Asymmetry in Galaxy Spin Directions: A Fully Reproducible Experiment Using HSC Data
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsSection 4 is scientifically irrelevant. Other results and the conclusions are to be clarified.
Comments on the Quality of English LanguageMinor corrections, including a few typos.
Author Response
Section 4 is scientifically irrelevant. Other results and the conclusions are to be clarified.
--Author response: I would like to thank you for the time you spent reading and commenting on the manuscript. Some explanations have been added about the method and results, and the new additions are highlighted in bold font. Namely, an analysis when the galaxies are assigned random directions of rotation has been added, while in the previous version just the maximum probability was specified. The full reproducibility of that experiment (code, data and step-by-step instructions) has also been added and is now available for the readers. Some description about the annotation method has been added too, although the full long technical description that also includes experimental analysis is available in the cited papers.
I agree that in a perfect scientific community there is no need for Section 4. The problem is that the “reproducibility crisis” in science is real, and science is being attacked by research results that cannot be reproduced, and therefore cannot be trusted. As the studies cited in the paper suggest, that is the majority of the papers being published today. It is not a convenient or ideal situation, but that is the reality we currently need to handle. It's a major problem in science. In particular, there are a few papers mentioned in Section 4 that suggest that my work is part of the reproducibility crisis. It is therefore important to address these papers and explain them. Ignoring these papers as if they did not exist would leave the reader wondering how they can be explained, and might also be seen as an ethical misconduct on my side, of intentionally ignoring papers I am aware of. I would certainly want to avoid a situation where I only discuss papers that agree with me, but avoid discussing those papers that claim the opposite. That is the importance of Section 4, and it is added to the fully reproducible nature of the paper, allowing every scientist to ensure that the results are fully reproducible and the data are exactly as described in the paper. A note about it has also been added to the Conclusion section in the revised version of the paper. In general, the reproduc8ibility crisis is a real threat to science, and it should be addressed.
Comments on the Quality of English Language
Minor corrections, including a few typos.
--Author response: Indeed a few typos were identified and corrected.
Reviewer 2 Report
Comments and Suggestions for AuthorsI read the manuscript with interest. I have some minor comments:
* Lines 124 to 127: I seem to find a contradiction between the sentences "galaxies rotating clockwise will be brighter" and "an excessive number of counterclockwise galaxies will be observed" when referring to the Northern end of the Galactic pole. Shouldn't it be "galaxies rotating counterclockwise will be brighter"? Similar for the Southern end of the Galactic pole.
* Table 1: it seems that the caption is not complete.
* Lines 164 to 168: "delta" should be Greek letter delta.
* Line 241: shouldn't "sueprclean" be "superclean"?
Author Response
I read the manuscript with interest. I have some minor comments:
--Author response: I would like to thank you for the time you invested in reading and commenting on the paper. The comments have been addressed as specified under each comment below. Some more information has also been added such as the analysis when assigning the galaxies with random directions of rotation. Subtracting two random vectors is not expected to lead to signal, so the analysis might seem trivial, but it is just another “sanity check”. It has also been made fully reproducible with code and data.
* Lines 124 to 127: I seem to find a contradiction between the sentences "galaxies rotating clockwise will be brighter" and "an excessive number of counterclockwise galaxies will be observed" when referring to the Northern end of the Galactic pole. Shouldn't it be "galaxies rotating counterclockwise will be brighter"? Similar for the Southern end of the Galactic pole.
--Author response: That is exactly correct. Galaxies that rotate in the opposite direction relative to the Milky Way should be brighter, and therefore in the Northern Galactic pole more galaxies that rotate *counterclockwise* are expected. That has been corrected in the text.
* Table 1: it seems that the caption is not complete.
--Author response: That is true. The caption was for some reason cut. It has been corrected in the revised version. Thank you!
* Lines 164 to 168: "delta" should be Greek letter delta.
--Author response: Thank you. That was an error with the latex notation and it has been corrected in the revised version.
* Line 241: shouldn't "sueprclean" be "superclean"?
--Author response: That was indeed a typo and it has been corrected. Thank you.
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsI think the paper would benefit, if the explicit explanations and comments from my report will be accounted. I am the first reader of the manuscript and my report briefly describes my impression of the presented results. The presentation should and can be improved in order to sounds more consistent and convincing. The ``reproducibility crisis'' is totally unnecessary in the overall context of the performed study.
Comments for author File: Comments.pdf
English is reasonable, only minor corrections will be sufficient.
