Monthly Archives: December 2016

Emergent Gravity: Verlinde’s Proposal

In a previous blog entry I give some background around Erik Verlinde’s proposal for an emergent, thermodynamic basis of gravity. Gravity remains mysterious 100 years after Einstein’s introduction of general relativity – because it is so weak relative to the other main forces, and because there is no quantum mechanical description within general relativity, which is a classical theory.

On reason that it may be so weak is because it is not fundamental at all, that it represents a statistical, emergent phenomenon. There has been increasing research into the idea of emergent spacetime and emergent gravity and the most interesting proposal was recently introduced by Erik Verlinde at the University of Amsterdam in a paper “Emergent Gravity and the Dark Universe”.

A lot of work has been done assuming anti-de Sitter (AdS) spaces with negative cosmological constant Λ – just because it is easier to work under that assumption. This year, Verlinde extended this work from the unrealistic AdS model of the universe to a more realistic de Sitter (dS) model. Our runaway universe is approaching a dark energy dominated dS solution with a positive cosmological constant Λ.

The background assumption is that quantum entanglement dictates the structure of spacetime, and its entropy and information content. Quantum states of entangled particles are coherent, observing a property of one, say the spin orientation, tells you about the other particle’s attributes; this has been observed in long distance experiments, with separations exceeding 100 kilometers.

400px-SPDC_figure.pngIf space is defined by the connectivity of quantum entangled particles, then it becomes almost natural to consider gravity as an emergent statistical attribute of the spacetime. After all, we learned from general relativity that “matter tells space how to curve, curved space tells matter how to move” – John Wheeler.

What if entanglement tells space how to curve, and curved space tells matter how to move? What if gravity is due to the entropy of the entanglement? Actually, in Verlinde’s proposal, the entanglement entropy from particles is minor, it’s the entanglement of the vacuum state, of dark energy, that dominates, and by a very large factor.

One analogy is thermodynamics, which allows us to represent the bulk properties of the atmosphere that is nothing but a collection of a very large number of molecules and their micro-states. Verlinde posits that the information and entropy content of space are due to the excitations of the vacuum state, which is manifest as dark energy.

The connection between gravity and thermodynamics has been around for 3 decades, through research on black holes, and from string theory. Jacob Bekenstein and Stephen Hawking determined that a black hole possesses entropy proportional to its area divided by the gravitational constant G. String theory can derive the same formula for quantum entanglement in a vacuum. This is known as the AdS/CFT (conformal field theory) correspondence.

So in the AdS model, gravity is emergent and its strength, the acceleration at a surface, is determined by the mass density on that surface surrounding matter with mass M. This is just the inverse square law of Newton. In the more realistic dS model, the entropy in the volume, or bulk, must also be considered. (This is the Gibbs entropy relevant to excited states, not the Boltzmann entropy of a ground state configuration).

Newtonian dynamics and general relativity can be derived from the surface entropy alone, but do not reflect the volume contribution. The volume contribution adds an additional term to the equations, strengthening gravity over what is expected, and as a result, the existence of dark matter is ‘spoofed’. But there is no dark matter in this view, just stronger gravity than expected.

This is what the proponents of MOND have been saying all along. Mordehai Milgrom observed that galactic rotation curves go flat at a characteristic low acceleration scale of order 2 centimeters per second per year. MOND is phenomenological, it observes a trend in galaxy rotation curves, but it does not have a theoretical foundation.

Verlinde’s proposal is not MOND, but it provides a theoretical basis for behavior along the lines of what MOND states.

Now the volume in question turns out to be of order the Hubble volume, which is defined as c/H, where H is the Hubble parameter denoting the rate at which galaxies expand away from one another. Reminder: Hubble’s law is v = H \cdot d where v is the recession velocity and the d the distance between two galaxies. The lifetime of the universe is approximately 1/H.


The value of c / H is over 4 billion parsecs (one parsec is 3.26 light-years) so it is in galaxies, clusters of galaxies, and at the largest scales in the universe for which departures from general relativity (GR) would be expected.

Dark energy in the universe takes the form of a cosmological constant Λ, whose value is measured to be 1.2 \cdot 10^{-56} cm^{-2} . Hubble’s parameter is 2.2 \cdot 10^{-18} sec^{-1} . A characteristic acceleration is thus H²/ Λ or 4 \cdot 10^{-8}  cm per sec per sec (cm = centimeters, sec = second).

