[astro-ph/0511628] Recent Supernovae Ia observations tend to rule out all the cosmologies

Authors:  R. G. Vishwakarma (Zacatecas University)
Abstract:  Dark energy and the accelerated expansion of the universe have been the direct predictions of the distant supernovae Ia observations which are also supported, indirectly, by the observations of the CMB anisotropies, gravitational lensing and the studies of galaxy clusters. Today these results are accommodated in what has become the `concordance cosmology': a universe with flat spatial sections t=constant with about 70% of its energy in the form of Einstein's cosmological constant \Lambda.
However, we find that as more and more supernovae Ia are observed, more accurately and towards higher redshift, the probability that the data are well explained by the cosmological models decreases alarmingly, finally ruling out the concordance model at more than 95% confidence level. This raises doubts against the `standard candle'-hypothesis of the supernovae Ia and their use to constrain the cosmological models.
[PDF]  [PS]  [BibTex]  [Bookmark]

Discussion related to specific recent arXiv papers
Garth Antony Barber
Posts: 59
Joined: July 19 2005
Affiliation: Published independent

[astro-ph/0511628] Recent Supernovae Ia observations tend to

Post by Garth Antony Barber » November 22 2005

Comments on this paper?

If the analysis stands up, then is it time for either a new 'standard candle', 'epicycle - to save the appearances', or paradigm?

Ben Gold
Posts: 81
Joined: September 25 2004
Affiliation: University of Minnesota
Contact:

[astro-ph/0511628] Recent Supernovae Ia observations tend to

Post by Ben Gold » November 22 2005

I don't really think it's the cosmology. For example, I've played with the Riess Gold+Silver sample a bit, and the goodness-of-fit you get in the end can depend on whether you bin the data or not and how you bin it. Relative likelihoods (i.e. the shape of the contours in parameter space) don't seem to change so much but the \chi^2/dof you get for the "best-fit" model certainly does. I can get P = 26-28% for the binned case whereas the unbinned gives only around 2%.

I take this as meaning that the errorbars on the individual supernovae are not quite right; there's some source of error that's still unmodeled. That the SNLS group used their data to fit for \sigma_{int} seems to be admitting as much; supernovae seem to be \emph{standardizable} but the physics isn't all there yet. This isn't really news, though, it's exactly why SNAP for example proposes observing ~10000 supernovae but only using 2000 for cosmology. The rest are all used for calibration.

Also I note that Vishwakarma never considers cosmologies that are both non-flat and have [tex]w\neq -1[/tex]; these can fit the supernovae data slightly better (although it's not clear that all such models are very realistic).

Syksy Rasanen
Posts: 119
Joined: March 02 2005
Affiliation: University of Helsinki

[astro-ph/0511628] Recent Supernovae Ia observations tend to

Post by Syksy Rasanen » December 02 2005

Regarding the intrinsic dispersion of the SN absolute magnitude [tex]\sigma_{int}[/tex], the author makes the strong claim that since the SNLS team keeps this as a free parameter, they can get a good fit for any model, even Omega_m=1. Is this criticism correct? (The implication would be that the SNLS analysis (in astro-ph/0510447) is useless for constraining alternatives to [tex]\Lambda[/tex]CDM.)

D. V. Ahluwalia-Khalilova
Posts: 4
Joined: May 28 2005
Affiliation: ASGBG/CIU University of Zacatecas
Contact:

Re: [astro-ph/0511628] Recent Supernovae Ia observations ten

Post by D. V. Ahluwalia-Khalilova » March 11 2006

Garth Antony Barber wrote:Comments on this paper?

If the analysis stands up, then is it time for either a new 'standard candle', 'epicycle - to save the appearances', or paradigm?
The problem of this paper is that it entirely neglects the dispersion in luminosities of SN 1a data. Granted, that ideally the dispersion should be an input rather than infered ... but that is what was done in earlier works, and only the Astier et al. work is a deviation from this approach. But including the SN 1a luminosity dispersion in some manner (even if not in an ideal manner, which at present may be difficult) is utterly important. The claim made by Vishwakarma is so strong, and the analysis he presents so flawed, that it should be entirely ignored except for emphasing that we need more data on SN 1a luminosity disperion.