Author Response
Thank you for the time you spent on reading and commenting on the paper. First of all, my apologies for not addressing your comments in the first round of the review. Your comments came as a separate file, and for some reason the existence of that file escaped my attention so I completely missed it in the on-line submission system. I apologize for that. In the second round I finally became aware of your specific comments, and therefore addressed them. The editor allowed me three days to complete the revisions, but I believe that all comments have been addressed, and many changes have been made to the manuscript. Response to the comments and the corresponding changes made to the manuscript based on them are specified below each comment. For convenience, changes are highlighted in the manuscript in bold font.
The revised version contains a few comments regarding the capabilities of the Ganalyzer algorithm. Unfortunately, the two my requests: one question and one comment, remained ignored, although they both relate the confidence of the results presented in the manuscript.
The importance of the discussed problems connected with the validity of the cosmological principle is unconditional. Therefore, the results of dedicated studies are to be pronounced explicitly and as precise as possible.
In this report I remind my queries:
- The asymmetry in the MS is classified only via a “binary” criterion – clockwise or counterclockwise (MW and OMW in the MS, Table 1). Generally, it does not look sufficient for a firm conclusion about the asymmetry. It is mentioned in Sec 2 (line 85) that “Ganalyzer provides the accuracy of virtually 100%”, however this means only the sign of galaxy spin: clockwise or counterclockwise. The accuracy of measurements of the galaxy’s spin direction with Ganalyzer remains unspecified in the MS. At the same time, the percentage of the difference between the numbers of MW and OMW is only 1.6%, and if the accuracy of the angle between the galaxies’ spins and the sight-lines towards them is > 2â—¦, the conclusion about the asymmetry turns to be elusive. In this regard, the difference between the assymetry index for samples with redshift range 0.0−0.1 and 0.1−0.2 shown in Fig 4 (upper and lower panels) would need special comments, because the accuracy of spin directions for the z = 0.1 − 0.2 sample seems to be 2 times of the one for the z = 0.0 − 0.1 sample as can be intuitively expected. To this question from my previous review I’ll add the next few points. As can be judged from the MS, the approach used in processing the statistical results on galaxy spins, is based on qualitative estimates inferred from the galaxies’ morphology, essentially within a binary (1,-1) approach. The reported excess of positives against negatives is very small, ≃ 1.6%, and in principle can be “flooded” beneath the inaccuracy of the exact direction of galaxies’ spin provided by the Ganalyzer from galaxies’ morphology. In this regard, one can mention that even for nearby galaxies at distances of 10–15 Mpc, the inclination angle of their disks is derived with the error from 0.5â—¦ to 5â—¦ depending on inclination and morphological features. Among possible sources are: the galaxies lopsidedness, spiral asymmetry, warps, vertical distibution of dust.
--Author response: Thank you for the comment. I believe the reason for the confusion between the sign of the galaxy spin directions and the accuracy of direction of the galaxy spin was bad wording that I used in the paper, and that has been corrected. The word “accuracy” is confusing here. It is the sign of the spin direction that is being studied in this paper. That description has been revised ensure that it is the direction of rotation that is being annotated, and that Ganalyzer rarely misannotate as can also be seen by browsing through the data.
The difference of 1.6% between MW and OWM galaxies is very much aligned with previous observations, starting with Galaxy Zoo 1 from about 15 years ago showing asymmetry of 1.5%-2% (although the small dataset did not allow statistical significance), and then other experiments. A discussion and comparison have been added to Section 3. Also, Table 1 now shows the (OMW-MW)/(OMW+MW) for the different redshift ranges in that table. It also shows that the value increases as the redshift gets higher, and that is also mentioned in the discussion.
It is correct that if the galaxies are tilted or if there is a difference in the line of sight it can lead to asymmetry. But if we assume that the inclination and spin direction of a galaxy are both random, that should impact galaxies that rotate clockwise in the same manner it impacts galaxies that rotate counterclockwise. The inclinations are random, and consistently the inclinations as observed from Earth are random. For that reason, the probability of a galaxy to appear to an Earth-based observer as a galaxy that rotate clockwise is the same as the probability of a galaxy to appear to be rotating counterclockwise. If the inclination or spin directions are somehow aligned, then you are correct that it can lead to asymmetry. But such alignment is not expected in such large fields. If the inclination and spin direction are both random, asymmetry is not expected because a galaxy will have equal chance to be rotating in either direction. That has been added to the paper. Moreover, as also mentioned in the paper, I am not sure at all that the observation is driven by the large-scale structure of the Universe. I suspect it could be driven by the internal structure of galaxies. That is, OMW galaxies are slightly brighter due to their rotational velocity, and therefore more of them can be seen from Earth. In that case, it is only the position of the galaxy relative to Earth that matters. I am not entirely sure that this is the case, but a note regarding that possibility is also in the paper. This entire discussion has been added to Section 4 in the revised version of the paper.
