Cosmological Simulations: Illustris Or Illusion?

Cosmological simulations can be quite useful.
They can lead to definitive predictions by which
one can test our cosmological knowledge.

A nice case in point is the definitive prediction
that there would be roughly 1,000 CDM dark matter
"subhalos" for every galaxy. This prediction was
generated by one of the first comprehensive
simulations (possibly the Millennium Simulation?).
Unfortunately, this vast population of "subhalos"
has been largely a "no show", as has "cold dark
matter" in general.

[Mod. note: not clear what you mean by inserting 'definitive' before
'prediction' here -- just habit, perhaps? Numerical models do not make
definitive predictions. -- mjh]

Cosmological simulations also offer considerable
potential for retrodictions. They show us how well
our model-building efforts to mimic nature are working.
The caveat here is that such retrodictions can create
a false sense of success if they are treated as
predictions or anything comparable to predictions.

Given a large number of dedicated workers and over
50 years of tinkering with a very complicated set
of models involving numerous adjustable parameters
and theoretical add-ons, is it surprising that the
set of models can be used to reproduce observations
fairly well? Hardly! The retrodictions are better at
illuminating the shortcomings of the models rather
than at demonstrating their accuracy in the description
of nature.

So the brand new Illustris Simulation is useful, but
one should note Joel Primack's pointed warning about how
this simulation is evaluated scientifically (see his
comments in Science).

In article , "Robert L.
Oldershaw" writes:

Cosmological simulations can be quite useful.
They can lead to definitive predictions by which
one can test our cosmological knowledge.

A nice case in point is the definitive prediction
that there would be roughly 1,000 CDM dark matter
"subhalos" for every galaxy. This prediction was
generated by one of the first comprehensive
simulations (possibly the Millennium Simulation?).
Unfortunately, this vast population of "subhalos"
has been largely a "no show", as has "cold dark
matter" in general.

[Mod. note: not clear what you mean by inserting 'definitive' before
'prediction' here -- just habit, perhaps? Numerical models do not make
definitive predictions. -- mjh]

Right.

Also, if it was generated by one of the first comprehensive simulations,
it has hardly been checked.

Also is the issue whether one can OBSERVE these halos. That depends on
the details of star formation, which is much trickier.

Also, less than two months ago I heard a talk which concluded that
inserting more and better physics into such simulations makes
ameliorates the problem.

On Tuesday, May 20, 2014 2:32:54 AM UTC-4, MODERATOR wrote:
[Mod. note: not clear what you mean by inserting 'definitive' before
'prediction' here -- just habit, perhaps? Numerical models do not make
definitive predictions. -- mjh]


The prediction of a vast population of CDM "subhalos"
fits my definition of a definitive prediction.

For over 30 years, in a large number of venues,
I have promoted the idea of definitive predictions.

DEFINITION: predictions that are (1) prior to relevant
observations, (2) feasibly testable, (3) quantitative,
(4) NON-ADJUSTABLE, and (5) unique to the theory being
tested.

[Mod. note: predictions made by numerical models in general, and the
ones we are discussing in particular, fail on, at a minimum, 4;
probably 1 and 5 as well -- mjh]

The term "definitive prediction" has been used and
defined explicitly in published papers and at many
commentary sites, and in letters to editors, etc.

But perhaps the message has not been received?

On Tuesday, May 20, 2014 2:15:48 PM UTC-4, Robert L. Oldershaw wrote:

DEFINITION: predictions that are (1) prior to relevant
observations, (2) feasibly testable, (3) quantitative,
(4) NON-ADJUSTABLE, and (5) unique to the theory being
tested.

Others have pointed out why such a definition is not particularly
useful especially when judging physics calculations in an
observational science like astrophysics.

But I want to point out that using this definition, the prediction of
inflation of the mass density of the universe is a perfect example of
a definitive prediction. Consider

(1) There are several varieties of inflation but they all predict that
Omega (the ratio of the observed mass density to the closure density)
should be observed to be exactly 1. This prediction is completely
non-adjustable.

(2) In 1979, when the first inflation models were developed, the
consensus value of Omega was 0.1 (Look at the (in)famous paper
written by 4 of the most noted astronomers of the time
http://adsabs.harvard.edu/abs/1974ApJ...194..543G ) So not only did
inflation give a definitive prediction, it gave a *bold* definitive
prediction, in that it went against the current consensus.

(3) From the 9 year WMAP results (Bennett et al 2013), the current
estimate of Omega is 1.0027 +/- 0.0039

Actually, I think other predictions of inflation are perhaps more
remarkable (e.g. for the index of the perturbation spectrum) but they
don't fit as nicely into the rules set up for a "definitive"
prediction.

--Wayne

In article , wlandsman
writes:

DEFINITION: predictions that are (1) prior to relevant
observations, (2) feasibly testable, (3) quantitative,
(4) NON-ADJUSTABLE, and (5) unique to the theory being
tested.

But I want to point out that using this definition, the prediction of
inflation of the mass density of the universe is a perfect example of
a definitive prediction. Consider

(1) There are several varieties of inflation but they all predict that
Omega (the ratio of the observed mass density to the closure density)
should be observed to be exactly 1. This prediction is completely
non-adjustable.

If Omega is understood to be what I call Omega + lambda, where Omega is
the density parameter for normal matter and lambda the cosmological
constant. Back when inflation was developed, most people assumed that
lambda was 0, so Omega+lambda=1 implied Omega=1.

Note that the idea of inflation was not adjusted or fudged with after it
was discovered that there is a significant positive cosmological
constant. The prediction was always Omega+lambda=1, but was sometimes
formulated differently because of the assumption lambda=0. (To be
picky, the prediction is not EXACTLY 1, but rather very close to 1. The
essential idea is that the radius of curvature is vastly larger than the
Hubble radius. For Omega+lambda=1, the radius of curvature is infinite.

