesis of Lorentz and FitzGerald. This is the hypothesis that looked so strange, and had to be introduced to explain the negative results of the experiment of Michelson. Here this hypothesis appears naturally as an immediate consequence of the principles assumed.

2. Let us consider a clock H’ which is at rest at the origin of S’, and runs P₀ times faster than one of the clocks used for the determination of physical time in the systems S or S’, ie when the two clocks are compared while at relative rest, clock H’ will indicate P₀ periods during the unit time indicated by the other clock. How many periods will clock H’ indicate during unit time if observed from the system S? 

Clock H’ will indicate the end of a period at the times in the form: tn’ = n/P₀.  Since we seek the time with respect to S, the first of the transformation equations will have to be written: t = β[t’ – vx’/c²]. And since clock H’ is at rest at the origin of S’, we must always have x’ = 0, which yields: tn = βtn’ = (β/P0)n. Observed from S, clock H’ thus indicates: P = P0/β = P0√(1-v²/c²) periods in a unit time. In other words, a clock moving uniformly with velocity v with respect to a reference system runs, as observed from this system, 1:√(1-v²/c²) times slower than an identical clock that is at rest with respect to this system.

3. Let us consider the equations of motion of a point moving in uniform translation with velocity u’ with respect to S’: x’ = uₓ’t’, y’ = uᵧ’t’, z’ = uz’t’.  If one replaces x’, y’, z’, t’ by their values as functions of x, y, z, t, one obtains x, y, z as functions of t and, hence, the components ux, uy, uz of the velocity u of the point with respect to the system S. In this way it is possible to obtain the formula that expresses the theorem of the addition of velocities in its general form, and one can immediately see that the law of the parallelogram of velocities is valid only in first approximation. In the special case when the velocity u’ has the same direction as the velocity v of the translation of S’ with respect to S, one easily obtains: u = (v + u’)/(1 + vu’/c²). This equation shows that if one adds two velocities, each smaller than the velocity of light in a vacuum, one always obtains a resultant velocity that is smaller than the velocity of light. For if one sets v = c – λ, u’ = c – μ, where λ and μ are positive and smaller than c, one gets: u = c[(2c – λ – μ)/(2c – λ – μ + λμ/c)] <c. When one adds the velocity of light c, and a velocity smaller than c, one always obtains the velocity of light.

Part 8: Remarks about Some Formal Properties of the Transformation Equations
Let us consider two coordinate systems S and S’ the origins of which coincide and which have the same orientation. There are two kinds of co-ordinate transformations in Newtonian mechanics that do not alter the laws of motion. These are: 1 A change in orientation of the system S’, with respect to the system S, about the common origin. This first transformation is characterized by equations linear in x’, y’, z’ and x, y, z, between the coefficients of which there exist relations, such that the condition: x’² + y’² + z’² = x² + y² + z² is identically satisfied. 

2 Uniform motion (translation) of the system S’, with respect to the system S. This second transformation is characterised by the equations: x’ = x + αt,  y’ = y + βt,  z’ = z + γt, where α, β, γ, are constants. For these two kinds of transformation, the condition t’ = t must be satisfied. In other words, time is an invariant under these two transformations. 

Combining these two transformations we obtain the most general transformation, by means of which one can transform the equations of mechanics without altering them. This transformation is characterised by the equation t’ = t, and by three equations that express x’, y’, z’ as linear functions of x, y, z. The coefficients of these three equations are connected by relations that, for t = 0, satisfy condition: x’² + y’² + z’² = x² + y² + z²  identically.

Let us now consider the most general co-ordinate transformation compatible with the theory of relativity. From what we have seen, this transformation is characterized by the fact that   x’, y’, z’, t’ must be linear functions of x, y, z, t, such that the condition: x’² + y’² + z’² – c²t’² =  x² + y² + z² – c²t²  will be satisfied identically. It should be noted that the transformations compatible with Newtonian mechanics can be obtained by setting c = ∞, in condition x’² + y’² + z’² – c²t’² = x² + y² + z² – c²t². Thus, if we take the same route as before, we arrive at the equations of ordinary kinematics if, instead of the principle of the constancy of the velocity of light, we assume the existence of signals whose propagation does not require any time. The group characterized by equation x’² + y’² + z’² – c²t’² = x² + y² + z² – c²t² contains the transformations that correspond to a change in the orientation of the system. These are the transformations compatible with the condition t = t’. 

The simplest transformations compatible with x’² + y’² + z’² – c²t’² = x² + y² + z² – c²t² are those for which two of the four coordinates of an elementary event remain invariant. Let us consider, for example, the transformations under which x and t do not change. Instead of the general condition x’² + y’² + z’² – c²t’² = x² + y² + z² – c²t², we will have the special condition: t’ = t, x’ = x, y’² + z’² = y² + z². To this condition corresponds a rotation of the system about the x-axis. If, on the other hand, we consider transformations under which two of the spatial coordinates, for example y and z, remain invariant, we will have instead of the general condition x’² + y’² + z’² – c²t’² = x² + y² + z² – c²t² the special condition: y’ = y, z’ = z, x’² – c²t’² = x² – c²t².

