betaFeedback

WorksPhysics and Astronomy

A Gravitomagnetic Bootstrap: from Newtonian gravity to general relativity via localizing rotational and conformal symmetries

Abstract

Newtonian gravity, in the light of the Equivalence Principle, is best understood as a localization of Galilean inertial structure — or at least, of the part of Galilean inertial structure that defines rectilinear acceleration. I consider what kind of theory can be constructed if we also localize rotational inertia and spatial and temporal scales. The resultant theory, given weak assumptions, is unique up to a constant with the dimensions of velocity and an overall sign. For an appropriate choice of these, it coincides with the first post-Newtonian approximation to general relativity, so that this construction gives a method to bootstrap from Newtonian gravity to general relativity without explicitly presupposing special relativity.

Introduction

In popular accounts, general relativity is often presented as a radical conceptual break from Newtonian gravity, replacing a theory of long-distance forces acting within a background of absolute space and time with a theory of dynamical spacetime geometry. But the central insight that led Einstein to general relativity — the Equivalence Principle, the equivalence between physics in an accelerating frame and physics under the action of a constant gravitational field — was not something Einstein learned through intuition or observation: it was an insight about Newtonian gravity itself.1

Einstein realized that Newtonian gravity could be better understood not as a theory of long-distance forces but as a theory that localized inertial structure. And since he had already concluded that Newton’s picture of inertial structure was not quite right and had to be replaced with the version implicit in *special relativity, the path to the general theory was “just” a matter of working out how to marry those two insights.

But it seems to be a matter of historical contingency that this insight into the nature of Newtonian gravity was gained only after physicists had already learned the limitations of Newton’s and Galileo’s conception of inertia and spacetime. (Indeed, Newton seemed at least to glimpse it: as argued by, e.g., Saunders [3], Corollary VI to the Principia is a clear precursor to the Equivalence Principle.) And had physicists come to it sooner, they might also have noticed that the localization of inertial structure implicit in Newtonian gravity is only partial. The standard of acceleration might be localized and made dependent upon the matter distribution, but the standards of rotation, and of spatial and temporal scales, remain global and absolute. And they might then have wondered what kind of theory could result if those aspects of inertia were also made dynamical.

This paper is an exploration of that possibility (an exploration, to be clear, that I intend as an anachronistic thought-experiment and not as a serious piece of counterfactual history). As I will show, any such theory is extremely tightly constrained, indeed so tightly constrained that it is almost uniquely fixed up to a new constant of nature with the dimensions of velocity. And to make the resultant theory self-consistent it is necessary to modify the Galilean symmetries of Newtonian physics in exactly the way that proved necessary in special relativity: that is, to replace them with the Lorentz symmetries, with the new constant of nature playing the role of the speed of light. The resultant extension of gravity can then be understood as the first post-Newtonian approximation to general relativity, and the conclusion is that we can bootstrap from the initial hypothesis of localized rotation and scale up to general relativity, in a way strikingly similar to the well-known bootstrap from magnetostatics to the Lorentz transformations and the Maxwell equations.

After reviewing the magnetostatic bootstrap (section 2), I begin (section 3) by considering what kind of theory can be made by localizing the Newtonian standard of unaccelerated motion, and argue that under weak assumptions the form of that theory is fixed uniquely up to the magnitude and sign of the gravitational constant. I then consider what kind of theories can be constructed under those same weak assumptions from localizing more general aspects of inertial structure. My first attempt (section 4) localizes both translational (that is: rectilinear-acceleration) and rotational structure, introducing a ‘Coriolis field’ to represent the latter; unfortunately, the resultant theory is inconsistent unless the Coriolis field is source-free. The second attempt (section 5) also localizes spatial scale structure; this turns out to fix the inconsistency in the previous theory and to produce a very rich extension of Newtonian gravity, uniquely defined up to the gravitational constant and a new constant with the dimensions of velocity. This extension, however, fails to be Galilean-covariant, but if we also localize the temporal scale (section 6) it can be made covariant under a new symmetry which agrees with Galilean boost symmetries in the limit of low velocities. The resultant theory (section 7) is the first post-Newtonian approximation to general relativity (section 8). I conclude (section 9) by exploring some conceptual, philosophical, and pedagogical lessons of this construction.

