Any complete standardized SNIa data out there?

An unrefereed paper astro-ph/0406437 appeared yesterday in which Ari
Brynjolfsson interprets SNIa data using a "plasma redshift" approach
which rejects time dilation as a redshift interpretation. I am not a
plasma man but have myself been considering some of the points raised
in that paper. This harkens back to a Bruno Leibundgut paper (ARAA
2001, 39,67) in which he addresses the problems with high-z SN
observations:
"Another discrepancy emerging is the color of the distant objects.
.... the distant SNe Ia appear clearly bluer than the nearby objects.
..... A striking discrepancy is that none of the slowly declining and,
hence, very luminous objects observed in the nearby sample have been
discovered at large distances. ... This is clearly contrary to what is
expected from a Malmquist bias ..."

In a nutshell, the problem is that SNIa are modelled to occur at
varying absolute brightness, characterized by the brightest having
brighter absolute magnitude (obviously), bluer colors and broader
light curve, ie, slower increase and decline in brightness.
Therefore, at very great distance (high z), one would expect a
preponderance of the brightest SN (due to Malmquist), thus the bluest
colors and the broadest light curves. As Leibundgut notes, the bluest
colors do occur, but the broadest light curves do not. The high-z SN
are characterized by blue colors but light curves which are *less than
average* compared with the low-z SN. How can this be?

The obvious departure point is the currently assumed
redshift-dependent light curve broadening. It is assumed, in all
today's projects, that the SN light curves are dilated by (1+z) so, as
an operational starting point, all the light curves are compressed
into the "rest frame" standard and analyzed from that point on. This
yields the Liebundgut conundrum.

Obviously a line of inquiry would be that the most distant SN are in
fact the intrinsically brightest (thus accounting for their blue
colors), that a Malmquist effect is allowing us to see only the
brightest, and that they thus must be visibly dimmer than currently
modelled. Brynjolfsson uses a "plasma red****" interpretation to
handle this situation, but my own interpretation is that a
Randall-Sundrum 5D geometry is diffusing light more broadly than
modelled by 3D-bound systems and that this accounts for both light
dimming at high-z and the old "numbers problem" that faint objects
seem over-numerous. This would produce the current observed
situation, of broad light curves and blue colors, but faint observed
brightness maxima, provided time dilation is discarded.

However, all our models are unimportant compared with the task of
reconciling our models to the available data. To that end what is
needed is a uniform table of SNIa data featuring redshift, blueness,
apparent magnitude and presumed stretch factor. The recent papers of
new SNIa data present their information very differently; it is
problematic to align these data into a consistent presentation. Does
anyone know of an on-line database which presents all these factors in
a consistent format?

My own premise is that the high-z SNIa data will be found to support a
Randall-Sundrum 5D geometry without any time dilation. I'm happy for
my premise to be shredded, er, disproven by the data, but first I need
that data. Does anyone have a pointer to uniform complete SNIa data
which includes absolute magnitude, blueness, and redshift?

thanks,
Eric

Eric Flesch
Wellington, New Zealand

(Eric Flesch) writes:

.... I'm happy for
my premise to be shredded, er, disproven by the data, but first I need
that data. Does anyone have a pointer to uniform complete SNIa data
which includes absolute magnitude, blueness, and redshift?

Ultra-short answer: no.

Longer answer: I think what you might hope to find is a collection
of slightly different information: the observed quantities.
It would be nice to have a table of:

- apparent magnitude(s) at maximum light, together
with passband(s) in which the magnitude(s) was measured
- estimate of redshift

If the table included at least two magnitudes at maximum light
in different passbands, you could calculate a color.

The redshifts may present some problems: in some cases, one can measure
the redshift of an unambiguous host galaxy and assign it to the supernova.
In other cases, however, there may be no visible host galaxy, or several
possible hosts, or the host may be too faint for a redshift measurement.
In some cases, astronomers estimate the redshift from a spectrum of
the supernova; you _might_ want to pick a sample of events for which
the redshift was determined in a uniform manner.