One can also define a cosmological acceleration scale simply by c \cdot H , the value for this is about 6 \cdot 10^{-8} cm per sec per sec (around 2 cm per sec per year), and is about 15 billion times weaker than Earth’s gravity at its surface! Note that the two estimates are quite similar.

This is no coincidence since we live in an approximately dS universe, with a measured  Λ ~ 0.7 when cast in terms of the critical density for the universe, assuming the canonical ΛCDM cosmology. That’s if there is actually dark matter responsible for 1/4 of the universe’s mass-energy density. Otherwise Λ could be close to 0.95 times the critical density. In a fully dS universe, \Lambda \cdot c^2 = 3 \cdot H^2 , so the two estimates should be equal to within sqrt(3) which is approximately the difference in the two estimates.

So from a string theoretic point of view, excitations of the dark energy field are fundamental. Matter particles are bound states of these excitations, particles move freely and have much lower entropy. Matter creation removes both energy and entropy from the dark energy medium. General relativity describes the response of area law entanglement of the vacuum to matter (but does not take into account volume entanglement).

Verlinde proposes that dark energy (Λ) and the accelerated expansion of the universe are due to the slow rate at which the emergent spacetime thermalizes. The time scale for the dynamics is 1/H and a distance scale of c/H is natural; we are measuring the time scale for thermalization when we measure H. High degeneracy and slow equilibration means the universe is not in a ground state, thus there should be a volume contribution to entropy.

When the surface mass density falls below c \cdot H / (8 \pi \cdot G) things change and Verlinde states the spacetime medium becomes elastic. The effective additional ‘dark’ gravity is proportional to the square root of the ordinary matter (baryon) density and also to the square root of the characteristic acceleration c \cdot H.

This dark gravity additional acceleration satisfies the equation g _D = sqrt  {(a_0 \cdot g_B / 6 )} , where g_B is the usual Newtonian acceleration due to baryons and a_0 = c \cdot H is the dark gravity characteristic acceleration. The total gravity is g = g_B + g_D . For large accelerations this reduces to the usual g_B and for very low accelerations it reduces to sqrt  {(a_0 \cdot g_B / 6 )} .

The value a_0/6 at 1 \cdot 10^{-8} cm per sec per sec derived from first principles by Verlinde is quite close to the MOND value of Milgrom, determined from galactic rotation curve observations, of 1.2 \cdot 10^{-8} cm per sec per sec.

So suppose we are in a region where g_B is only 1 \cdot 10^{-8} cm per sec per sec. Then g_D takes the same value and the gravity is just double what is expected. Since orbital velocities go as the square of the acceleration then the orbital velocity is observed to be sqrt(2) higher than expected.

In terms of gravitational potential, the usual Newtonian potential goes as 1/r, resulting in a 1/r^2 force law, whereas for very low accelerations the potential now goes as log(r) and the resultant force law is 1/r. We emphasize that while the appearance of dark matter is spoofed, there is no dark matter in this scenario, the reality is additional dark gravity due to the volume contribution to the entropy (that is displaced by ordinary baryonic matter).


Flat to rising rotation curve for the galaxy M33

Dark matter was first proposed by Swiss astronomer Fritz Zwicky when he observed the Coma Cluster and the high velocity dispersions of the constituent galaxies. He suggested the term dark matter (“dunkle materie”). Harold Babcock in 1937 measured the rotation curve for the Andromeda galaxy and it turned out to be flat, also suggestive of dark matter (or dark gravity). Decades later, in the 1970s and 1980s, Vera Rubin (just recently passed away) and others mapped many rotation curves for galaxies and saw the same behavior. She herself preferred the idea of a deviation from general relativity over an explanation based on exotic dark matter particles. One needs about 5 times more matter, or about 5 times more gravity to explain these curves.

Verlinde is also able to derive the Tully-Fisher relation by modeling the entropy displacement of a dS space. The Tully-Fisher relation is the strong observed correlation between galaxy luminosity and angular velocity (or emission line width) for spiral galaxies, L \propto v^4 .  With Newtonian gravity one would expect M \propto v^2 . And since luminosity is essentially proportional to ordinary matter in a galaxy, there is a clear deviation by a ratio of v².