Notation: Roman indices range over (1,2,3)(1,2,3) and are raised and lowered freely; repeated indices are summed over; if vv is a vector, v2≡v⋅vv^2 \equiv v \cdot v; if TT is a tensor and vv a vector, (T⋅v)i≡Tijvj(T \cdot v)_i \equiv T_{ij}v_j. For readability, most of the purely mathematical results used are proved in the Appendix, which is largely self-contained.

Recap: the magnetostatic bootstrap

(This is all well-known textbook material and I claim no originality.)

Start with the equations for (electro)-magnetostatics, the (classically exact) theory of how test charges move in the presence of time-independent charges and currents. These are the Lorentz force law,

d(mv)dt=q(E+v×B),(1)\frac{\mathrm{d}(m\mathbf{v})}{\mathrm{d}t}=q(\mathbf{E}+\mathbf{v}\times\mathbf{B}), \tag*{(1)}

and the static Maxwell equations,

∇⋅E=ρϵ0(2)\nabla\cdot\mathbf{E}=\frac{\rho}{\epsilon_0} \tag*{(2)}
∇⋅B=0(3)\nabla\cdot\mathbf{B}=0 \tag*{(3)}
∇×E=0(4)\nabla\times\mathbf{E}=0 \tag*{(4)}
∇×B=μ0J.(5)\nabla\times\mathbf{B}=\mu_0\mathbf{J}. \tag*{(5)}

Under these equations, given two infinitesimal pieces of moving charge with velocities v1\mathbf{v}_1, v2\mathbf{v}_2 and charges q1q_1, q2q_2, the electric force between them scales like (q1q2/r2)(1/ϵ0)(q_1q_2/r^2)(1/\epsilon_0) and the magnetic force scales like (q1q2/r2)(μ0∣v1∣∣v2∣)(q_1q_2/r^2)(\mu_0|\mathbf{v}_1||\mathbf{v}_2|), so that μ0ϵ0\mu_0\epsilon_0 must have the dimensions of inverse velocity squared, μ0ϵ0=1/c2\mu_0\epsilon_0=1/c^2 for some velocity cc.

Taking the divergence of (2.2d) gives 0=μ0∇⋅J0=\mu_0\nabla\cdot\mathbf{J}. Local conservation of charge is expressed by ∂tρ+∇⋅J=0\partial_t\rho+\nabla\cdot\mathbf{J}=0, so that assuming charge is conserved, ∂tρ=0\partial_t\rho=0. Allowing time-varying charge density thus requires adding some time-dependent term to the right-hand side of (2.2d) whose divergence is μ0∂tρ\mu_0\partial_t\rho, and it is easy to see from (2.2a) that this modification must be

∇×B=μ0J+μ0ϵ0∂tE.(6)\nabla\times\mathbf{B}=\mu_0\mathbf{J}+\mu_0\epsilon_0\partial_t\mathbf{E}. \tag*{(6)}

Our theory may or may not be Galilean-invariant, but at any rate failure of Galilean invariance had better appear only at order 1/c21/c^2 or higher, otherwise experiments in labs on different sides of the planet would give detectably different results. If for the moment we assume exact invariance of the Lorentz force law under x→x+utx\to x+ut, this requires the transformation law

E→E−u×B(7)\mathbf{E}\to\mathbf{E}-\mathbf{u}\times\mathbf{B} \tag*{(7)}
B→B.(8)\mathbf{B}\to\mathbf{B}. \tag*{(8)}

Under this transformation we have

∇×E→∇×E−∇×(u×B)=∇×E−u∇⋅B+(u⋅∇)B(9)\nabla\times E \to\nabla\times E - \nabla\times(u \times B) = \nabla\times E - u\nabla\cdot B + (u \cdot\nabla)B \tag*{(9)}

so that (2.2c) is not Galilean-invariant even if (2.2b) is assumed. Fixing this requires replacing (2.2c) with

∇×E+∂tB=0(10)\nabla\times E + \partial_t B = 0 \tag*{(10)}

which is the last Maxwell equation.

