A Space & astronomy forum. SpaceBanter.com

Go Back   Home » SpaceBanter.com forum » Astronomy and Astrophysics » Research
Site Map Home Authors List Search Today's Posts Mark Forums Read Web Partners

Any complete standardized SNIa data out there?



 
 
Thread Tools Display Modes
  #21  
Old September 6th 04, 01:12 PM
sean
external usenet poster
 
Posts: n/a
Default

[[Mod. note -- Long lines wrapped and excessively-quoted text trimmed;
posters, please do this yourself. -- jt]]

(Steve Willner) wrote in message
...
(sean) wrote in message
...
The method I used was to take individual
SN lightcurves from Knops paper pages 11-12( R.A. Knop
arXiv:astro-ph/0309368 vl 12 Seop 2003) and compare them with ones
supplied in Adam G. Reiss` paper pages 23-25 (arXiv:astro-ph/9810291
vl 19 Oct 1998).


[[ ... trimmed by moderator ... ]]

As you can see the Knop mags are in a different `mag` format from
Reiss`.


Yes, as noted above, Knop et al. use linear flux density. One
magnitude is close to a factor of 2.5, as you no doubt know.



No I didnt but, thanks for that information. I assume then that a
decay in linear from 1 to 0.4 is equivelent to a decay in log of 1 mag
as you suggest.

[[Mod. note -- Magnitudes are defined so that a difference of 5 magnitudes
is a factor of 100 in (linear) flux density. Thus a difference of 1
magnitude is a factor of 100**0.2 = 2.512, and more generally a factor
of X is a magnitude difference of 2.5*log10(X).
-- jt]]

I`m not sure what you mean "by consistent with 1" ? Is that when you
compare the ratio between the days it takes a low redshift SN to
decay by one mag as compared to how many days a high redshift SN
decays the same mag? So if its `1` then thats a 1/1 ratio ie they both
decay at the same rate ? .


I checked 1995E (Reiss pg 23 of his paper..... arXiv:astro-ph/9810291
vl 19 Oct 1998).. and I seem to get completely different results from
reading the data that you have.
I agree with you that it takes only about 17 days to decay by one
mag in the B band lightcurve of 1995E.
However with the I band in 1997ek (Knop) I found that the 0.4 reading
is clearly stated in the tables as being 22 days and not 29 days, as
you claim, after the highest (peak luminosity ) reading for SN1997ek .
This is the HST reading of 1.54 on julien day 50846 which on the graph
is represented in linear as 0.4 mag. (It is clearly the brightest mag
reading in I band for that SN despite being placed at about 8 days
past peak on the graph.I assume thats for convenience to fit a
theoretical template. Nonetheless it is the brightest mag for I band
and thus for me the decay is 22 days from 1 to 0.4 )

So the ratio I get is 22 days for high redshift as compared to 17 days
for low redshift which is much closer to 1 than you have calculated.



And to explain this small difference without invoking any time
dilation at all , I do the following calculation..The high redshift I
band (Knop ) which you calculate as being an emission wavelength of
438nm is compared to the Reiss B band , but , remembering that Reiss B
band is still a 0.1 redshift, this means that the low redshift 1995E
B band actually was an emission wavelength of 390nm. Thats a 50nm
difference between the two that are being compared incorrectly as like
for like. And if one then notes that the time it takes a low redshift
530nm V band from Reiss 30 days to decay by one mag on average from
his survey as compared to the 17 days for the average B band to
decay 1 mag then it becomes obvious that the shorter the wavelength
the shorter the time it takes to decay one mag for the same SN. In
other words extrapolating a 12 day difference in decay rates for a low
redshift 530nm band to 430nm band one could say that it would
probably be even six day less for a 390nm when compared to a 438nm
lightcurve for both to decay by one mag . That means a 390nm
emission wavelength of a SNtype 1a actually should decay about 6 days
faster over 1 mag then a 438 nm emission wavelength. . And thats what
we see as the 438nm lightcurve (originally I band observed at
redshift0.85) decays 1 mag in 22 days whereas the 390nm lightcurve
(originally a B band observed at 0.1 redshift) decays at 17 days .
Thats a 5 day difference which is close to the 6 day I calculated
above meaning no time dilation is observed. So to me the apparent time
dilation is only an artifact of incorrectly comparing different
emission wavelength lightcurves together as like for like.


Hopefully this also answers Craigs concerns except maybe his
following question about filter bandwidths ..."How do you address the
fact that the filter bandpasses of the optical instruments involved
are broad, not narrow, and therefore are not necessarily compatible
with associating them with a single wavelength?"....
I cant answer that except maybe to ask him the same question regarding
the pro time dilation arguments use of the same data I use.
thanks
Sean
  #22  
Old September 8th 04, 01:20 PM
Steve Willner
external usenet poster
 
Posts: n/a
Default

(sean) wrote in message ...
I`m not sure what you mean "by consistent with 1" ?


My comment was "not consistent with 1," i.e., the data are NOT
consistent with absence of time dilation. I probably confused you by
mistyping one of the numbers. The one-mag decay time of 1998ek in I
looks like about 29 days from the graph in Fig. 1 of astro-ph/0309368;
I typed 39 but used the correct 29 in my next paragraph. This
compares to 17 days for 1995E in B band. There is no way 17 and 29
are equal, no matter how much you stretch the error bars.

Is that when you
compare the ratio between the days it takes a low redshift SN to
decay by one mag as compared to how many days a high redshift SN
decays the same mag? So if its `1` then thats a 1/1 ratio ie they both
decay at the same rate ? .


Yes exactly. If there were no time dilation, the decay times would be
the same, i.e. have a ratio of 1.