Moncy Vilavinal John
Posts: 3
Joined: March 21 2006
Affiliation: St. Thomas College, Kozhencherry, Kerala, India

[astro-ph/0511628] Recent Supernovae Ia observations tend to

Post by Moncy Vilavinal John » March 21 2006

Astier et. al. and also Alhuwalia-Khalilova think that it is natural to evaluate σint (an unknown contribution to the errorbars, which is termed as intrinsic dispersion of SN absolute magnitudes) by requiring beforehand a reduced χ2
=1. But has such a procedure of evaluating the errorbars by assuming χ2
=1 been performed in any other experiment in physics? Can somebody please enlighten me on this?

Since it does not seem to be a common practice, I feel that the authors ought to have highlighted its novelty. After all, these are considered to be the most credible data, which also have revolutionized the present understanding on the cosmos.

It deserves to be mentioned that intrinsic dispersion is inferred in all the previous SNe data and not just in Astier et. al., contrary to what Alhuwalia-Khalilova claims.

Mike Hudson
Posts: 5
Joined: September 09 2005
Affiliation: U Waterloo
Contact:

Re: [astro-ph/0511628] Recent Supernovae Ia observations ten

Post by Mike Hudson » March 30 2006

Moncy Vilavinal John wrote: But has such a procedure of evaluating the errorbars by assuming χ2 =1 been performed in any other experiment in physics? Can somebody please enlighten me on this?
This is not uncommon in peculiar velocity work. For example, see the VELMOD method of Willick and Strauss et al 97, ApJ, 486, 629. There they put all parameters, including the Tully-Fisher scatter, into a likelihood expression (rather than a [tex]\chi^2[/tex]). This is really the proper way to do this, because it simultaneously gives errors on [tex]\sigma_{int}[/tex] and the cosmological parameters. Ideally Astier et al would have done it this way as well. However, Astier et al's analysis is not invalid.

R G Vishwakarma
Posts: 4
Joined: November 13 2006
Affiliation: Unidad Academica de Matematicas, Universidad Autonoma de Zacatecas, Zacatecas, Mexico

Re: [astro-ph/0511628] Recent Supernovae Ia observations ten

Post by R G Vishwakarma » November 16 2006

Garth Antony Barber wrote:Comments on this paper?

If the analysis stands up, then is it time for either a new 'standard candle', 'epicycle - to save the appearances', or paradigm?

I have been informed by a friend about the discussion on my paper at this cite. Hereby, I would like to register my humble replies and clarifications in the following.

The analysis done by me in this paper does not employ anything new. It is
the standard methodology followed by everybody and should be considered seriously. My results are also supported by some recent observations which go against the `standard candle' hypothesis of SNe of type Ia, I mention the following two:

(i) Observation of SNLS-03D3bb, which is a recently observed high redshift (z=0.244) type Ia SN with extreme unusual features and no
obvious analogue at low redshifts. It does not obey the usual lightcurve
shape-luminosity relationship for SNe Ia that allows them to be calibrated
as standard candles [astro-ph/0609616].


(ii) Observations of two supernova remnants DEM L238 and DEM L249 made with the Chandra and XMM-Newton X-ray satellites in the Large Magellanic Cloud (astro-ph/0608297). While the presence of Fe-rich gas at the centres of these objects suggests that they are remnants of type Ia supernova explosions, the standard model of type Ia supernova remnants cannot explain the presence of relatively dense supernova ejecta with long
ionization timescales.

The emerging picture is also supported by Middleditch (astro-ph/0608386)
who argues that SNe Ia seem to be affected by some systematic effects which alone, without invoking any dark energy, could make them too
faint for their redshifts. It is argued that it may be impossible to get a
clean sample of SNe Ia which are free from this kind of effects (MNRAS 2004, 348, 261; A&A 2005, 431, 757).

R G Vishwakarma
Posts: 4
Joined: November 13 2006
Affiliation: Unidad Academica de Matematicas, Universidad Autonoma de Zacatecas, Zacatecas, Mexico

Re: [astro-ph/0511628] Recent Supernovae Ia observations ten

Post by R G Vishwakarma » November 16 2006

Ben Gold wrote:I don't really think it's the cosmology. For example, I've played with the Riess Gold+Silver sample a bit, and the goodness-of-fit you get in the end can depend on whether you bin the data or not and how you bin it. Relative likelihoods (i.e. the shape of the contours in parameter space) don't seem to change so much but the \chi^2/dof you get for the "best-fit" model certainly does. I can get P = 26-28% for the binned case whereas the unbinned gives only around 2%.