However, (2.2a) and (2.3) are invariant under Galilean transformations only up to terms of order 1/c21/c^2. To preserve exact invariance of the Maxwell equations under boosts we have to replace the Galilei group with the Lorentz group and change the transformation law for magnetic fields to

B→B+μ0ϵ0u×E(11)B \to B + \mu_0\epsilon_0 u \times E \tag*{(11)}

(note that again this is the Galilean form up to a term of order 1/c21/c^2). The Lorentz force law is in turn invariant under these transformations only if we assume mass depends on velocity like m(v)=m0(1−v2/c2)−1/2m(v) = m_0(1-v^2/c^2)^{-1/2}. Now we are done: we have proceeded from magnetostatics and the assumption of leading-order Galilean symmetry, to the full Maxwell equations, the Lorentzian symmetries of special relativity, and the velocity-dependence of inertial mass.

Newtonian gravity as localized inertia

Now let us forget what we know of electrodynamics and relativity, and return to Newtonian gravity.

Newtonian dynamics and global inertial structure

Start with some collection of NN point particles with masses m1,…,mNm_1,\ldots,m_N moving in Euclidean space, interacting by position-dependent forces: the force exerted by particle nn on particle kk is FknF_{kn}, acts along the line between nn and kk, and has magnitude dependent only on the distance ∣xn−xk∣\lvert x_n-x_k\rvert between the spatial locations of the particles. By Newton’s Third Law, Fkn=−FnkF_{kn}=-F_{nk}, and the equations of motion for the system are

d2xn(t)dt2=1mn∑k≠nFnk(xn(t)−xk(t)).(12)\frac{\mathrm{d}^2 x_n(t)}{\mathrm{d}t^2} = \frac{1}{m_n}\sum_{k\ne n} F_{nk}(x_n(t)-x_k(t)). \tag*{(12)}

This system of equations is Galilean-covariant: solutions are preserved under the transformations

x(t)→Rx(t)+a+ut;(13)x(t) \to Rx(t) + a + ut; \tag*{(13)}
t→t+τ(14)t \to t+\tau \tag*{(14)}

for any time-independent rotation matrix R\mathbf{R}, vectors a,u\mathbf{a},\mathbf{u}, and scalar τ\tau. They are not, however, invariant under the wider class of time-dependent translations x→x+a(t)\mathbf{x} \to\mathbf{x}+\mathbf{a}(t) (for an arbitrary smooth vector-valued function a(t)\mathbf{a}(t)), nor the still-wider class of time-dependent rotations-plus-translations, where R\mathbf{R} is allowed to depend on time. Insofar as we take the symmetries of the equations as characterizing the structure of spacetime, this is to say that the implied setting of this dynamical system is Galilean (or neo-Newtonian) spacetime, not Maxwellian spacetime (which allows arbitrary time-dependent translations) or Leibnizian spacetime (which also allows time-dependent rotations).2^{2}

Another way to characterize the covariance of these equations is that they hold in arbitrary inertial frames, where the inertial frames are those in which force-free bodies move in straight lines at constant speed; given one inertial frame, performing a Galilean transformation yields another such frame. As written, they do not hold in accelerated frames — but formally we can fix this easily enough, just by introducing a new vector field G(t)\mathbf{G}(t) and rewriting the equations as

d2xn(t)dt2−G(t)=1mn∑k≠nFnk(xn(t)−xk(t)).(15)\frac{\mathrm{d}^{2}\mathbf{x}_{n}(t)}{\mathrm{d}t^{2}}-\mathbf{G}(t)=\frac{1}{m_{n}}\sum_{k\ne n}\mathbf{F}_{nk}\bigl(\mathbf{x}_{n}(t)-\mathbf{x}_{k}(t)\bigr). \tag*{(15)}

