A deviation from General Relativity on cosmic scales...

[Orion1]

We consider evidence for deviations from GR on cosmic scales in the growth of large scale structure using ISW, galaxy distribution and weak lensing shear correlations
and a two parameter model: \gamma describing CDM growth, and \eta describing how the two Newtonian potentials are related. We find the COSMOS weak lensing data shows a significant preference for a deviation from GR, with \eta < 1 at the 98% level at 1 < z < 2. The COSMOS data gives the first indication that dark energy might be a modification to GR, rather than \Lambda. It will be extremely interesting to see if this signature is seen in other lensing datasets, such as from the CFHTLS survey data [39], and those to be obtained with upcoming weak lensing surveys from the Dark Energy Camera (DES) and the Large Synoptic Survey Telescope (LSST).
The weak lensing technique can also be used to measure two different effects of gravity. General relativity calls for gravity's curvature of space to be equivalent to its curvature of time. Light should be influenced in equal amounts by both.

When the COSMOS data was released in 2007, the team - led by Massey - assumed these two factors were equivalent. Their analysis revealed that gravitational tugs on light were stronger than anticipated, but they put this down to a slightly higher concentration of ordinary and dark matter in the survey's patch of sky than had been predicted.

To look for potential deviations from general relativity, Bean reanalysed the data and dropped the requirement that these two components of gravity had to be equal. Instead the ratio of the two was allowed to change in value. She found that between 8 and 11 billion years ago gravity's distortion of time appeared to be three times as strong as its ability to curve space. An observer around at the time wouldn't have noticed the effect because it only applies over large distances. Nonetheless, "there is a preference for a significant deviation from general relativity", says Bean

A deviation from General Relativity?

Reference:
A weak lensing detection of a deviation from General Relativity on cosmic scales (http://arxiv.org/pdf/0909.3853v2)
New Scientist - Rethinking relativity: Is time out of joint? (http://www.newscientist.com/article/mg20427314.400-rethinking-relativity-is-time-out-of-joint.html)
Six Puzzles for LCDM Cosmology (http://arxiv.org/pdf/0811.4684v2)

[LithiumHelios]

Some discussion of this over on Cosmic Variance (http://blogs.discovermagazine.com/cosmicvariance/2009/10/12/a-new-challenge-to-einstein/) (Sean Carroll).

[Nabeshin]

I know she revised her initial estimate of confidence from something like 99.99% to what it stands right now at 98%. It will be interesting to see if she cleans things up once again, which is my suspicion.

[Garth]

The result is interesting, and it may be pertinent to mention the value eta = 1/3, comes naturally from the 2002 version of Self Creation Cosmology.

i.e. when the Robertson parameters alpha = 1 and gamma = 1/3.

http://arxiv.org/abs/gr-qc/0302088 (eq. 60).

It would seem though that is not the case for z < 1.

Garth

[Wallace]

Yes, the strangest part of this result, if one was to accept it at face value, is that it implies that the law of gravity itself changes over time, in a rather significant way. Given that the COSMOS team have indicated the photometric redshift calibrations for high redshift are probably a bit dodgy (they are being improved with a follow up spectroscopic survey z-COSMOS) I'd be very wary of giving too much weight to this result yet, since the signal of the deviation only occurs at the very highest redshifts, were the photometric errors will be greatest.

The z-COSMOS team have stated that they are doing a similiar analysis using the updated redshifts, so when that is published the picture should become clearer.

[atyy]

Given that the COSMOS team have indicated the photometric redshift calibrations for high redshift are probably a bit dodgy (they are being improved with a follow up spectroscopic survey z-COSMOS) I'd be very wary of giving too much weight to this result yet, since the signal of the deviation only occurs at the very highest redshifts, were the photometric errors will be greatest.

Wouldn't the observational error estimates take care of that?

[Wallace]

In principle, but error estimation is astronomy/cosmology is often quite tricky. It's those pesky unknown unknowns that are the killer. Quite often you will see results presented with a statistical error but no systematic error, because it is simply unknown. For instance, Hubble's original 'Hubbles constant' was orders of magnitude away from the current best value, yet the 'error' on the value in the original publication was much smaller than this difference, because Hubble didn't realise (not that he could have!) that he had an enormous systematic error lurking in the Cephied period-luminosity calibration.

In this particular case, the COSMOS team have stated that there are problems with the high-z photo-z's originally published in terms of an un-moddelled systematic. Photo-z's are a bit of a nightmare, the term 'catastrophic error' is a technical term in that field. The z-COSMOS follow up survey will significantly reduce this systematic.