These are what we encountered in the preceding section while investigating a system in uniform motion parallel to the x-axis of an identically oriented system at rest. The two systems of equations differ only by a change of sign in the third condition. But even this difference can be made to disappear if one chooses, with Minkowski, to take ict instead of t as a variable, where i is the imaginary unit. In that case this imaginary temporal coordinate plays the same role in the transformation equations as the spatial coordinates. If we set x = x₁, y = x₂, z = x₃, ict = x₄, and consider x₁ x₂ x₃ x₄ as the coordinates of a point in a four-dimensional space such that to each elementary event there corresponds a point in this space, we reduce everything that happens in the physical world to something static in the four-dimensional space. In that case the condition x’² + y’² + z’² – c²t’² = x² + y² + z² – c²t² will be written as:  x’₁² + y’₂² + x’₃² + x’₄² = x₁² + x₂² + x₃² + x₄². 

This is the condition that corresponds to a rotation without relative translation of a four-dimensional coordinate system. The principle of relativity demands that the laws of physics not be altered by the rotation of the four-dimensional coordinate system to which they are referred. The four coordinates x₁ x₂ x₃ x₄ must appear in the laws symmetrically. To express the different physical states, one can use four-dimensional vectors which behave in the calculations in a manner analogous to ordinary vectors in three-dimensional space.

4.2.3 Volume 3: The Swiss Years: Writings 1909-1911, Doc 17: The Theory of Relativity, 1911
How to put kinematics back on its feet? The answer is the very same circumstances that led us into so many difficulties in the past, lead us to a negotiable path by putting aside the arbitrary assumptions mentioned above. For it turns out that precisely those two seemingly incompatible axioms, which were imposed on us by experience, namely the principle of relativity and the principle of constancy of the velocity of light, lead us to a perfectly definite solution of the space-time transformation problem.

One arrives at results that, in part, run very much counter to customary conceptions. First, things purely kinematic. Since we defined the coordinates, and the time, in a definite way in physical terms, all relationships between these will have a perfectly definite physical content. We obtain the following: If we have a solid body that is moving uniformly with respect to the coordinate system k, which we take as the basis for our analysis, then this body appears contracted by a perfectly definite ratio in the direction of its motion, as compared with the shape it has when it is in a state of rest, with respect to this system. If we denote the velocity of motion of the body by v, and the velocity of light by c, then each length measured in the direction of motion, and equal to l in a motionless state, will be diminished because of the body’s motion relative to the non-comoving observer to the length l.√(1 –  v2/c2).

If the body has a spherical shape in the state of rest, it will have the shape of a flattened ellipsoid if we move it in a certain direction. When its velocity reaches the velocity of light, it will collapse to a plane. However, judged by a comoving observer, the body retains, before and after, its spherical shape; on the other hand, to the observer moving with the body, all non-comoving objects appear, in exactly the same way, contracted in the direction of the relative motion. This result loses very much of its oddness, if one considers that this assertion about the shape of a moving body has a complicated meaning, since this shape can be ascertained only with the aid of determinations of time. The feeling that this concept, “the shape of the moving body,” has an immediately obvious meaning is due to the fact that in our day-to-day experience we are accustomed to encountering only such velocities of motion that are practically infinitely small compared with the velocity of light.

And now a second purely kinematic consequence of the theory that strikes us as even more peculiar. We imagine that there is given a clock capable of indicating the time of a reference system k, provided that it is arranged at rest relative to this system. It can be proved that this same clock, when set into uniform motion relative to the reference system k, runs slower, as judged from the system k, in such a way that when the time reading of the clock has increased by 1, the clocks of the system k indicate that, with respect to the system k, there has elapsed the time 1/√(1-v²/c²). Were we to succeed in making the clock move with the velocity of light, the hands of the clock, as judged from k, would move forward infinitely slowly.

The thing is at its funniest when one imagines that the following is being done: One imparts to this clock a very great velocity (almost equal to c), then lets it fly on in uniform motion, and after the clock has covered a long stretch, one imparts to it a momentum in the opposite direction, so that it returns to the point from which it has been launched. It then turns out that the positions of the clock’s hands have hardly changed during the clock’s entire trip, while an identically constituted clock, that remained at rest at the launching point during the entire time, changed the setting of its hands quite substantially. It should be added that whatever holds for this clock, which we introduced as a simple representation of all physical phenomena, holds also for closed physical systems of any other constitution.

4.2.4 Volume 4: The Swiss Years: Writings 1912-1914, Doc 1: Manuscript of the Special theory of Relativity, 1912

Part 5: Principle of the Constancy of the Velocity of Light
By processing a number of equations, some of which include the propagation of light in empty space, Einstein asserts that: Hence in accordance with Lorentz’s theory we can proclaim the following, which we call “the principle of the constancy of the velocity of light”: “There exists a coordinate system with respect to which every light ray propagates in vacuum with the velocity c.” This principle contains a far-reaching assertion. It asserts that the propagation velocity of light depends neither on the state of motion of the light source nor on the states of motion of the bodies surrounding the propagation space. The question as to what extent this principle can be considered certain is of fundamental significance for the theory of relativity. For the time being we will content ourselves with the realization that this principle is demanded by Lorentz’s theory.