I checked 1995E (Reiss pg 23 of his paper..... arXiv:astro-ph/9810291
vl 19 Oct 1998).. and I seem to get completely different results from
reading the data that you have.
I agree with you that it takes only about 17 days to decay by one
mag in the B band lightcurve of 1995E.


OK. We agree here.

However with the I band in 1997ek (Knop) I found that the 0.4 reading
is clearly stated in the tables as being 22 days and not 29 days, as
you claim, after the highest (peak luminosity ) reading for SN1997ek .
This is the HST reading of 1.54 on julien day 50846 which on the graph
is represented in linear as 0.4 mag.


I agree with your date (actually 50846.7) at which the decay is one
magnitude. I'm looking at Table 11.

(It is clearly the brightest mag
reading in I band for that SN despite being placed at about 8 days
past peak on the graph.I assume thats for convenience to fit a
theoretical template. Nonetheless it is the brightest mag for I band
and thus for me the decay is 22 days from 1 to 0.4 )


I have no idea what you mean by this. The brightest single
measurement is at day 50817.65, a value of 5.89 +/- 1.23. There are a
bunch of other measurements near the same time. I haven't bothered to
pull out a calculator, but by eye, they average to something a bit
above 4, and this is the time of maximum. The time difference is
50846.7-50817.65=29 days, just what I got by reading off the graph.
For "by eye" work, the graph is better because some of the points have
already been averaged, and the graph includes the standard decay
curve.

So the ratio I get is 22 days for high redshift as compared to 17 days
for low redshift which is much closer to 1 than you have calculated.


Let's see... 22 days from 50846.7 (when we agree the decay has reached
one magnitude)... you are claiming that maximum was at 50824.7.
There's an HST measurement of 3.83 at that date, but it is clearly
well after maximum. Look again at all those measurements near
50846.7.

A more extreme example is 1998as, which was not observed in I until
well past maximum. If you took the first measurement to be the time
of maximum, you would get a wrong answer, as you can see from the R
measurements.

And to explain this small difference without invoking any time
dilation at all , I do the following calculation..The high redshift I
band (Knop ) which you calculate as being an emission wavelength of
438nm is compared to the Reiss B band , but , remembering that Reiss B
band is still a 0.1 redshift,


That is 0.01 for 1995E, as stated in the earlier post.

this means that the low redshift 1995E
B band actually was an emission wavelength of 390nm. Thats a 50nm
difference between the two that are being compared incorrectly as like
for like. And if one then notes that the time it takes a low redshift
530nm V band from Reiss 30 days


How are you getting 30 days? Looks like about 18 days to me;
certainly not more than 20 (again based on 1995E). The V maximum is
two or three days later than the B maximum, but V is a magnitude below
its peak no later than 21 days after the B maximum.

If you want to do all this "right," you need to do a large sample and
take account of filter bandwidths, exact redshifts (K-correction), and
all the uncertainties. However time dilation is obvious on a casual
glance.
  #23  
Old September 8th 04, 03:08 PM
Bjoern Feuerbacher
external usenet poster
 
Posts: n/a
Default

Steve Willner wrote:
[snip]

Sorry to step into your discussion, but I've also got a problem with
the time dilation of SN light curves, and you seem to be quite
knowledgeable about this.

There is a paper on the preprint archive:
astro-ph/0404207
by a Jerry W. Jensen, where he argues that the stretch in the light
curves is not due to time dilation, but due to the supernovae having
higher magnitudes and therefore longer decay times (if I understand him
correctly). He also seems to suggest that what is assumed to be distant
supernovae are actually hypernovae (thus explaining their higher
magnitudes). Some of the things he proposes look rather cranky, but
others seem to be valid.

This was also discussed in some detail on the badastronomy.com bulletin
board:
http://www.badastronomy.com/phpBB/viewtopic.php?p=227836&highlight=malmquist+bias+su pernova&sid=dc9b062af8a301dd9f0be4c7a6c78c10#22783 6
But most of his arguments there were not refuted by the posters, and I
am unsure if he has a point or not.

Could you please help me out?


Bye,
Bjoern
  #24  
Old September 11th 04, 01:50 PM
sean
external usenet poster
 
Posts: n/a
Default

(Steve Willner) wrote in message ...

I have no idea what you mean by this. The brightest single
measurement is at day 50817.65, a value of 5.89 +/- 1.23. There are a
bunch of other measurements near the same time. I haven't bothered to
pull out a calculator, but by eye, they average to something a bit
above 4, and this is the time of maximum. The time difference is
50846.7-50817.65=29 days, just what I got by reading off the graph.
For "by eye" work, the graph is better because some of the points have
already been averaged, and the graph includes the standard decay
curve.

This is what I am not so convinced about. There are 13 readings
averaged out to 4.8 by my calculations so I`m not sure whether the
single peak measurement point at 1.2 on the graph represents the
`averaged` reading of 4.8 or the one peak reading of 5.89.Anyways
this is shown on the graph as 1.2 mag at day 50817 and my guess is
1.2 probably represents the averaged lower mag value of 4.8.
One problem I have is that the 1.2 `reading` on the graph does not
decay by one mag to 0.4 as you suggest but rather to 0.48 linear
(using the calculation 1.2/2.5=0.48) . Giving a time in days for
0.48 mag is difficult to do accurately on the graph, but from the
graph 0.48 occurs at about 21-22 days after peak, not 29 days!

Using the peak reading of 5.89 mag from the tables I think
a 1 mag decay must also be in the order of about 21 days but its
difficult for me to convert 5.89 to the linear scale used on the
graph so if you could help me on that it would confirm my rough
calculation of 21 days for mag 5.89 or1.2 to decay by 1 mag.
So thats about 21/18=1.16. Nowhere near 1.85 that time dilation
needs and essentially 1 within error margins which means ..no time
dilation.
IncidentallyI have gone through several of the low redshift SN`s
from the tables and I get an average for the lightcurves from the
redshift range around 1995E as being...