With this information, you could then compare the measurements to
your favorite models of a) supernova properties and b) cosmological
parameters. You asked in your message for a table of "absolute magnitudes"
of supernovae, but of course, those depend on the cosmology one adopts
to convert the apparent to absolute magnitudes.

I share your wish for such a compilation of data, but I wouldn't bet
on finding one any time soon :-(

Michael Richmond

(Michael Richmond) wrote in message ...
(Eric Flesch) writes:

.... I'm happy for
my premise to be shredded, er, disproven by the data, but first I need
that data. Does anyone have a pointer to uniform complete SNIa data
which includes absolute magnitude, blueness, and redshift?

Ultra-short answer: no.

Longer answer: I think what you might hope to find is a collection
of slightly different information: the observed quantities.
It would be nice to have a table of:

- apparent magnitude(s) at maximum light, together
with passband(s) in which the magnitude(s) was measured
- estimate of redshift

If the table included at least two magnitudes at maximum light
in different passbands, you could calculate a color.

The redshifts may present some problems: in some cases, one can measure
the redshift of an unambiguous host galaxy and assign it to the supernova.
In other cases, however, there may be no visible host galaxy, or several
possible hosts, or the host may be too faint for a redshift measurement.
In some cases, astronomers estimate the redshift from a spectrum of
the supernova; you _might_ want to pick a sample of events for which
the redshift was determined in a uniform manner.

With this information, you could then compare the measurements to
your favorite models of a) supernova properties and b) cosmological
parameters. You asked in your message for a table of "absolute magnitudes"
of supernovae, but of course, those depend on the cosmology one adopts
to convert the apparent to absolute magnitudes.

I share your wish for such a compilation of data, but I wouldn't bet
on finding one any time soon :-(

Michael Richmond



I have the same request for such a compilation of data.

In the following paper... 12 sept 03 by Knopf et al... the HST high
redshift survey data is presented already K corrected to support the
time dilation argument. Yet I am convinced that if I had access to the
`un-K corrected` data (I assume this is the same as what eric
requests) I could present a good case to support the alternative
argument that the SN data is not time dilated;
I would do this by taking all the HST survey observed wavelengths
lightcurves in the R and I bands (un k corrected) and calculate what
the emmision wavelength for each observed band would have been .

So for instance...For the 2 observed HST R and I band lightcurves
for a z=0.78 SN , I would recalculate these as being after redshift is
taken out lets say ,... V and R band EMISSION wavelength lightcurves.
If my argument that NO time dilation was occuring then these two
corrected to emmision wavelength lightcurves should match perfectly
the same wavelength lightcurves as found in the low redshift Reiss
lightcurves. That is the HST emmision lightcurves of V and R (observed
as R and I) should match the observed V and R lightcurves from the
Reiss low redshift survey.
I have already tried this approach using the k corrected data from
Knopf et al HST survey and found that my argument for no time dilation
is supported by the data despite it being K corrected. All I need is
final confirmation that there is no time dilation of SN lightcurves
by getting access to non k corrected data from the HST survey.
Is it possible that the data isnt available simply because it would
prove damaging to the BBT ?
Michael in his post asks for apparent magnitude AND an estimated
redshift as does Eric . You almost have this in the HST survey ( that
is you have estimated redshift but Knopf only supplies the apparent
magnitude , after he has k corrected the original observed data )and
obviously the aparent magnitudes ARE available as Knopf could not
have calculated his K corrected lightcurves without it.
So For Eric , Michael and myself, Sean,.. Does anybody know how to
access the NASA HST deep field survey apparent magnitude SN data?

Sean

Here are links to various papers, discussions and sites which concern
the interpretation of SN1a light curves. Some are recycled from the
thread "Plasma redshift, coronal heating . . ." which groups.google
won't let me write to, since the last activity was over a month ago.
See also the "Supernova & GRB time dilation" thread which I can't write
to either.

A paper by Jerry W. Jensen which I think is highly significant:

Supernovae Light Curves: An Argument for a New Distance Modulus
http://arxiv.org/abs/astro-ph/0404207 2004 April 6

He contends that the conventional interpretation (such as by the
researchers at the Supernovae Cosmology Project
http://www-supernova.lbl.gov) of supernovae light curves is flawed. His
argues that his corrections to the conventional analysis show that there
is there is no time dilation - and therefore no reason to believe the
Universe is expanding according to the Big Bang Theory.