Part 7: Apparent Incompatibility of the Principle of the Constancy of the Speed of Light with the Relativity Principle
Lorentz’s theory seemed incompatible with the relativity principle.  For we have seen that, according to Lorentz’s theory, there exists a reference system K in which every light ray propagates in vacuum with the velocity c.  It would appear that a system in uniform translational motion with respect to K could not have this property.  For if again we choose K’ such that its origin glides along the x axis of K with velocity v, and if we choose a ray of light that propagates along the x axis of K with the velocity c with respect to K, then according to the law of the parallelogram of velocities this same light ray propagates along the x’ axis with velocity c-v with respect to K’.  It would follow that a law of nature, namely the principle of the constancy of the velocity of light, is indeed valid with respect to K but not with respect to K’. Thus the principle of the constancy of light (in vacuum) seems to be incompatible with the relativity principle. 

Strictly speaking, however, we have only learned that the following three things are incompatible with one another:
(a) the relativity principle
(b) the principle of the constancy of the velocity of light (Lorentz’s theory)
(c) the transformation equations or the law of the parallelogram of velocities

One arrives at the theory that is now called “the theory of relativity” by keeping
(a) and (b) but rejecting (c). But whoever has not yet studied the electrodynamics
of moving bodies thoroughly will be inclined to keep (a) and (c) but give up (b).  

Therefore, we will first draw several consequences starting out from the latter standpoint in order to justify in that way the theory of relativity.”

[There then follows commentary on Fizeau’s experiment with the conclusion that:] On the other hand, since the Maxwell-Lorentz theory requires the principle of the constancy of the (vacuum) velocity of light, there is every reason to stick with it. It turns out that the result of the Fizeau experiment can be derived quantitatively from (a) and (b). [Because] If we take this standpoint [keeping (a) and (c)], then Fizeau’s experiment forces us to conclude that in a transparent medium, various velocities of light must be possible. For if the velocity of light in the medium, as assessed by an observer moving with the medium, were always equal to Vo, then, according to the parallelogram law of velocities, the velocity of light as assessed by an observer not moving with the medium would be Vo + nl, We are thus forced to assume that, in Fizeau’s experiment, the velocity of light relative to the moving medium is, in this medium, different from what it would be in the same medium if the latter were at rest.  

Part 8: The Physical Meaning of Spatial and Temporal Determinations
We now seek to solve the dilemma that consists in the incompatibility of postulates (a), (b), and (c) by relinquishing (c) but keeping (a) and (b).  We realize that the postulate (c) can be given up, by investigating the meaning that the spatial and temporal determinations have in physics.

Physical meaning of spatial determinations:
The propositions of Euclidean geometry acquire a physical content through our assumption that there exist objects that possess the properties of the basic structures of Euclidean geometry. We assume that appropriately constructed edges of solid bodies not subjected to external influences have the definitional properties that straight lines have (material straight line), and that the part of a material straight line between two marked material points has the properties of a segment. Then the propositions of geometry turn into propositions concerning the arrangements of material straight lines and segments that are possible when these structures are at relative rest. By virtue of its properties, every material straight line can be extended. By repeatedly laying a shorter segment on a longer one and counting these operations, one can measure the latter by means of the former; ie the length of the latter can be expressed by a number if the former (the measuring rod) is regarded as being given once and for all. The position of a material point on a material straight line on which a fundamental point is given once and for all can be defined by its distance from the latter, that is, by a number (coordinate). 

From among all the material straight lines at rest relative to one another, we think of three that are perpendicular to one another and intersect in one point as being distinguished, and from among all the material straight line segments we likewise imagine a (transportable) segment, which we will think of as being used for measuring all lengths (measuring rod). We call these structures, taken together, “the coordinate system.” Geometry teaches how we can express the position of every material point with respect to such a coordinate system K by means of three numbers (coordinates). The shape and position of a material structure at rest relative to K are given by the totality of the positions (coordinates) of all of its material points with respect to K.

The physical meaning of temporal determinations:
If we consider what happens physically with respect to the coordinate system K, then we know that spatial coordinates x, y, z do not suffice for the determination of the physical variables. The specification of a fourth basic variable, the temporal coordinate, is also needed. This coordinate as well we will define in such a way that each temporal determination shall appear as the result of a precisely defined measurement procedure. We imagine a completely isolated physical system that repeatedly assumes a specific state Z. Then the state Z is always followed by the states Z’, Z”, etc., until the state Z is reached again. The system changes its state periodically. We can then count how often the system, which we shall call a “clock,” assumes the state Z; we will call this number the “temporal determination” of the clock.

We imagine that such a clock is permanently arranged at the point of origin O of K, and that the spatial dimensions of the clock are so small that, from a geometrical point of view, it may be treated as a “point.” By means of the determination of this clock, every event that is spatially infinitely close to the coordinate origin can be assigned a temporal determination, the “time coordinate,” or, in brief, the “time” of the event, if we have arbitrarily fixed the zeroth temporal determination of the clock. We cannot evaluate directly the time of an event taking place at another point A(x, y, z) of K by means of the clock set up at the origin of K; for we possess no means, at first, for deciding which temporal determination of the clock set up at the origin of K is “simultaneous” with the event, ie is to be assigned to it. 