 Apparent distribution of spoofed dark matter,  for a given ordinary (baryonic) matter distribution

When one moves to the scale of clusters of galaxies, MOND is only partially successful, explaining a portion, coming up shy a factor of 2, but not explaining all of the apparent mass discrepancy. Verlinde’s emergent gravity does better. By modeling a general mass distribution he can gain a factor of 2 to 3 relative to MOND and basically it appears that he can explain the velocity distribution of galaxies in rich clusters without the need to resort to any dark matter whatsoever.

And, impressively, he is able to calculate what the apparent dark matter ratio should be in the universe as a whole. The value is \Omega_D^2 = (4/3) \Omega_B where \Omega_D is the apparent mass-energy fraction in dark matter and \Omega_B is the actual baryon mass density fraction. Both are expressed normalized to the critical density determined from the square of the Hubble parameter, 8 \pi G \rho_c = 3 H^2 .

Plugging in the observed \Omega_B \approx 0.05 one obtains \Omega_D \approx 0.26 , very close to the observed value from the cosmic microwave background observations. The Planck satellite results have the proportions for dark energy, dark matter, ordinary matter as .68, .27, and .05 respectively, assuming the canonical ΛCDM cosmology.

The main approximations Verlinde makes are a fully dS universe and an isolated, static (bound) system with a spherical geometry. He also does not address the issue of galaxy formation from the primordial density perturbations. At first guess, the fact that he can get the right universal \Omega_D suggests this may not be a great problem, but it requires study in detail.

Breaking News!

Margot Brouwer and co-researchers have just published a test of Verlinde’s emergent gravity with gravitational lensing. Using a sample of over 33,000 galaxies they find that general relativity and emergent gravity can provide an equally statistically good description of the observed weak gravitational lensing. However, emergent gravity does it with essentially no free parameters and thus is a more economical model.

“The observed phenomena that are currently attributed to dark matter are the consequence of the emergent nature of gravity and are caused by an elastic response due to the volume law contribution to the entanglement entropy in our universe.” – Erik Verlinde


Erik Verlinde 2011 “On the Origin of Gravity and the Laws of Newton” arXiv:1001.0785

Stephen Perrenod, 2013, 2nd edition, “Dark Matter, Dark Energy, Dark Gravity” Amazon, provides the traditional view with ΛCDM  (read Dark Matter chapter with skepticism!)

Erik Verlinde 2016 “Emergent Gravity and the Dark Universe arXiv:1611.02269v1

Margot Brouwer et al. 2016 “First test of Verlinde’s theory of Emergent Gravity using Weak Gravitational Lensing Measurements” arXiv:1612.03034v


The Curiously Tangential Dwarf Galaxies

There are some 50 or so satellite galaxies around the Milky Way, the most famous of which are the Magellanic Clouds. Somewhat incredibly, half of these have been discovered within the last 2 years, since they are small, faint, and have low surface brightness. The image below shows only the well known ‘classical’ satellites. The satellites are categorized primarily as dwarf spheroidals, and most are low in gas content.


Image credit: Wikipedia, Richard Powell, Creative Commons Attribution-Share Alike 2.5 Generic

“Satellite galaxies that orbit from 1,000 ly (310 pc) of the edge of the disc of the Milky Way Galaxy to the edge of the dark matter halo of the Milky Way at 980×103 ly (300 kpc) from the center of the galaxy, are generally depleted in hydrogen gas compared to those that orbit more distantly. The reason is the dense hot gas halo of the Milky Way, which strips cold gas from the satellites. Satellites beyond that region still retain copious quantities of gas.” – Wikipedia article

In a recent paper “The tangential velocity excess of the Milky Way satellites“, Marius Cautun and Carlos Frenk find that a sample of satellites (drawn from those known for more than a few years) deviates from the predictions of the canonical Λ – Cold Dark Matter (ΛCDMcosmology. (Λ refers to the cosmological constant, or dark energy).

“We estimate the systemic orbital kinematics of the Milky Way classical satellites and compare them with predictions from the Λ cold dark matter (ΛCDM) model derived from a semi-analytical galaxy formation model applied to high resolution cosmological N-body simulations. We find that the Galactic satellite system is atypical of ΛCDM systems. The subset of 10 Galactic satellites with proper motion measurements has a velocity anisotropy, β = −2.2 ± 0.4, that lies in the 2.9% tail of the ΛCDM distribution. Individually, the Milky Way satellites have radial velocities that are lower than expected for their proper motions, with 9 out of the 10 having at most 20% of their orbital kinetic energy invested in radial motion. Such extreme values are expected in only 1.5% of ΛCDM satellites systems. This tangential motion excess is unrelated to the existence of a Galactic ‘disc of satellites’. We present theoretical predictions for larger satellite samples that may become available as more proper motion measurements are obtained.”

