3.1. Tolman–Ehrenfest and Davies–Unruh Effect
Assume there is a static space–time manifold
and two observers at space–time points,
locally at rest; partially synchronize their thermal clocks. In a static metric, the eigen-time interval for an observer, who is locally at rest, is
. Condition (4), together with (3), then directly translates into
If we set
, then (5) is the condition for thermal equilibrium in a gravitational field. The position-dependence of temperature is also known as
Tolman–Ehrenfest effect [
6]. Equally, in spirit, we can ask that two thermal clocks in Minkowski space–time are synchronized, where one is in relative motion at constant speed
to the other. With
, there holds for time in the moving coordinates
. Hence, if
denotes the temperature in the moving coordinates, relation (4) translates (with the same
as in (5)) into
Relation (7) corresponds to the original formula for relativistic temperature transformation in [
17]. After these introductory examples, we now consider a uniformly accelerated observer in Minkowski space–time, who moves with acceleration
in
-direction, say. We chose a co-moving coordinate system, which is defined in the wedge, limited by
, and given by the transformations
The eigen-time element is
These are so-called Rindler coordinates [
18]. At
the observers are at rest in their respective local inertial frames, and ratio (3) takes the form
A thermal clock in the local inertial frame at
and another one stationary at
say, are synchronized by (4), if
Hence, if
, which we from now on assume, if not stated otherwise,
With
, we arrive at
So far, we have worked with a single acceleration or chart. We can generalize (12) by demanding it to hold for different accelerations
, which is the basis for the generalization to non-flat manifolds like in (5). For our purpose, we consider a specific second system, namely, a single oscillator of frequency
with energy of one photon
. For its energy above the ground state, we have
With
and
being the wavelength, the acceleration is
, and the left-hand side of (12) turns into
is independent of
and synchronizing
implies
From (15), we get for the corresponding temperature
Expression (16) is the
Unruh–Davies temperature [
4,
5]. From Equation (15), there immediately results a number of well-known temperature expressions, if the acceleration
is chosen appropriately. In case of a Schwarzschild black hole of mass
, for instance,
is he surface gravity on the horizon
, where
denotes the gravitational constant, and we get the Hawking temperature
Similar expressions hold for Reissner–Nordström or Kerr black holes, if the corresponding surface accelerations are plugged into (16).
3.2. Bekenstein-Type of Bounds on Entropy
Relations (15) and (16) can be used to derive some further interesting results. Let the total energy within a ball of radius
in flat space be
, as measured in the local rest frame of a single photon clock. Clearly the total dissipation out of this region for any thermodynamic entropy
must satisfy
Hence, with (16) for
, we directly get the Bekenstein-bound [
7]
Instead of the photon rest-frame, we chose the local rest-frame of an observer, who is (weakly) gravitationally attracted by a mass
, relatively at rest at
. In the Newtonian limit with potential
there is the local acceleration
, where
is the gravitational constant. In the presence of matter, we have to slightly change the synchronization Equation (15) by adding an additional energy quantum to the photon-clock in vacuum (hence we actually synchronize with a two-photon clock, one energy quantum for space and one for matter), leading to
Like in (18), there must hold
With the total energy
, we get from (20)
and, finally, with the definition of the Planck length
and
, denoting the surface of the sphere with radius
, there directly results the area law [
8]
Estimate (23) remains true, and can be derived from (16) for the surface area of a Schwarzschild black hole, if we set with denoting the Schwarzschild radius.
3.3. De-Sitter Vacuum
We now consider de-Sitter space–time, representing the vacuum with energy density
, where
denotes the cosmological constant. It is well known that this space–time has a horizon
with surface-acceleration
given by the Hubble constant
. There holds
We have, in analogy to (18) for any thermodynamic entropy
within the region bounded by the horizon,
In the static de-Sitter metric (the line-element is
)
Hence, by help of relation (24)
By (16), we arrive at
and, finally, with the Planck length
, we get an area law
For radii
, however, (26) does not, a priori, reduce to a simple expression. An interesting question is, what the estimate looks like far away from the horizon in the visible universe, i.e., for a ball of radius
. Here, the volume of the ball is almost Euclidean. By Hubble’s law, we have for the surface acceleration
and we get, in analogy to (25) and (27),
Finally, again by (24), we get the traditional area law
It is interesting that the static de-Sitter coordinate frame represents the perspective of a non-co-moving observer, for whom there is a velocity of the “river of space” [
19]. Yet, the metric looks entirely static. One can indeed ask how, in a vacuum, one should differentiate between
and
in the absence of anything to compare. There is the ansatz to attribute the scale factor
to the entropy instead, and define
[
20]. In Equation (30), we would then start with the expression
. This ansatz can be useful, if we are in the Schwarzschild–de-Sitter space–time, which is defined by a gravitating mass
and the expanding vacuum, in combination (the line-element is
). For consistency’s sake, the thermal clocks of two observers, one locally accelerated by
and the other by
, also taking into account the vacuum expansion, must then march in step (5), i.e., with
Therefore, with Hubble’s law and
,
which leads to
Expression (34) is the empirically quite well-tested MOND (Modified Newtonian Dynamics) acceleration [
20,
21]. It can be shown that the effect of the vacuum expansion will only be relevant for distances
, where (29) does not hold, and where the vacuum acceleration is bigger than
[
22]. From this perspective, the correction (34) to the gravitation law is ultimately due to
.
We have seen, so far, that the synchronization principle is a very powerful tool to offer a simple derivation of some key relations in space–time physics. We now want to show that the Einstein equations can also be derived from this principle.
3.4. Einstein Equations
We want to come back to the local situation of weak gravitational attraction and use Equation (20).
With
, we write
By using (14), we arrive at
Equation (37), which holds on a hypersurface in the Newtonian limit, can be covariantly generalized, if we assume a static (or conformal) static metric. We sketch the derivation in, e.g., [
23]. We consider an asymptotically flat, static background with global time-like Killing vector field
. The generalization of Newton’s potential
can be defined by
The exponential
is the red-shift factor, which defines a foliation of space–time in space-like surfaces
of constant red-shift. In this set-up, a particle on the corresponding Killing world-line will have a four-acceleration perpendicular to
given by
The left-hand side of (37) formally turns into the more general expression
Relation (40) is (modulo constants) exactly the expression for the Komar-mass
, [
24], and we indeed get, by accounting for the constants, the equivalent equation to (37):
Re-expressed in terms of the Killing vector
, there holds
By Stokes theorem and the identity
(42) turns into
where
is a volume bounded by
. Since by (43), the Ricci tensor
equals zero in a massless region, relation (43) holds for any boundary surface
of
as long as
comprises all the matter. By writing the right-hand side as an appropriate integral over components (the energy field should be source free) of the energy-stress tensor
, we get
Finally, by considering an infinitesimally small region and imposing that, if matter
crosses the screen, then the Komar-integral changes by that amount, and (44) can be shown to hold for all (approximate) Killing vectors
and screens
with normal vector
. Therefore,
A similar approach was taken in [
25] by using null-screens.