MOND — Modified Newtonian Dynamics

The problem MOND was built to solve

By 1980 the rotation-curve evidence for missing mass in spiral galaxies had become overwhelming. Vera Rubin, Kent Ford, and collaborators had measured the orbital velocities of stars and gas at radii well beyond the visible disk of dozens of galaxies; instead of falling off like 1/√r the way Kepler's laws demand once you are outside the bulk of the mass, the velocities stayed essentially constant. The standard inference, then as now, was that each galaxy is embedded in an extended halo of unseen mass whose gravity continues to dominate well past the optical edge.

The dark-halo picture is internally consistent and has since been written into ΛCDM with great care. But in 1983 the Israeli physicist Mordehai Milgrom asked a different question. The transition from "Keplerian" to "flat" in observed rotation curves does not happen at a fixed radius; it happens, across an enormous range of galaxy types and sizes, at a fixed acceleration — about 10⁻¹⁰ m/s². Wherever the Newtonian acceleration computed from visible matter drops below that value, the rotation curve refuses to fall. What if, Milgrom asked, the universal threshold is telling us something about the law of motion itself, rather than about a coincidental conspiracy of dark-matter distributions?

The modified law

Milgrom proposed a minimal alteration. Newton's second law in the standard regime reads F = ma. At accelerations far below a critical value a_0, replace it with

F = m a²/a_0           (deep-MOND regime, a ≪ a_0)

To stitch the two regimes together smoothly, introduce a dimensionless interpolation function μ(x) with μ(x) → 1 for x ≫ 1 and μ(x) → x for x ≪ 1. The full Milgrom law is then

μ(a / a_0) · a = g_N
where g_N = GM/r² is the Newtonian acceleration from the baryons.

The numerical value of the constant is fitted from rotation-curve data and converges on a_0 ≈ 1.2 × 10⁻¹⁰ m/s². Suggestively, this is of the order of c·H_0 — the speed of light times the Hubble constant — and roughly c²/(radius of the observable universe). Whether that coincidence reflects deep cosmological structure or is numerical happenstance has been argued about for forty years.

How MOND produces flat rotation curves

Consider a test star on a circular orbit at radius r around a galaxy of total baryonic mass M, far enough out that its Newtonian acceleration g_N = GM/r² is well below a_0. In the deep-MOND regime the law gives

a²/a_0 = GM/r²
a = √(GM a_0) / r

For a circular orbit, the centripetal requirement is a = v²/r. Equating,

v²/r = √(GM a_0) / r
v² = √(GM a_0)
v = (GM a_0)^(1/4)

The orbital velocity is independent of r. Outside the bulk of the visible mass, every star — at every radius — orbits at the same speed. The rotation curve is asymptotically flat by construction, with the flat value set entirely by the total baryonic mass and the universal constant a_0. No halo profile is fitted. No dark matter is invoked.

The baryonic Tully-Fisher relation falls out

Take the flat-velocity result, v⁴ = GMa_0, and read off what it says about a sample of galaxies. The relation between asymptotic flat velocity and total baryonic mass is

M = v⁴ / (G a_0)

Every galaxy that is in the deep-MOND regime far from its centre must obey this exact scaling, with a slope determined by fundamental constants and a normalisation determined by a single fitted number — a_0. This is the baryonic Tully-Fisher relation, observationally established over five decades in galaxy mass with a scatter of about 0.1 dex. In MOND the relation is a theorem with no free parameters; in ΛCDM it is a fortuitous outcome of complicated galaxy-formation feedback that must conspire across enormous ranges of halo mass, environment, and merger history to land on a single tight line.

The radial acceleration relation — MOND, written across the sky

In 2016, Stacy McGaugh, Federico Lelli, and James Schombert published the most stringent galaxy-scale test of any gravity theory to date. They took 153 nearby disk galaxies from the Spitzer Photometry and Accurate Rotation Curves (SPARC) database, computed the Newtonian acceleration g_bar from the visible baryons at every radius in every galaxy, and plotted it against the measured total radial acceleration g_obs inferred from the rotation curve.

The result is striking. Every galaxy — dwarf, giant, gas-rich, gas-poor, low-surface-brightness, high-surface-brightness — and every radius within every galaxy lies on a single tight curve. The scatter is 0.13 dex, and most of that is consistent with observational uncertainty. The functional form is

g_obs = g_bar / [1 − exp(−√(g_bar / a_0))]

with a_0 = (1.20 ± 0.02) × 10⁻¹⁰ m/s². This is exactly the MOND interpolation function in one of its standard forms. At high g_bar, g_obs equals g_bar — Newtonian gravity. At low g_bar, g_obs approaches √(g_bar a_0) — deep MOND. The transition is at a_0.

