1ES1011+496, EBL studies with 2014 obs. Ready for Pub. Board

[david.paneque]

Posted on behalf of Adiv Gonzalez

This manuscript is now ready for review within the MAGIC Publication Board.

1ES1011_EBLPaper.pdf

1ES1011+496, EBL studies with 2014 obs. Manuscript for review within Publication Board

(237.2 KiB) Downloaded 165 times

[ulisses.barresdealmeida]

Dear Authors,

as per request of David, I have read the paper about the measurement of the EBL density using the 1101 flare data and I think it is in very good shape and requiring no great modifications. That said, please find below my questions and comments/suggestions. You can find attached an annotated pro file for your convince.

As you will see, there is just one or two major comments which I would like you to take into consideration (item 12, mostly, and 8 as well).

Best Wishes and thanks for the good job,
Ulisses.


1. In the abstract/results (last sentence), when you quote the results obtained for the EBL density, please compare with previous knowledge or say how this result expand what we know, i.e is is a more stringent measurement of the EBL density at this range? has it expanded the frequency range of knowledge of the EBL with this precision, etc.

2. please, slightly rewrite the sentence marked in green on page 2. It is a bit confused as to the language (it is too long). Maybe separate in two sentences.

3. please, develop a bit what you mean by "limited hardness", so that the uninformed reader does not have to go to the literature just to get basic info as to what is the motivation of this assumption and typical values adopted to the maximum spectral indices, if it is possible to quote some typical, generalised numbers.

4. again, when you mention the inflection of the spectra, please mention that this is due to a peak in the EBL density at these energies and contextualise why physically the EBL has a peak at this region. Again, tho sis to give some "landscape" of the problem for the less informed reader.

5. You say that the CGRH is "the energy at which..." If it i a horizon, it should be "the distance at which optical depth = 1", of course this horizon is a function of E. So, it seems the definition you present is inverted. In the same paragraph, instead of "doing a prediction", say "made an extrapolation".

6. In the beginning of section 2, please define the acronym IACT.

7. please, use \noindent after formulas in the text

8. Figure 3 clearly shows how the photon index determination degrades when the source goes into lower flux. For the measure of the EBL nevertheless you used all data. Why not attain oneself to the highest states only?

9. In section 3, when you discuss the factors which influence the blazer spectra (at source), maybe you could show a parameterisation of the constraints or limits to the index which are imposed by the SSC production model you mention, in both regimes? This is just to have a physical outlook which also justifies the presentation of the spectral shapes used for the middling afterwards, PWL, LP, etc...

10. Regarding the use of different models as shown in figure 4, would it be feasible to present a summary table using all models, instead of simply following the argumentation which led you to adopt one specific for the density measurement?

11. Also, did you try to test on just the best part of the sample (higher flux, lower error bars)? Following figure 3, and not expecting much dependence on the density measurement with estimates at the two epochs, I would verify if it is best to us sonly part of the sample data and leave the rest out.

12. You rightly observe that there are reasons not to consider the PWL model, since it would wash out intrinsic curvature that could affect the EBL measurement later. In principle the argumentation is fairly logical and one could go along with that alone. Nevertheless, one knows that it can be dubious to base a choice of a model in such genre of argumentation and it would be ideal to apply an appropriate statistical test. There is analysis (actually a family of analysis) which can be done to tackle the problem of deciding the on the best model when comparing likelihood functions drawn from different number of parameter functions. If this was not consider, please add this to your analysis and maybe you will reach a decision about which model from figure 4 to adopt simply based on these novel statistical considerations - which would-be much straightforward and superior in statistical terms to what was done by Abramowski et al. (2013).

I am referring to what is sometimes called Akaike information criteria (AIC), and relate methods. If you cannot find your way through it in the web, please let me know and I try to provide you with some inputs.

13. When you discuss the behaviour of figure 5, I would like to see, if possible, the behaviour on the evolution of log L vs \alpha if it is taken away this convexity condition, to understand better its effect on the modelling.

14. again, once you derive \alpha, again, unless you exclude statistically via AIC, it would be nice to presente the \alpha_0 as measured with other methods in a table, just for completion.

15. very good treatment of systematics!