I band decays 1 mag in 38 days
R band " " " " 30 days
V band " " " " 22 days
440nm " " " " 18 days*
B band " " " " 17 days

I agree with you that it would be to do a compilation and
average out all the SN`s etc etc ,but I think it is important, at
least, to get 1997ek right.
Unfortunately 1997ek is one of only a couple where there is *any*
chance of defining a peak reading that has been observed rather
than inferred by the template.
1997eq is viable as is 1997ez. 1998aw,2000fr and1998bi all have
HST readings near the template peak but its not clear
whether they happen before or after a `real` peak by as much as
5-10 days and whether the actual peak is indeed higher by a mag
or so than the template peak. This makes them almost impossible
to use because the error margin is on 10`s of days and useless
for any definitive pro or anti time dilation argument. The rest
of the high redshift Knops have no observations near peak and are
essentially useless in my opinion unless one relies on the
templates.
And the templates are not an accurate enough guide as one
only has to see how in SN1997eq and ek the actual peak
measurements are well above the inferred template peak of 1
mag linear. So much so that using the templates gives a ratio of
29/18 whereas using actual observations one gets 21/18

Incidentally,are you sure that 1995E is 0.01 redshift? It seems
to be in the mid range of `heliocentric redshifts` (on table 3
page 51 Reiss) at 3.54 in a range between 3.1 - 4.5 (or 0.01 - 0.1)
but unfortunately I dont know how to convert `heliocentric` to z.

A more extreme example is 1998as, which was not observed in I until
well past maximum. If you took the first measurement to be the time
of maximum, you would get a wrong answer, as you can see from the R
measurements.

I get a bit different for the R band from 1998as than you do.
1998as R band converts to restframe... 600/1.35=444nm. And my reading
from the graph of R band 1998as is: That its about a 20 day decay time
from 1 to 0.4 (and thats ignoring the fact that the peak is actually
higher than 1) And as I`ve pointed out before the comparison 444nm
restframe lightcurve from the 0.01 1995E lightcurves shows about
a 19 day decay for 1 mag. And so if one then compares the two ....
Thats 20/19=1.05 which is essentially showing: No time dilation
for a z=0.35 SN R band lightcurve compared to an equivelent restframe
lightcurve. Remember the ratio should be 1.35 if there *was* time
dilation
Sean
If you want to do all this "right," you need to do a large sample and
take account of filter bandwidths, exact redshifts (K-correction), and
all the uncertainties. However time dilation is obvious on a casual
glance.

  #25  
Old September 15th 04, 09:29 AM
Steve Willner
external usenet poster
 
Posts: n/a
Default

In article ,
(sean) writes:
[Referring to Knop et al., 2003 ApJ 598, 102 or the preprint thereof]
There are 13 readings
averaged out to 4.8 by my calculations so I`m not sure whether the
single peak measurement point at 1.2 on the graph represents the
`averaged` reading of 4.8 or the one peak reading of 5.89.


It's pretty obvious if you compare the table and the graph and note
the open/filled circles, as the figure caption explains. The open
circle just below 1.2 represents the 13 "BTC" measurements, the
filled circles at 1.0 are the two HST measurements (3.83, 3.89 in
Table 11), and the open circle at 0.8 with the big error bar
represents the WIYN measurements. Thus "1.0" on the figure is
something like 3.9 in the table (probably the 3.89). One magnitude
fainter is 1.56 in the table, very close to 1.54 measured at day
50846.7. (I didn't pick this SN as an example at random!)

this is shown on the graph as 1.2 mag at day 50817


Not "mag"; linear flux density.

and my guess is
1.2 probably represents the averaged lower mag value of 4.8.


Close enough; I get 5.1. Did you weight the measurements when you
took the average?

One problem I have is that the 1.2 `reading` on the graph does not
decay by one mag to 0.4 as you suggest but rather to 0.48 linear
(using the calculation 1.2/2.5=0.48) .


Look at the error bars! The "1.2" is highly uncertain. The two HST
measurements around the same time have much smaller uncertainties.
Those are what establish the peak of the light curve, which
conveniently is put at 1.0 in the graph.

Giving a time in days for
0.48 mag is difficult to do accurately on the graph, but from the
graph 0.48 occurs at about 21-22 days after peak, not 29 days!


Again not "mag," and you are looking for 0.40, not 0.48.

Unfortunately 1997ek is one of only a couple where there is *any*
chance of defining a peak reading that has been observed rather
than inferred by the template.


There's nothing wrong with using the templates. In fact that's the
"right" way to do things because the time of maximum, where the light
curve is flat, is hard to determine. Fortunately, the authors have
done the work for you. If you don't like the template, look at the
time to decline an additional 0.78 mag (from 1.54 to 0.75 in Table
11), 12.1 days. The corresponding time for 1995E is about 5 days.

There are many more supernovae published than in the Knop et
al. paper. All we are looking for here is a sanity check. Probably
a hundred or more astronomers spend most of their time working on SN
light curves. If there were no time dilation, don't you think one of
them would have noticed by now?

And the templates are not an accurate enough guide as one
only has to see how in SN1997eq and ek the actual peak
measurements are well above the inferred template peak of 1


Only if you ignore the error bars.

Incidentally,are you sure that 1995E is 0.01 redshift? It seems
to be in the mid range of `heliocentric redshifts` (on table 3
page 51 Reiss) at 3.54 in a range between 3.1 - 4.5 (or 0.01 - 0.1)
but unfortunately I dont know how to convert `heliocentric` to z.