The new system of equations is invariant under the wider Maxwell symmetry group

x(t)→Rx(t)+a(t)(16)\mathbf{x}(t)\to\mathbf{R}\mathbf{x}(t)+\mathbf{a}(t) \tag*{(16)}
t→t+τ(17)t\to t+\tau \tag*{(17)}
G(t)→RG(t)+d2a(t)dt2.(18)\mathbf{G}(t)\to\mathbf{R}\mathbf{G}(t)+\frac{\mathrm{d}^{2}\mathbf{a}(t)}{\mathrm{d}t^{2}}. \tag*{(18)}

A useful way to think about this is that a Galilean-covariant dynamical system has an implied inertial structure, which suffices to distinguish unaccelerated and unrotated motion from accelerations and rotations. In introducing G\mathbf{G}, we have represented the accelerational part of the inertial structure by an inertial force: the inertial force on each body is equal to its mass times G\mathbf{G}, it can be treated just as the ‘real’ forces Fnk\mathbf{F}_{nk}, and the inertial frames are just those in which G\mathbf{G} vanishes.

Localizing translational structure

In itself, this move tells us little substantive about the dynamical system: the move is simply a special case of Kretschmann’s famous observation [7, 8] that any theory can be put in a generally covariant form with enough mathematical care. And the symmetry here is purely formal, and does not correspond to any physical symmetry (that is, to any symmetry that can be performed on an isolated subsystem of a larger system, as in the case of Galileo’s ship2).

However, the same mathematics can be reinterpreted in a way which does tell us something substantive. To begin, let’s define a force field as a field on a region of spacetime which induces some force on each body in the region, where that force might depend on the body’s location and velocity and also on intrinsic properties like charge or mass. And a force field is universal if the force it induces on a (localized) body brings about an acceleration that depends perhaps on the body’s location and/or velocity but at any rate not on any other feature of the body.

One such force field is specified by our vector function G(t)\mathbf{G}(t): if we declare that it induces a force

Fn[G,t]=mnG(t)(19)F_n[\mathbf{G},t] = m_n\mathbf{G}(t) \tag*{(19)}

on the nnth particle, then the combined effect of that force on the system of particles (together with the interparticle forces) is to induce exactly the equations of motion (3.3). Which is to say: a spatially invariant, velocity-independent universal force is indistinguishable from the inertial force induced by a time-dependent translation into an accelerating reference frame; put another way, universal forces of this form can be reinterpreted as part of the inertial structure. This is a form of Einstein’s Equivalence Principle (see [10]) or of Newton’s Corollary VI (see [3]).

Now suppose that we allow G\mathbf{G} to vary with space as well as time — but that we require its spatial variation to be slow enough in the vicinity of the particles that the relative accelerations between particles nn and pp induced by the universal force are negligible compared to those induced by interparticle forces: that is,

G(xn)−G(xp)≪1mn(∑k≠nFnk)−1mp(∑k≠pFpk)(20)\mathbf{G}(\mathbf{x}_n) - \mathbf{G}(\mathbf{x}_p) \ll\frac{1}{m_n}\left(\sum_{k\ne n} \mathbf{F}_{nk}\right) - \frac{1}{m_p}\left(\sum_{k\ne p} \mathbf{F}_{pk}\right) \tag*{(20)}

(the exact conditions for this will depend on the configuration of the particles, on the form of the interparticle forces, and of course on our tolerance for approximation). Under these conditions, the particles move under their interparticle accelerations together with a shared universal acceleration G(X)\mathbf{G}(\mathbf{X}), where X\mathbf{X} is some point characterizing the overall location of the particle swarm (say, the center of mass). We now have a stronger result (a form of the Strong Equivalence Principle; see again [10]):

Any universal and velocity-independent force field is indistinguishable locally from a shift of inertial structure or a shift to a rectilinearly accelerating reference frame, provided it is probed on scales where the interparticle accelerations of the probing matter due to internal forces are large compared to the relative acceleration of different particles induced from their differing locations and the spatial variation of the universal force field.