16. What is the potential physical implications of measuring a same EBL density as Abramowski et al. (2013), now up to z = 0.212? Any consequences for the evolution of the EBL with redshift / cosmic time, and therefore for the evolution of galaxy and structures? If the redshift range is still too small for any conclusion of the kind, please state and sign to what extension one would like to go to be able to probe this with EBL measurements at this energy band, to contextualise the work in a broader physical scenario.

[ulisses.barresdealmeida]

Please, find here the correspondent annotated pdf for my comments to the first draft sent to internal review.

Best Wishes,
Ulisses.

[stefano.covino]

Hello everybody,

and here my comments. As a matter of fact, I do agree that the paper is well written, clear and concise. Definitely ready for submission. My comments are mainly minor typos, and only in few cases I have some concern about what we write or about how we derive some result. Nothing, however, that cannot be solved with a careful rephrase or some simple additional consideration.

Good work,
Stefano






Abstract

. Not sure indeed. But in the "Results" item I would say "we conclude" rather than "concluded".


Introduction

. Page 1, redshift = 0.212 (in the text we have =212).

. Page 2 and later in the paper, we should check that we insert the hyphen any time it is needed. For instance in "gamma-ray", etc.

. Page 1, The EBL is the diffuse rational that came... the past here sounds strange to me. I would have used "come", but of course English is not my mother tongue!

. page 1, right column, (from THE X-ray Timing Explorer)

. Page 1, right column, [...] no evidence of variability [...], we should add here on which timescale.



Results

Page 3, right column. Fig. 1, rather than Fig 1.

Page 3, just after Eq. 1. Here I have my first concern. The fit quality we report is rather poor, and with these parameters reporting the errors should not be statistically allowed. I think we should at least mention the matter and discuss whether it might have consequences for us.


EBL measurement

Page 4, Fig. 4 caption. The tested models should be five. Right?

Page 4, right column, [...] the two hypotheses [...]. not hypothesis.

Page 4, right column. This is only a feeling, but the fit to the straight line in Fig. 3 looks better than the reported Chi2. I am just asking to check.

Page 5, left column. Here I essentially agree with Ulisses' comment 12, and this is my second (a bit more) serious concern. I understand that what we write is reasonable, but for a generic reader it sounds rather unsatisfactory. We should devote some more discussions to the PWL case.


Systematic uncertainties

Page 5. Right column. Possibly we should use {\rm scale} when we report \alpha.



Conclusions

Page 7, Fig. 8. This plot is amazing, yet I'm afraid very difficult to read. No way to publish it in landscape using an entire page?

[adiv.gonzalezmunoz]

Dear Authors,

as per request of David, I have read the paper about the measurement of the EBL density using the 1101 flare data and I think it is in very good shape and requiring no great modifications.
Thanks!
That said, please find below my questions and comments/suggestions. You can find attached an annotated pro file for your convince.

As you will see, there is just one or two major comments which I would like you to take into consideration (item 12, mostly, and 8 as well).

Best Wishes and thanks for the good job,
Ulisses.


1. In the abstract/results (last sentence), when you quote the results obtained for the EBL density, please compare with previous knowledge or say how this result expand what we know, i.e is is a more stringent measurement of the EBL density at this range? has it expanded the frequency range of knowledge of the EBL with this precision, etc.
Well, we didn't want to make any claim saying that our measurement is the most constraining or something like that, basically because we are taking only one source of systematic uncertainty, while in the HESS paper the took (or made up) several sources of systematics. Although from the final plot it looks that it is more precise measurement, I thing the novelty is in the fact that the measurement is from only 1 source and at higher redshift.

2. please, slightly rewrite the sentence marked in green on page 2. It is a bit confused as to the language (it is too long). Maybe separate in two sentences.
Done.

3. please, develop a bit what you mean by "limited hardness", so that the uninformed reader does not have to go to the literature just to get basic info as to what is the motivation of this assumption and typical values adopted to the maximum spectral indices, if it is possible to quote some typical, generalised numbers.
Done.

4. again, when you mention the inflection of the spectra, please mention that this is due to a peak in the EBL density at these energies and contextualise why physically the EBL has a peak at this region. Again, tho sis to give some "landscape" of the problem for the less informed reader.
Done.