Also I note that Vishwakarma never considers cosmologies that are both non-flat and have [tex]w\neq -1[/tex]; these can fit the supernovae data slightly better (although it's not clear that all such models are very realistic).
It is already well known that the information drawn from a binned data depends on how the bins are selected, and also to the weights chosen, if any. This happens because quite some information is lost in the process of binning. For example, there are many SNe at z~1 and beyond shown in the top panel of figure 7 of astro-ph/0402512 of Riess et al (two out of these SNe at z~1.3 have been observed quite precisely with significantly small error bars), which do not look consistent with a later epoch of deceleration. However, these features are completely lost in the binned data shown in the bottom panel of the same figure. Thus preference should be given to fit the full sample rather than the binned ones, if possible. It should be noted that in the fittings, I have considered individual SNe from the full samples, without any binning.

It should also be noted that I did consider non-flat models and with variable w in my calculations, but the fit does not improve significantly, apart from the fact that they are not consistent with the CMB measurements. So, I did not mention them in order to avoid clumsiness.

R G Vishwakarma
Posts: 4
Joined: November 13 2006
Affiliation: Unidad Academica de Matematicas, Universidad Autonoma de Zacatecas, Zacatecas, Mexico

Re: [astro-ph/0511628] Recent Supernovae Ia observations ten

Post by R G Vishwakarma » November 16 2006

Ben Gold, Syksy Rasanen, D. V. Ahluwalia-Khalilova, Mike Hudson wrote: Ben Gold:
I take this as meaning that the errorbars on the individual supernovae are not quite right; there's some source of error that's still unmodeled. That the SNLS group used their data to fit for σint seems to be admitting as much; supernovae seem to be standardizable but the physics isn't all there yet...

Syksy Rasanen:
Regarding the intrinsic dispersion of the SN absolute magnitude [tex]\sigma_{int}[/tex], the author makes the strong claim that since the SNLS team keeps this as a free parameter, they can get a good fit for any model, even Omega_m=1. Is this criticism correct? (The implication would be that the SNLS analysis (in astro-ph/0510447) is useless for constraining alternatives to [tex]\Lambda[/tex]CDM.)

D. V. Ahluwalia-Khalilova:
But including the SN 1a luminosity dispersion in some manner (even if not in an ideal manner, which at present may be difficult) is utterly important.


Mike Hudson:
This is not uncommon in peculiar velocity work. For example, see the VELMOD method of Willick and Strauss et al 97, ApJ, 486, 629. There they put all parameters, including the Tully-Fisher scatter, into a likelihood expression (rather than a χ2). This is really the proper way to do this, because it simultaneously gives errors on σint and the cosmological parameters. Ideally Astier et al would have done it this way as well. However, Astier et al's analysis is not invalid.

The analysis followed by Astier et al assumes that the theory (LCDM) explains the data well (by requiring \chi^2=1) and just goes on estimating the parameters of the model. This is fine if we have belief in the theory tested by other independent ways, and this procedure is followed in many problems of physics.

But, what are the other observations which predict dark energy independently and conclusively? The only other precise observations are the anisotropy measurements of CMB made by the WMAP projects. However, taken at their face values, the only apparent prediction of the WMAP observations is a flat geometry, and the decelerating models like the EdS (\Omega=1) also explain them successfully (see for example, Blanchard, astro-ph/0502220). Observations on gravitational lensing, quasars, galaxy clusters, etc are not precise enough. Age considerations depend on the observations of H_0 and \Omega_m which have wide degeneracy. Moreover, many of them come from the SNe Ia observations themselves. (There are also some observations which give significantly low H_0, rendering the EdS model worth considering.) In fact, a conclusive prediction of dark energy comes from the SNe Ia only. Moreover, there are also claims against the LCDM model from time to time, though they are ignored. One can find a recent one in astro-ph/0611355, for example. So, if we do not have a conclusive independent evidence for dark energy, what is the meaning of the parameters estimated by Astier et al, in the absence of a credible goodness-of-fit probability?

It should be noted that the absolute magnitude is measured from the low-z SNe. Hence the intrinsic dispersion of the absolute magnitude can be estimated from the scatter of the Low-z SNe. In fact all the other observations, made before SNLS, already include this term in their error bars estimated by reasonable ways.