He offers an explanation of the cosmological redshift with a theory
known as CREIL - Coherent Raman Effects on Incoherent Light. But maybe
a plasma redshift theory could explain it to.

This paper refers to "Malmquist Type II Bias" which is explained in a
1997 paper by P. Teerikorpi. (See especially page 109 and the example
near the bottom of page 112.)

Observation Selection Bias Affecting the Determination of the
Extragalactic Distance Scale
http://nedwww.ipac.caltech.edu/level...orpi/paper.pdf

An earlier paper by Jerry Jensen and Jacques Moret-Baily explains CREIL:

Propagation of electromagnetic waves in space plasma
http://arxiv.org/abs/astro-ph/0401529 2004 January 25

There have been lively discussions of Jerry Jensen's paper at
http://www.badastronomy.com . (Ideally, I think, these would occur on
the moderated, public, fully archived and searchable sci.astro.research
- but maybe the moderation and propagation delays make the instant
response of the Web-based discussion server more attractive.)

Against the Mainstream: Cutting the Cord on the Big Bang
http://www.badastronomy.com/phpBB/viewtopic.php?t=14433

and an earlier one

Bad Supernova Data Reduction
http://www.badastronomy.com/phpBB/viewtopic.php?t=14269

These discussions give more insight into the critique Jenson makes of
the conventional SN1a approach. In the earlier discussion, on 15 June,
he says that he has been writing to the conventional researchers for
over a year before April 2004 and "My emails ... are never answered if I
ask anything more than the most innocent questions."

I think his critique of the conventional SN1a approach is very credible.
I won't try to summarise his critique here.

Ari Brynjolfsson has published a second paper - with a new analysis of
supernovae light curves, according to his plasma redshift theory, again
showing no time dilation.

Plasma Redshift, Time Dilation, and Supernovas Ia
http://arxiv.org/abs/astro-ph/0406437 2004 June 19


According to Eric Flesch (in the "Plasma redshift ..." thread), a paper
with unadjusted light curves for 11 SNe at redshifts 0.35 to 0.86:

http://arxiv.org/abs/astro-ph/0309368

A page about SN1a light curves and spectra, which Gordon D. Pusch cited
when debating Eric Flesch's views about these light curves:

http://www.nd.edu/~kkrisciu/supernovae.html

The Online Supernova Spectrum Archive:

http://bruford.nhn.ou.edu/~suspect/

I believe anyone who is interested in SN1a observations and
interpretations will find Jerry Jenson's paper and the ensuing
discussion fascinating. I think this critique can be completely
decoupled from CREIL or whatever theories may explain the redshift of
galaxies and quasars without the Big Bang.


- Robin http://astroneu.com

(Robin Whittle) wrote in message ...
Here are links to various papers, discussions and sites which concern
the interpretation of SN1a light curves. Some are recycled from the
thread "Plasma redshift, coronal heating . . ." which groups.google
won't let me write to, since the last activity was over a month ago.
See also the "Supernova & GRB time dilation" thread which I can't write
to either.

A paper by Jerry W. Jensen which I think is highly significant:

Supernovae Light Curves: An Argument for a New Distance Modulus
http://arxiv.org/abs/astro-ph/0404207 2004 April 6

Hi Robin
Thanks for al the reference,..
On the point made on page 3 of Jensens paper of 06 04 2004 where he
says....

" Even though the variance in magnitudes is relatively small, the
light-curves (in days) vary significantly.
Longer light-curves correlate with higher magnitude SNe Ia. (Goldhaber
1996, Hamuy 1996.)"