5. You say that the CGRH is "the energy at which..." If it i a horizon, it should be "the distance at which optical depth = 1", of course this horizon is a function of E. So, it seems the definition you present is inverted. In the same paragraph, instead of "doing a prediction", say "made an extrapolation".
Well, the definition we use is what one finds in the literature. I agree that seems as if the definition is inverted, but the dependency of the optical depth it is always given as a function of E, for a given redshif. Saying the opposite would require to visualize the dependency of the optical depth as function of the redshift, for a given energy.

6. In the beginning of section 2, please define the acronym IACT.
It is already defined in the introduction, but I guess does not hurt putting it again.

7. please, use \noindent after formulas in the text
Done.

8. Figure 3 clearly shows how the photon index determination degrades when the source goes into lower flux. For the measure of the EBL nevertheless you used all data. Why not attain oneself to the highest states only?
Although it goes to a lower flux, it is still higher that previous measurements, and since the method works better for higher statistics, we wanted to take advantage of those photons. If we have attained only to the highest states, the significance would have been lower (and actually I made that test)

9. In section 3, when you discuss the factors which influence the blazer spectra (at source), maybe you could show a parameterisation of the constraints or limits to the index which are imposed by the SSC production model you mention, in both regimes? This is just to have a physical outlook which also justifies the presentation of the spectral shapes used for the middling afterwards, PWL, LP, etc...
Sorry, I'm not quite sure what do you mean by showing a parametrization. I know that the explanation is just qualitative, but the main idea is: from the SSC model the spectrum is always steeper with the energy and even if its a pure PWL, the index cannot be harder than 1.5.

10. Regarding the use of different models as shown in figure 4, would it be feasible to present a summary table using all models, instead of simply following the argumentation which led you to adopt one specific for the density measurement?
I don't think that would be strictly necessary. Saving ourselves for using tables is the main purpose of the plots. An even if we put the tables I think the argumentation is still necessary.

11. Also, did you try to test on just the best part of the sample (higher flux, lower error bars)? Following figure 3, and not expecting much dependence on the density measurement with estimates at the two epochs, I would verify if it is best to us sonly part of the sample data and leave the rest out.
Yes, I tried using only the best part of the sample and the significance of the measurement was not as high as using the average spectrum putting all the data together.

12. You rightly observe that there are reasons not to consider the PWL model, since it would wash out intrinsic curvature that could affect the EBL measurement later. In principle the argumentation is fairly logical and one could go along with that alone. Nevertheless, one knows that it can be dubious to base a choice of a model in such genre of argumentation and it would be ideal to apply an appropriate statistical test. There is analysis (actually a family of analysis) which can be done to tackle the problem of deciding the on the best model when comparing likelihood functions drawn from different number of parameter functions. If this was not consider, please add this to your analysis and maybe you will reach a decision about which model from figure 4 to adopt simply based on these novel statistical considerations - which would-be much straightforward and superior in statistical terms to what was done by Abramowski et al. (2013).

I am referring to what is sometimes called Akaike information criteria (AIC), and relate methods. If you cannot find your way through it in the web, please let me know and I try to provide you with some inputs.
I checked what I think is the basic form of the AIC, which is defined as AIC=2k-2log L or asymptotically equivalent to AIC=2k+Chi². If I apply blindly the test to the data we have, the outcome is that the best model at the maximum L (or min Chi²) is the PWL, basically because all models have the same Chi² at the minimum. At alpha=0, then, the LP is favoured. Please let me know if there is a more proper statistical test, since this basic AIC I think does not add much info.

13. When you discuss the behaviour of figure 5, I would like to see, if possible, the behaviour on the evolution of log L vs \alpha if it is taken away this convexity condition, to understand better its effect on the modelling.
The plots became more symmetric, therefore, the statistical uncertainty on the right side increases

14. again, once you derive \alpha, again, unless you exclude statistically via AIC, it would be nice to presente the \alpha_0 as measured with other methods in a table, just for completion.
Well, from the plot you can tell that the alpha_0 will be practically in the same place for all the models. The difference will be in the uncertainties.

