----------------------------------------------------------------------
--- Log opened Mon Jan 21 00:00:04 2008
[...]
13:59 -!- nicolas [...] has joined #cosmo
14:05 -!- You're now known as boud
14:05 -!- mode/#cosmo [+o nicolas] by boud
14:05 <@boud> nicolas salut :)
14:05 <@nicolas> bonjour
14:06 <@nicolas> (non je ne me suis pas trompé j'ai eu peur d'un
quelconque décalage horaire)
14:06 <@boud> :P
14:06 <@nicolas> je t'ai envoyé un mail tout ? l'heure aussi
14:06 <@boud> je vois que bruno est venu a 11:23, sans doute
pour tester, et il a quitte a 11:27
14:07 -!- Bruno [...] has joined #cosmo
14:07 < Bruno> hello there
14:07 <@boud> Bruno hi :)
14:07 <@nicolas> Hello
14:08 < Bruno> Hello Nicolas, . Gaudin, i presume?
14:08 <@nicolas> yes
14:09 < Bruno> well gentlemen, thanks for being willing to be
interviewed
14:09 <@boud> alstublieft
14:09 * boud not sure of the spelling :P
14:09 < Bruno> ?t's perfect, polite version even
14:10 <@boud> :)
14:10 < Bruno> this channel is new to me, do you know if a log
is automatically saved?
14:10 <@boud> i was just going to say that - i am logging
14:10 <@boud> one or both of you can log too, depending on what
client you use
14:11 <@boud> we should decide if we want the log to be private
among us, or should it be considered public
14:12 <@boud> bruno ?
14:12 < Bruno> anything is fine by me, but obviously I will quote
(very little) parts of it in the story, OK
14:12 < Bruno> ?
14:13 <@nicolas> it is boud who decides, I just follow the discuss
..
14:13 <@boud> ok, in that case i propose that after the end of the
discussion, i post a copy somewhere "public" e.g. to the
http://cosmo.torun.pl/mailman/listinfo/cosmo-media mailing list
14:13 < Bruno> ok
14:14 <@boud> if someone accidentally says personal/private type
info, please say so and i'll remove it from the log before posting
14:14 < Bruno> sure
14:14 <@boud> ok, i'm ready for questions...
14:15 < Bruno> well, is the universe finite or not?
14:15 <@boud> i don't know
14:15 <@boud> i would *like* to know that and we have done work
towards answering that question, but that's not the same thing
14:16 < Bruno> what would you argue, after publishing/posting
your last paper?
14:17 <@boud> scientifically, i would say just what is said in
our paper
14:17 <@boud> i guess you want a "popular science" compressed
version...?
14:19 <@boud> maybe something like "the evidence seems to be
cumulating in favour of a finite (PDS) universe"
14:20 < Bruno> well, let me put it this way: the diagrams in
the paper look pretty clear to me
14:20 <@boud> thanks :)
14:20 < Bruno> what would they look like if there was no PDS
signal at all?
14:21 <@boud> to give a fully quantitative answer to that question
would/will require a lot of work
14:22 < Bruno> but would you expect something like a pretty evenly
distributed cloud of dots?
14:22 <@boud> intuitively, yes
14:22 <@boud> probably not totally evenly distributed
14:23 <@boud> but i didn't expect such a strong concentration
14:23 < Bruno> right
14:23 <@boud> however, my intuition does not always much a full,
detailed calculation
14:24 < Bruno> in what essential way does this result differ
from the 2004 result?
14:24 <@boud> probably several ways:
14:25 <@boud> (1) we use a much larger amount of the data (temperature
fluctuations in WMAP map) and should be less sensitive to small
errors
14:26 <@boud> (2) by allowing for arbitrary twist angle, we are
able to use a fairly simple method of estimating the probability
14:26 <@boud> (2 continued) in the 2004 result, we didn't calculate
a formal probability or other statistical statement - in this
case we did
14:27 <@boud> (3) using the new method, the 2004 result is not
a stable solution and leads (through the Markov Chains) to the
new result
14:28 <@boud> ok i think that's the main 3 differences
14:29 <@boud> correction to a line above: "match" not "much"
- however, my intuition does not always match a full, detailed
calculation
14:30 < Bruno> Roger
14:32 < Bruno> could you please explain a bit on (3): has the
PDS orientation moved since then? If so, how?
