We have in MOND a formula that has had repeated predictive successes. Many of these have been true *a priori* predictions, like the absolute nature of the baryonic Tully-Fisher relation, the large mass discrepancies evinced by low surface brightness galaxies, and the velocity dispersions of many individual dwarf Spheroidal galaxies like Crater 2. I don’t see how these can be an accident. But what we lack is an underlying theoretical basis for the observed MONDian phenomenology: Why does this happen?

One apparently promising idea is the emergent gravity hypothesis by Erik Verlinde. Gravity is not a fundamental force so much as a consequence of microscopic entanglement. This manifests on scales comparable to the Hubble horizon, in particular, with an acceleration of order the speed of light times the Hubble expansion rate. This is close to the acceleration scale of MOND.

An early test of emergent gravity was provided by weak gravitational lensing. It does remarkably well at predicting the observed lensing signal with no free parameters. This is promising – perhaps we finally have a physical basis for MOND. Indeed, as Milgrom points out, the equivalent success was already known in MOND.

Weak lensing occurs deep in the MOND regime, at very low accelerations far from the lensing galaxy. In that regard, the results of Brouwer et al. can be seen as an extension of the radial acceleration relation to much lower accelerations. In this limit, it is fair to treat galaxies as point masses – hence the similarity of the solid and dashed lines in the figure above.

For rotation curves, it is not fair to approximate galaxies as point masses. Rotation curves are observed in the midst of the stellar and gaseous mass distribution. One must take account of the detailed distribution of baryons to treat the problem properly. This is something MOND is very successful at.

Emergent gravity converges to the same limit as MOND in the point mass case, which holds for any mass distribution once you get far enough away. It is not identical for finite mass distributions. When one solves the equation of emergent gravity for a finite mass distribution, you get one term that looks like MOND, which gives the success noted above. But you also get an additional term that depends on the gradient of the mass distribution, dM/dr.

The additional term that emerges for extended mass distributions in emergent gravity lead to different predictions than MOND. This is good, in that it makes the theories distinguishable. This is bad, in that MOND already provides good fits to rotation curves. Additional terms are likely to mess that up.

And so it does. Two independent studies recently come to this conclusion: one including myself (Lelli et al. 2017) and another by Hees et al. (2017). The studies differ in their approach. We show that the additional term in emergent gravity leads to the prediction of a larger mass discrepancy than in MOND, driving one to abnormally low stellar mass-to-light ratios and degrading the empirical radial acceleration relation. Hees et al. make detailed rotation curve fits, showing that the dM/dr term over-amplifies bumps & wiggles in the rotation curve. It has always been intriguing that MOND gets these right: this is a non-trivial success to reproduce.

The situation looks bad for emergent gravity. One caveat is that at present we only have solutions for emergent gravity in the case of a spherical cow. Conceivably a better treatment of the geometry would change the result, but it won’t eliminate the dM/dr term. So this seems unlikely to help with the fundamental problem: this term needs not to exist.

Perhaps emergent gravity is a clue to what ultimately is going on – a single step in the right direction. Or perhaps the similarity to MOND is misleading. For now, the search for a satisfactory explanation for the observed phenomenology continues.