In which I summarise the current debate over whether this apparent "new law of nature" means anything or not, and conclude (spoiler !) that it probably doesn't.
It's been known for many years that there's a correlation between how fast matter in a galaxy rotates and how much gas and stars it contains. The problem is that there shouldn't be a nice relation, because galaxies seem to be dominated by dark matter. This strange relationship between normal and dark matter, which should be independent of each other, has been shown in different ways over the years. Recently it was claimed that all of these are just manifestations of a deeper underlying "law" : the Mass Discrepancy Acceleration Relation (or technically the Radial Acceleration Relation, but whatever).
The problem is that the expected acceleration of matter (based on its mass and standard Newtonian gravity) correlates very, very well with its actual acceleration. If there is, as we think for many other reasons, actually a huge amount of unseen "dark matter" present, then it shouldn't do that. Could it be that the dark matter theory is wrong ?
YES ! But it probably isn't. Although modified theories of gravity do predict this relation - indeed, did predict it 30 years ago - it seems that it also occurs in standard dark matter simulations without any difficulties. You can see a comparison of observations (left) and the simulations (right) in the example image below. Since both modifying gravity and using dark matter give the same result, this discovery is probably useless. Weird, but useless.
Read on for tales of academics behaving badly, pretty pictures of galaxies, lots of and lots of graphs, and a surprised dog...
Placeholder post intended to be replaced with a slightly better summary.
Thursday, 3 August 2017
Monday, 31 July 2017
Ghostly giants versus huge dwarves - the battle continues
Ghostly giants versus huge dwarves - the battle continues
Ultra-diffuse galaxies are currently the trendiest thing in extragalactic astronomy. The same size as the Milky Way but a thousand times fainter, these things may be extremely boring to look at but they break all the rules of galaxy formation. Or do they ? Two years after first hitting the headlines, I review the current state of the art on the most important question : how massive are they ? Are they actually just huge dwarves or are they genuinely ghostly giants ? What have we really learned so far and where are we going ?
With star-studded cast of a giant squid, Father Ted, Gimli, Ghostbusters and for some reason James Corden...
Placeholder post intended to be replaced with a better summary.
Ultra-diffuse galaxies are currently the trendiest thing in extragalactic astronomy. The same size as the Milky Way but a thousand times fainter, these things may be extremely boring to look at but they break all the rules of galaxy formation. Or do they ? Two years after first hitting the headlines, I review the current state of the art on the most important question : how massive are they ? Are they actually just huge dwarves or are they genuinely ghostly giants ? What have we really learned so far and where are we going ?
With star-studded cast of a giant squid, Father Ted, Gimli, Ghostbusters and for some reason James Corden...
Placeholder post intended to be replaced with a better summary.
Thursday, 27 July 2017
Yet another Dark Matter versus MOND paper
Well... partially. Mainly this paper is about whether the standard dark matter model can explain some of the recently discovered "ultra faint" dwarf galaxies in the Local Group. Cold Dark Matter (CDM) models predict that really small dark matter halos, which have low rotational speeds, shouldn't be able to form any stars at all. They shouldn't have enough mass to pull in enough gas to form stars... but that's just what these new galaxies seem to have done.
In this paper, the authors use new simulations with much higher resolution and more complex physics than previous efforts. They show that similar objects can form through tidal stripping (i.e. gravitational effects), which can remove large fractions of the dark matter and stars but, interestingly, doesn't change the size of the galaxy. So apparently weird new objects (like the "giant" Crater II) may make sense after all.
They also show more quantitatively that the main objections to this idea aren't as fatal as might be thought. First, tidal stripping is thought to be very disruptive, with stars and gas and dark matter flying all over the place. And it is... but not for very long. The damage is done only when the dwarf galaxy flies closest to its bigger, scarier neighbour. The rest of the time the tidal effects don't really do a lot, and afterwards the galaxy can settle down into a nice round shape, like those that are observed. In essence it's like that bit in Jaws where the decapitated head suddenly appears - everyone jumps, but you don't go running off down the street screaming your head off for ten minutes afterwards. Usually.
The second objection is that these tidally-disrupted galaxies should only be found near their larger neighbours. But because the scary moment of disruption is short-lived, the galaxy can continue on its merry way back out to the hinterlands. So it's no surprise that the observed galaxies are found quite far away from the Milky Way.
All well and good, but now we reach the controversial stuff....
First, there were several recent papers claiming the discovery of a "new law of nature", where the gravity from dark matter appeared to be closely correlated with the gravity from normal matter. Which seemed a bit odd, but then there were several other papers which said, "yeah, but we see this is dark matter simulations anyway, shut your ugly face, it's not very interesting." Or words to that effect.
This paper shows that the scatter in this "mass discrepancy acceleration relation" is much higher for faint galaxies than was previously reported. They also claim that their simulations explain this. I think this is true, but the agreement is not that great. It's OK, but not particularly impressive. More interesting is simply the claim about the size of the scatter, which I find fairly convincing. Our resident tame MOND expert says that the original papers already discussed this, but I don't think this is true. However, they do mention that they only use the highest-quality data available. A better objection might be whether these new galaxies are rotating at all or if the stars are just moving on random orbits - this might not change the conclusions, but it should be discussed.