14:33 <@boud> the coordinates of the face centres of the fundamental
domain of the PDS (the dodecahedron) are listed in the abstracts
of the two articles
14:33 <@boud> l = galactic longitude, b = galactic latitude
14:33 <@boud> they are two very different sets of coordinates
14:34 <@boud> also, in the 2004 article the twist was -36 degrees;
in the new article it is +39 \pm 3 degrees
14:35 <@boud> i'm not totally sure if this is what you mean by
"moved" ?
14:36 < Bruno> well, they didn't really move, probably, but is
there an explanation why the two calculations came out so differently?
14:36 < Bruno> please replace 'probably' with 'of course'
14:36 <@boud> i think that (1) is probably the reason
14:36 < Bruno> right
14:37 < Bruno> what does alpha =~20 mean (if we suppose there
is a PDS universe)?
14:38 <@boud> this is discussed in section 4.2
14:38 <@boud> it means that this method probably does not give
an accurate estimate of alpha - the matched circle size
14:39 <@boud> have a look at Fig 9
14:39 < Bruno> OK
14:39 <@boud> let's suppose that the alpha in the diagram is
the true alpha and that the PDS model is correct
14:40 < Bruno> yes
14:41 <@boud> the matched circle in this diagram consists (in
the plane of the paper or your screen) of just two points
14:41 <@boud> the two points where the two spheres intersect
14:41 < Bruno> hm i understand
14:41 <@boud> ok so far ?
14:41 <@boud> ok
14:41 <@boud> now look at the pair of points P and P'
14:42 <@boud> they are also "fairly" close to one another
14:42 <@boud> they're not exactly at the same point, but they're
close
14:43 <@boud> so, again under the assumption that we have the
correct model here with the correct alpha, the temperature fluctuations
at P and P' should be "nearly" the same - they should be well
correlated
14:43 < Bruno> ok
14:43 < Bruno> is alpha is small
14:43 < Bruno> IF alpha is small, i mean
14:43 <@boud> yes
14:43 <@boud> if alpha is big, the argument is wrong
14:44 < Bruno> ok
14:44 <@boud> now, the whole idea of using the cross-correlation
is that we will include pairs which are "close" such as P and
P', not just points which are "exactly" at the same position
14:45 <@boud> now imagine shifting the left sphere a bit to the
left - keep the right sphere stationary
14:45 < Bruno> ok, i thought you were matching up rings, or is
that the same?
14:46 < Bruno> OK, i shifted the left sphere
14:46 <@boud> "rings" was what i gave the NS reporter at the
last minute when she wanted to write "circles"
14:47 <@boud> i'll get back to "rings" in a moment...
14:47 < Bruno> these reporters...but that's what I understood
from the 1999 scientific american paper by luminet as well,
14:47 <@boud> now that you have shifted the left sphere further
to the left, alpha is smaller
14:47 < Bruno> OK
14:47 < Bruno> yes..
14:48 <@boud> and P and P' are still close - in fact, they are
even closer to one another than before
14:48 < Bruno> uhuh
14:48 <@boud> since our method uses all pairs of points which
are "close", we will find a good correlation for this new, smaller
alpha, even though it's smaller than the true alpha
14:49 < Bruno> I see
14:50 <@boud> so if the true alpha is, say, 40 degrees, then
we should find high cross-correlations not only for alpha = 40
deg, but also for many smaller values of alpha
14:50 <@boud> and also for alpha values a little (but not too
much) bigger
14:51 <@boud> in the figure caption we write this as "approximately
matched annuli" or "approximately matched discs" - if the true
alpha is small enough, then it the full discs (including points
P and P') will be matched
14:51 <@boud> and that gets us to "rings" - "ring" is what i
suggested to the NS reporter as a less frightening (to the lay
public) word than "annuli"
14:52 <@boud> and in fact the latin root of "annulus" is "ring"
(in modern French we have "anneau" for english "ring")
14:53 <@boud> we haven't made a proper model of this, so that
why we say our estimate of alpha should be considered "~" as in
"with a big uncertainty"
14:54 < Bruno> allright, so but to make sure I understand, does
the MCMC look for matched rings/annuli or disks then? (sorry to
be dim, if I am)
14:55 <@boud> (you're not dim - it's the fault of the NS
reporter/method-of-journalism...)