In a dark-matter universe, the radial-acceleration relation is surprising. The total acceleration at any point in any galaxy is the vector sum of contributions from baryons and a dark halo of unspecified mass, concentration, triaxiality, and orientation. Halo properties depend on assembly history, mergers, and feedback. The expectation is significant galaxy-to-galaxy scatter. Yet the data show essentially none. ΛCDM advocates have constructed explanations — feedback-driven self-regulation, halo-baryon coupling, observational selection — but no consensus account exists for why the relation should be as tight as it is.

The predictive track record

The strongest practical argument for MOND is that it has repeatedly predicted what new data would show, before they were taken. The pattern is unusual in galactic astrophysics, where most "theories" are fits to existing observations.

• Low-surface-brightness galaxy rotation curves. By the early 1990s, deep-MOND predictions had been made for diffuse galaxies whose entire visible structure lies below a_0. When 21-cm observations finally caught up, the predicted rotation curves were correct to within photometric uncertainties — including the precise asymptotic velocities.
• Renzo's rule. Every feature in the surface-brightness profile of a galaxy has a corresponding feature in its rotation curve. In MOND this is automatic — baryons are the only source of gravity. In a dark-halo model the rule has to emerge from a fine balance between disk and halo gravity.
• Galaxy-by-galaxy curve shapes. Given only the photometric mass profile of a single galaxy, MOND predicts the entire rotation curve with one free parameter — the stellar mass-to-light ratio — within the range expected from stellar-population modelling. ΛCDM galaxy fits typically have three or more free parameters per galaxy (halo mass, concentration, mass-to-light) and still struggle with diversity at the dwarf scale.
• Baryonic Tully-Fisher zero-point. The normalisation of the v⁴ ∝ M relation, computed from a_0, agrees with observation. No further tuning is needed.

Where MOND breaks down

For all its galaxy-scale successes, MOND has serious problems at every other scale.

Galaxy clusters

Apply MOND to the hot X-ray-emitting gas in a galaxy cluster, and use the cluster's observed temperature profile to compute the expected enclosed mass. The result still falls short by a factor of two to three of what is required to bind the cluster. MOND, in other words, needs its own dark matter at the cluster scale — typically attributed to ordinary baryons in some unseen form (cold molecular clouds, faint stars, sterile neutrinos at the 2-eV scale). The need to invoke an unseen component at clusters is uncomfortable for a theory whose central appeal is the elimination of unseen components.

The Bullet Cluster

In 2006 Douglas Clowe and collaborators presented a striking image of two clusters that had collided and passed through each other. The hot gas, which contains most of the baryonic mass, was decelerated by ram pressure and lags behind. Stars and galaxies, being effectively collisionless, sailed through. Weak gravitational lensing maps the total mass distribution — and the lensing peaks lie not on the gas but on the dispersed galaxy components. Mass and light are spatially offset.

For collisionless cold dark matter this is straightforward: the dark matter passed through with the galaxies. For a force law that follows baryons, it is awkward. MOND can be tuned to fit the lensing pattern only by invoking, again, an unseen mass component (sterile neutrinos at ~2 eV are the standard candidate) — undermining the original motivation.

The cosmic microwave background

The angular power spectrum of CMB temperature anisotropies — particularly the relative heights of the second and third acoustic peaks — is sensitive to the ratio of baryonic to total matter density in the early universe. ΛCDM, with cold dark matter Ω_c ≈ 0.265 and baryons Ω_b ≈ 0.049, reproduces the observed peak heights to within fractions of a percent. A baryon-only model predicts the second peak too tall and the third peak too small. No MOND-only theory has so far reproduced the observed CMB peak pattern; the relativistic extensions either appeal to a hidden non-baryonic component (which defeats the purpose) or struggle to fit at all.

Gravitational lensing of clusters

Strong-lensing analyses of massive clusters (e.g. Abell 1689) routinely require more mass than is visible by a factor of order three to five — again, even in MOND. Combined with the Bullet Cluster offset, the lensing evidence is widely regarded as the strongest empirical case for actual dark matter rather than modified gravity.

MOND versus ΛCDM at a glance

Observational test	MOND	ΛCDM
Flat galaxy rotation curves	Forced by the law (v⁴ = GMa₀)	Requires tuned halo profile per galaxy
Baryonic Tully-Fisher v⁴ ∝ M	A one-line theorem	Emerges from feedback, with scatter
Radial-acceleration relation (RAR)	Predicted exactly, 0.13 dex scatter	Surprisingly tight; explanations debated
Low-surface-brightness rotation curves	Predicted before measurement	Diversity problem at dwarf scale
Cluster mass budgets	Still need unseen mass (factor 2–3)	Cleanly fit by NFW haloes
Bullet Cluster mass–light offset	Awkward; needs hidden component	Natural for collisionless DM
CMB acoustic peak heights	Pure-baryon fits fail; extensions strained	Fits with no free parameters
Large-scale structure growth	Lacks a robust relativistic completion	Linear-perturbation theory works
Big Bang nucleosynthesis	Consistent (baryons only)	Consistent
Solar system / lab gravity	Identical to Newton (a ≫ a₀)	Identical to Newton

Worked example: the Milky Way at the Sun's orbit

The Sun orbits the Galactic centre at r ≈ 8.2 kpc with circular velocity v ≈ 230 km/s. Let us compute g_N from visible matter and check that we are in the transition regime, not deep MOND.