As for the accuracy of the spin directions changes in the range of 0.1<z<0.2, I indeed did not think about it but re-did the inspection of random galaxies by selecting 100 random galaxies from the dataset such that 50 had redshift of <0.1 and 50 had redshift >0.1 to test for their accuracy. All galaxies had the correct spin directions. That showed that the correctness of the annotation does not degrade as the redshift gets higher. The dataset is obviously open to the public. A discussion about it has been added to the Data section.
The change with the redshift is puzzling, but with JWST it can be verified very easily. JWST deep imaging shows an extreme asymmetry that can be also noticed by the naked eye, and without using any Ganalyzer or other algorithms (Shamir, 2024, PASA, 41, e038). That increase is also aligned with previous studies using SDSS, but the observation with JWST is the most clear and most obvious. That discussion has been added to the end of the Results section in the revised version.
- Section 4 is totally devoted to discussion of the so-called “reproducibility crisis” appeared in the scientific community during the recent decade when different fields of science faced problems with big data. The author explicitly aims this discussion to reply to a few publications, that critisized his papers with an alternative conclusion of lacking definite indication of the galaxy’s spin asymmetry. The replies are worded though in a manner far from being scientifically correct. The author baselessly claims these publications to contain irreproducible results, or even “dubious results”. In order to convince potential readers in these claims, the author presents an extended discussion of these issues in Sec 4, line 268 –290. My understanding is that the discussion in Sec 4 is written in an unacceptable form, its content is irrelevant to the problem under discussion and should be removed. This my note from the first review stays valid, with only a few additional points.
- i) The MS title and its abstract announce that the author presents new results in support of the galaxy spin asymmetry as derived from the HSC (Subaru Hyper Suprime-Cam) survey. The results complement the previous author’s study, are interesting by themselves and deserve clear presentation. On the contrary, a considerable part of the MS contains discussion of the “reproducibility crisis” problem, which seems not only irrelevant but rather confusing and destracting the reader.
- ii) The results of the paper by Patel & Desmond that partly provoked the author to
discuss the “reproducibility crisis”, are obtained with making use a code publicly available on https://github.com/harrydesmond/GalaxySpinAnisotropy, and the link is explicitly given in the end of the abstract of their paper [MNRAS 534, 1553–1560 (2024)].
iii) The discussion of aspects connected with the “reproducibility crisis” given by the
author in Sec 4 is interesting by itself, but I am not sure that the Symmetry journal is the right place for this, and definitely not the paper devoted to study validity of the cosmological principle.
- iv) Moreover, the discussion of the “reproducibility crisis” is presented not quite in an acceptable form. The impression which follows from the arguments given in Sec 4 (lines 302 to 323) is that the authors of Ref [61] published a wrong or even a falsified results, whereas the author of the refereed MS carries the truth. This discussion does not follow the scientific standars, and moreover sounds ethically inacceptable.
- v) There is an option for the author to show that the results presented in Ref [61] are not correct or irrelevant, is to carefully reproduce their Bayesian multivariate analysis for seven different samples presented by other authors (listed in their Table 1), and to demonstrate the reproducible results. This would be the only correct approach. In this regard, the last note in the abstract “The paper also discusses the impact of the “reproducibility crisis” on the study of the symmetricity of the distribution of galaxy rotations as observed from Earth” is irrelevant.
--Author response: Thank you for the detailed comment. Section 4 has been removed as you proposed, and also the last sentence in the abstract was removed as you also mentioned. I believe that the reproducibility crisis is a real problem in science, but at the same time I also accept your position that this paper might not be the optimal place to discuss it. The section has been therefore removed, with just a very brief summary that is based on references to previous papers made in the Introduction.
Although it is no longer relevant to the paper, I would note that the term “dubious research” was not proposed by me, but by the institution at which that research was done. I also shared a link to the communications in the previous version of the paper. The communications also include the authors declining to explain why the reproduction of the results provide different results than reported in the paper. Yet, I can agree that the reproducibility problem is outside the scope of this paper. It might require a new paper dedicated entirely to the reproducibility problem, but perhaps it is true that it is not the topic of this paper that deals with showing the asymmetry. It has therefore been removed other than brief mentioning in the Introduction section.