14:56 <@boud> the MCMC is certainly *not* defined to look for
matched annuli/disks
14:57 <@boud> however, to *some* extent what it really does is
probably partly equivalent to looking for matched annuli/disks
14:58 <@boud> it looks for high correlations based on the set
of all points which are "close" under a PDS assumption at a given
orientation, circle size and twist
14:59 <@boud> "close" can be in any direction, but of course
it's related to the PDS geometry/topology
15:01 < Bruno> close in the sense you just explained, right?
15:01 <@boud> yes
15:02 <@boud> figs 2 and 3 show this
15:03 < Bruno> OK, i'm getting there
15:04 <@boud> i think the idea of "matched annuli/disks" is a
useful idea, but there would be several complications to relate
it to what we did - it could probably be the subject of a whole
new article...
15:06 < Bruno> allright, so i'll probably stick with something
more abstract But the idea that two intersecting spheres would
yield a matching ring, even in the PDS model is OK?
15:06 <@nicolas> so boud, the corellation fonction, evaluating
between two "distance" does that all points are in an annuli,
no ?
15:06 <@nicolas> (I am not sure for my english, sorry)
15:09 <@boud> nicolas : i'm not totally sure - maybe, maybe not...
15:10 <@boud> bruno: i agree that two intersecting spheres for
the PDS model should yield "approximately" matching rings, but
there is a complication here:
15:10 <@boud> start with two spheres - these define a pair of
matching circles
15:11 < Bruno> yes..
15:11 <@boud> now we allow these circles to thicken - so they
become annuli
15:11 <@boud> however...
15:12 <@boud> what we see in Fig 9 is that they thicken slowly
outwards, but faster inwards, as we increase from "zero separation"
to "small separation = 'close'"
15:13 <@boud> if we imagine painting some lines along the matched
circles (like in the middle of a road), and then we look at the
"nearly" matching annuli, we will see that the "middle of the
road marker" is not in the middle of the annulus, it's closer
to the outside limit of the annulus
15:14 < Bruno> ok
15:18 < Bruno> ok, it's intricate,I understand. Now a perfect
match in the CMB would mean that the SLS at t=380.000 years doesn't
fit in the universe dodecahedron (assuming PDS), so it 'crosses
itself', yes?
15:21 <@boud> i think that's close enough to correct, except
that better than "universe dodecahedron" would be "fundamental
domain" or "one copy of the fundamental spherical dodecahedron",
though i realise that that might confuse your non-specialist readers...
15:21 <@boud> here, "spherical" means "inside of a 3-space with
positive curvature"
15:21 <@boud> it doesn't mean "like a 2-sphere"
15:22 < Bruno> yes sir, i have to cut corners somewhere, not
only for myself
15:22 <@boud> you could probably say "fundamental dodecahedron"
15:22 < Bruno> I will
15:23 < Bruno> Another thing, I know it's early days, but any
word from A&A? And responses from other colleagues?
15:23 <@boud> we're waiting for a referee report
15:24 <@boud> other colleagues have said they're interested and
are reading our paper...
15:24 <@boud> the time scale for this is months, not days
15:25 < Bruno> i know
15:25 <@boud> (it takes time to properly analyse stuff, check
calculations, recheck, etc. etc. and then write it up consistently
with what you've really done...)
15:26 < Bruno> now is it a problem that you have phi=39 plm 2
instead of 36?
15:27 <@boud> not at all
15:27 <@boud> one sigma difference is reasonable - that's what
uncertainty estimates are all about
15:27 < Bruno> OK
15:28 <@boud> if it turns into 39 pm 0.01 , then of course it
would be a problem for the model
15:28 < Bruno> Right, that would really be something exotic.
Now what's next? wait for Planck?
15:30 <@boud> well, we do have some ideas of other things we
could check, but they're not published yet, and will require a
bit of work...
15:30 <@boud> and of course, Planck will come up with some very
useful new data
15:30 <@boud> getting back to your question about what colleagues
think...