Total baryonic mass interior to the solar orbit is roughly M ≈ 8 × 10¹⁰ M☉. Then

g_N = GM/r²
    = (6.67 × 10⁻¹¹) (1.6 × 10⁴¹) / (2.5 × 10²⁰)²
    ≈ 1.7 × 10⁻¹⁰ m/s²

So g_N is roughly equal to a_0 — the solar neighbourhood sits right at the MOND transition. Newtonian and MOND predictions for v² there differ by about 30 percent, which is exactly where the discrepancy is. Out at r = 20 kpc the Galactic g_N has dropped well below a_0 and the rotation curve is dominated by the modified-gravity correction. In MOND this is just where you would expect the asymptotic v⁴ = GMa_0 to kick in; in the ΛCDM model it is where the dark halo must take over from the disk.

Relativistic completions and theoretical variants

• Bekenstein's TeVeS (2004). Tensor-Vector-Scalar gravity. The metric tensor g_μν, a unit timelike vector U_μ, and a scalar field φ together produce a relativistic theory whose weak-field, non-relativistic limit reduces to Milgrom's law. Predicts gravitational lensing at the galaxy scale that matches MOND-fitted rotation curves. Has difficulty with the second and third CMB acoustic peaks and with the Bullet Cluster offset.
• Moffat's MOG / STVG. Scalar-Tensor-Vector Gravity. An additional repulsive vector field at short range plus an enhanced effective gravitational constant at long range. Reproduces galaxy rotation curves; partial success on clusters with no dark matter; mixed results on the CMB.
• QUMOND (Quasi-linear MOND). A reformulation by Milgrom (2010) that makes numerical simulations of MOND galaxies tractable. Used in the few existing MONDian N-body codes.
• Aether-scalar tensor theories (AeST). Skordis & Złośnik (2021) proposed an extension that promises to reproduce the CMB peak pattern. Active research front; not yet conclusive.
• Superfluid dark matter (Khoury 2016). A hybrid: dark matter genuinely exists, but inside galaxies it condenses into a superfluid whose phonons mediate a MOND-like force. Aims to inherit MOND's galaxy successes and ΛCDM's cluster and CMB successes.
• Emergent gravity (Verlinde 2017). A thermodynamic / entropic derivation that recovers a MOND-like effect in the appropriate limit from cosmological dark energy. Speculative.

The current empirical status

As of the mid-2020s the situation can be summarised as follows. At the scale of individual galaxies — and especially in their outer, low-acceleration regions — MOND wins. The radial-acceleration relation, the baryonic Tully-Fisher law, Renzo's rule, and the predictive track record on diffuse rotation curves are not in serious dispute, and they are not naturally produced by current cold-dark-matter simulations. At the scale of galaxy clusters, the Bullet Cluster, the CMB, and large-scale structure, ΛCDM wins decisively. MOND requires either substantial additional unseen mass at those scales or a relativistic completion that no one has yet succeeded in writing down.

What this most likely indicates is not that one framework is simply right and the other simply wrong, but that some deeper structure is at work — perhaps a dark sector whose internal physics produces MOND-like phenomenology inside galaxies (the superfluid-dark-matter or self-interacting-dark-matter route), perhaps a relativistic modified gravity not yet discovered, perhaps both. The fact that an empirical regularity as tight as the radial-acceleration relation exists at all is information about the universe that any successful theory will eventually have to explain.

Common pitfalls

• Equating MOND with "rejecting dark matter". MOND in its full form rejects dark matter at the galaxy scale but still requires unseen mass at the cluster scale. Calling MOND a "no dark matter" theory is overstatement.
• Treating a_0 as a free parameter. A_0 is fitted once, from rotation curves, and then fixed forever. Every subsequent prediction across thousands of galaxies uses the same number. It is not retuned per galaxy.
• Confusing flat rotation curves with the entire problem. Flat rotation curves are necessary but not sufficient evidence for dark matter — they can be produced by an extended halo or by MOND. The cluster, lensing, Bullet Cluster, and CMB tests are what distinguish the two.
• Forgetting that the solar system is Newtonian. At a = GM_☉/r² for Earth's orbit, the acceleration is about 6 × 10⁻³ m/s² — fifty million times a_0. MOND in the solar system is indistinguishable from Newton, and lunar laser ranging and planetary ephemerides have set no constraint that disfavours it.
• Assuming TeVeS = MOND. TeVeS is one relativistic completion of MOND, not the only one. Its specific predictions for cosmology and lensing are not automatically inherited by every MOND-like theory.