Therefore, in order to evaluate the events occurring at A, we imagine a clock of exactly the same constitution as that at O set up at A, with the zeroth temporal determination of this clock chosen arbitrarily. By this means we arrive at a temporal valuation of all events occurring at A. But with a procedure of this kind there is, as yet, no connection between the temporal determinations concerning the places O and A; for these temporal determinations to fulfill the demands we impose on temporal determinations in physics, we still have to indicate a procedure by which the clock at A can be “regulated according to the one at O.” To that end we imagine some physical arrangement that makes it possible to send “signals” both from O to A and from A to O in such a way that the signal O – A and the signal A – O are to be conceived, for reasons of symmetry, as completely identically constituted processes.

Suppose a signal that was sent from O at the determination to of the clock at O (O-time) arrives at A at the A-time tₐ. Let a signal sent from A at the A-time tₐ‘ arrive at O at the O-time to‘. We stipulate that the clock at A should be regulated in such a manner that we shall always have tₐ – t₀ = t‘₀– tA‘ . If this is possible and is accomplished, then we say that the clock at A is synchronised with the clock at O. Using the same procedure, clocks arranged at rest at other points B, C, etc., and of identical constitution as the above-mentioned clocks, can also be synchronized with the clock at O. If, upon the completion of this procedure, two other arbitrary clocks, eg those at points A and B, are synchronised in accordance with the same definition, then it is finally no longer of consequence that we privileged the clock at O by regulating all of the other clocks according to it. Each clock is synchronized with each of the others.

We define the time coordinate of an event taking place at an arbitrary point of K (point event) as the simultaneous reading of the clock set up at this point and regulated according to the given procedure. Two point events (occurring at different points) are simultaneous if their time coordinates are equal.

It is of importance that the determination of a time coordinate, and thereby also the determination of simultaneity, has meaning only if the coordinate system K to which that determination refers is indicated. Because for the measurement of a time coordinate, or the verification of simultaneity, we need a system of identically constructed clocks that are set up at rest relative to K. We will designate the coordinate system K, together with the measuring rod and the regulated system of clocks at rest relative to K, as the “reference system Σ.” 

We have now completely defined the physical meaning of spatial and temporal determinations with respect to the reference system Σ.  Let us now introduce a second reference system, Σ‘, which is in uniform translational motion with respect to the first one.  Spatial and temporal determinations with regard to this second reference system Σ‘ can be interpreted in exactly the same way as with regard to the original system Σ, and we will stipulate that the measuring rod and the clocks of Σ‘ are constituted in the same way as the measuring rod and clocks of Σ ,when these objects are compared in a state of relative rest.

The space-time coordinates of a specific event with respect to Σ(x, y, z, t), and those of the same event with respect to Σ‘(x’, y’, z’, t’) are totally independent of one another by the terms of their definition, even if the state of motion of Σ‘ is given relative to Σ. Hence it does not follow from these definitions that two point events that are simultaneous with respect to Σ must also be simultaneous with respect to Σ‘.

Let us consider, further, a body in arbitrary motion with respect to Σ.  Obviously, its shape and position with respect to Σ is determined by the totality of the spatial coordinates of the material points of the body at a specific time t (of Σ). A corresponding remark can be made about the shape and position of the same body with respect to Σ‘ at a specific time t’ (of Σ‘).  The fact that the definitions of the space-time coordinates with respect to Σ and Σ‘ are independent of each other entails that no relationships between the shape of the body with respect to Σ and that with respect to Σ‘ can be given on the basis of these definitions.

If one takes into account these two consequences, one finds oneself unable to derive the [standard] transformation equations.  If we replace these equations with transformation equations that are consistent not only with the relativity principle but also with the principle of the constancy of the velocity of light, we arrive at the theory presently designated as the “theory of relativity.” Only in this way can we reconcile Lorentz’s eminently fruitful extension of Maxwell’s theory with the relativity principle.

Part 9: Derivation of the Lorentz Transformation
The principle of the constancy of the velocity of light demands the existence of a reference system Σ relative to which every light ray propagates in vacuum with velocity c. According to the relativity principle, all reference systems Σ‘ in uniform translational motion relative to Σ must possess the same property. 

What kind of transformation equations must obtain between the space-time coordinates x, y, z, t (with respect to Σ) and the space-time coordinates x’, y’, z’, t’ (with respect to Σ‘) of the same point event so that the principle of the constancy of the velocity of light would hold with respect to both systems?

The functions that are being sought must be completely linear functions of these variables. We demand this in order to preserve the homogeneity properties of physical space. If one did not make this assumption, then bodies that are at rest, congruent, and identically located with respect to Σ‘ would be differently shaped or located when referred to Σ; or clocks that are at rest and identically constructed with respect to Σ‘ would have different or time-dependent rates when referred to Σ. 