15:31 <@boud> there's a bit of discussion on cosmocoffee about
a totally unrelated paper about results from the ACBAR CMB experiment:
http://cosmocoffee.info/viewtopic.php?t=1038
15:31 <@boud> they don't actually mention what is one of the
more interesting results from the new ACBAR paper
15:32 <@boud> you'll find it on page 14 at the end of section
8.2 of the paper: http://cosmocoffee.info/arxivref.php?file=pdf/0801.1491
15:34 <@boud> they continue to agree with previous results that
the best estimate for the curvature of the Universe is a small
positive curvature: they get Omega_total = 1.03 (errors: +0.06,
-0.04)
15:35 <@boud> this is consistent with a flat universe, but it
also seems that now for several years, people doing this sort
of work continually get Omega_total in the range 1.015 to 1.02
to 1.03
15:36 < Bruno> which is consistent with PDS..
15:36 <@boud> yes - on page 19 we discuss this
15:36 <@boud> section 5.7
15:37 <@boud> especially if the true alpha is about 40-50 degrees
- which gives about 1.015 to 1.018 for Omega_total
15:38 <@boud> as we discussed earlier, in our own work we don't
have a good constraint on alpha, but an alpha value somewhere
near the maximum alpha values in Fig. 8 would be reasonable
15:38 <@boud> would be a reasonable interpretation
15:40 < Bruno> Also 40-50 degrees. Do you suppose we'll know
for (reasonably) sure after Planck? Some other time limit?
15:42 < Bruno> would 40-50 degrees say anything about the size
of the fundamental dodecahedron?
15:42 <@boud> i'll answer the last question first: yes most definitely
15:43 <@boud> alpha is the angular radius of the matched circles
15:43 <@boud> equation (15) gives you the exact formula you need
15:44 <@boud> rSLS is close to 9.5h^{-1} Gpc (where you can
put in h = 0.71 = H_0 / 100km/s/Mpc is the Hubble constant)
15:45 <@boud> it only changes slowly if we change Omega_total
and/or Omega_matter
15:45 < Bruno> but would be the size at t=380 000 years, right?
15:46 <@boud> yes
15:46 <@boud> well, i'm not sure if it's exactly 380 000 years,
i don't usually use a number there...
15:47 < Bruno> Ok,but it has expanded a lot since then, hasn't
it?
15:47 <@boud> each local region has locally expanded by a factor
of about 1100, yes
15:48 <@boud> however, all of this work is done in comoving coordinates
15:48 <@boud> this is a choice of coordinates in which the universe
is static
15:48 <@boud> all the expansion is then put into a variable we
call "the scale factor" - usually written a(t)
15:49 <@nicolas> and comoving coordinates = real coordinates
only actually
15:50 <@boud> from equation (15), since you have rSLS and alpha,
you can deduce R_C
15:50 < Bruno> am I terribly wrong in saying that the fundamental
dodecahedron has grown 1100-fold (if it exists)?
15:52 <@boud> you're correct in saying that, but the same thing
applies to the SLS, it too has grown something like 1100-fold
since when the Universe was ~ 4 x 10^5 years old
15:52 < Bruno> OK
15:53 < Bruno> Well, I think I have probed as far as is possible/useful/bearable
for you
15:53 <@boud> nicolas - i would be careful with the word "real"
here - from a point of view of local physical experiments, comoving
coordinates are just a theoretical construct, and "physical/local/proper"
coordinates are more "real"
15:54 <@boud> well, i was just going to finish with eq. 15...
15:54 <@boud> to talk about a "size" of the fundamental dodecahedron,
there are a few different ways you can define "size"
15:55 <@boud> but one way is to take the distance from one face
centre to the opposite face centre
15:55 < Bruno> yes..
15:55 <@boud> if we are thinking in R^4, then this is an angle
of pi/5 (36 degrees)
15:56 <@boud> so then this "size" is pi/5 * R_C
15:56 <@boud> so then you have it in some length units (e.g.
Gpc)
15:58 < Bruno> and it is...
15:59 <@boud> if i've done it right just now in octave, then
for alpha = 40-50 degrees i get 13-15 h^{-1} Gpc
16:00 <@boud> if you want you can put h=0.71 in that, as some
people do nowadays, but i prefer the old style of using h^{-1}Gpc
as the "length unit"
16:01 <@boud> the diameter of the SLS is around 19 h^{-1} Gpc
16:01 < Bruno> You mean 19 Gpc?