This could be a simple case of the higher mag SN being closer and less
redshifted. Reiss` IRVB examples show that the redder end of the
emmision spectra of a SN have slower decay rates (which also,
importantly, appear to be more time dilated). In which case at any one
observation frequency one would be seeing a longer emmision wavelength
from a higher mag SN than for a lower mag SN. Thus the observed higher
mag SN lightcurve would `appear` to be more time dilated when in fact
the apparent time dilation is an effect of observing the the lower
redshifted SN (that is the higher mag SN )at a longer wavelength than
the higher redshifted SN

According to Eric Flesch (in the "Plasma redshift ..." thread), a paper
with unadjusted light curves for 11 SNe at redshifts 0.35 to 0.86:

http://arxiv.org/abs/astro-ph/0309368

This is the paper I referred to in my previous post. The lightcurves
graphs in the paper on pages 11 and 12 are *definitely* k corrected.
I`m not 100 % sure but the supplied tables of data for each SN (later
in the paper) for the same 11 SN lightcurves also appear when
plotted against the lightcurves to be identical. This means that
either the data is k corrected in the tables OR the k correction is
so minimal that there is a neglible difference between the k corrected
and the un corrected data which I doubt is the case. Having said that,
on reading the paper it does make clear that the plotted 11
lightcurves graphs are k corrected but the tables dont appear to have
any mention of a k-correction applied. So its confusing.

Maybe someone on this newsgroup who is more familiar with the
terminology would be able to tell me whether the tables of data for
the 11 lightcurves starting on page 43 of Knopf et al`s 12 september
2003 paper are k corrected or not.
( http://arxiv.org/abs/astro-ph/0309368 )
Sean

In article ,
you write:
Maybe someone on this newsgroup who is more familiar with the
terminology would be able to tell me whether the tables of data for
the 11 lightcurves starting on page 43 of Knopf et al`s 12 september
2003 paper are k corrected or not.
( http://arxiv.org/abs/astro-ph/0309368 )

Knop, not Knopf. Now published in 2004 ApJ 598, 102.

The lightcurves in the tables are k-corrected. I'm not sure why you
think that makes any difference, but if you think so, you could
probably work out how the k-correction was done and back it out. At
the very least, you should be able to estimate how big the correction
is and how much it changes during the time a given SN is observed.
(I haven't tried this, but I certainly hope and expect the authors
have given relevant details, probably via references.)

Trying to understand and critique the way these SN observations have
been processed and interpreted is a big task. I don't have time at
present, but I would want to read this paper (astro-ph/0309368) to about
page 25 and the relevant parts of every paper the authors cite. Then, I
would compare my understanding of this process with the account given by
Jerry Jensen. Only then could I evaluate Jerry's critiques. Trying to
poke into the beast from the outside with a few quick stabs is unlikely
to result in real enlightenment. It needs to be fully disassembled and
all its bits laid out in the sunshine.

Here is my understanding of these graphs of the lightcurves (Fig 1 and
2). The description on page 10 does not clearly describe the graphs:

Appendix A tabulates all of the lightcurve data for the eleven HST
supernovae in this paper. The lightcurves for these supernovae (and
the F675W WFPC2 image nearest maximum light) are shown in Figures 1
and 2.

However the description in Appendix A is more detailed:

Tabulated below are lightcurve data for the eleven HST supernovae
presented in this paper. For each event, there are two lightcurves,
one for R-band and one for I-band. All photometry has been
color-corrected to the standard Bessel filters as described in
section 3, using color corrections which assume the lightcurve
parameters in Table 3.

This indicates that the graphs are directly derived from the tabular
data in Appendix A.

My understanding is that these show the exact observer epochs of each
observation, specifying which telescope was used, with a flux for each
observation. These fluxes are relative to a zeropoint for each table,
which I guess is intended to be something like the flux observed from
the area of the supernova before and after its light was detectable.

These flux figures are not, however, what was observed - they are "color
corrected" versions of the original observations. The observations were
made through various colour filters, depending on the telescope. There
were various reasonably predictable errors due to absorption in the
Earth's atmosphere for the ground-based telescopes.

Ideally, the way to observe each SN at a given redshift would be through
a filter specifically made for that redshift, so that the telescope's
filter response, convolved with its detector response, when receiving
the redshifted light, selects the same spectrum of photons as a standard
filter and detector would with no redshift. This is impossible for
various reasons, and so they use a complex "color-correction" algorithm
to convert their observations into what they think would be observed
through R and I filters if there had been no redshift. (From what I
know, this is before and totally separate from the "K-correction",
whatever that is.)