The second controversial point was the claim that they have discovered something which poses a "possibly insurmountable challenge" to Modified Newtonian Dynamics, the main alternative to dark matter. Strong words. Strong, angry words - which have been used before against MOND... but with admirable if cockroach-like tenacity the bloody thing keeps coming back.
MOND has this funny thing called the "external field effect". If there's another galaxy nearby, acceleration in a smaller galaxy reverts back to standard Newtonian behaviour. When you account for this, it seems that the velocities in the faint galaxies are in strong disagreement with MOND's predictions (they're also in strong disagreement even if you ignore it).
The tame MOND expert agrees with the method used, but disputes the conclusion. He's probably right that the conclusion is too strong, although personally I find claims of "a new law of nature" to be far worse on that front. His objection is that the fraction of binary stars in a galaxy can change the overall measured velocity dispersion, or it could be that even those the galaxies aren't in the process of being tidally disrupted right now, they're still out of equilibrium. It's possible, but I'm not convinced this can really explain the strong systematic offset seen in the data. It needs to be quantified though.
The paper concludes by worrying that this model requires some pretty dramatic tidal disruption to explain the galaxies - they've have to have lost 99% of their original mass (whether this is common in their simulations is not clear). But with measurements of the 3D velocities of the galaxies it would be possible to trace their orbits back in time and work out whether they really have experienced close encounters with the giant galaxies of the Local Group. So, as usual, watch this space.
https://arxiv.org/abs/1707.03898
In this paper, the authors use new simulations with much higher resolution and more complex physics than previous efforts. They show that similar objects can form through tidal stripping (i.e. gravitational effects), which can remove large fractions of the dark matter and stars but, interestingly, doesn't change the size of the galaxy. So apparently weird new objects (like the "giant" Crater II) may make sense after all.
They also show more quantitatively that the main objections to this idea aren't as fatal as might be thought. First, tidal stripping is thought to be very disruptive, with stars and gas and dark matter flying all over the place. And it is... but not for very long. The damage is done only when the dwarf galaxy flies closest to its bigger, scarier neighbour. The rest of the time the tidal effects don't really do a lot, and afterwards the galaxy can settle down into a nice round shape, like those that are observed. In essence it's like that bit in Jaws where the decapitated head suddenly appears - everyone jumps, but you don't go running off down the street screaming your head off for ten minutes afterwards. Usually.
The second objection is that these tidally-disrupted galaxies should only be found near their larger neighbours. But because the scary moment of disruption is short-lived, the galaxy can continue on its merry way back out to the hinterlands. So it's no surprise that the observed galaxies are found quite far away from the Milky Way.
All well and good, but now we reach the controversial stuff....
First, there were several recent papers claiming the discovery of a "new law of nature", where the gravity from dark matter appeared to be closely correlated with the gravity from normal matter. Which seemed a bit odd, but then there were several other papers which said, "yeah, but we see this is dark matter simulations anyway, shut your ugly face, it's not very interesting." Or words to that effect.
This paper shows that the scatter in this "mass discrepancy acceleration relation" is much higher for faint galaxies than was previously reported. They also claim that their simulations explain this. I think this is true, but the agreement is not that great. It's OK, but not particularly impressive. More interesting is simply the claim about the size of the scatter, which I find fairly convincing. Our resident tame MOND expert says that the original papers already discussed this, but I don't think this is true. However, they do mention that they only use the highest-quality data available. A better objection might be whether these new galaxies are rotating at all or if the stars are just moving on random orbits - this might not change the conclusions, but it should be discussed.
The second controversial point was the claim that they have discovered something which poses a "possibly insurmountable challenge" to Modified Newtonian Dynamics, the main alternative to dark matter. Strong words. Strong, angry words - which have been used before against MOND... but with admirable if cockroach-like tenacity the bloody thing keeps coming back.
MOND has this funny thing called the "external field effect". If there's another galaxy nearby, acceleration in a smaller galaxy reverts back to standard Newtonian behaviour. When you account for this, it seems that the velocities in the faint galaxies are in strong disagreement with MOND's predictions (they're also in strong disagreement even if you ignore it).
The tame MOND expert agrees with the method used, but disputes the conclusion. He's probably right that the conclusion is too strong, although personally I find claims of "a new law of nature" to be far worse on that front. His objection is that the fraction of binary stars in a galaxy can change the overall measured velocity dispersion, or it could be that even those the galaxies aren't in the process of being tidally disrupted right now, they're still out of equilibrium. It's possible, but I'm not convinced this can really explain the strong systematic offset seen in the data. It needs to be quantified though.
The paper concludes by worrying that this model requires some pretty dramatic tidal disruption to explain the galaxies - they've have to have lost 99% of their original mass (whether this is common in their simulations is not clear). But with measurements of the 3D velocities of the galaxies it would be possible to trace their orbits back in time and work out whether they really have experienced close encounters with the giant galaxies of the Local Group. So, as usual, watch this space.
https://arxiv.org/abs/1707.03898
Saturday, 22 July 2017
There... are... THREE... clouds !