15. very good treatment of systematics!
Thanks!

16. What is the potential physical implications of measuring a same EBL density as Abramowski et al. (2013), now up to z = 0.212? Any consequences for the evolution of the EBL with redshift / cosmic time, and therefore for the evolution of galaxy and structures? If the redshift range is still too small for any conclusion of the kind, please state and sign to what extension one would like to go to be able to probe this with EBL measurements at this energy band, to contextualise the work in a broader physical scenario.
Ok, that'll be included.

[adiv.gonzalezmunoz]

Hello everybody,

and here my comments. As a matter of fact, I do agree that the paper is well written, clear and concise. Definitely ready for submission.
Thanks!

My comments are mainly minor typos, and only in few cases I have some concern about what we write or about how we derive some result. Nothing, however, that cannot be solved with a careful rephrase or some simple additional consideration.

Good work,
Stefano






Abstract

. Not sure indeed. But in the "Results" item I would say "we conclude" rather than "concluded".
OK

Introduction

. Page 1, redshift = 0.212 (in the text we have =212).
Done

. Page 2 and later in the paper, we should check that we insert the hyphen any time it is needed. For instance in "gamma-ray", etc.
OK, I have checked where should or shouldn't put the hyphen (in my understanding) but I guess this should be checked by the native speakers of the collaboration.

. Page 1, The EBL is the diffuse rational that came... the past here sounds strange to me. I would have used "come", but of course English is not my mother tongue!
Same as the previous point.

. page 1, right column, (from THE X-ray Timing Explorer)
Done.

. Page 1, right column, [...] no evidence of variability [...], we should add here on which timescale.
Done.


Results

Page 3, right column. Fig. 1, rather than Fig 1.
Done.

Page 3, just after Eq. 1. Here I have my first concern. The fit quality we report is rather poor, and with these parameters reporting the errors should not be statistically allowed. I think we should at least mention the matter and discuss whether it might have consequences for us.
Well, indeed the fit is rather poor, but the fact that the fit to the de-absorbed spectrum is better just with a power law goes in the direction that there is some "features" that we cannot fit in the observed spectrum. As can be seen in Fig. 4, the fit does not improve using more complex functions.

EBL measurement

Page 4, Fig. 4 caption. The tested models should be five. Right?
Yes, 5.

Page 4, right column, [...] the two hypotheses [...]. not hypothesis.
Done.

Page 4, right column. This is only a feeling, but the fit to the straight line in Fig. 3 looks better than the reported Chi2. I am just asking to check.
I checked and the data is consistent.

Page 5, left column. Here I essentially agree with Ulisses' comment 12, and this is my second (a bit more) serious concern. I understand that what we write is reasonable, but for a generic reader it sounds rather unsatisfactory. We should devote some more discussions to the PWL case.
I already answered to Ulisses concerning that point. I'm still think that we cannot use more statistical arguments because the statistics is blind to the physics of the problem. The explanation may needs to be more extended, but we cannot follow blindly what the statistical test tell us without taking into account the physics that we know about the problem.

Systematic uncertainties

Page 5. Right column. Possibly we should use {\rm scale} when we report \alpha.
OK


Conclusions

Page 7, Fig. 8. This plot is amazing, yet I'm afraid very difficult to read. No way to publish it in landscape using an entire page?
I agree that the figure looks kinda small, but I guess publishing it on an entire page will depend on the journal.

[stefano.covino]

Hello everybody,

I am indeed satisfied by the replies. However, please allow me to better explain my point about the fit quality. I do agree it is a minor point, and it does not affect the whole paper. Yet, what I meant is simply that if the fit quality is poor, associating errors to the fit parameters, and basing consideration on these uncertainties, is not an allowed procedure. I would consider this thing in the discussion, although in our case I do not think it is a problem.

Stefano

[ulisses.barresdealmeida]

Dear all,

I reply per points, but first of all, I would just like to stress that some observation in the discussion should be made towards Covino's point on the fit quality and parameter derivation.