So a theory characterized by a universal velocity-independent force field GG is indistinguishable from a theory in which inertial structure — or at least that part of inertial structure which distinguishes uniform rectilinear acceleration from inertial motion — is localized and can vary from place to place. And by ‘indistinguishable’ I mean not just that no experiment can distinguish the possibilities, but that they are identical at the level of mathematical formalism. The same vector field GG can be interpreted either as a universal force on Galilean spacetime, or as a local standard of what motions count as force-free.

I am inclined to say (see [11]) that the question of which interpretation is ‘correct’ is a pragmatic rather than factive matter; following Knox [1] I am inclined to add that the pragmatically sensible choice is to take spacetime structure as by definition given by the motions of bodies under inertial forces, so that the ‘correct’ spacetime structure is the one in which inertial motion is localized, and not the unobservable background Galilean spacetime. But my concern here is less whether GG has a univocal interpretation, more what kind of dynamical theory we can build that incorporates such a universal force field.

Deriving Newtonian gravity

Attempts to lock down the form of a physics theory are usually heuristic rather than deductive: one states reasonable-sounding assumptions and sees what follows from them, without a claim that the assumptions are logically necessary. With this modest understanding of the game we are playing, let us make the following assumptions about our theory.

  1. The force field GG is determined by the matter distribution (that is, by where the matter is, how fast it is going, and perhaps other intrinsic properties like mass, charge or shape). This does not entirely preclude the possibility that GG has independent degrees of freedom (analogous to electromagnetic radiation) but it does presume that we can identify a regime where those additional degrees of freedom are absent.

  1. The dynamics for GG are local at least inasmuch as they are given by some local differential equations that relate GG and its spatial and temporal derivatives to some local function of the matter distribution. Furthermore, as a simplifying ansatz rather than for any deeper reason, we will assume these equations are first order and that the matter distribution occurs linearly in them. Given the additional assumption of Euclidean symmetry which we make below, this amounts to saying that the equations constrain the divergence, curl, and rate of change of GG, equating them to matter densities and currents (or to zero, in the absence of matter).

  1. The equations in (2) are solved for empty space in accelerated coordinates: that is, for uniform GG.

(4) Given some collection of particles very distant from other particles, we can treat that collection as an isolated system and solve for the G\mathbf{G} associated to that system, under the assumption that it tends to zero at large distances. This is basically a scale separation assumption: there may in addition be a ‘background’ G\mathbf{G}-field attributable to distant matter, but provided that it varies slowly enough that it can be treated as constant over the extent of the system, it can be transformed away by boosting to an accelerating reference frame.

(5) The equations of the theory are covariant under the Maxwell group, with the transformation law for the field being (3.4), the one appropriate for an inertial field.

(6) The overall dynamics of matter under gravity conserves momentum (if we assumed an action formalism we could regard this as a consequence of (5), given that momentum is the Noether invariant associated to spatial translation symmetry, but we will not do so).

Assumptions (1)–(6) pretty much pin down the form of the theory. To see this, firstly note that uniform fields have vanishing curl and divergence. By (1)–(3), that fixes the field equations in vacuum to be

∇⋅G=0(21)\nabla\cdot\mathbf{G} = 0 \tag*{(21)}
∇×G=0(22)\nabla\times\mathbf{G} = 0 \tag*{(22)}

Now, suppose we have some region containing matter, with mass density ρ\rho. The force per unit volume on some small chunk of the matter is

f=ρG.(23)\mathbf{f} = \rho\mathbf{G}. \tag*{(23)}

That is: the force on a small chunk of matter depends only on its mass and not on its velocity or anything specific to the type of matter the chunk is made up of (such as electric charge or chemical composition). If we then consider (say) two rocks interacting under gravity, then the gravitational force on each rock depends only on that rock’s mass, and so in order to conserve the total momentum of the system, the gravitational force exerted by each rock can depend only on its mass. This entails that the gravitational force must be sourced by ρ\rho alone, and given our other assumptions, the only available field equations are

∇⋅G=−4πGNρ(24)\nabla\cdot\mathbf{G} = -4\pi G_N \rho \tag*{(24)}
∇×G=0(25)\nabla\times\mathbf{G} = 0 \tag*{(25)}

We can readily verify that the theory is indeed Maxwell-covariant. The constant GNG_N is a free parameter, and indeed nothing about our arguments requires it to be positive: a localized theory of acceleration must be universally repulsive or universally attractive, but empirical data is required to fix which.

