Tuesday 12 August 2014

X-ray bananas

This year's discoveries follow the well-known 5-stage Kübler-Ross pattern: 1) announcement, 2) excitement, 3) debunking, 4) confusion, 5) depression.  While BICEP is approaching the end of the cycle, the sterile neutrino dark matter signal reported earlier this year is now entering stage 3. This is thanks to yesterday's paper entitled Dark matter searches going bananas by Tesla Jeltena and Stefano Profumo (to my surprise, this is not the first banana in a physics paper's title).

In the previous episode, two independent analyses  using public data from XMM and Chandra satellites concluded the presence of an  anomalous 3.55 keV monochromatic emission from galactic clusters and Andromeda. One possible interpretation is a 7.1 keV sterile neutrino dark matter decaying to a photon and a standard neutrino. If the signal could be confirmed and conventional explanations (via known atomic emission lines) could be excluded, it would mean we are close to solving the dark matter puzzle.

It seems this is not gonna happen. The new paper makes two claims:

  1. Limits from x-ray observations of the Milky Way center exclude the sterile neutrino interpretation of the reported signal from galactic clusters. 
  2. In any case, there's no significant anomalous emission line from galactic clusters near 3.55 keV.       

Let's begin with the first claim. The authors analyze several days of XMM observations of the Milky Way center. They find that the observed spectrum can be very well fit by known plasma emission lines. In particular, all spectral features near 3.5 keV are accounted for if Potassium XVIII lines at 3.48 and 3.52 keV are included in the fit. Based on that agreement, they can derive strong bounds on the parameters of the sterile neutrino dark matter model: the mixing angle between the sterile and the standard neutrino should satisfy sin^2(2θ) ≤ 2*10^-11. This excludes the parameter space favored by the previous detection of the 3.55 keV line in  galactic clusters.  The conclusions are similar, and even somewhat stronger, as in the earlier analysis using Chandra data.

This is disappointing but not a disaster yet, as there are alternative dark matter models (e.g. axions converting to photons in the magnetic field of a galaxy) that do not predict observable emission lines from our galaxy. But there's one important corollary of the new analysis. It seems that the inferred strength of the Potassium XVIII lines compared to the strength of other atomic lines does not agree well with theoretical models of plasma emission. Such models were an important ingredient in the previous analyses that found the signal. In particular, the original 3.55 keV detection paper assumed upper limits on the strength of the Potassium XVIII line derived from the observed strength of the Sulfur XVI line. But the new findings suggest that systematic errors may have been underestimated.  Allowing for a higher flux of Potassium XVIII, and also including the 3.51 Chlorine XVII line (that was missed in the previous analyses), one can a obtain a good fit to the observed x-ray spectrum from galactic clusters, without introducing a dark matter emission line. Right... we suspected something was smelling bad here, and now we know it was chlorine... Finally, the new paper reanalyses the x-ray spectrum from Andromeda, but it disagrees with the previous findings:  there's a hint of the 3.53 keV anomalous emission line from Andromeda, but its significance is merely 1 sigma.

So, the putative dark matter signals are dropping like flies these days. We urgently need new ones to replenish my graph ;)

Note added: While finalizing this post I became aware of today's paper that, using the same data, DOES find a 3.55 keV line from the Milky Way center.  So we're already at stage 4... seems that the devil is in the details how you model the potassium lines (which, frankly speaking, is not reassuring).


Kevork Abazajian said...

Note the Boyarsky et al. 1408.2503 Milky Way Galactic Center (GC) analysis isn't a question of a robust detection, but a consistency check with the cluster and M31 detections. What Boyarsky et al. are saying is that the GC has a 3.5 keV line in the GC consistent with the dark matter in the field of view of the observation. Jeltema & Profumo 1408.1699 didn’t seriously consider that the 3.5 keV line they detected in the GC could be consistent with the dark matter in the field of view because they already “knew” a priori that the line "had to be" from potassium, then fixed the potassium line flux in the data, placing limits with this assumption. This led to Jeltema & Profumo’s unjustified conclusion that the GC observations exclude the prior cluster detections that are in fact consistent with the new potential detection in this Boyarsky et al. analysis, a detection Jeltema & Profumo missed.

Kevork Abazajian said...