There's plenty of scope for problems here. Any such correction must
involve assumptions about the spectrum of the SN. In particular, it
involves assumptions about how luminosity integrated through a filter
response over one range of wavelengths (as observed) correlates with the
luminosity over another range of wavelengths (as we would like to
observe).

This is tricky enough, but as Sean wrote earlier, there is the
additional problem of the spectrum changing over time as the SN grows
generally brighter and dimmer. It seems likely (I haven't checked) that
the light-curve in redder wavelengths would generally be longer in time,
especially in the tail, since temperatures would be dropping and the
long wavelengths would remain strongly in the resulting Planck curve for
a longer period of time than the short wavelengths.

To correct for this, it would not be enough to consider an isolated
measurement. The researchers would need to know where in the
light-curve this measurement was taken, to place it in some model of
cooling-induced spectral changes. If (and I am not saying they do)
there was some assumption at this point about time dilation of the light
received from high redshift objects, then there could be trouble. The
whole approach is dodgy anyway, since it involves assumptions about the
SN which may not be correct. As Jerry Jensen points out, there are
questions about whether we are really always looking at what we
understand as SN type 1a. Even if these objects really do fit this
formal definition, there are lots of questions about how these things
vary from SN to SN, what the statistics of these variations are etc.

Extinction in our galaxy, the host galaxy and all places in-between also
need to be considered at some point when trying to determine the peak
absolute magnitudes of these SNe.

I have not chased the references to see exactly how they do this - but
it seems that the *input* parameters for this color-correction process
are found in Table 3.

All photometry has been color-corrected to the standard Bessel filters
as described in section 3, using color corrections which assume the
lightcurve parameters in Table 3.

My initial understanding of Table 3 is that it contains the redshift of
each SN and the *outputs* of various stages of their correction process.
Unless the sole input to the color-correction process is the redshift,
then from my initial reading, "using color corrections which assume the
lightcurve parameters in Table 3." doesn't make sense.

I would need to read a lot of material to understand exactly how they
color-correct the fluxes we see in Appendix A and the graphs of Fig 1
and 2. However, it is my impression that these tables and graphs give
the exact epochs of the observations, unaltered in any way. To check
this, consider Table 1, the right column, for SN 1997ek. These are HST
observations on 7 days, starting with 5 January 1998 and ending with 16
November 1998. These observations span 315 days. Now look on page 43
at Table 11. The earliest HST observation is on Julian day 50818.93 and
the last on 51134.26. The difference is 315.33 days. So the table
epochs seem to be raw.

The epochs in the graph seem to directly reflect the table epochs. I
haven't noticed if they specify what epoch on Earth they regard as the
peak of the light-curve (which is something which would result from a
lot of crunching, but I have looked at the top right graph in Fig 1 -
the I-band filter light curve (as calculated via color-correction, I
think) - of SN 1997ek at redshift 0.86. There are two black dots (HST
observations) near the peak of their artificial light-curve line and
three to the right as it gets dimmer. One more is in the non-linear
right end of the graph, where the days 150 to 550 after the assumed peak
are compressed horizontally. I assume this is really the two HST
observations on Julian days 51126.93 and 51134.26.

The left-most black dot looks like it is aligned with 0 on the
horizontal scale. I will assume this first observation is on day
50818.93. On that basis, the first 5 observations would have these
times from 0:

Julian day Relative to Flux
50818.93

50818.93 0 3.83
50824.78 5.85 3.89

50846.74 27.81 1.54
50858.84 39.91 0.75
50871.95 53.02 0.46

Within the limits of the clarity of the graph, the black dots are in
exactly the right place for these figures.

So the $13.7 billion question is whether the light curves for the high
redshift SNe have longer times than those with low redshifts!

A wag might point out that the 1998be curve (z=0.644) is clearly shorter
than that of 1997as (z=0.355) - while if the SNe themselves were
identical explosions, and if the redshift is caused by them moving away
from us, then the resulting time-dilation would cause our observation of
the 1998be curve to be 21.3% longer.