If we consider some distribution of matter such that the background gravitational field can be idealized as constant over the extent of the matter, and we then perform a coordinate transformation to eliminate that background field, we can adopt boundary conditions where G\mathbf{G} tends to zero at spatial infinity; with this boundary condition, G\mathbf{G} is given explicitly by the familiar Newtonian expression

G=∇Φ;Φ(x)=GN∫d3y ρ(y)∣x−y∣.(26)\mathbf{G} = \nabla\Phi;\quad\Phi(\mathbf{x}) = G_N \int\mathrm{d}^3 y \,\frac{\rho(\mathbf{y})}{\lvert\mathbf{x} - \mathbf{y} \rvert}. \tag*{(26)}

For a point source, of course, this reduces to the inverse square law: the gravitational force due to a particle of mass mm has magnitude GNm/r2G_Nm/r^2 at a distance rr from that particle.

To conclude: on fairly weak and plausible additional assumptions, and up to the sign and magnitude of Newton’s constant, the form of Newtonian gravity is pretty much determined once we require it to be the theory of a universal velocity-independent force — or equivalently, once we require it to be a theory that localizes the part of global inertial structure that distinguishes uniform acceleration from inertial motion. And the overall system has as its symmetries the Maxwell group of Euclidean transformations, time translations, and time-dependent spatial translations.

Continua, extended bodies, and the Equivalence Principle

My account so far has moved back and forth between characterizing matter as discrete particles and as a continuum, and it will be helpful for later to pause and say a little about how these are related. In astrophysics, ‘particle’ does not refer to genuinely pointlike singularities, nor ‘continuum’ to matter genuinely continuous at all scales. Rather: these are intrinsically approximate and scale-dependent notions. In both cases, the important notion is scale separation: a body of matter can be treated as a point particle if the scale on which it is extended is very small compared to the other dynamically relevant scales in the problem (such as: how quickly the background fields vary). Similarly, a collection of many such particles can be treated as a continuum if we can identify ‘representative volume elements’ (RVEs), volumes large enough to contain a great many particles but still small compared to the other relevant scales of the problem.

(This scale relativity of matter is a commonplace of astrophysics. On kiloparsec scales, the Galaxy can be treated as a fluid of stars. Zoom in, and that fluid resolves into individual stars which can be treated as point particles. Zoom in further, and those point particles resolve into a fluid of plasma. Zoom in further still, and that fluid resolves into electrons and nuclei, albeit at this scale classical mechanics itself breaks down.)

Thus understood, the continuum dynamical equations can be understood as consequences of the discrete ones. In point-particle mechanics, the basic dynamical principles are that force is rate of change of momentum and that individual particle mass is conserved. The continuum versions of these are

∂tJ+∇⋅T=f(27)\partial_t \mathbf{J} + \nabla\cdot\mathbf{T} = \mathbf{f} \tag*{(27)}
∂tρ+∇⋅J=0(28)\partial_t \rho+ \nabla\cdot\mathbf{J} = 0 \tag*{(28)}

where ρ\rho, J\mathbf{J}, T\mathbf{T}, and f\mathbf{f} are respectively the mass density, momentum density, stress tensor, and external force density. If we understand the continuum as a swarm of particles then these are defined by