Now,

(1) Independently if the systematics of HESS is better treated than ours or know, and if thus we cannot consider our results are de facto more stringent, still, whatever is the novelty of the result should be made explicit in the abstract in opposition to what already has been done in the field. Unless there is nothing that it really adds either as a result or as an experimental achievement. In our case, I believe that you should stress than the fact that this is an achievement for a single source constraint, from a source of the highest redshift. Nowhere in the abstract this most remarkable feature of the paper is made explicit.

(5) One thing is how the plot is visualised. The other thing is the definition of horizon. The horizon is a distance not an energy. Of course the function can be inverted, but this does not change the definition of the word "horizon", which is associated to a distance. Please I think you should rewrite.

(8) OK.

(9) OK, you convinced me this is fine.

(10) I agree with you, the argumentation should remains even with the table, but I would like to see the numbers comparing the different models. It just adds to the model. Is this to be a letter? If not, then we don't have to be necessarily economic in the use of tables.

(11) OK.

(12) No, the AIC you used is a proper statistical account of the problem, and I believe sufficient - unless someone better informed than me would have reasoned to suggest a different test. Now, once you found consistent in this, please evaluate if it is worth mentioning or not that you performed this on the paper. I leave this at your discretion.

(13) OK.

(14) You are right.

(16) thanks. I would like to read this bit once included, but that is all.

I am happy with everything... apart from (5) and (10).

Thanks and best wishes (and sorry for my delay).
Ulisses.

[adiv.gonzalezmunoz]

Hi! sorry for the delayed reply.

Stefano
I rephrased the text to clarify that even when the fit to a log-par is not very good, is the best fit using simple functions with not so many parameters.

Ulisses
Regarding the definition of the horizon, Daniel Mazin insisted that the way it is written is the correct definition.

About the table, since is to strength the argumentation of the election of the model I would include the likelihood probabilities for the case with alpha=0 and alpha=1.07. Would that be OK?

I will upload the new version of the draft shortly


Adiv

[david.paneque]

Hi Adiv,
sounds good. With those changes, you address all the remarks from the two referees.
When can you have a new version of the manuscript ?
Once you have it, you can already circulate the paper through magic priv and give the nominal 2 weeks for further comments.
cheers,
D.

[dario.hrupec]

A few comments on style
===================

* in Introduction, paragraph 6, lines 3-5 *
"The CGRH is defined as the energy at which the optical depth of the photon-photon pair production becomes unity as function of redshift."
this sounds unclear to me, let me propose e.g.
"By definition, the CGRH is the energy at which the optical depth of the photon-photon pair production is qual to 1. The optical depth is a function of redshift."

* in Abstract, line 2 *
(VHE, E>100GeV) ---> (VHE, E > 100 GeV)
E in italic; spaces

* in Introduction, paragraph 1, line 3 *
at redshift =0.212 ---> at redshift 0.212
I don't think "=" is necessary

* in Introduction, paragraph 1, line 7 *
(VHE, E>100GeV) ---> (VHE, E > 100 GeV)
E in italic, spaces

* in Introduction, paragraph 4, line 6 *
(E < 25 GeV)
E in italic

* in Introduction, paragraph 4, last line *
the EBL of 3 ± 1 nW... ---> the EBL of (3 ± 1) nW...
parenthesis

* in Introduction, paragraph 5, last line *
of 15 ± 2 stat ± 3 sys nW... ---> of (15 ± 2 stat ± 3 sys) nW...
parenthesis

* in Introduction, paragraph 7, lines 6-7 *
* ALSO in Introduction, paragraph 7, line 16 *
* ALSO in Introduction, paragraph 8, line 5 *
F(>200 GeV)
F in italic

* in Results, paragraph 2, line 2 *
E > 200 GeV
E in italic

* in Results, equation 1 *
differentials (letters d) in upright style, not italic

* in Results, figure 1 *
E > 200 GeV
E in italic

* in Results, figure 2 *
differentials (letters d) in upright style, not italic
units (TeV, cm, s) in upright style, not italic

* in EBL measurement, paragraph 1, equation 2 and line after it*
differentials (letters d) in upright style, not italic

* in EBL measurement, paragraph 3, lines 10 and 13*
Eest
E in italic

* in Conclusions, last line*
z=0.212 ---> z = 0.212
z in italic; spaces