But these curves - the lines themselves - are the product of a lot of
interpretation, as are the corrected fluxes. All this is based on
sparse observations - and we don't know for sure when the lightcurves
really started. Maybe the colour-correction process for high-redshift
objects does result in raised flux levels at the tail of the presumed
curve - that would make sense if Sean's suggestion is a factor.

I wouldn't attempt to divine any more insight without looking at in
detail at how the observations are "corrected". Ideally, I would want
the original fluxes.


- Robin http://astroneu.com

In my previous message I gave an inadequate description of how I
thought the supernovae light curve data was processed. The
researchers use the term "color-correction" and "K correction"
separately. I am not sure what the former is, but it the term
"K correction" is a specific transformation, as described in:

The K correction
David W. Hogg, Ivan K. Baldry, Michael R. Blanton,
Daniel J. Eisenstein
http://arxiv.org/abs/astro-ph/0210394

- Robin http://astroneu.com

(Steve Willner) wrote in message ...
In article ,
you write:
Maybe someone on this newsgroup who is more familiar with the
terminology would be able to tell me whether the tables of data for
the 11 lightcurves starting on page 43 of Knopf et al`s 12 september
2003 paper are k corrected or not.
( http://arxiv.org/abs/astro-ph/0309368 )

Knop, not Knopf. Now published in 2004 ApJ 598, 102.

The lightcurves in the tables are k-corrected. I'm not sure why you
think that makes any difference,

In my two previous posts in this thread I explained why it is
neccesary to access non k corrected data but Ill try here again as
briefly as possible.

If one were able to observe a high redshift SN in B and R bands `close
up` one would get a two emmision lightcurves ,..one B one R like so...


. . . .
. . . .
. . . .
. . . .
.. . . .

B R

(Notice B is steeper decay and R is less so and appears to be time
dilated. Reiss` survey shows IRBV has this progression of faster
decays for shorter wavelength lightcurves)

Here on Earth we see those two emmision bands redshifted to longer
wavelengths ,lets say from BR to RI. So the light, we observe in
filters R and I was initially emitted in B and R as illustrated
above. My point is that IF the redshift is not caused by a BBT
expansion there should not be any time dilation of the observed
lightcurves of high redshift SN.
This can checked and verified or dismissed simply by taking the two
observed lightcurves in R and I of a distant SN and calculating back
what the emmision wavelength for the observed R and I bands would be .
In this case as I have said they would be emmited in B and R bands.
SO,... One simply has to take the two Uncorrected R and I band
lightcurves (observed from the high redshift SN) and overlay them with
the B and R band lightcurves from Reiss` survey (Reiss survey being
low redshift gives us a good representation of what emmision
lightcurves of SN for IRBV are).

*IF* they match then this is conclusive proof there is no time
dilation of high redshift SN ligntcurves contrary to accepted wisdom.
I have done so with k corrected HST lightcurves and there is no sign
of time dilation using this above method. However I realize that to
prove my point absolutely I must do the same with *uncorrected* high
redshifted lightcurves. Hence my desire to find uncorrected data .

probably work out how the k-correction was done and back it out. At
the very least, you should be able to estimate how big the correction
is and how much it changes during the time a given SN is observed.
(I haven't tried this, but I certainly hope and expect the authors
have given relevant details, probably via references.)

I believe I *may* have found some uncorrected data in the SCP
database. In that site there are downloadable GIF lightcurves of high
redshifted SN and no mention of any k correction applied.
Unfortunately an email to the site requesting confirmation of whether
or not the SCP raw lightcurves are k corrected has not been responded
to. Can I then assume the online SCP lightcurves NOT k corrected?
Sean

Sean wrote:

I believe I *may* have found some uncorrected data in the SCP
database.

I can't find any such .gifs with Google. Can you give the URL of these
files or of wherever you are looking?

Can I then assume the online SCP lightcurves NOT k corrected?