Radial velocities are easy, we get those from redshifts. Tangential velocities are much tougher, but can be obtained from relatively nearby objects by measuring their proper motions. That is, how much do their apparent positions change on the sky after many years have passed. It’s all the more tough when your object is not a point object, but a fuzzy galaxy!

For a ‘random’ distribution of velocities in accordance with ΛCDM cosmology, one would expect the two components of tangential velocity to be each roughly equal on average to the radial component, and thus 2/3 of the kinetic energy would be tangential and 1/3 would be radial. But rather than 33% of the kinetic energy being in radial motion, they find that the Galactic satellites have only about 1/2 that amount in radial, and over 80% of their kinetic energy in tangential motion.

Formally, they find a negative velocity anistropy, β, which as it is defined in practice, should be around zero for a ΛCDM distribution. They find that β differs from zero by 5 standard deviations.

One possible explanation is that the dwarf galaxies are mainly at their perigee or apogee points of their orbits. But why should this be the case? Another explanation: “alternatively indicate that the Galactic satellites have orbits that are, on average, closer to circular than is typical in ΛCDM. This would mean that MW halo mass estimates based on satellite orbits (e.g. Barber et al. 2014) are biased low.” Perhaps the Milky Way halo mass estimate is too low. Or, they also posit, without elaborating, do the excess tangential motions “indicate new physics in the dark sector”?

So one speculation is that the tangential motions are reflective of emergent gravity class of theories, for which dark matter is not required, but for which the gravitational force changes (strengthens) at low accelerations, of order c \cdot H, where H is the Hubble parameter, and the value works out to be around 2 centimeters per second per year. And it does this in a way that ‘spoofs’ the existence and gravitational affect of dark matter. This is also what is argued for in Modified Newtonian Dynamics, which is an empirical observation about galaxy light curves.

In the next article of this series we will look at Erik Verlinde’s emergent gravity proposal, which he has just enhanced, and will attempt to explain it as best we can. If you want to prepare yourself for this challenging adventure, first read his 2011 paper, “On the Origin of Gravity and the Laws of Newton”.

Modified Newtonian Dynamics – Is there something to it?

You are constantly accelerating. The Earth’s gravity is pulling you downward at g = 9.8 meters per second per second. It wants to take your velocity up to about 10 meters per second after only the first second of free fall. Normally you don’t fall, because the floor is solid due to electromagnetic forces and also it is electromagnetic forces that give your body structural integrity and power your muscles, resisting the pull of gravity.

You are also accelerating due to the Earth’s spin and its revolution about the Sun.


International Space Station, image credit: NASA

Our understanding of gravity comes primarily from these large accelerations, such as the Earth’s pull on ourselves and on satellites, the revolution of the Moon about the Earth, and the planetary orbits about the Sun. We also are able to measure the solar system’s velocity of revolution about the galactic center, but with much lower resolution, since the timescale is of order 1/4 billion years for a single revolution with an orbital radius of about 25,000 light-years!

It becomes more difficult to determine if Newtonian dynamics and general relativity still hold for very low accelerations, or at very large distance scales such as the Sun’s orbit about the galactic center and beyond.

Modified Newtonian Dynamics (MOND) was first proposed by Mordehai Milgrom in the early 1980s as an alternative explanation for flat galaxy rotation curves, which are normally attributed to dark matter. At that time the best evidence for dark matter came from spiral galaxy rotation curves, although the need for dark matter (or some deviation from Newton’s laws) was originally seen by Fritz Zwicky in the 1930s while studying clusters of galaxies.


NGC 3521. Image Credit: ESA/Hubble & NASA and S. Smartt (Queen’s University Belfast); Acknowledgement: Robert Gendler 


Galaxy Rotation Curve for M33. Public Domain, By Stefania.deluca – Own work,

If general relativity is always correct, and Newton’s laws of gravity are correct for non-relativistic, weak gravity conditions, then one expects the orbital velocities of stars in the outer reaches of galaxies to drop in concert with the fall in light from stars and/or radio emission from interstellar gas, reflecting decreasing baryonic matter density. (Baryonic matter is ordinary matter, dominated by protons and neutrons). As seen in the image above for M33, the orbital velocity does not drop, it continues to rise well past the visible edge of the galaxy.