The table is labelled "log cz." Take 10^(cz) -- sometimes called
"antilog" -- then divide by c=3E5. You have to know the units meant
are km/s. You can check by looking up the galaxy velocity (NGC 2441,
3470 km/s). So yes, I'm sure, z=0.012. I didn't pick this SN at
random, either.

In every post, you have made elementary mistakes. I am not sure what
to suggest, but you need to learn how to interpret data if you want
to pursue this project (or indeed any other in astronomy).

--
Steve Willner Phone 617-495-7123

Cambridge, MA 02138 USA
(Please email your reply if you want to be sure I see it; include a
valid Reply-To address to receive an acknowledgement. Commercial
email may be sent to your ISP.)
  #26  
Old September 16th 04, 10:30 AM
sean
external usenet poster
 
Posts: n/a
Default

(Steve Willner) wrote in message ...


It's pretty obvious if you compare the table and the graph and note
the open/filled circles, as the figure caption explains. The open
circle just below 1.2 represents the 13 "BTC" measurements, the
filled circles at 1.0 are the two HST measurements (3.83, 3.89 in
Table 11), and the open circle at 0.8 with the big error bar
represents the WIYN measurements. Thus "1.0" on the figure is
something like 3.9 in the table (probably the 3.89). One magnitude
fainter is 1.56 in the table, very close to 1.54 measured at day
50846.7. (I didn't pick this SN as an example at random!)


Close enough; I get 5.1. Did you weight the measurements when you
took the average?


I got 4.8 because I used only 12 measurements from the same day
or..day 50817 as I have previously pointed out , whereas you have
included one measurement from the previous day which was one day
before peak
The first point I would like to clarify is this: Am I correct in
assuming that the lightcurve templates used in the graphs are made
by introducing a stretch factor `s` into each lightcurve template?
If so then these templates are not rest frame, but rest frame +
time dilation. Which would make them biased towards time dilation.
However the terminology is complex in the paper and I need a second
opinion on that.

Look at the error bars! The "1.2" is highly uncertain. The two HST
measurements around the same time have much smaller uncertainties.
Those are what establish the peak of the light curve, which
conveniently is put at 1.0 in the graph.


You are completely wrong here and appear to have made yet
another `elementary error`.If you want to look at just
the HST measurements you`ll notice that the first HST reading
is 3.8. The second HST reading is the highest at 3.89 (1.0)
Thats the peak observation. And the next or third HST reading,
is at 1.59 (0.4) and its 22 days after the peak reading !
Thats 22/19=1.15 Or.. no time dilation seeing as you need to
have 1.85 to show time dilation using HST data.
Regarding the other ground based measurements. You ignore the
fact that its not 1 single ground based measurement at 1.2
with `error margins`. But rather, *13* different, seperate
observations over 2 days that average out to 1.2. And if you
say that 3.9 is 1.0 linear flux then at least 3 of those 13
observations do not include 1.0 within their error margins
( 3 are above 5.0 with error margins less than 1.5 each which
puts 3.9(or1.0) outside the error margins of those 3 seperate
observations). Trying to pass off 13 seperate observations
averaging 1.2 linear flux over 2 days, of which 3 do not
include 1.0 within the low end of error margins, as being
most likely 1.0 can only be described as incorrect or at best
highly improbable.
Furthermore you havent commented on the R band measurements
from 1998as (roughly comparable to rest frame B band) which
very clearly give a 20 day decay for 1 mag from 1 to 0.4 as
compared to an expected 19. that gives a 20/19=1.05 ratio. There
is no way even with error margins taken into account that
these observations can support anything but
a `no time dilation `argument.

Or for instance the I band 1998ba where a 1 mag decay from
1.0 to 0.4 gives about 32 days which compares at z=0.43 to
a 569nm lightcurve which is about 27 days so thats 32/27=1.18
Thats much closer to 1 (no time dilation)than the expected
1.43 expected from time dilation .
And thats using the template!

Or take 1998as I band. Using the template its 34/31= 1.09.
MUCH closer to the no time dilation 1.0 than an expected
1.35 for time dilation

1998aw I band template is 33/27= 1.2. Thats still closer
to 1.0 than a expected 1.44 for time dilation

1997ez I band is 30/20=1.5 against the expected 1.78 for
time dilation

1997eq is 30 days for a 1 mag decay from the template, or
using the actual peak observation of 1.15 flux density its
about 25-27 days. Thats 26/22=1.18 or using the template
one gets 30/22=1.36. Time dilation needs 1.54. Both
arguments can get support from this one,

1998ax I band template gives 1.48 against the time dilations
expected 1.5. A notable exception supporting time dilation.

2000fr I band using the template is about 30/22=1.36 and it
should be 1.54

1998bl I band using the template is 25/22=1.13 against time
dilations expected 1.74. Much closer to 1

1998be I band using template is 25/20=1.25 against time dilations
1.64 Closer to 1 than 1.64.

1998ay I band using the template is 31/21=1.47 against time
dilations 1.64

There's nothing wrong with using the templates. In fact that's the
"right" way to do things because the time of maximum, where the light
curve is flat, is hard to determine. Fortunately, the authors have
done the work for you. If you don't like the template, look at the
time to decline an additional 0.78 mag (from 1.54 to 0.75 in Table
11), 12.1 days. The corresponding time for 1995E is about 5 days.