(2) In 1979, when the first inflation models were developed, the
consensus value of Omega was 0.1 (Look at the (in)famous paper
written by 4 of the most noted astronomers of the time
http://adsabs.harvard.edu/abs/1974ApJ...194..543G )

Interestingly, one of the authors, David Schramm, later became a
defender of the Omega=1 idea, where Omega here is just the matter, not
the "total Omega". This is ironic because it later turned out that the
analysis of the paper you cite, namely, that Omega is much less than 1,
has stood the test of time; this could be reconciled with inflation
because of the discovery of the cosmological constant. Schramm died
around the time this was happening.

So not only did
inflation give a definitive prediction, it gave a *bold* definitive
prediction, in that it went against the current consensus.

Right.

Actually, I think other predictions of inflation are perhaps more
remarkable (e.g. for the index of the perturbation spectrum) but they
don't fit as nicely into the rules set up for a "definitive"
prediction.

I think the prediction that n is approximately 1 but slightly less than
1 is pretty good. This was a firm prediction made at a time when
observations could say nothing about this.

In article ,
"Robert L. Oldershaw" writes:
DEFINITION: predictions that are (1) prior to relevant
observations, (2) feasibly testable, (3) quantitative,
(4) NON-ADJUSTABLE, and (5) unique to the theory being
tested.

I'm probably wasting bandwidth to bother with this, but the above is
ridiculous.

1 is irrelevant, as I wrote before.

2 is an essential part of the definition of a theory, though I'd omit
"feasibly." A prediction is still a prediction, and a theory can
still be considered, even if our current technology doesn't allow a
given test. Of course until the test is done, ot can't give evidence
either for or against the theory.

3 is normal, though if there were some non-quantitative but testable
prediction, it would still be evidence.

4 is ridiculous as stated unless there's a special definition of
"non-adjustable." The relevent question is how many parameters there
are relative to the number of observables. Newtonian theory, for
example, requires seven free parameters for each planet (less a few
for Earth given freedom to define the AU and year, etc.).

5 is impossible. In general there will be an infinite number of
"theories" that could give any specific prediction, but most of them
are silly because they have more free parameters than observables.

All in all, the key point is whether a theory makes accurate
predictions with few free parameters compared to observables.
Concordance cosmology stands up very well in that scale; that's why
it got its name.

--
Help keep our newsgroup healthy; please don't feed the trolls.
Steve Willner Phone 617-495-7123
Cambridge, MA 02138 USA

On Saturday, May 24, 2014 8:47:26 AM UTC-4, Steve Willner wrote:
[Mod. note: entire quoted article trimmed -- mjh]

The above opinions cannot go unanswered.

As I have pointed out before at SAR, Einstein's prediction for the
eclipse experiment fulfilled all 5 requirements of a definitive
prediction.

Throughout his career Einstein insisted on such definitive predictions
as the only sure path toward accurate understanding of nature, and
equally importantly, the only way to avoid long detours into various
pseudo-science cul-de-sac situations.

#4 is not "ridiculous" to a scientist like Einstein, or Galileo, or
Planck, or Schrodinger, or...

If theoretical physics is having trouble coming up with definitive
predictions, then it is a failing of currently fashionable theoretical
physics and not a problem with the time-tested scientific method.

Here is an archetypal quotation from Linde concerning the multiverse
"theory", delivered at a recent conference on BICEP2.

"If you cannot disprove [multiverse "theory"], you have this powerful
weapon of thinking about and explaining things around you in an
anthropic way."

No, that is not a misquote. HE ACTUALLY SAID THAT!

I cannot believe that more scientists are not worried about current
trends away from the time-tested scientific method and towards
Aristotelean "explanations", which are just-so stories without
adequate empirical testing.

[Mod. note: reformatted -- mjh]

In article , "Robert L.
Oldershaw" writes:

"If you cannot disprove [multiverse "theory"], you have this powerful
weapon of thinking about and explaining things around you in an
anthropic way."

No, that is not a misquote. HE ACTUALLY SAID THAT!

What's the problem? Science can never prove anything, it can only
disprove hypotheses. So, he is saying that as long as the idea of the
multiverse has not been disproven, it offers a means to an anthropic
explanation for various things (strengths of fundamental forces etc).
In other words, he is saying that not only is this idea not disproven,
but it has some evidence in support of it. It is easy to come up with a
theory which has not been disproven (and perhaps cannot be disproven);
it is more interesting if there is some evidence in favour of the theory
(though this can never prove the theory).

On Tuesday, May 27, 2014 2:55:34 PM UTC-4, Phillip Helbig---undress to reply wrote:
What's the problem? Science can never prove anything, it can only
disprove hypotheses. So, he is saying that as long as the idea of the
multiverse has not been disproven, it offers a means to an anthropic
explanation for various things (strengths of fundamental forces etc).
In other words, he is saying that not only is this idea not disproven,
but it has some evidence in support of it. It is easy to come up with a
theory which has not been disproven (and perhaps cannot be disproven);
it is more interesting if there is some evidence in favour of the theory
(though this can never prove the theory).

Excuse my candor, but I think the above is the "What Me Worry"
response made famous by MAD Magazine.

Feel free to adopt the untestable multiverse stuff and anthropic
reasoning, but I think that would be tantamount to throwing science
under the bus.

In article , "Robert L.
Oldershaw" writes:

Feel free to adopt the untestable multiverse stuff and anthropic
reasoning, but I think that would be tantamount to throwing science
under the bus.

YOU say it is untestable, ignoring claims that it is. Please rebut
them.