Regarding the cluster analyses, Jeltema & Profumo do not get a consistent picture of the lines in a single-temperature model of the cluster(s). However, the X-ray astronomer team led by Esra Bulbul in 1402.2301 get a consistent picture in detailed multi-temperature plasma models. In their extensive modeling of the full spectra, they can constrain the K XVIII line's flux that can partially mimic the 3.57 keV line flux. Jeltema & Profumo's paper does no detailed modeling, but makes statements based in pairs of line ratios in single temperature plasmas. Jeltema & Profumo therefore then discard a key ratio because they cannot get a set of pair ratios correct in single temperature models. The single-temperature line ratio reasoning is the basis of their claim that there are significant systematic uncertainties, no limit can be placed on the 3.52 keV K XVIII flux, which could be the source of the 3.57 keV line, if one includes a systematic offset due to the energy shift. The Bulbul et al. team finds no way of having such a high K XVIII even in multitemperature, varying abundance models.

Joseph Conlon said...

Just a comment on the Cl XVII - this line at 3.5 keV is a 3 p to 1s transition, and the same element has a 2 p to 1 s transition which is expected to be a factor of ten stronger. So if Cl XVII was a significant contaminant in the Bulbul analysis, you would expect another, stronger, line at 2.96 keV from the 2p transition, but they make no mention of such a line.

The Bulbul paper makes a similar comment about a Cl XVI line at 3.5 keV.

Anonymous said...

The bump at 3.55 keV in Fig#1 of the Boyarsky et al. paper and the bump in Figs 7,10,11 of Bulbul et al. are most likely not emission from Potassium or Chlorine because the width of the bumps is ~0.1 keV.

This is quite a broad line, and one would need a large number of emission lines just magically placed with the largest in the middle in order to explain this bump with emission lines from K or Cl.

While it's not clear exactly what this bump is, it's pretty clear that it's not instrument signal (because the emission shows up at different values for higher z galaxies) and it's pretty clear that it's actually coming from the galaxy because there is virtually no emission from the blank sky and there's a larger signal coming from more massive galaxies.

A 7.1 keV dark matter particle is right in the middle of range of values consistent both with DM halos and LymanAlphaForest.
But we'll just have to wait for more data to roll in before there's any real "detection of DM decay." Astro-H should be able to confirm or refute the claim that the bump is due to DM decay. So, let's hope that everything goes well for the Astro-H team.

Alex said...

Suppose, for the sake of argument, that a 7.1 keV sterile neutrino exists. Is there any laboratory experiment that could detect it?

Or, suppose it were a 7.1 keV axion. Could that sort of thing be detected in some sort of laboratory experiment, given plausible estimates of the axion-photon coupling?

I do a lot of work on extracting weak signals from noisy background, albeit in a completely different context, so I'm optimistic that this problem can be solved. But it's going to require a whole bunch of people who enjoy geeking out on fine details of plasma physics and models of galactic cores.

Jester said...

In principle, you can probe a keV mass sterile neutrino by searching for kinks in the beta decay spectrum. But it will take a lot of time for the experiments to reach the interesting parameter space. In the short run we are stuck with astrophysics. As pointed out above, the astro-H satellite with its much better resolution should clarify the situation in a few years.

Alex said...

What would be the difference in signatures between a 7.1 keV sterile neutrino and a 7.1 keV axion? Would the axion signal depend on local magnetic field and dark matter density, while the sterile neutrino signal would only depend on dark matter density?

Jester said...

Exactly, for axions the signal also depends on the galactic magnetic field, and this way you can reconcile the existence of a signal from some sources (like Andromeda) and the lack thereof from other sources (like the Milky Way center).
As for probing axions in the lab, I have no idea whether the parameter space relevant for the x-ray signal can be reached in the near future.

Alex said...

Maybe a better question is whether axions in the relevant parameter range could be produced in the sun in numbers sufficient for a directional detector in a lab.

Anonymous said...

Dark matter is most likely a fermion particle with rest mass between 2-10 keV. The reason we know that it needs to be a fermion is that Fermi pressure is required to keep it from clumping together in the center of galaxies. This forces dark matter to be a fermion and have a rest mass in the low keV range.
As such, boson particles (such as axions) can't be dark matter. And equally important, this rules out hypotheical GeV fermions.

Alex said...

Why wouldn't angular momentum be sufficient to prevent clumping?

And I am not a cosmologist, but a lot of them seem to consider the axion a plausible candidate. Why is the fermionic requirement not universally agreed on?

Jester said...

Eddie's statement is a peculiar opinion, certainly not universally accepted :)

Stefano Profumo said...

Stefano Profumo here.

Thank you for your review of our work. I have one remark on your "Note Added": in our study, we do report a line at ~3.5 keV in XMM data from the Galactic center in our sec. 2, with basically an identical flux to what subsequently reported by Boyarsky et al. We indicate such detection and provide the associated best-fit values for the sterile neutrino mixing angle (see our sec. 2.3). We actually calculate the dark matter "implications" in an even more detailed way than Boyarsky et al (including for example the non-trivial masks structure shown in Fig.2).