16:01 <@boud> no, i mean 19 h^{-1} Gpc
16:02 <@boud> h = H_0 / 100km/s/Mpc is the Hubble constant
16:02 < Bruno> OK, I understand
16:02 <@boud> for most of the XX century, there was a very big
uncertainty in the value of h
16:03 <@boud> and it cancels out of many calculations
16:03 <@boud> so in many cases, we don't actually need to use
a number there
16:04 <@boud> this is \LaTeX notation h^{-1} = 1/h
16:04 < Bruno> I get it
16:06 < Bruno> Boud, thanks for your patient explanation, I will
do my best to put (some of) it to good use in the piece
16:06 <@boud> ok, my pleasure :)
16:07 < Bruno> one more (slightly off-topic) thing:
[off-topic thing removed from log]
16:12 < Bruno> OK, boud, I'll be in touch then, Nicolas, goodbye
and good luck to you both in getting to grips with the universe
16:12 <@boud> ok, tot siens
16:13 <@nicolas> goodbye
16:13 < Bruno> tot ziens, oant sjen (Frisian)
16:14 -!- Bruno [...] has quit
----------------------------------------------------------------------
hi cosmo-media,
A science journalist wants an interview, most probably about
http://arXiv.org/abs/0801.0006, so that s/he can choose one sentence
to publish from a telephone conversation of probably 10-20 minutes or
more....
The journalist refuses to use electronic communication, even jabber!
S/he will probably telephone tomorrow Friday 4 Jan 2008 at 15:00.
S/he also wants our comments on Niarchou & Jaffe (NJ07):
http://arxiv.org/abs/astro-ph/0702436v2http://space.newscientist.com/article/mg19225811.300-fold-testament-what-sh…
My initial thoughts are:
# practical: to get the references to 0702436, download the source
and paste the .bbl file into the .tex file, then latex && latex && dvips
* the NJ07 method is less physically direct than ours, since it
focuses on spherical harmonics on the 2-sphere rather working directly
in comoving 3-space;
* NJ07 is a perturbation simulation approach: the limitations
of this approach are described in Section 1.2.1 of
Roukema (2000) MNRAS, 312, 712, arXiv:astro-ph/9910272;
* specifically, they assume that even on the largest scales, the
function P(k) is a power law k^n function and that the phases of the
perturbations are Gaussian distributed, which at the largest scales
in a multiply connected universe is an assumption with neither
observational nor theoretical support - it might be right, it might
be wrong;
* the closest papers to their method are probably Caillerie et
al. 2007 arXiv:0705.0217 which works directly with the eigenmodes of
the PDS, up to k = 230, where k = l \frac{R_C}{c/H_0}, and and Aurich
2005b arXiv:astro-ph/0412569; NJ07 use up to "beta" = 41 which seems
to more or less correspond to k, i.e. they have 5 times less
resolution than Caillerie et al.; Caillerie et al's conclusion
is that the PDS fits WMAP better than the infinite flat model;
Aurich et al. say that it's not yet possible to make "firm"
conclusions.
* apart from the normalisation of their probabilities, their Fig. 5
looks like it agrees with Aurich et al. and Caillerie et al for
Omega_{tot} \sim 1.015-1.018
However, the journalist probably wants "sound bites" for the
"human interest factor" rather than "just science". My suggestions
for sound bites if it's about NJ07:
* "NJ07 have done some interesting work, but their method requires
making some assumptions about theoretical models which have no
physical motivation to be correct on the largest scales of a PDS
universe. We avoid making this type of assumption and use the
observations directly."
* "The apparent space associated with the PDS is the three-dimensional
space called the hypersphere or 3-sphere. So we work in the
hypersphere, embedded in Euclidean 4-space, which makes it relatively
easy to work with. NJ07 work primarily on the 2-sphere instead of on
the 3-sphere. This could cause some problems."
* "We agree with their final phrase 'the issue of the topology of the
Universe is far from settled'."
As for sound bites about our own work.... i don't have any good ideas
so far.
pozdr
boud
