8.5 Scholium

Considering that the first Scholium of Newton's Principia begins with the famous assertion "absolute, true, and mathematical time...flows equably, without relation to anything external", it's ironic that Newton's theory of universal gravitation can be interpreted as a theory of variations in the flow of time. Suppose in Newton's absolute space we establish the Cartesian coordinates x,y,z, and then assign a fourth coordinate, t, to every point. We will call this the coordinate time parameter, but we don't necessarily identify this with the "true time" of events. Instead we postulate that the true lapse of time along an incremental timelike path is dt, given by

From the Galilean standpoint, we assume that a single set of assignments of the time coordinate t to events corresponds to the lapses of proper time dt along any and all paths, which implies that g₀₀ = 1 and k = 0. However, this can only be known to within some observational tolerance. Strictly speaking we can say only that g₀₀ is extremely close to 1, and the constant k is very close to zero (in conventional units of measure).

Using indices with x⁰ = t, x¹ = x, x² = y, and x³ = z, we can re-write (1) as the summation

where

Now let's define a four-dimensional array of number representing the second partial derivatives of the g_bd as a function of every pair of coordinates x^a, x^c

Also, we define the "contraction" of this array (using the summation convention for repeated indices) as

Since the only non-zero components of R^a_bcd are R⁰₀cd, it follows that the only non-zero component of R_ab is

If we assume g₀₀ is independent of the coordinate t (meaning that the metrical configuration is static), the first term vanishes and we find that R₀₀ is just the Laplacian of g₀₀. Hence if we take our vacuum field equations to be R_mn = 0, this is equivalent to requiring that the Laplacian of g₀₀ vanish, i.e.,

For convenience let us define the scalar j = g₀₀/2. If we consider just spherically symmetrical fields about the origin, we have j = j (r) and so

and similarly for the partials with respect to y and z. Since we have

and similarly for the y and z partials. Making these substitutions back into the Laplace equation gives

This is simple linear differential equation has the unique solution dj /dr = J/r² where J is a constant of integration, and so we have j = -J/r + K for some constants J and K.

Incidentally, it's worth noting that this applies only in three dimensions. If we were working in just two dimensions, the constant "2" in the above equation would be "1", and the unique solution would be dj /dr = J/r, giving j = J ln(r) + K. This shows that Newtonian gravity "works" only with three space dimensions, just as general relativity works only with four spacetime dimensions.

Now that we've solved for the g₀₀ field we need the equations of motion. We assume that objects in gravitational free-fall follow geodesics through the spacetime, so the equations of motion are just the geodesic equations

where x^a denote the quasi-Euclidean coordinates t,x,y,z defined above. Since we have assumed that the scale factor k between spatial and temporal coordinates is virtually zero, and that g₀₀ is nearly equal to unity, it's clear that all the speed components dx/dt, dy/dt, dz/dt are extremely small, whereas the derivative dt/dt is virtually equal to 1. Neglecting all terms containing one or more of the speed components, we're left with the zeroth-order approximation for the spatial accelerations

From the definition of the Christoffel symbols we have

and similarly for the Christoffel symbols in the y and z equations. Since the metric components are independent of time, the partials with respect to t are all zero. Also, the metric tensor g_mn and its inverse g^mn are both diagonal and the non-zero components of the latter are virtually equal to 1, 1/k, 1/k, 1/k. All the mixed components of g^mn vanish, so we are left with

and similarly for G ^y_tt and G ^z_tt. As a result, the equations of motion in the weak slow limit are closely approximated by

We've seen that the Laplace equation requires g_tt to be of the form 2K - 2J/r for some constants K and J in a spherically symmetrical field, and since we expect dt/dt to approach 1 as r increases, we can set 2K = 1. With g_tt = 1 - 2J/r we have

and similarly for the partials with respect to y and z. Therefore the approximate equations of motion in the weak slow limit are

If we set J/k = -m, i.e., to the negative of the mass of the gravitating source, these are exactly the equations of motion for Newton's inverse-square attraction. Interestingly, this implies that precisely one of J,k is negative. If we choose to make J negative, then the gravitational "potential" has the form g_tt = 1 + 2|J|/r, which signifies that the potential would increase as we approach the source, as would the rate of proper time along a stationary worldline with respect to coordinate time. In such a universe the value of k would need to be positive in order for gravity to be attractive, i.e., in order for geodesics to converge on the gravitating source. On the other hand, if we choose to make J positive, so that the potential and the rate of proper time decrease as we approach the source, then the constant k must be negative. Referring back to the original line element, this implies an indefinite metric. Naturally we can scale our units so that |k| = 1, but the sign of k is significant. Thus from the observation that "things fall down" we can nearly infer the Minkowski metrical structure of spacetime.

The fact that we can derive the correct trajectories of free-falling objects based on either of two diametrically opposed assumptions is not without precedent. This is very closely related to how Descartes and Newton were able to deduce the correct law of refraction based on the assumption that light travels more rapidly in denser media, while Fermat deduced the same law from the opposite assumption.

In any case, taking k = -1 and J = m, we see that Newton's law of gravitation in the vacuum is R_mn = 0, closely paralleling the vacuum field equations of general relativity, which represents the vanishing of the Laplacian of g₀₀/2. At a point with non-zero mass density we simply set this equal to -4pr to give Poisson's equation. Hence is we define the energy-momentum array

we can express Newton's geometrical spacetime law of gravitation as

This can be compared with Einstein's field equations

Of course the "R" and "T" arrays in Newton's law are based on simple partial derivatives, rather than covariant differentiation, so they are not precisely identical to the Ricci tensor and the energy-momentum tensor of general relativity. However, the definitions are close enough that the tensors of general relativity can rightly be viewed as the natural generalizations of the simple Newtonian arrays. The above equations show that the acceleration of gravity is proportional to the rate of change of g_tt as a function of r. At any given r we have dt/dt = , so g_tt corresponds to the squared "rate of proper time" (with respect to coordinate time) at the given r. It follows that our feet are younger than our heads, because time advances more slowly as we get closer to the center of the field. So, despite Newton's conception of the perfectly equable flow of time, his theory of gravitation can well be interpreted as a description of the effects of the inequable flow of time. In essence, the effect of Newtonian gravity can be explained in terms of the flow of time being slower near massive objects, and just as a refracted ray of light veers toward the medium in which light goes more slowly (and as a tank veers in the direction of the slower tread-track), objects progressing in time veer in the direction of slower proper time, causing them to accelerate toward massive objects.

Return to Table of Contents