I don't think anything like this can be assumed. I am struggling to
understand how Knop et al. and the other major papers in this field,
process their data. Here is my current, partial, probably inadequate
and/or faulty understanding:


Referring to Knop et al. http://www.arxiv.org/abs/astro-ph/0309368
the flux (linear brightness) values used in Figs 1 and 2, seem to
directly correspond to the values in Appendix A, which can also be found at:

http://brahms.phy.vanderbilt.edu/deepsearch/hstpaper/

These have times (horizontal) exactly as observed (as I noted in an
earlier post). However the values (vertical) are not the raw observed
fluxes - they are the product of some corrections. I found it it
confusing trying to figure out exactly how they arrived at these "color
corrected" flux values:

(p8) For both high- and low-redshift supernovae, color corrections
and K-corrections are applied . . .

Many paragraphs of details follow - how they figured out the best way of
deciding on a curve to represent the total light curve, as it would be
with a given filter if observed near to the SN, based on their limited
number of observations at various redshifts with various telescopes and
filters. My understanding of "fitting" means using the MINUIT program:

http://wwwasd.web.cern.ch/wwwasd/cernlib/
http://wwwasdoc.web.cern.ch/wwwasdoc...t/minmain.html

to find the values of various variables which optimise (typically
minimise) the output of some Fortran function. By coding up some stuff,
with light-curve tables and corrected observational data, MINUIT will
find how best to place a light-curve, with starting time, width, and
maximum flux to the observed data points. Its a dodgy business, but
they need a light curve to figure out what the maximum flux was, or
would have been if they had observed it at maximum. They also need this
light-curve to do various corrections which involve assumptions about
the spectrum of the light at any particular point in time.

On page 10:

The final results of the light-curve fits, including the effect
of color corrections and K-corrections are listed in Table 3 ...

For each SN, Table 3 contains the stretch factors and a calculated
maximum observed magnitude - the magnitude that would (they intend) have
been observed through a standard filter on Earth when we received the
light from when the SN was at its brightest. (BTW, "magnitude" drives
me nuts - brighter is a lower value, so "maximum magnitude" applied
numerically means dimmer, whilst in a meaningful sense about brightness,
means a lower number.)

Column d has this brightest magnitude corrected for host-galaxy
extinction. This figure, together with the redshift, goes into their
cosmological analysis.

The SNe were observed with differing redshifts, on various telescopes
with differing filters. Ideally, they would have measured the light
curve using a single filter from a location close to each SN.

Ordinarily, colour correction - which I understand to be converting a
flux (number of photons) value measured with one filter to a value which
is what we think would have been observed through another filter -
involves precise knowledge of the two filter response curves, and an
assumption about the spectrum of light being observed. By "ordinarily",
I mean that the spectrum is known. However, the SN spectrum changes
over time, so in order to perform color or K corrections, from one
filter and observed redshift to some other filter and desired redshift,
they need to know the spectrum. Since they don't know the spectrum at
all at that time of the photometric measurement, they have to try to
estimate it from the generalised spectrum of a similar SN at this
particular point in its brightening and dimming light curve. I think
this means they need to estimate where in the history of the SN light
curve this observation is, in order to do their corrections. An
illustration of the way the SN spectrum starts off bluer and ends redder
is at:

http://supernova.lbl.gov/public/figures/saul_sm.mpg

So I think that the sum total of their corrections (For Knop et al. in
two stages - color and K) to the flux of each observation needs to be
done after they have figured out where in the light-curve, in the SN's
timeframe, each observation lies.

Generally, as I understand it, what I described above as color
correction is in fact K-correction:

The K correction
Hogg et al http://arxiv.org/abs/astro-ph/0210394

but Knop et al. have two stages. If all their observations were with a
single telescope and filter, they probably would have done the
corrections in one "K-correction" step. However, they are mixing
different telescopes, with different filters, and SN of different
redshifts. Section 2.3 explains their process.