ρ×volume of RVE=∑nmn(29)\rho\times\text{volume of RVE} = \sum_n m_n \tag*{(29)}

References

  1. [1]Knox E 2014 British Journal for the Philosophy of Science 65 863–880
  2. [2]Wallace D 2020 British Journal for the Philosophy of Science 71 1–32
  3. [3]Saunders S 2013 Philosophy of Science 80 22–48
  4. [4]Earman J 1989 World Enough and Space-Time (Cambridge, Mas-sachusetts: MIT Press)
  5. [5]Carter B and Khalatnikov I M 1994 Reviews in Mathematical Physics 6 277–304
  6. [6]Duval C 1993 Classical and Quantum Gravity 10 2217–2222 (Preprint 0903.1641)
  7. [7]Kretschmann E 1917 Annalen der Physik 53 575–614
  8. [8]Norton J D 1993 Reports on Progress in Physics 56 791–858
  9. [9]Wallace D 2022 Observability, redundancy and modality for dynamical symmetry transformations The Philosophy and Physics of Noether’s Theorems: A Centenary Volume ed Read J, Roberts B and Teh N (Cambridge: Cambridge University Press) pp 322–353DOI
  10. [10]Lehmkuhl D 2022 The equivalence principle(s) The Routledge Companion to Philosophy of Physics (London: Routledge) pp 125–154DOI
  11. [11]Wallace D 2022 Philosophical Perspectives 36 345–378
  12. [12]Costa L F and Natário J 2016 American Journal of Physics 84 388–395
  13. [13]Heaviside O 1893 The Electrician 31 281–282, 359 in two parts; reprinted as Appendix B of Electromagnetic Theory, vol. 1
  14. [14]Maxwell J C 1865 Philosophical Transactions of the Royal Society of London 155 459–512
  15. [15]McDonald K T 1997 American Journal of Physics 65 591–592
  16. [16]Di Francesco P, Mathieu P and Sénéchal D 1997 Conformal Field Theory Graduate Texts in Contemporary Physics (New York: Springer)
  17. [17]Brown H R 2005 Physical Relativity (Oxford: Oxford University Press)
  18. [18]Weinberg S 1972 Gravitation and Cosmology: Principles and Applications of the General Theory of Relativity (New York: John Wiley and Sons)
  19. [19]Gerber P 1898 Zeitschrift für Mathematik und Physik 43 93–104
  20. [20]Wells J D 2012 Effective Theories in Physics: From Planetary Orbits to Elementary Particle Masses (Berlin, Heidelberg: Springer)
  21. [21]Carroll S 2003 Spacetime and Geometry: an Introduction to General Relativity (San Francisco, CA.: Addison Wesley)
  22. [23]GRAVITY Collaboration, Abuter R et al. 2020 Astronomy & Astrophysics 636 L5
  23. [24]Damour T, Soffel M and Xu C 1991 Physical Review D 43 3273–3307
  24. [25]Poisson E and Will C M 2014 Gravity: Newtonian, Post-Newtonian, Relativistic (Cambridge: Cambridge University Press)
  25. [26]Worrall J 1989 Dialectica 43 99–124
  26. [28]Stanford K 2006 Exceeding Our Grasp: Science, History, and the Problem of Unconceived Alternatives (Oxford: Oxford University Press)
  27. [29]Egg M 2026 Realism about Newtonian gravity British Journal for the Philosophy of Science, forthcomingDOI
  28. [30]Stachel J 2007 The story of Newstein or: Is gravity just another pretty force? The Genesis of General Relativity (Boston Studies in the Philosophy of Science vol 250) ed Janssen M, Norton J D, Renn J, Sauer T and Stachel J (Dordrecht: Springer) pp 1962–2000DOI
  29. [31]Lovelock D 1971 Journal of Mathematical Physics 12 498–501
  30. [32]Wald R M 1986 Physical Review D 33 3613–3625
  31. [33]Burgess C P 2004 Living Reviews in Relativity 7 5
  32. [34]Costa L F O and Natário J 2014 General Relativity and Gravitation 46 1792
  33. [35]Bell J S 1976 Progress in Scientific Culture 1 1–13 reprinted in J.S. Bell, Speakable and Unspeakable in Quantum Mechanics (Cambridge: Cambridge University Press, 1984), pp. 67-80; page numbers refer to this reprint.
  34. [36]Binney J and Tremaine S 2008 Galactic Dynamics 2nd ed (Princeton: Princeton University Press)

Paper details

vlnc:W_w4yn55vy
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 1

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Claude Opus 5.5

View statement
Looking for related papers…
Contents