(Theres nothing wrong with using the templates provided they dont
have any `s` factored in . Do they?)
1.54 in ek table 11 is at 29 days (thats 440nm restframe) . To
compare correctly one must go to 29 days in 1995E tables B band
(425 nm restframe being the closest available) or day 9801. There
is no reading there, but a speculative estimate would give it about
19.1 mag. Then if one tries to find 1 mag decay from that one has
to find where 20.1 is on 1995 E tables and there are no readings
for anything lower or later than 19.8 mag!
How did you come up with a 5 day decay for 1995E B band from 19.1
mag to 20.1 mag when there is no available data?
(The V band or 524nm rest frame from 1995E takes 23 days to decay
1 mag from day 29 but there is no comparable 524 nm restframe data
from 1997ek table 11 to compare with as the ek I band is only
440nm restframe.In fact the last reading of 19.86 in the B band
from 1995E which is a bit less than a mag decay from 19.01 on
day 29 is already at about day 46 or 17 days later. Which is
more,rather than less, than the equivelent decay you mention
in 1997ek for the same start time of 29 days.)

There are many more supernovae published than in the Knop et
al. paper. All we are looking for here is a sanity check. Probably
a hundred or more astronomers spend most of their time working on SN
light curves. If there were no time dilation, don't you think one of
them would have noticed by now?


Well if you cant notice despite all the verifiable data I have
supplied to back a `no time dilation` argument, then I dont expect
them to.Thats why I posted here on this newsgroup. Somebody has to
"notice" sooner or later as the vast bulk of observations show
there is no time dilation of SN`s, unless one ignores the data.
Can you explain why 1998as R and I band data show virtually no
time dilation? Can you explain why 1998aw,1998bl and 1998be are
almost at 1 and show very little time dilation? And can you explain
why the rest are all, except 1.. 1998ax, no closer than 1/3 to time
dilation and more often closer to 1 or `no time dilation`?
All above calculations use the templates by the way.
Sean
  #27  
Old September 18th 04, 01:03 PM
Steve Willner
external usenet poster
 
Posts: n/a
Default

(sean) wrote in message ...
I got 4.8 because I used only 12 measurements from the same day
or..day 50817 as I have previously pointed out , whereas you have
included one measurement from the previous day which was one day
before peak


Near peak, the brightness is hardly changing, so it doesn't much
matter which points you include to determine the maximum brightness.

The first point I would like to clarify is this: Am I correct in
assuming that the lightcurve templates used in the graphs are made
by introducing a stretch factor `s` into each lightcurve template?


Before we get to that, let's take a step back and see where we are.
What I've done is tell you how to interpret the data well enough for
a "sanity check" to see that time dilation is obvious. If you want to
do a real analysis, you are going to have to learn a lot more about
data analysis in general and astronomical data in particular: I'd say
about the equivalent of two years of graduate school. That is not
something you will get on Usenet.

If so then these templates are not rest frame, but rest frame +
time dilation. Which would make them biased towards time dilation.


The templates obviously include time dilation; otherwise they wouldn't
fit the data! This does not make them biased. I suggested using the
templates because the work is done for you, and it is easy to see what
is going on. However, if you don't like that, what you need to do is
find the slope of the light curve on the part well after maximum,
where the brightness (in magnitude units) is declining by a constant
amount per day. Compare the slopes for near and distant SNe at the
same magnitudes relative to maximum light. For example, you might
compare the time to go from (max plus 0.75 mag) to (max plus 1.5 mag).
The problem with this is that the distant SNe cannot always be
followed to 1.5 mag fainter than maximum, so I suggested the templates
as a shortcut. They do, after all, incorporate everything we know
about light curves based on nearby SNe.

There is nothing magic about the specific numbers 0.75, 1.5 mag, but
it looks to me as though they ought to work OK. The actual fitting
procedure is left to you, but be sure you take the error bars into
account.

SW Look at the error bars! The "1.2" is highly uncertain. The two
HST
SW measurements around the same time have much smaller uncertainties.
SW Those are what establish the peak of the light curve, which
SW conveniently is put at 1.0 in the graph.

You are completely wrong here and appear to have made yet
another `elementary error`.


This is not a very good way to solicit my help, and in view of the
past history of this thread, you would be wise to adopt a more
skeptical attitude towards your own conclusions. I am beginning to
suspect that you are advocating a preconceived notion rather than
trying to understand the data.

If you want to look at just
the HST measurements you`ll notice that the first HST reading
is 3.8. The second HST reading is the highest at 3.89 (1.0)
Thats the peak observation.


Please look again at what I wrote about error bars and (in the bit you
snipped) about the light curve being flat near maximum. You cannot
simply take the largest single measurement and call it the "time of
maximum." For 1997ek, all the measurements near "day 0" are
consistent with constant brightness lasting 8 days. The "time of
maximum" is in there somewhere. You can either determine it from the
template or else bypass determining the time of maximum altogether by
dealing only with the slope.

And the next or third HST reading,
is at 1.59 (0.4) and its 22 days after the peak reading !


If we use your definition of peak, it is the "5.89" at day 50817.6, 29
days before the HST "1.54" value. (As noted above, this is not a
useful way to determine the time of maximum.)

Regarding the other ground based measurements. You ignore the
fact that its not 1 single ground based measurement at 1.2
with `error margins`. But rather, *13* different, seperate
observations over 2 days that average out to 1.2. And if you
say that 3.9 is 1.0 linear flux then at least 3 of those 13
observations do not include 1.0 within their error margins


This is normal (pun not intended, but I'll leave it in). The error
bars given are standard errors, often called "one sigma." If the
measurement errors are Gaussian (and they should be close to that),
about 1/3 of all measurements will be outside the +/- one sigma
boundaries. This is undergraduate-level data analysis. You will have
to learn at this level before even starting those two years of
graduate school I mentioned.

Furthermore you havent commented on the R band measurements
from 1998as (roughly comparable to rest frame B band) which
very clearly give a 20 day decay for 1 mag from 1 to 0.4 as
compared to an expected 19.