We can further demand that the transformation equations be homogeneous, because all that is needed for this is that the path described by the origin of Σ‘ with respect to Σ  pass through the origin of Σ, and that the origin of the time scales in Σ and Σ‘ be chosen in such a way that the clocks located at the origins of the systems Σ and Σ‘ both read zero at the moment when the two points coincide.

According to the principle of the constancy of the velocity of light, then the spatial points that are just reached by the signal at times t and t’ with respect to Σ  and Σ‘, respectively, will be determined by the equations: √(x² + y² + z²) = ct  &  √(x’² + y’² + z’²) = ct’ This means that the equations: √(x² + y² + z²) -ct = 0  &  √(x’² + y’² + z’²) – ct’ = 0 must be equivalent.

The transformations must therefore make the equation:  λ²(x² + y² + z² -c²t²) = (x’² + y’² + z’² – c²t’²) into an identity, where all we know about the factor λ²  for the time being is that it must not vanish. But one can see that λ² must be independent of x, y, z, t, for otherwise the right hand side divided by λ² could not be an homogeneous, complete function of second order in x, y, z, t, after the substitution is carried out.  We then have: x² + y² + z² -c²t² = x’² + y’² + z’² – c²t’²  

If one introduces the variable u = ict’ or u’ = ict’ in place of the time variables, where i denotes the imaginary unit, one obtains: x² + y² + z² + u² = x’² + y’² + z’² + u’². As is well known this choice of time variables derives from Minkowski. By means of it, the equation: x² + y² + z² -c²t² = x’² + y’² + z’² – c²t‘², is brought into a form into which the spatial co-ordinates and the temporal co-ordinate enter the same manner.

The transformations we seek are exactly the same as those we have to apply to the spatial coordinates when passing from an orthogonal coordinate system to another one with the same origin, the only difference being that here one deals with a four-dimensional manifold rather than with a three-dimensional manifold.

The simplest Lorentz transformations are those in which, in addition to the time coordinate, only one of the spatial coordinates (e.g., the x-coordinate) undergoes a transformation, then the transformation equations assume the form: x’ = (x -vt)/√(1 – v²/c²), y’ = y, z’ = z, u’ = (t-(v/c²)x)/√(1 – v²/c²)

 If one performs several Lorentz transformations in succession, one obtains again a Lorentz transformation; i.e the set of the Lorentz transformations forms a group.  We can now also formulate the relativity principle in the following way: The theory of relativity requires that systems of equations in physics turn into systems of equations of the same form if one transforms them by means of the Lorentz transformation. Simple calculation shows that the fundamental equations of classical mechanics do not have this property. Thus, they are not compatible with the theory of relativity. 

Part 10: The Physical Content of the Lorentz Transformation
For x’ and t’ to be real in the Lorentz transformation equations then v must be <c. The translational velocity of Σ‘ relative to Σ must be smaller than c. Thus, according to the theory of relativity, it is impossible in principle for a body (coordinate system) to move with superluminal velocity. If the coefficients in the Lorentz transformation were also dependent on x, etc, and t, then it would be not only differences of the spatial and temporal coordinates that entered into propositions about moving bodies and clocks, but their values as well. This is out of the question from a physical standpoint.

Further, in λ²(x² + y² + z² -c²t²) = (x’² + y’² + z’² – c²t‘²) we arbitrarily set λ² = 1. Had we not done this, we would have obtained: x’=λ(v)[(x -vt)/√(1 – (v/c²)x], y’=λ(v)y, z’=λ(v)z, t’=λ(v)[(t – v²/c²)/ )/√(1 – v²/c²)] instead of: x’ = (x -vt)/√(1 – v²/c²), y’ = y, z’ = z, u’ = (t-vx/c²)/√(1 – v²/c²)

For, since λ is independent of x and t while the special Lorentz transformation is defined by v alone, λ can only depend on v alone. If we introduce a third system E” that possesses the velocity (-v) in the X direction relative to E’, we obtain transformation equations between E and E” which means the coordinate axes of E and E” are permanently coincident. Since the coordinates are measured with identical measuring rods in both systems, we must have x” = x, etc. Hence λ(v)λ(-v) = 1. Furthermore, transformation shows that 1/λ(v) is the length, with respect to E, of a rod of length 1, that moves with E’ and is oriented parallel to the y’ axis. From a physical standpoint the length of a rod moving perpendicularly to its extension must be independent of the orientation of the motion, which means that we must have λ(v) = λ(-v).  From these two equations, it follows that λ = ±1. The special case v = 0 shows that the choice of the positive sign is the only one that is possible.

4.2.5 Volume 4: The Swiss Years: Writings 1912-1914: Document 8: Relativity and Gravitation, Reply to comment by M Abraham, 1912
The theory presently designated as “the theory of relativity” rests on two principles that are totally independent of one another, namely

1 The principle of relativity (with respect to uniform translation)

2 The principle of the constancy of the velocity of light

From these two principles it is possible to develop the theory currently known as “the theory of relativity”.  