Minor comments of my previous review are also ignored. I would remind them:
- In Fig. 1 a short quantitative explanation of the deficient number of galaxies is
worth to be given. More precisely: what is the limiting stellar magnitude of galaxies in this sample? What is its estimated completeness? Whether the gap in the 1st bin and the decline at higher redshift bins do correspond to the expected behavior? Comments are essential for the results to sound convincing.
--Author response: Yes. A figure has been added to the revised version to show the magnitude. As the new graph shows, the vast majority of the galaxies are with (r) magnitude of 16-18, with very rare cases of galaxies dimmer than 19 (0.1% of the galaxies). The new figure and the discussion have been added.
- Fig. 2 is not commented and discussed in the text. At the same time, a few comments on the distribution of galaxies versus RA would make sense: the lack of galaxies in RA range between 45â—¦ – 105â—¦ and between 255â—¦ – 315â—¦ does not look obvious, and when compared to the map on Fig 3 it seems inconsistent. Explicit explanation are obviously needed.
--Author response: Yes. The RA distribution is driven by the footprint of HSC DR3. RA ranges that are outside of HSC DR3 footprint do not contain galaxies, and therefore also in this dataset they do not have galaxies. That has been added to the revised version, with a brief explanation in the figure caption and in the text.
- Lines 191 to 195 and Fig 5: in the Fig 5 captions the author note that images shown in the upper and lower panels look roughly the same. The term roughly seems to be important and needs comments: the minimum of σ of the low redshift sample lies exactly between the minumum and maximum of σ of the sample with higher redshifts, i.e. the difference between the distribution on sky is ∼ 30â—¦ in RA. Possible reasons for such a strong deviation and conclusions about the accurecy and confidence of the results are needed.
--Author response: The caption of Figure 5 does not argue that they look roughly the same, but that the *number of galaxies* in the two experiments is roughly the same. The number of galaxies in each experiment and the fact that the datasets are not overlapping have been added to the caption to avoid confusion. Also, a discussion has been added to the text where the figure is discussed. It adds a comparison of the most likely axes to show that the two datasets provide a consistent location of the dipole axis, and well within the 1 sigma. In the original version of the manuscript the most likely axes of the experiments were not specified. It also adds information about the increase with the redshift, and especially in JWST. JWST allows to see the asymmetry by the naked eye, and without the need for any algorithms. A short statement and a reference to the paper has been added.
The two datasets are independent, so it is not clear what the difference between the sigma means. The higher redshift has a maximum sigma of about twice the maximum sigma observed in the dataset of the low redshift galaxies, but that is a coincidence and does not have a special meaning that can be interesting. It is definitely not an average or anything of the kind. The two datasets are analyzed completely independently.
- Sec 3, lines 152 – 170: the χ2 test is formally applicable for normal distributions. This circumstance is worth mentioning.
--Author response: That has been added. If the sign of the spin direction is random as the inclination is random then normal distribution can be assumed.
- In its current form the “Conclusion” section contains a generic discussion of a relevance of the galaxy spin measurements to the parity violation problem and related topics, but does not actually summarize the specific results obtained with the Subaru HSC sample. It would make sense to define this section as “Discussion”, and add the section describing briefly the results obtained within the Subaru HSC survey and their comparison with the previous ones. The paragraph on lines 370 – 377 should be removed as irrelevant to the problem under consideration.
--Author response: Thank you for the comment. It’s a good point that the Conclusion section can be separated into Discussion. That has been done. The Conclusion section is a shorter section that just briefly summarizes the results. The Discussion section is somewhat longer, and describes the interpretations and existing cosmological theories that can be related to the observation. The paragraph about cosmic expansion has been removed.
I still think the presented paper in this form cannot be accepted in the Symmetry journal. Additional consideration is needed after the mentioned issues are accounted. The results as such deserve this.
--Author response: Once again I apologize for ignoring the comments from your previous report. The comments came as an additional file, which I failed to notice in the submission system. Given your response in this round I realized that I missed something in the previous round. But now the paper has changed substantially, and the comments have been addressed.
Round 3
Reviewer 1 Report
Comments and Suggestions for AuthorsThe presentation is greatly improved, although a few points needs rewording and additional explanations in order to make the paper more friendly for readers. Several minor points can be met in text and need a careful verification. The problem deserves a proper regard.