I make it about 22 days. Certainly not as short as 17, which is what
1995E shows at B. In fact, 17*(1.35/1.01) = 22.7, which looks like
pretty good agreement to me. I don't see how you can possibly make
the decay as short as 20 days, let alone 17.

There
is no way even with error margins taken into account that
these observations can support anything but
a `no time dilation `argument.


I would take out the "no." This again gives me the impression that
your mind is made up, regardless of what the data show.

Or for instance the I band 1998ba where a 1 mag decay from
1.0 to 0.4 gives about 32 days which compares at z=0.43 to
a 569nm lightcurve which is about 27 days


The I light curve isn't well sampled; there are no measurements at all
near peak. For R, the decay time from template-fitting is 20 days; we
expect 24 days from 1995E. However, there's a big error bar on the
first measurement, meaning the time of maximum is poorly determined.
This is one where you have to look at the slope of the decay. And of
course for a sanity check, one is best advised to look at the SNe with
the highest redshifts.

I have neither time nor desire to deal with all the others. I have
given you a roadmap; it's up to you to use it or not.
  #28  
Old September 20th 04, 02:47 PM
Craig Markwardt
external usenet poster
 
Posts: n/a
Default

(sean) writes:
....
The first point I would like to clarify is this: Am I correct in
assuming that the lightcurve templates used in the graphs are made
by introducing a stretch factor `s` into each lightcurve template?
If so then these templates are not rest frame, but rest frame +
time dilation. Which would make them biased towards time dilation.
However the terminology is complex in the paper and I need a second
opinion on that.



In a given analysis, there is only one light curve template, and it is
based on the nearby supernova remnant sample. The "stretch factor" s
is introduced to match an individual light curve to the template, with
the understanding that individual supernova light curves can be
slightly slower or faster than the mean.

There would be evidence of bias if the fitting of the light curve
profiles treated the nearby and high-z supernovae separately. BUT,
they were NOT. Both the samples were fitted with the same function,
including a stretch factor. [ref. 1, Tables 1 & 3]

There would be evidence of bias if the fitting of the light curve
profiles allowed only a stretch but not a compression. BUT, they did
NOT. Both stretched and compressed light curves were permitted in the
fitting, and in fact some moderate-z light curves evolved faster than
local ones as a result. [ref. 1, Tables 1 & 3] However, on *average*,
the high-z supernova light curves evolve much more slowly. [ref. 1,
Fig. 3]

Thus, your claim of bias is unsubstantiated; if the high-z light
curves evolved faster on average than low-z ones, this would have been
detected.

And further, your approach of examining single light curves is a path
to problems. Supernova light curves are understood to be variable,
and so your technique is susceptible to picking outliers.

CM

References
1. Goldhaber, et al. 2001, ApJ, 558, 359
  #29  
Old September 20th 04, 02:51 PM
sean
external usenet poster
 
Posts: n/a
Default

(Steve Willner) wrote in message ...

Before we get to that, let's take a step back and see where we are.
What I've done is tell you how to interpret the data well enough for
a "sanity check" to see that time dilation is obvious.

I`ve done that and the results show,.. no time dilation is obvious.
And to back up that argument I`ve supplied the relevent data
again below.
The templates obviously include time dilation; otherwise they wouldn't
fit the data! This does not make them biased. I suggested using the
templates because the work is done for you, and it is easy to see what
is going on.

True the data can be made to fit the template if one uses the lower
end of the error margins, but they dont make as good a fit to the
`time dilation` template as they do to a `no time dilation`
interpretation. Look at 1998ek . The averaged measurements are
1.2 linear with an error margin that includes 1.0 at its lower end,
which is admittedly what a time dilation argument needs. Nonetheless
the middle of the error margin is 1.2 And 1.2 gives the 22/18.5 =1.18
ratio that supports a no time dilation argument. And so, undeniably,
within error margins a `no time dilation` is *also* supported and
because it uses the middle of the error margin range , supported
more strongly. On the whole the templates consistently under
estimate the peak observations where they are available. If one were
to remove the time dilation from the templates the peaks in the new
undilated templates would always be higher and that in turn would
give shorter time decays for a one mag decay from peak etc which in
turn would bring down the already favorable results closer to
`No time dilation`
However, if you don't like that, what you need to do is
find the slope of the light curve on the part well after maximum,
where the brightness (in magnitude units) is declining by a constant
amount per day. Compare the slopes for near and distant SNe at the
same magnitudes relative to maximum light. For example, you might
compare the time to go from (max plus 0.75 mag) to (max plus 1.5 mag).
The problem with this is that the distant SNe cannot always be
followed to 1.5 mag fainter than maximum, so I suggested the templates
as a shortcut. They do, after all, incorporate everything we know
about light curves based on nearby SNe.

There is nothing magic about the specific numbers 0.75, 1.5 mag, but
it looks to me as though they ought to work OK. The actual fitting
procedure is left to you, but be sure you take the error bars into
account.

OK. I`ll use peak + 1mag to peak +2mag seeing as you say its not
that important to use 0.75 and 1.5 mag. And what are the results?...
just as I expected.
1997ek (440nm restframe) I band takes 14 days to decay from
peak+1 mag to peak +2mag
1995S 440nm Restframe takes 12 days to decay from peak+1 mag
to peak +2 mag
Thats 14/12= 1.16 Now considering it should be 1.85 for a time
dilation argument I would say that 1.16 definitely supports a
`no time dilation` conclusion. What can I say Steve, no matter
what test you set, it always ends up supporting a
`no time dilation` outcome.

This is not a very good way to solicit my help, and in view of the
past history of this thread, you would be wise to adopt a more
skeptical attitude towards your own conclusions. I am beginning to
suspect that you are advocating a preconceived notion rather than
trying to understand the data.