Also, it is worth emphasizing that the limits we derive in sec. 2.3 (ii) are somewhat "immaterial" (albeit technically correct) to the message we intend to convey, given that the focus of our paper is to question whether or not an excess exist in any of the systems that have been observed so far. For example, we do not even feature these limits in the abstract of our paper. The gist of our study is the following: before claiming any excess that invokes "new" physics, we must ascertain whether, within systematics, known physical processes might potentially account for observations. This is nothing more nothing less than Occam's razor. Our conclusions are that in all cases systematics are currently large enough that no firm excess can be claimed.

One more comment, about M31: Boyarsky et al agree with us that there is no statistically significant line between 3 and 4 keV ("In particular, in the case of M31 there are no strong astrophysical lines between 3 and 4 keV"). They argue that a line emerges when broadening the energy range. We believe that the reason for this is simply that in that case the modeling of the continuum deviates from a simple power-law, multiple lines need to be added in, and spurious residuals appear at around 3.5 keV.

I will give a few thoughts below on some of the comments about our study that have appeared on this blog.

Stefano Profumo said...

Kevork Abazajian is wrong in several of his comments:

1. In his first comment on this blog, Abazajian claims that "Jeltema & Profumo 1408.1699 didn’t seriously consider that the 3.5 keV line they detected in the GC could be consistent with the dark matter in the field of view because they already “knew” a priori that the line "had to be" from potassium". Abazajian's statement is wrong. We do discuss in detail the implications for a dark matter interpretation of the line in section 2.3 (iii). As explained above, we give an even more accurate account of this possibility than in Boyarsky et al.

2. Abazajian then argues again that we did not detect the 3.5 keV line in the XMM GC data (" a detection Jeltema & Profumo missed."). This is, again, wrong. We give full details of such detection in section 2 of our paper.

3. Abazajian is wrong in stating that "Regarding the cluster analyses, Jeltema & Profumo do not get a consistent picture of the lines in a single-temperature model of the cluster(s)". He evidently does not understand what we did. Line ratios are temperature-dependent, and we employ the measured fluxes for S XVI, Ca XIX, and Ca XX to predict, as a function of temperature, what the flux of K XVIII could be (irrespective of obtaining a full model for the relative line fluxes of all lines, which is a daunting task even for one cluster only, imagine 73!). As we explain, for any temperature between 1 and 8 keV, both Ca lines predict large enough fluxes of K XVIII plus Cl XVII to explain (within a factor 2) the flux at 3.5 keV for the cases of Perseus and the Coma+Oph+Cen stack; we show that there might still be a somewhat marginal excess for the large stack, but only for large enough temperatures (~5 keV), though, we find, smaller than what Bulbul et al claim. This is not true using the S line, albeit this line seems to be a systematically dim line compared to e.g. the observed Ca lines. Notice that this exercise contains an additional potential systematic in that relative metal abundances are not reliably known, especially for the case of stacked cluster observations! Thus, unlike what Abazajian appears to argue, we are not after "a consistent picture"; rather we use a detailed, model-independent estimate of the range of possible K XVIII fluxes. What apparently Abazajian misses is that Bulbul et al only use the central values of the temperatures they derive in their models to predict the K XVIII flux (instead of exploring the temperature range covered by their uncertainties), and then append a somewhat arbitrary factor 3 to their predictions to account for systematics. We make the key point that to assess the presence of a putative excess, one should look at the *maximal*, not at the *median* prediction for the background. Independent of using multi-temperature models, what Abazajian does not seem to understand is that 2 out of 3 of the brightest lines quoted in Bulbul et al predict *for any temperature between 1 and 8 keV* enough photons at 3.5 keV to explain the observed line (with the caveat above for the large stack)!

Joe Conlon comments about the Cl line. As we discussed with Joe over email, looking for the 2.96 keV line would be a nice cross check to examine whether indeed Cl contributes significantly at 3.5 keV. We did not re-perform the cluster data analysis, relying instead on the detail results presented by Bulbul et al. Unfortunately, Bulbul et al do not quote any quantitative information regarding the 2.96 keV Cl line. It is important to note that we find that even without the Cl contribution the 3.5 keV flux from clusters observations would still be compatible with K only emission within systematics. Also, we find that the Cl adds at most a factor of 2 (at 8 keV) to the line flux; at lower temps it does not add much at all.

Eddie Devere comments on the width of the line. In all cases (our analysis and Boyrasky's) that width is compatible with a K or K+Cl given instrumental energy resolution and photon counts.

Anonymous said...