Comments for author File: Comments.pdf
Author Response
I would like to thank you again for the time you put in reviewing the changes and commenting on them. Responses to the comments are specified below, and changes made to the manuscript are highlighted in bold font.
After the second revision the MS sounds much better and clearer. Only a few points remained to be clarified. Among them:
lines 93–95 along with 99–100: the accuracy in the inferred asymmetry still remains the issue, though not answered explicitly. The easest way to answer this qustion would be to add in the author’s statement “At the same time, the algorithm also rejects galaxies that the algorithm cannot determine their direction of rotation” the quantity which characterizes the condition(s) when the algorithm rejects the galaxies. The referring to human eyes helps, but does not give a clearer understanding. Indeed, the human eye acuity is capable to resolve a point-like source of 20× 20′′ angular size. This seems to be insufficient to resolve even a typical galaxy at z ∼ 0.1 with an angular size of ∼ 3′′, while its features with angular sizes of <∼0.03′′ needed to exclude possible effects from galactic warps, lobsidedness, dust asymmetry in galactic disks will be totally lost. As mentioned early in the previous report, contaminations from these effects restrict the accuracy in determining exact inclinations even for nearby, within 10–15 Mpc, spiral galaxies. I think, explicit pronouncation of this issue would considerably improve presentation of the manuscript and its understanding.
--Author response: The annotation algorithm was described more thoroughly in previous papers, but it is definitely possible to add more information to this paper to make it easier for the reader. First of all, there is no involvement of the human eye in the process, and all of the annotation is done fully automatically. The human eye is misleading for this purpose, and only used to validate the annotation. Testing the annotations by eye might be interesting, and the readers can also test some random galaxies just to ensure that the data is fully consistent with the human eye. But the human eye is not what guarantees the correctness of the algorithm. That note has been added to the paper.
I would also mention that with the power of HSC, galaxies far above z of 0.1 can be identified by eye, and the open data shows that.
A more detailed description has been added about the conditions of rejection of galaxies, close to the sentence you specified. That includes an explanation of the threshold for rejecting a galaxy (by the regression coefficient), the specific minimum absolute value of the coefficient for rejecting a galaxy, and the frequency of galaxies that were rejected. The information also exists in the cited papers, but it can be added to this paper too for the convenience of the reader.
So the consistency of the annotation is based on symmetric analysis, and the manual inspection is just an additional method of observing the results. The evaluation of the algorithm does not depend in the manual inspection, and in fact it does not have any impact on the algorithm because all galaxies are classified correctly when compared with the human eye. But if for some reason a rare type of galaxies can confuse the algorithm, it is expected that galaxies of that type will be distributed equally between galaxies that rotate clockwise and counterclockwise. That note has also been added to the paper.
Figure 2 looks isolated from the bulk manuscript, although it would be essential to comment the gaps shown in the plot with the notes on lines 111 - 112 with regard to the RA ranges [45â—¦ −120â—¦] and [255â—¦ −330â—¦], along with the peaks of σ in these areas shown in Figures 4 & 6: as seen, the peaks of σ in Figs 4 & 6 fall directly into these gaps. In addition, “the most likely dipole axis” mentioned by the author in line 200 α = 94â—¦, δ = 52â—¦, falls nearly the center of the gap. In order for a reader to avoid misunderstanding these issues should be clearly explained.
--Author response: Yes. The analysis works by fitting the entire dataset into a dipole model. It is not direct measurement as done in (Shamir, 2022, MNRAS), but fitting of an entire model. It therefore does not depend on the population of galaxies in one specific part of the sky. Obviously, more galaxies and larger footprint will improve the accuracy of the dipole location, but there is nothing in the analysis that would shift the most likely dipole into a part of the sky that has no galaxies in it. Also, previous experiments with other telescopes that have different footprints showed that the most likely dipole axis fell right inside highly populated parts of the sky. An explanation has been added to Section 3 (in bold font) the revised version. A very quick note has been added to Section 2 where the RA distribution is discussed, with a reference to Section 3. It cannot be discussed in Section 2 because in that section discusses the data, before the analysis is introduced to the reader, and therefore the explanation is done mostly in Section 3.
line 253: the term “slighly brighter” would sound more convincing if accompanied with a numerical etimate.
--Author response: The expected magnitude difference is ~0.006, and that has been added to the revised version. The full analysis that shows why the expected difference is 0.006 is provided in the reference added in the end of the sentence.
The presented version of the paper can be accepted in the Symmetry journal after minor revision which would make the presentation more consistent and readable.