R G Vishwakarma
Posts: 4
Joined: November 13 2006
Affiliation: Unidad Academica de Matematicas, Universidad Autonoma de Zacatecas, Zacatecas, Mexico

Re: [astro-ph/0511628] Recent Supernovae Ia observations ten

Post by R G Vishwakarma » November 16 2006

D. V. Ahluwalia-Khalilova wrote: The problem of this paper is that it entirely neglects the dispersion in luminosities of SN 1a data. Granted, that ideally the dispersion should be an input rather than infered ... but that is what was done in earlier works, and only the Astier et al. work is a deviation from this approach. But including the SN 1a luminosity dispersion in some manner (even if not in an ideal manner, which at present may be difficult) is utterly important. The claim made by Vishwakarma is so strong, and the analysis he presents so flawed, that it should be entirely ignored except for emphasing that we need more data on SN 1a luminosity disperion.

One should not write comments just for writing comments. Also, one must read the papers carefully before commenting. It must be noted that my claims (that the recent observations seem to rule out all the cosmologies at fairly high confidence levels....) are deduced from those observations only which, unlike the SNLS data, have already included the dispersion in the SN absolute magnitude (estimated by reasonable methods) in their error bars. I have considered Astier et al work for comments and discussion only.

The analysis I have done is the standard one used by everybody. However, if somebody finds some FLAW in it, it should be brought before the community rather than declaring a FATWA against it.

Similar conclusions have also been drawn in one of the papers from the SN-Cosmology-Project group itself (A&A 2006, 447, 31) and also in (A&A 2005, 429, 807) from a smaller sample of data. However, this is more evident in my analysis from a bigger sample of data.

John Middleditch
Posts: 3
Joined: August 23 2006
Affiliation: Los Alamos National Laboratory

[astro-ph/0511628] Recent Supernovae Ia observations tend to

Post by John Middleditch » November 17 2006

Hi All,

Most of this dicussion is a bit beyond my expertise, so all I can contribute is
generalities on Ia's:

1. The strongest argument for Ia cosmology is the fact that the
dispersion is so small, notwithstanding the "outliers" which have
been occasionally excluded under the assumption that there's
some firm distribution of "ordinary" Ia's, and that there is no
continuous distribution for the "outliers".

2. The deduction that devastates #1 (and indicts the low dispersion) is
that Ia's are not single-degenerate, but are instead double-degenerate
(core-collapse) events. This means that, if we could resolve a Ia at the
level of SN 1987A, it would look like 87A, except that the thermonuclear
ball would be much brighter, and the bipolar ejection would consist of faster
moving cones. This is supported by several facts:

A. If Ia's were single degenerate, then there should be some H and He in their
spectra, but there is _absolutely_ no indication of that (2006, A&A, 443, 649 on
SN 2001el).

B. Their progenitors would have to be cataclysmic variables, a population
which is unsuited to produce progenitors of Ia's (2005, ApJ, 629, L85).

C. The degree of continuum and line polarization and magnitude of velocities
are both _inversely_ correlated to Ia absolute magnitude, a circumstance
which fits the modified image of 87A very well.

D. Core-collapse in Ia's is needed to produce the observed abundance of
Zinc, assuming that the polar cones of DD-induced events are the site of
the r-process, as seem likely (ignoring wild-eyed ideas such as fallback-
induced r-process in collapsars, etc. ).

E. SNLS-03D3bb (SN 2003fg) pretty much requires a DD mechanism.
Occam's Razor would then say that nearly all, if not all, of these have
to be DD-induced. But we know from 87A's neutrinos (and the likely 2 ms
pulsar remnant) that the DD mecanism results in core-collapse.

1.3 degenerate solar masses + 1.3 degenerate solar masses = 1.4 solar mass
neutron star
+ 1.2 solar masses
in ejecta

When Bob Kirshner gave a talk on Ia cosmology at Los Alamos on
2005, August 9, he had the picture of 87A as a screen-saver at the
end of the talk (I had a 3' by 3' poster ready in case he didn't!).
I then went through the argument about polarizations and
velocities, etc, using a flashlight to point out the geometry of 87A
as it would apply to Ia's, and his answer was: "Well, gee, we got the
right answer!" But have they really?

Garth Antony Barber
Posts: 59
Joined: July 19 2005
Affiliation: Published independent

[astro-ph/0511628] Recent Supernovae Ia observations tend to

Post by Garth Antony Barber » November 17 2006

Might there also be systematic cosmological effects, such as an evolving average metallicity, that affects Ia luminosities?