But you have made mistakes so its not unfair of me to point them
out.
How else am I to argue the case for no time dilation?
Here for instance, You just made another in your latest post. You
claim below that 5.89 at day 50817.6, is 29 days before the HST
1.54 value.
The suggestion you are trying to make is that there is a 1
mag difference between 5.89 and 1.54! But thats a mistake because
it`s more than a 1 mag difference. It even contradicts a statement
you make in an earlier post where you claim that it takes 29 days
to decay 1 mag from 3.89 to 1.54 for the same SN! Surely you cant
accuse me of being rude or unhelpful when I point out obvious
errors like this?
And you too,are also advocating a `preconceived notion` except
yours is: time dilation. Are you not?
Anyways I`m sorry. I very much appreciate your input , dont get
me wrong. After all ,in a previous post it was *your* suggestion
that the best way to test for time dilation was to compare time
decays from peak mag between low and high redshift SN`s. And
thats exactly what I have done. And it helps my argument for
`no time dilation`. So I am in debted to you for giving me
an extra method to support a `no time dilation` interpretation
of data. Unfortunately you now say its *no*t an acceptable method.
In case you have some time ,these are the results of comparing
high redshift Sn`s I band lightcurves with lower redshift
comparable lightcurves. Notice how most support `no time dilation`

I band 1998ba where a 1 mag decay from
1.0 to 0.4 gives about 32 days which compares at z=0.43 to
a 569nm lightcurve which is about 27 days so thats 32/27=1.18
Thats much closer to 1 (no time dilation)than the expected
1.43 expected from time dilation .
And thats using the template!

Or take 1998as I band. Using the template its 34/27.5= 1.24
about 1/2 way but admittedly slightly favoring time dilation

1998aw I band template is 33/27= 1.2. Thats still closer
to 1.0 than a expected 1.44 for time dilation

1997ez I band is 30/20=1.5 against the expected 1.78 for
time dilation

1997eq is 30 days for a 1 mag decay from the template, or
using the actual peak observation of 1.15 flux density its
about 25-27 days. Thats 26/22=1.18 or using the template
one gets 30/22=1.36. Time dilation needs 1.54. Both
arguments can get support from this one,

1998ax I band template gives 1.48 against the time dilations
expected 1.5. A notable exception supporting time dilation.

2000fr I band using the template is about 30/22=1.36 and it
should be 1.54

1998bl I band using the template is 25/22=1.13 against time
dilations expected 1.74. Much closer to 1

1998be I band using template is 25/20=1.25 against time dilations
1.64 Closer to 1 than 1.64.

1998ay I band using the template is 31/21=1.47 against time
dilations 1.64

If you want to look at just
the HST measurements you`ll notice that the first HST reading
is 3.8. The second HST reading is the highest at 3.89 (1.0)
Thats the peak observation.


Please look again at what I wrote about error bars and (in the bit you
snipped) about the light curve being flat near maximum. You cannot
simply take the largest single measurement and call it the "time of
maximum." For 1997ek, all the measurements near "day 0" are
consistent with constant brightness lasting 8 days. The "time of
maximum" is in there somewhere. You can either determine it from the
template or else bypass determining the time of maximum altogether by
dealing only with the slope.
If we use your definition of peak, it is the "5.89" at day 50817.6, 29
days before the HST "1.54" value. (As noted above, this is not a
useful way to determine the time of maximum.)

As I mentioned before,...
The 5.89 reading at day 50817.6 is much more than 1 mag above 1.54.
So its irrelevent whether it takes 29 days to decay from 5.89 to
1.54 as a 1 mag decay from 5.89 would be to something much higher
than 1.54 and therefore less than 29 days.
I make it about 22 days. Certainly not as short as 17, which is what
1995E shows at B. In fact, 17*(1.35/1.01) = 22.7, which looks like
pretty good agreement to me. I don't see how you can possibly make
the decay as short as 20 days, let alone 17.

OK maybe 21 at a pinch and for you I`ll allow 22. But dont forget
that its being compared against a 444nm lightcurve , and not a B
band... (B band is 430/1.01=424nm restframe and V band
is 530/1.01=524nm). And according to Reiss` low redshift survey
424nm (B) takes 17 days to decay 1 mag and 524nm(I) takes approx
22 days to decay 1 mag from peak For 1995E and thus a 444nm
lightcurve is about 18 . So if the the 1998as R band restframe
of 444nm (600/1.85=444nm) decay is about 22.0 days decay for a
1 mag decay from peak then the ratio is at the outside...
22/18 = 1.2! Now thats a lot closer to 1.0 (no time dilation)
than 1.85(time dilation) no matter what spin one puts on it.
I also noticed that you managed to find one of the shortest
low redshift V band decays available at 21 days. Notice below
the other SN`s around the same 1995E redshift are all more
than 21 days. If I took lets say 1995S as an example then it
would be the following ratio 22/18.4=1.19. Thats even closer
to 1.0 than 1.85 and still a valid ratio as its using 1995S
data for B and V band.(the second column is heliocentric redshift)
1995al=3.18 = 23 days for B band
1995ac=3.1 = 22 days " "
1995E =3.5 = 21 days " " or 18 days for 440nm
1995S =3.6 = 23 days " " or 18.4 days for 440nm
1995bd=3.6 = 22 days " "
The I light curve isn't well sampled; there are no measurements at all
near peak. For R, the decay time from template-fitting is 20 days; we
expect 24 days from 1995E. However, there's a big error bar on the
first measurement, meaning the time of maximum is poorly determined.