It is common knowledge that it is impossible to base a theory of the transformation laws of space and time on the principle of relativity alone. As we know, this is connected with the relativity of the concepts of “simultaneity” and “shape of moving bodies.” To fill this gap, I introduced the principle of the constancy of the velocity of light, which I borrowed from Lorentz’s theory of the stationary luminiferous ether, and which, like the principle of relativity, contains a physical assumption that seemed to be justified only by the relevant experiments (experiments by Fizeau, etc).

But what about the limits of the validity of the two principles?  As I have already emphasised, we have not the slightest reason to doubt the general validity of the principle of relativity.  On the other hand, I am of the view that the principle of the constancy of the velocity of light can be maintained only insofar as one restricts oneself to spatio-temporal regions of gravitational potential.  This is where, in my opinion, the limit of the validity of the principle of the constancy of the velocity of light, though not the principle of relativity, and therewith the limit of the validity of our current theory lies.

In my opinion, the present theory of relativity will always retain its significance as the simplest theory for the important limiting case of spatio-temporal events in the presence of a constant gravitational potential.

4.2.6 Volume 4: The Swiss Years: Writings 1912-1914, Doc 21: Theory of Relativity, 1913.
The phenomena of interference and refraction of light compelled physicists to view light as a wave-like process. Until the end of the last century, it was thought that light consisted in mechanical oscillations of a hypothetical medium, the ether. When the electromagnetic theory of light gained the upper hand, this conception changed, only insubstantially, in that light was no longer conceived as a motion of the ether, but rather as an electromagnetic process in the ether. Still, the conviction persisted that, in addition to ponderable matter, there must exist a second matter, the ether, considered the carrier of light. This conception led to the question of how this ether behaves, mechanically, with respect to matter.

Does ether take part in the motions of ponderable matter? Fizeau demonstrated that the hypothesis, according to which the luminiferous ether simply takes part in the motions of matter, is untenable. The next simplest hypothesis is that the luminiferous ether does not participate at all in the motions of matter. According to Lorentz’s theory, the electromagnetic laws of the ether are independent of the state of motion of matter. Matter interacts with the ether, only in the sense that matter is to be conceived as the carrier of electric masses, the motions of which are produced and influenced by electromagnetic processes in the ether.

Physical phenomena depend only on the motions of bodies relative to each other, ie from the physical standpoint, absolute motion does not exist. More precisely, wherever spatial determinations play a role in physics, they always represent determinations concerning the relative position of some object, or marker, relative to a solid body. In a theory, the coordinate system is the representative of this solid body. 

Let us take some simple empirical law in which spatial determinations appear, eg Galileo’s familiar law of inertia: a material point not acted upon by external forces moves uniformly in a straight line. It is clear that this law cannot hold true, if the motion is referred to an arbitrarily moving coordinate system (eg one undergoing arbitrary rotation). We must therefore formulate Galileo’s fundamental law in the following way: It is possible to choose a coordinate system K, that is in such a state of motion that every freely moving material point moves rectilinearly and uniformly relative to it. Naturally, the law also holds then for all other coordinate systems at rest with respect to K.

If Galileo’s fundamental law were not valid for any coordinate system that is in motion relative to K, then the state of motion of K would be privileged, with respect to all other states of motion. We could then designate this state of motion as that of absolute rest. However, a simple argument shows that every freely moving material point satisfies Galileo’s fundamental law, not only with respect to K, but also with respect to every coordinate system K’, in uniform translational motion relative to K. The laws of mechanics are just as valid relative to such systems K’, as they are relative to K. There exists a whole class of coordinate systems moving uniformly, relative to one another, that are strictly equivalent when it comes to formulating the laws of mechanics. But this equivalence of the systems K and K’, that are moving uniformly relative to each other, is not limited to mechanics. So far as our experience extends, this equivalence holds generally. The assumption of the equivalence of all such systems K, K’, which rules out the privileging of one state of motion over all others, we will designate as the “relativity principle”.

According to Lorentz’s theory, the motion of matter does not result in any motion of the luminiferous ether. Instead, the parts of the latter are at rest relative to each other. If we choose a coordinate system K, that is at rest relative to the ether, then this coordinate system is privileged, with respect to all coordinate systems K’, that are in motion relative to K. Thus, the theory does not agree with the relativity principle. We can also carry out this argument without making use of the concept of the luminiferous ether. According to Lorentz’s theory, there exists a coordinate system K, relative to which every light ray propagates in a vacuum with the definite constant velocity c. If we refer such a light ray to a coordinate system K’, that is in motion relative to K, moving, say, in the direction of the propagation of the light, then we feel intuitively compelled to assume that this same light ray has a different propagation velocity relative to K’. 

Thus, one would have to conclude that, in contradiction to the relativity principle, the coordinate system K is privileged, with respect to all coordinate systems K’, that are in motion relative to it. The fundamental assertion of Lorentz’s theory, that every light ray in a vacuum always propagates (at least with respect to a certain coordinate system K) with the definite constant velocity c, we will call the principle of the constancy of the velocity of light. The previously indicated difficulty with Lorentz’s theory consists in the fact that the principle of the constancy of the velocity of light seems to be incompatible with the relativity principle.