Garth

Tommy Anderberg
Posts: 47
Joined: November 24 2005
Affiliation: independent

Re: [astro-ph/0511628] Recent Supernovae Ia observations ten

Post by Tommy Anderberg » November 17 2006

John Middleditch wrote:SNLS-03D3bb (SN 2003fg) pretty much requires a DD mechanism.
Hi! I had a hunch you would say that. What do you think of Howell's high-spin suggestion?

Something else: I recently stumbled upon the following counterargument to your astro-ph/0608386, on Wikipedia of all paces:
his hypothesis that SNe Ia and Ic are the same objects viewed from different orientations fails to explain the dramatic difference between the sites and hosts of the two types of SNe. SNe Ic live on bright, star forming regions and are never seen in elliptical hosts (old stellar populations), whereas SNe Ia avoid the sites of active star formation and frequently occur in elliptical hosts.
Care to comment?

Tommy Anderberg
Posts: 47
Joined: November 24 2005
Affiliation: independent

Re: [astro-ph/0511628] Recent Supernovae Ia observations ten

Post by Tommy Anderberg » November 21 2006

Tommy Anderberg wrote:I recently stumbled upon the following counterargument to your astro-ph/0608386, on Wikipedia of all paces:
his hypothesis that SNe Ia and Ic are the same objects viewed from different orientations fails to explain the dramatic difference between the sites and hosts of the two types of SNe. SNe Ic live on bright, star forming regions and are never seen in elliptical hosts (old stellar populations), whereas SNe Ia avoid the sites of active star formation and frequently occur in elliptical hosts.
Speaking of the devil, Penn State issued this press release (yes, I know, I know...) today:
NGC 1316, a massive elliptical galaxy about 80 million light years way, has recently merged with a spiral galaxy. Mergers do indeed spawn supernovas by forcing the creation of new, massive stars, which quickly die and explode. Yet all four supernovas in NGC 1316 appear to be Type Ia, a variety previously not associated with galaxy mergers and massive star formation. Scientists are intrigued and are investigating whether the high supernova rate is a coincidence or a result of the merger.

John Middleditch
Posts: 3
Joined: August 23 2006
Affiliation: Los Alamos National Laboratory

[astro-ph/0511628] Recent Supernovae Ia observations tend to

Post by John Middleditch » November 22 2006

Apologies for not getting back to cosmocoffee before now -- I'm new
at this.

Yes indeed, in ellipticals the overlayer of a double-degenerate (DD) white dwarf-
white-dwarf (WD-WD) merger-induced core-collapse event is usually not sufficient to
hide the thermonuclear ball from near-merger-pole views. This is in line with Ia's in ellilpticals being less luminous, as noticed in at least three recent papers. Thus the
Ia/c equivalance is mostly a consideration for galaxies other than ellipticals. The
overlayers of CO-CO WD merger-induced SNe in ellipticals may frequently be SO
insufficient, that the core-collapse event produces no distantly observable SN
(see, e.g., Fynbo et al., astro-ph/0608313, re: SN-less GRBs).

As for spirals and other galaxies, consider that modest-mass common-envelope (CE)
WR stars occur frequently in that population, as discovered by DeMarco and others.

Now we have to consider how frequently do CE WR stars merge to produce
Type Ia SNe, vs SNe of 40 solar mass single WR stars. The rarity but short
lifetime of the latter vs the commonness and longer lifetime of the former would
argue that at worst, CE WR stars produce SNe just as frequently as
40 solar mass singles.

So the questions all this begs are:

1. Where ARE these SNe?
2. What do they look like?
3. How are they represented in SNe Ic?

The only possible answer is that we've been seing them as Ic's
all along, with the extreme velocities a result of LITTLE mass
ejected, rather than A LOT.

The DD paradigm is a very powerful, and almost always unavoidable mechanism
to invoke when considering the origin of SNe, particularly when 87A was
very likely due to this process. The ONLY known exception in recent
history is SN 1986J, which has 200 times the Crab nebula luminosity
at 15 GHz. If we are ever lucky enough to get a Ia as near as the SMC or
LMC or even closer, the neutrinos will literally choke our detectors.

As a corollary, the DD paradigm makes many, if not all, calculations of
SNe invalid, as these have been calibrated to DD SNe, so none have yet
dealt with the implied mixing and modified core-collapse.

The latest version of 0608386 will have the updated arguments in it, and,
as the report on my latest submission to ApJL s overdue, I may well put this up
tomorrow.

Guess I'll have to go nail Wikipedia as well.

Post Reply