The R band (restframe 419nm) yes, I agree, is 20 days using the
template and that compares approximately with the B band from SN1995E
(rest frame 424nm) which is about 18.5 days so one gets the ratio
20/18.4 or 1.08! Almost 1.0 or no time dilation . Thats supposed to
be 1.43 if there were time dilation. And the error bars are smaller
than those on SN1997ek I band measurements which you seemed quite
happy to use.
This is one where you have to look at the slope of the decay. And of
course for a sanity check, one is best advised to look at the SNe with
the highest redshifts.OK I`ll try the peak+x to peak+y method you now prefer on SN1998ba

R band (restframe 419nm ). It takes 12 days to decay from peak+1 mag
to peak+2 mag. (thats 0.4 to 0.16 on the graph)
And if the B band from 1994S is restframe 425nm and takes 10 days for
peak+1 to peak+2 then thats roughly comparable. So the ratio
is 12/10=1.2. Well thats about half way between both arguments
as it should be 1.43 for time dilation.
Then again dont forget the template `peak` is at the lower end of
the peak observations `error margin` so a time dilation argument
could use a peak ,within error margins, of about 1.1. That brings
forward the peak+1 mag to peak+2 mag timeframe on the lightcurve
which becomes steeper closer to peak. That in turn will reduce the
time decay and bring the ratio down closer to 1. By how much I dont
know, But as the 1.2 ratio is closer to 1.0 already I think this
favors no time dilation anyways , although by a smaller margin
than most of the others.
Using the I band (569 restframe) template one gets about a 22 day
decay for peak+1 to peak+2 decay. The comparable decay from 1994S
is 20 days for 525nm so we are looking at about at least an even
ratio of 22/22=1 if not less than 1.0. Thats very much no time
dilation. Mixed results there but still *definitely* favoring NO time
dilation, especially in I band.So what do you say? Am I wrong above?
If you think I am, dont just say "your wrong". Prove it by trying the
calculations yourself using peak +1 to peak+2 on the same SN and
then post your calculations. I`m proving a `no time dilation`
argument using only *your methods* and accepted data from NASA etc.
You can say that undergraduates study this or that for so many
years but it doesnt alter the fact that an undergraduate, or Steven
Hawking would both come to the same conclusions I have. That is IF
they use the data and do the simple calculations, which you
yourself have suggested,or at least agreed upon, correctly.One
doesnt have to be a graduate in astrophysics to see there is
a stronger argument for No time dilation then there is for
time dilation
Sean
  #30  
Old September 22nd 04, 12:35 PM
sean
external usenet poster
 
Posts: n/a
Default

Craig Markwardt wrote in message ...
In a given analysis, there is only one light curve template, and it is
based on the nearby supernova remnant sample. The "stretch factor" s
is introduced to match an individual light curve to the template, with
the understanding that individual supernova light curves can be
slightly slower or faster than the mean.
There would be evidence of bias if the fitting of the light curve
profiles treated the nearby and high-z supernovae separately. BUT,
they were NOT. Both the samples were fitted with the same function,
including a stretch factor. [ref. 1, Tables 1 & 3]
There would be evidence of bias if the fitting of the light curve
profiles allowed only a stretch but not a compression. BUT, they did
NOT. Both stretched and compressed light curves were permitted in the
fitting, and in fact some moderate-z light curves evolved faster than
local ones as a result. [ref. 1, Tables 1 & 3] However, on *average*,
the high-z supernova light curves evolve much more slowly. [ref. 1,
References
1. Goldhaber, et al. 2001, ApJ, 558, 359

It is Goldhaber himself who has put the time dilation into the SN
data by introducing a stretch s proportional to redshift. Take out
the stretch from the data (and the k correction to B,and the
template fits to R or parab-18) and the apparent time dilation will
disappear.
The closest available data that I can find, supplied by Goldhaber
seems to be fig 1a and 1b which do not have the k correction or
stretch but do have a template fit to R.? If there were a
lightcurve or table of the SN data without this template fit to
parab-18 for both the C-T and SCP SN data, I am sure that the
apparent `time dilation` would disappear and what would remain
would be only a range of lightcurves reflecting the various different
restframe emission wavelengths of the SN`s studied. However, its
not that clear to me what the filter bands the observations were
made in but I believe that its R band observed for all the SCP
SN`S and B band observed for all the C-T SN`s. These then appear
to have been fitted to R and B band templates respectively. You
may possibly be able to clarify this part of the process a
bit more.
One question here regarding 2 SN`s from Knop may provide an answer
as to why even with the s factored in, Knops` results still favor
no time dilation...Why do you 1998ay and 1998be have different
`s` values applied? The two lightcurves graphs are
distinctinctly different from each other with 1998be having
faster decays in both I and R. The stretch value s in the
table 3 Knop is also different for each. 1998ay is s=1.04
while 1998be is s=0.8. Yet they are both the same redshift
at z=0.64 . I would have thought that the same stretch factor s
would be applied to both SN`s seeing as they both have the
same redshift.
Sean
(My apologies to Steve W. In my last post I incorrectly stated
that 1998as is z=0.85 when in fact its 0.35. Sorry)
 




Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
French's Primordial Study and Schramm & Turner, 1997 greywolf42 Astronomy Misc 19 July 11th 04 06:43 PM
FAQ-2-B: sci.space.tech reading list dave schneider Technology 11 June 10th 04 03:54 AM
Spirit has a mind of its own? Jon Berndt Space Shuttle 33 January 28th 04 05:48 AM
FAQ-2-B: sci.space.tech reading list dave schneider Technology 23 January 21st 04 12:42 AM
A single data point. Rich SETI 2 October 8th 03 06:02 AM


All times are GMT +1. The time now is 02:36 AM.


Powered by vBulletin® Version 3.6.4
Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
Copyright ©2004-2024 SpaceBanter.com.
The comments are property of their posters.