wadawson / merging-cluster-dynamics-paper Goto Github PK

Referee comment:

The question of changing cluster masses during a merger
(due to accretion, dynamical friction, tidal stripping, and, in the
case of SIDM, momentum transfer) could, in principle, have a
significant impact on the accuracy of results from the model.
Appropriately, this topic is discussed in the paper, and the author
provides reasonable arguments as to why these effects aren't important
(at least for the case sigma=0). However, much of the argument is
based on the agreement between the model predicted and N-body
simulation parameters for the one case considered, while, as noted by
the author, Farrar & Rosen find that dynamical friction leads to a 10%
reduction in the inferred collision velocity for the subcluster. This
apparent disagreement is a bit puzzling, and a possible explanation is
not identified. Just as a check, can the author verify that the halo
mass parameters given in section 2.3 are at t_obs, and not the initial
mass profiles (which, again, are likely to be somewhat different)?

Currently the note at the end of section 5.2 states that my python code implementing the discussed Monte Carlo method is available upon request. I need to update this to reflect the codes availability at git://github.com/MCTwo/MCMAC.git and the this paper's availability at XXX once I make it public.

Referee comment:

The values of M_200_2 differ slightly in the left and right
"panels" of Table 3. It is not clear why this should be the case
(typo?).

Need to investigate.

Referee comment:

In the first paragraph of section 5.2, the author briefly
discusses including a prior to allow for the possibility of a non-zero
impact parameter, but indicates that it is not clear what prior to
use. As a point of potential interest, Randall et al. (2002)
(references by the author elsewhere) gives a probability distribution
for the dimensionless spin parameter lambda and describe how to derive
the impact parameter from lambda.

This is interesting I need to go back and reread Randall et al. (2002) with this in mind.

Referee comment:

On a related note, claiming in the first place that the method
is within 10% agreement with N-body simulations (plural) is
over-stating things a bit. Once again, all that has been looked at is
one snapshot from one simulation of one system. Ideally, one would
compare results from the model with simulations for a variety of input
parameters (mass ratios, merger velocities, concentration parameters,
etc.) at a distribution of times (and a distribution of viewing
angles) during the simulations to better characterize the accuracy of
the model. I appreciate that running a series of N-body simulations
for comparison is beyond the scope of this paper (and, in some sense,
defeats the purpose of the model), but on the other hand it has yet to
be shown that the model agrees with simulations to within 10% in
general. If the author wishes to strengthen their case a bit further,
I note that comparing results from the model to the one simulation
considered (Springel & Farrar, 2007) over a spread of snapshot times
is in principle easily done, as compared with running new simulations
for comparison (assuming that merger parameters at other times in the
simulation are available).

This should be feasible since Springel and Farrar plot these values over time/space, it will require me running my analysis over a range of parameters but this is possible. I just need to think about how best to do it and how long it will take.

We got ~47 new cluster redshifts in the north/south subcluster areas which we could incorporate. I have not looked at how this changes the redshift estimates of the Musket Ball yet.

Background

This issue was actually identified after a comment by Marusa Bradac after rev0 of the paper was submitted to ApJ. Since Barrena et al. 2002 report the redshifts of the main and bullet subclusters in terms of velocity (km/s) I needed to convert this to actual redshift values. I applied the full relativistic conversion formalism, but Barrena et al. (2002) actually just used v = c*z and later when calculating the velocity difference of the two subclusters used (v2-v1)/(1+z). The end result being that I overestimated the observed relative line of sight velocity difference.

To Do

• Calculate correct redshifts for each subcluster
• Correct values in table 2
• Update Table 3 results
• Update bullet cluster related figures (2, 3, 7, 8, 10-15)
• Verify text of section 3.2.1 is still correct
• Verify text of section 3.2.2 is still correct
• Verify text of section 5 is still correct

With about 600 redshifts in the area I am not sure of the best way to release them. I need to check with ApJ about the recommended way. I could imagine including a representative subsample in the paper and the full sample online.

Perhaps I should also include a note that the raw and reduced 1D & 2D spectra are available upon request.

To Do List:

• Estimate position uncertainty for objects with -99 for object match (take 68% confidence interval of matchdelta values for objects with matches)
• Estimate the redshift uncertainty for objects with -99 redshift uncertainty (take 95% confidence interval of other redshift uncertainty)
• Get R-band magnitude and uncertainty for matched objects
• Specify source for each redshift (e.g. DEIMOS 2011a)
• Add DEIMOS 2011 catalog
• Add DEIMOS 2013 catalog
• Add all other DLS redshifts in the area

This is a request of the referee:

In the abstract (and elsewhere), results from this method
are said to be in agreement with N-body simulations to within 10%.
While the phrasing given is technically accurate, given the context it
is easily misinterpreted as meaning that, given a set of observables,
the inferred parameters of the merger agree with those from numerical
simulations matched to a particular observation at any time t during
the merger (at least that was what I came away with from my initial
reading of the abstract). The test that was actually done was to take
known parameters from the merger (including d_3D(t_obs), which is not
an observable), run this single case through the model (with no Monte
Carlo analysis to consider errors), and show that the output
v_3D(t_col) and TSC_o are within 10% of the values found in the
simulation for the given merger state only (i.e., at t_obs). Although
the way this is currently described in the paper is fair, the point
could be clarified with some minor changes in language.

Referee comment:

The abstract could also clarify that the model as it's
presented is only applicable to dissociative mergers (since it assumes
that the impact parameter b~0), although to be fair it could in
principle be extended to include more general cases.

Referee comment:

On more than one occasion, the author claims that for SIDM
mergers the offset between the galaxies and the dark matter increases
with time after core passage (i.e., t_coll) until the subcluster
reaches the turn around radius. I suspect that this is not generally
the case, and that there is a similar "slingshot" effect with the SIDM
as observed with ram pressure stripped gas in merging galaxies and
subclusters (i.e., at some time after t_coll and before reaching d_max
the gravitational influence of the subcluster DM halo on the subcluster
galaxies leads to a decrease in the separation). Has this been
carefully checked, or simply assumed?

I need to find all "occasions" in the paper and think about editing.

• 2nd paragraph of introduction
• 1st paragraph of section 4.2
• 3rd paragraph of section 5

In short, this has simply been assumed. I attempted to check this by running a simple multibody simulation that containted two galaxy particles and two dark matter particles (one for each subcluster). The galaxy particles only interacted via the gravitational force, and the dark matter particles interacted via gravity and a self-interaction force. After running this simplistic simulation through it was clear that the results were unphysical, largely due to treating the galaxies as a single coherent body and lack of dissapative effects. My collaboration is working on fullblown self-intearcting dark matter merger simulations but these results are still a ways off.

Referee comment:

Third paragraph in the introduction: when discussing various
methods for constrained v_3D for merging clusters the author might
include the use of merger cold fronts via a comparison of the pressure
at the stagnation point and in the free streaming regions (e.g. A3667,
Vikhlinin et al. 2006, and several others), although the importance of
projection and other systematic effects for this method are unclear.

I need to perform a bit of literature review for this comment.

Referee comment:

Section 5.2 typo: "in a manner similar to..."

Redshift and catalog help. Also consider adding Brian for his redshift help.