If one maintains, in accordance with Lorentz’s theory, that there exists a privileged coordinate system K, in which the velocity of light in a vacuum equals c, then one cannot assume that the earth is at rest relative to this coordinate system. Because one cannot then assume that the (stationary) ether participates in the motion of the earth around the sun. A great number of experiments were carried out to demonstrate this relative motion. But it proved wholly impossible to demonstrate experimentally such a privileged direction. However, most of these negative findings did not prove anything against the theory. In a very ingenious theoretical investigation, Lorentz showed that, to a first degree of approximation, the relative motion is without effect on the way rays proceed in arbitrary optical experiments. Only one optical experiment [Michelson], employing a method so extremely sensitive that its negative outcome could not be accounted for, even by Lorentz’s theoretical analysis. 

To reconcile the negative result of this experiment with the theory, Lorentz and FitzGerald set forth the hypothesis that the stone plate, together with all the objects mounted on it, experience a minuscule contraction in the direction of the earth’s motion, the magnitude of which is such that the expected effect is compensated by an opposite effect resulting from this contraction. This manner of providing theoretical justification for experiments with a negative outcome, by means of hypotheses invented ad hoc, is very unsatisfactory. 

It becomes hard to avoid the idea that no reality attaches to this relative motion of the earth with respect to the system K, ie that it is, in principle, impossible to demonstrate such a relative motion. In other words: we arrive at the conviction that the relativity principle is universally, and strictly, valid. On the other hand, the foundation of Lorentz’s theory, and therewith, the principle of the constancy of the velocity of light, seem to be incompatible with the relativity principle. But anyone who has made an attempt to replace Lorentz’s theory with another one, that would do justice to the experimental facts, would have to admit that, given the current state of our knowledge, this undertaking seems quite hopeless.

Given this state of affairs, one has to ask oneself, again, whether Lorentz’s theory, and the principle of the constancy of the velocity of light, are really incompatible with the relativity principle. Close examination shows that the two principles are compatible with each other,  and that Lorentz’s theory does not contradict the relativity principle. However, our conception of space and time must be subjected to a fundamental revision. 

Further, it is easy to see that we must also abandon the idea of introducing a luminiferous ether into the theory. Because, if every light ray in vacuum is supposed to propagate with the velocity c, with respect to K, then we must conceive of this luminiferous ether as being everywhere at rest with respect to K. However, if the laws of propagation of light with respect to the system K’ (in motion relative to K), are the same as those with respect to K, then we would have to assume, with as much right, the existence of a luminiferous ether that is at rest with respect to K’. Since it is absurd to assume that the luminiferous ether is simultaneously at rest with respect to both systems, and since it would be hardly less absurd for the theory to privilege one of the two (or of infinitely many) physically equivalent systems over the others, one must dispense with that concept, which was, anyway, just a useless accessory of the theory ever since the mechanical interpretation of light had been abandoned. 

As far as its interpretation in theoretical physics is concerned, the coordinate system is nothing but a rigid measuring framework, on which to mark off the values of the spatial coordinates, with the help of rigid rods. We now have to pose the question of the physical meaning of the temporal determinations that usually occur in physics, in association with spatial determinations.

We designate as a clock a system that automatically repeats the same process. The number of processes of this kind that have already taken place, counting from an arbitrarily chosen one, is the temporal determination of the clock. The temporal determination of the clock that is simultaneous with an event, is called the time of the event, as measured by the clock. Suppose that a clock is set up at the origin of our coordinate system, and that some event takes place at a location immediately adjacent to this origin. Then we are in a position to specify the clock time that is simultaneous with the event, ie the time of the event (referred to our clock). 

But, if the location of the event is far away from the location at which the clock is set up, then we cannot ascertain directly the clock reading that is simultaneous with the event. For an observer standing next to the clock cannot perceive the event directly, but only via some intermediary process (signal) induced by the event, and propagated to the observer (eg via light rays). The observer determines only the time of arrival of the signal, but not the time of the event. He could ascertain the latter only if he knew the length of time during which the signal was en route. But it is impossible in principle to ascertain this length of time by means of the clock set up at the origin of K. Only the times of events occurring in the immediate vicinity of the clock can be ascertained directly by means of that clock.

If we also had a clock at the location at which the [distant] event took place (we will assume that this clock is of exactly the same constitution as the other one), and if an observer stood there who determined the time of the event from this clock, even this would not yet do us any good. For we are, at present, unable to determine the temporal reading of the first clock, that is simultaneous with the temporal determination read off the other clock. From this we see that, for a definition of time, a physical definition of simultaneity is also needed. If the latter is given, then the physical definition of time that we are seeking is complete. In other words, we also need a rule by which to regulate the other clock according to the first clock. 

Let some means be given for sending signals from the origin of the system K to the other location, and vice versa, such that the signals either way are completely equivalent physical processes. Then, the other clock should be adjusted such that the time taken for the signal to travel in each direction is the same. The process can then be repeated with other clocks.  