Just maybe a stupid question. Is there any instrumental possibility to measure a flux do monochromatic neutrino at this energy? I guess no as the flux seems very small (10-6 photons per sec per cm2), but it will be (independently on the presence or not of this signal) a clear way to distinguish between different models (sterile neutrino, axion, axino, annihilating DM..)

andrew said...

Another blow to an 3.55 keV monochromatic X-ray line as evidence of dark matter comes from:

Michael E. Anderson, Eugene Churazov, and Joel N. Bregman, "Non-Detection of X-Ray Emission From Sterile Neutrinos in Stacked Galaxy Spectra" (18 Aug 2014). http://arxiv.org/abs/1408.4115

Kevork Abazajian said...

1 & 2: The unjustified conclusion in the Jeltema & Profumo manuscript relies on the GC 3.5 keV flux being fit to the K line and then fixed to place the constraints. Their conclusions claim that with this GC analysis "we rule out a dark matter decay origin for both the clusters and the M 31 observations," which is clearly incorrect. Boyarsky et al. 1408.2503 show explicitly how the GC spectrum is not only not "ruling out" a dark mattter decay origin, but consistent with it.

3. I do understand the modeling of the emission from the cluster plasmas. Unlike Bulbul et al., Jeltema & Profumo do not do detailed plasma emission modeling, but argue based on line ratios in single temperature plasmas. This will be sorted out in the literature.

Lastly, the recent manuscript arxiv:1408.4388 shows that the basic method of the Jeltema & Profumo M31 analysis used too narrow of an energy window (only 3-4 keV) with a simple power law model. This was responsible for their failure to detect the 3.5 keV line in M31.

Theo Nieuwenhuizen said...

Dear Jester, thanks for offering this platform to discuss the matters at a serious level. In mean time two negative results have appeared: from dwarfs and by stacking 81 and 89 galaxies observed with different instruments, but masking the galactic centers. Neither of the approaches gives a positive result. Do you consider to have reached stage 5 or are we still in stage 4?

Theo Nieuwenhuizen said...

Apart from the paper 1408.4115 on dwarfs, mentioned by andrew, also 1408.3531 "Constraints on 3.55 keV line emission from stacked observations of dwarf spheroidal galaxies" by D. Malyshev, A. Neronov, D. Eckert seems to make the DM interpretation of the 3.55 keV line unlikely.

Joseph Conlon said...


These papers rule out sterile neutrinos, or any model where dark matter decays directly to photons.

They don't refute the Bulbul analysis, as they involve different objects (galaxies rather than clusters)

In some models (eg our DM -> ALP -> photon model) the absence of a signal from dwarfs/almost all galaxies is actually a requirement.

Just to show this was indeed a pre-diction, this is what we said in advance:

"In environments.... such as dwarf galaxies, the line should be suppressed" 1403.2370

"We note that...the above points make M31 an unusually favourable galaxy for observing a 3.55 keV line. For general galaxies in this scenario the signal strength of the 3.55 keV line would be much lower than for galaxy clusters" 1404.7741

Theo Nieuwenhuizen said...

Adam/Jester, can you please correct my last post:
Apart from the paper 1408.4115 on dwarfs,
=> Apart from the paper 1408.4115 on stacked galaxies with masked centers,
Thanks, Theo

Kevork Abazajian said...

Note that the conservative constraint from "Constraints on 3.55 keV line emission from stacked observations of dwarf spheroidal galaxies" by D. Malyshev, A. Neronov, D. Eckert, 1408.3531 is less constraining than that from our M31 analysis from 2013, Horiuchi+, 1311.0282.

Kevork Abazajian said...

There is a comment on the Jeltema & Profumo "bananas" paper (JP) by the Bulbul et al. Harvard-Goddard team at arXiv:1409.4143.

Bulbul et al. go through the careful modeling of the problem I had pointed out regarding the simplistic single-temperature line-ratio "argument" that JP stated. JP did no actual modeling. Importantly, Bulbul et al. show the line ratios JP use are incorrect and make the single-temperature arguments of JP even more flawed in a multi-temperature case of a galaxy cluster.

Kevork Abazajian said...

(1) The JP Galactic Center analysis is clearly flawed in making the assumption of all of the 3.5 keV flux coming from K XVIII, and then placing constraints from the Galactic Center after this assumption, (2) the Boyarsky team showed how their M31 analysis is flawed in using much too narrow of an energy window in their line search modeling [arXiv:1408.4388], and (3) the Bulbul et al [1409.4143] team showed that they use simplistic models with incorrect line ratios in their X-ray cluster modeling. There's nothing left correct in the "bananas" JP paper.