To first order, assuming a roughly spherical distribution of matter, the square of the velocity at a given distance from the center is proportional to the mass interior to that distance divided by the distance (signifying the gravitational potential), thus

   v² ~ G M / r

where G is the gravitational constant, and M is the galactic mass within a spherical volume of radius r. This potential corresponds to the familiar 1/r² dependence of the force of gravity according to Newton’s laws.  In other words, at the outer edge of a galaxy the velocity of stars should fall as the square root of the increasing distance, for Newtonian dynamics.

Instead, for the vast majority of galaxies studied, it doesn’t – it flattens out, or falls off very slowly with increasing distance, or even continues to rise, as for M33 above. The behavior is roughly as if gravity followed an inverse distance law for the force (1/r) in the outer regions, rather than an inverse square law with distance (1/r²).

So either there is more matter at large distances from galactic centers than expected from the light distribution, or the gravitational law is modified somehow such that gravity is stronger than expected. If there is more matter, it gives off little or no light, and is called unseen, or dark, matter.

It must be emphasized that MOND is completely empirical and phenomenological. It is curve fitted to the existing rotational curves, rather successfully, but not based on a theoretical construct for gravity. It has a free parameter for weak acceleration, and for very small accelerations, gravity is stronger than expected. It turns out that this free parameter, a_0 , is of the same order as the ‘Hubble acceleration’ c \cdot H. (The Hubble distance is c / H and is 14 billion light-years; H has units of inverse time and the age of the universe is 1/H to within a few percent).

The Hubble acceleration is approximately .7 nanometers / sec / sec or 2 centimeters / sec / year  (a nanometer is a billionth of a meter, sec = second).

Milgrom’s fit to rotation curves found a best fit at .12 nanometers/sec/sec, or about 1/6 of a_0 . This is very small as compared to the Earth’s gravity, for example. It’s the ratio between 80 years and one second, or about 2.5 billion. So you can imagine how such a variation could have escaped detection for a long time, and would require measurements at the extragalactic scale.

The TeVeS – tensor, vector, scalar theory is a theoretical construct that modifies gravity from general relativity. General relativity is a tensor theory that reduces to Newtonian dynamics for weak gravity. TeVeS has more free parameters than general relativity, but can be constructed in a way that will reproduce galaxy rotation curves and MOND-like behavior.

But MOND, and by implication, TeVeS, have a problem. They work well, surprisingly well, at the galactic scale, but come up short for galaxy clusters and for the very largest extragalactic scales as reflected in the spatial density perturbations of the cosmic microwave background radiation. So MOND as formulated doesn’t actually fully eliminate the requirement for dark matter.


Horseshoe shaped Einstein Ring

Image credit: ESA/Hubble and NASA

Any alternative to general relativity also must explain gravitational lensing, for which there are a large number of examples. Typically a background galaxy image is distorted and magnified as its light passes through a galaxy cluster, due to the large gravity of the cluster. MOND proponents do claim to reproduce gravitational lensing in a suitable manner.

Our conclusion about MOND is that it raises interesting questions about gravity at large scales and very low accelerations, but it does not eliminate the requirement for dark matter. It is also very ad hoc. TeVeS gravity is less ad hoc, but still fails to reproduce the observations at the scale of galaxy clusters and above.

Nevertheless the rotational curves of spirals and irregulars are correlated with the visible mass only, which is somewhat strange if there really is dark matter dominating the dynamics. Dark matter models for galaxies depend on dark matter being distributed more broadly than ordinary, baryonic, matter.

In the third article of this series we will take a look at Erik Verlinde’s emergent gravity concept, which can reproduce the Tully-Fisher relation and galaxy rotation curves. It also differs from MOND both in terms of being a theory, although incomplete, rather than empiricism, and apparently in being able to more successfully address the dark matter issues at the scale of galaxy clusters.


Wikipedia MOND entry:

M. Milgrom 2013, “Testing the MOND Paradigm of Modified Dynamics with Galaxy-Galaxy Gravitational Lensing”

R. Reyes et al. 2010, “Confirmation of general relativity on large scales from weak lensing and galaxy velocities”

“In rotating galaxies, distribution of normal matter precisely determines gravitational acceleration”