Einstein on time symmetry in electrodynamics

Zur Theorie der Lichterzeugung und Lichtabsorption, Annalen der Physik 20 (1906), 203:

Obwohl die Maxwellsche Theorie auf Elementarresonatoren nicht anwendbar ist, so ist doch die mittlere Energie eines in einem Strahlungsraume befindlichen Elementarresonators gleich derjenigen, welche man mittels der Maxwellschen Theorie der Elektrizität berechnet.


Über die Entiwickelung unserer Anschauungen über das Wesen und die Konstitution der Strahlung, Deutsche Physikalische Gesellschaft, Verhandlungen 7 (109), 491, Beck's translation:

The basic property of the wave thoery that gives rise to these difficulties seems to me to lie in the following. While in the kinetic theory of matter there exists an inverse process for every process in which only a few elementary particles take part, e.g., for every molecular collision, according to the wave theory this is not the case for elementary radiation processes. According to the prevailing theory, an oscillating ion produces an outwardly propagated spherical wave. The opposite process does not exist as an elementary process. It is true that the inwardly propagated spherical wave is mathematically possible; however, its approximate realization requires an enormous amount of emitting elementary structures. Thus, the elementary process of light radiation as such does not possess the character of reversibility. Here, I believe, our wave theory is off the mark. Concerning this point the Newtonian emission theory of light sems to contain more truth than does the wave theory, since according to the former the energy imparted at emission to a particle of light is not scattered throughout the infinite space but remains available for an elementary process of absorption. Keep in mind the lwas of production of secondary cathode rays by X-rays.

If primary cathode rays impinge upon a metal plate $P_1$, they produce X-rays. If these impinge upon a second metal plate $P_2$, cathode rays will be produced once again, their velocity being of the same order of magnitude as that of the primary cathode rays. As far as we know today, the velocity of the secondary cathode rays depends neither on the distance between the plates $P_1$ and $P_2$ upon which they impinge nor on the intensity of the primary cathode rays, but exclusively on the velocity of the primary cathode rays. Let us for once assume that this is strictly valid. What will happen if we let the intensity of the primary cathode rays, or the size of the plate $P_1$ upon which they impinge, decrease to such a degree that the impinging of an electron of the primary cathode rays can be conceived as an isolated process? If the above is really true, then, because the velocity of the secondary rays is independent of the intensity of the primary rays, we will have to assume that on $P_2$ (as a result of the impinging of the above electron on $P_1$) either nothing is being produced or that a secondary emission of an electron occurs on it with a velocity of the same order of magnitude as of the electron impinging on $P_1$. In other words, the elementary radiation process seems to proceed such that it does not, as the wave theory would require, distribute and scatter the energy of the primary electron in a spherical wave propagating in all directions. Rather, it seems that at least a large part of this energy is available at some location of $P_2$ or somewhere else. The elementary process of radiation seems to be directed. Furthermore, one gets the impression that the process of X-ray production in $P_1$ and the process of secondary cathode ray production in $P_2$ are essentially inverse processes.


Zum gegenwärtigen Stand des Strahlungsproblems, Physikalische Zeitschrift 10 (1909), 186, my translation:

It is however not correct that the solution $f_3$ [ $=a_1f_1 + a_2f_2$ ] is a more general solution than $f_1$ [the retarded potential], and that one specializes the theory by setting $a_1=1$, $a_2=0$. If one puts
$f(x,y,z,t) = f_1,$
then it follows that the electromagnetic effects at the point x,y,z is calculated from the motions and configurations of the electric quantities that take place before the moment t. If one puts
$f(x,y,z,t) = f_2,$
then one uses, for determination of electromagnetic effects, those motions and configurations that take place after the moment t.

In the first case one calculates the electromagnetic field from the totality of the emitting processes, in the second case from the totality of the absorbing processes. If the entire process takes place in a space that is bounded on all sides, then one can represent it equally well in the forms

$f=f_1$
and
$f=f_2.$

If now we consider a field that is emitted from the finite into the infinite, we naturally can use only the form

$f=f_1,$

for even the totality of the absorbing processes cannot be brought under consideration. This has to do with an irrefutable paradox of the infinite. Both methods of representation are applicable no matter how distant one imagines the absorbing bodies to be. Therefore, one cannot conclude that the solution $f=f_1$ is more special than the solution $a_1f_1 + a_2f_2$, where $a_1 + a_2 = 1$.
That a body does not "receive energy from the infinite without some other body losing a corresponding quantum of energy" cannot in my opinion be invoked as an argument. First of all we cannot, if we want to stay with experience, speak of the infinite, instead only of spaces that lie beyond the space under consideration. Furthermore, we can infer from the unobservability of such a process a irreversibility of elementary electromagnetic processes only to the extent that a irreversibility of elementary movements of atoms follows from the second law of thermodynamics.


Letter to Paul Bernays, 1916 (Beck translation):

The orientational sense of time exhibited by living organisms everywhere is intimately connected with the second law. It primarily involves processes of diffusion, irreversible chemical processes, heat conduction, viscous currents, etc. It is entirely correct that this temporal bias of events finds no expression in the fundamental laws we use as a basis. But the theory of relativity shares this circumstance with classical mechanics, likewise with conventional electrodynamics and optics. The second law is understood such that a very improbable state is set for one temporal limit (lower t-limit) of a four-dimensional region under consideration; for the region's upper t-limit the probability considerations then yield a state of greater probability. The puzzle is thus transferred into the boundary conditions and therefore avoids "explanation" by means of the equations.