As regards the given definition, one has to be careful about one thing in particular. We use for the definition of time, a system of clocks that are at rest relative to the system K. Hence this definition has meaning only with respect to a coordinate system K, in a specific state of motion. If in addition to the coordinate system K, we introduce another system K’ that is in uniform translational motion relative to K, then we can define a time with respect to K’ just as well as we have done previously with respect to K. But it is not, a priori, evident that agreement can be produced between the readings of these two systems of clocks. There is no reason a priori, why two events that are simultaneous with respect to K, must also be simultaneous with respect to K’. This is what one understands by the “relativity of time”.

Thus, it turns out that the principle of constancy of the velocity of light, and the relativity principle, are incompatible with each other only as long as one holds to the postulate of absolute time, ie to the absolute meaning of simultaneity. But if one admits the relativity of time, then the two principles turn out to be compatible with each other; proceeding from these two principles, one arrives at the theory designated as the “theory of relativity”.

The fundamental problem connected with this conception is the following: We are given two coordinate systems K and K’. Let K’ be in a state of uniform translation with respect to K, and let v be the velocity of this motion. We are given the location and time of an arbitrary event (ie the coordinates x, y, z, and the time t) with respect to K. One has to determine location and time (x’ y’, z’, t’) with respect to K’. Conventional kinematics solves this problem by means of the following equations: x’ = x – vt, y’ = y, z’ = z, t’ = t. The last of these equations expresses the assumption that temporal determinations have a meaning independent of the state of motion (assumption of “absolute time”). 

But these equations also have hidden in them another implicit assumption. The state of motion, of the two systems K and K’, as these appear when observed from K. Let us now consider a point P’ on the x’ axis, whose distance from O’ is equal to l’. That means: an observer moving with K’, must lay his meter stick l’ times along the x’ axis, in order to get from O’ to P’. But observers who are at rest in system K, will have to proceed differently in order to evaluate the distance O’P’. They determine those spatial points in the system K, at which O’ and P’ are located, at a specified time (of the system K). The distance l between these two points, subsequently determined by laying the meter stick along the x axis of K, is the length we are seeking. 

One sees that the two procedures are fundamentally different, so that it is, a priori, possible that their numerical results, l and l’, differ from each other. This shows that one cannot reject, a priori, the possibility that the concept of spatial distance might also possess only a relative meaning. Thus, in addition to the “relativity of time”, we must also admit a “relativity of lengths”. This shatters the foundation on which the original transformation equations for spatial coordinates, and time values, are based. In the theory of relativity, the place of those equations is taken by equations that simultaneously satisfy the principle of relativity and the principle of constancy of the velocity of light. One finds the new equations by formulating mathematically the requirement that every light ray should propagate with the same velocity c in both systems K and K’. In this way one arrives at the transformation equations: x’ = (x -vt)/√(1 – v²/c²), y’ = y, z’ = z, t’ = (t-(v/c²)x)/√(1 – v²/c²)

The last of these equations shows that, in general, the equality of time values (simultaneity) of two events with respect to K, does not result in the equality of time values (simultaneity) of the same events with respect to K’. Thus, the absolute meaning of simultaneity is lost.

How great is the length of a rod, as observed from the system K, that is at rest with respect to K’, is oriented parallel to the x’ axis, and possesses the length l’ with respect to K’? The first of the equations provides the answer. For both ends of the rod, that is, the x coordinates of the rod ends, satisfy the equation:

 x₁ = (x₁ -vt)/√(1 – v²/c²), x₂’ = (x₂ -vt)/√(1 – v²/c²), so: x₂’ – x₁’ = (x₂ – x₁)/√(1 – v²/c²)

or: l = l’[√(1 – v²/c²)]

This means the following. If a rod possesses the length l’ when measured at rest, then, if it moves with velocity v along its axis, it will possess the smaller length l = l’[√(1 – v²/c²)] for a non-comoving observer, whereas for a comoving observer, it will always, have the length l’. The greater the velocity v of the moving rod, the smaller the length. If v approaches the velocity of light c, then the length of the rod approaches the value zero. For values of v that exceed the velocity of light, our result becomes meaningless; such velocities of motion are not possible according to the theory of relativity. One sees that the above mentioned hypothesis of Lorentz and FitzGerald, which was advanced in order to explain Michelson’s experiment, follows as a consequence of the theory of relativity. According to the latter, bodies at rest relative to K, evaluated from K’, will display exactly the same contraction as bodies at rest in K’, if these are evaluated from K.

Let a clock with a second hand be located at the origin of K’. For this clock we always have x’ = 0, and the clock strikes the seconds at times t’ = 0, 1, 2, 3, etc. The first and the fourth of our equations yield the following values for the times t of these strokes: t = 0/√(1 – v²/c²), 1/√(1 – v²/c²), 2/√(1 – v²/c²), 3/√(1 – v²/c²), etc

Thus, evaluated from K, the time between two strokes of the clock is equal to: t = 1/√(1 – v²/c²), and thus longer than one second. A clock traveling with the velocity v, runs slower, judged from a non-comoving system, than the same clock when it does not travel. Generalizing, one can conclude: Every event in a physical system slows down if the system is set into translational motion. But this slowing occurs only from the standpoint of a non-comoving coordinate system.

Scroll to Top