In order to combine data from different telescopes, color corrections
were applied to remove the differences in the spectral responses of
the filters relative to the Bessell system (1990PASP..102.1181). For
the ground-based telescopes, the filters are close enough to the
standard Bessell filters that a single linear color term (measured at
each observatory with standard stars) suffices to put the data onto
the Bessell system, with most corrections being smaller than 0.01
magnitudes. The WFPC2 filters are different enough from the
ground-based filters, however, that a linear term is not sufficient.
Moreover, the differences between a SN Ia and standard star spectral
energy distribution are significant. In this case, color corrections
were calculated by integrating template SN Ia spectra (described
below) through the system response.

So this first stage seems to be a way of coping with the various
telescopes and filters - one would think this has nothing to do with the
redshift of the observed light. They then use a "single series of
K-corrections" - I guess one particular correction for each SN's
redshift - applied to each one of that SN's color corrected flux levels.
This seems to be for simplicity, to reduce the number of K corrections
and the possibility of errors.

The color correction to the nearest standard Bessell filter followed
by a K-correction to a rest-frame filter is equivalent to a direct
K-correction from the observed filter to the standard rest-frame
filter. In practice, we perform the two steps separately so that all
photometry may be combined to provide a lightcurve effectively
observed through a standard (e.g. R-band) filter, which may then be
fit with a single series of K-corrections. The data tabulated in
Appendix A have all been color-corrected to the standard Bessell
filters.

So here they say that Appendix A, and as far as I can tell, the flux
levels in Figures 1 and 2, have just been subject to this first level of
correction - color correction - and by implication not K-correction,
which is to cope with the different redshifts of the SNe. (Why do they
publish this half-way corrected stuff, and apparently graph it? I would
have thought that the most important thing was the result of the
K-correction if they are only going to publish one set of figures.
Maybe they sometimes use "color correction" to include both the first
"color correction" stage and the second "K-correction" stage.)

But if "color correction" is just to cope with differences between
telescope filters, with nothing to do with redshift etc. why in Appendix
A do they write that this requires something to do with fitted light curves:

All photometry has been color-corrected to the standard Bessel Filters
as described in section 3, using color corrections which assume the
lightcurve parameters in Table 3.

??

I can understand them using redshift and assuming time-dilation of high
redshift SN in order to do their K-corrections. In order to correct a
particular flux level, they need to know the spectrum at that time,
which they don't have - so I think they would figure out where in the
assumed light-curve the observation is, in the time-frame of the source,
and use that as part of their K-correction from one filter function to
another.

I can't understand, by reading Knop et al. exactly how the data was
transformed - and I couldn't clearly see where Jerry Jensen's charge of
assuming time-dilation in the high redshift observations was justified
in this paper. But Knop et al. cites (page 10) Nugent, Kim & Perlmutter
(http://arxiv.org/abs/astro-ph/0205351) for how they did the color and K
corrections. I haven't read this yet. However, I have reason to
believe Jerry's charge sticks in the general procedures used by these
researchers. For instance, on page 3 of:

Measurements of Omega and Lambda from 42 High-Redshift Supernovae
Perlmutter et al. http://arxiv.org/abs/astro-ph/9812133

they state clearly that they assume time-dilation in the observations of
high redshift SNe:

For the supernovae discussed in this paper, the template must
be time-dilated by a factor 1 + z before fitting to the observed
lightcurves to account for the cosmological lengthening of the
supernova timescale

Also, in page 16 of the frequently cited:

Measurements of the Cosmological Parameters Omega and Lambda from the
First 7 Supernovae at z = 0.35
S. Perlmutter et al. http://arxiv.org/abs/astro-ph/9608192

there is another telling sentence:

The rising slope of the template light curve before rest-frame day -10
...

This convinces me that they are assuming time-dilation in how the light
from high-redshift SN arrives on Earth.


I have a long way to go before I am confident of understanding how the
observations are transformed into the data which are published. But
it seems that Jerry's first charge of assuming time-dilation is correct.

This is bad science if the purpose is to test whether or not there is in
fact time-dilation in high redshift objects. Maybe they think they
settled the matter once and for all in 1996:

Observation of Cosmological Time Dilation using Type Ia Supernovae as
Clocks
G. Goldhaber et al. http://arxiv.org/abs/astro-ph/9602124

- Robin http://astroneu.com