Thanks everyone who voted on yesterday's poll (https://plus.google.com/u/0/+RhysTaylorRhysy/posts/h4ePJaDsHXf). The final score was :
1 group : 0%
2 groups : 7%
3 groups : 71 %
4 groups : 16%
More : 7%
... and the correct answer is.... there areFOUR LIGHTS three groups !
This diagram shows the location of galaxies in a particularly complicated part of the Virgo cluster, where, in addition to the main cluster, it's believed there are two other infalling sub-groups. Distance measurements to individual galaxies are hard but velocity (redshift) measurements are much easier. The distance measurements which we do have indicate that these three different groups are at different distances. It's also possible to see this using the velocity measurements, with each group centred on a different overall velocity (though with lots of scatter).
A colleague of mine disputed whether you'd really pick out three distinct groups without already knowing the distances. I thought it looked clear enough, but heck, this is the age of social media so why not test it ? The diagram I showed you yesterday was the same as the one below, just stripped of all distance information and rotated to a random angle. I deliberately gave you absolutely no information on the problem and didn't specify what size features count as significant - I wanted to see what you'd naturally guess without over-thinking what counts as a group.
Turns out the answer's three. Since I told you absolutely nothing, that makes the conclusion much stronger - not only can you pick out the three groups, but that's what most people naturally actually do. So you definitely don't need the distance measurements to realise there are three distinct groups here.
Clusters are messy places, so it's completely understandable that you might think there are more than three groups from this limited information. Even with the distance assignments, you can see that some objects at similar velocities apparently below to different groups. That's probably not actually the case - more likely, the distance assignments just didn't take this velocity information into account. There's clearly a lot of scatter in all cases - these aren't nice spheroidalish distributions - this is the real universe, messy and ugly and with galaxies lying in awkward locations. It's quite possible that a few of them do lie at quite different distances to the others; whether they would constitute separate groups or not is another matter (though not a terribly interesting one).
1 group : 0%
2 groups : 7%
3 groups : 71 %
4 groups : 16%
More : 7%
... and the correct answer is.... there are
This diagram shows the location of galaxies in a particularly complicated part of the Virgo cluster, where, in addition to the main cluster, it's believed there are two other infalling sub-groups. Distance measurements to individual galaxies are hard but velocity (redshift) measurements are much easier. The distance measurements which we do have indicate that these three different groups are at different distances. It's also possible to see this using the velocity measurements, with each group centred on a different overall velocity (though with lots of scatter).
A colleague of mine disputed whether you'd really pick out three distinct groups without already knowing the distances. I thought it looked clear enough, but heck, this is the age of social media so why not test it ? The diagram I showed you yesterday was the same as the one below, just stripped of all distance information and rotated to a random angle. I deliberately gave you absolutely no information on the problem and didn't specify what size features count as significant - I wanted to see what you'd naturally guess without over-thinking what counts as a group.
Turns out the answer's three. Since I told you absolutely nothing, that makes the conclusion much stronger - not only can you pick out the three groups, but that's what most people naturally actually do. So you definitely don't need the distance measurements to realise there are three distinct groups here.
Clusters are messy places, so it's completely understandable that you might think there are more than three groups from this limited information. Even with the distance assignments, you can see that some objects at similar velocities apparently below to different groups. That's probably not actually the case - more likely, the distance assignments just didn't take this velocity information into account. There's clearly a lot of scatter in all cases - these aren't nice spheroidalish distributions - this is the real universe, messy and ugly and with galaxies lying in awkward locations. It's quite possible that a few of them do lie at quite different distances to the others; whether they would constitute separate groups or not is another matter (though not a terribly interesting one).
Monday, 3 July 2017
The busy life of an astronomer
EWASS is over and I'm still alive; the 1200-strong horde of barbarous astronomers has been sent back to the hellish netherworld from whence they came. I finally booked my summer holiday to see the total solar eclipse from Grand Teton (hello darkness, bye bye money - also, Expedia seems designed to make the whole process as nerve-inducing as watching a horror movie while receiving random electric shocks). My flat has been restored to something approaching normality. Today we have a scientific visitor hanging around after EWASS. Tonight/early tomorrow morning a friend arrives for a week of holiday and on Wednesday I'm giving a 90 minute public talk.
Wheeee.....
Wheeee.....
Friday, 30 June 2017
Conference concert
Conference concert in the Rudolphinium. And guess who got to sit dead centre in the third row from the front ? Me, that's who. I win, bitches.
And with that I withdraw once more for one final day of non-stop astronomy...
[I found out later that this was entirely down to luck, as the seats were given out completely at random]
Sunday, 25 June 2017
EWASS begins
EWASS. the European Week of Astronomy and Space Science, has now commenced. For the next week I'll be fighting off the invading horde of 1100 barbarous astronomers and the week after that will be spent recovering. I'll be online, but intermittently.
http://eas.unige.ch/EWASS2017/about.jsp
http://eas.unige.ch/EWASS2017/about.jsp
In Theory
Alternative title : Ten Times Scientists Didn't Use The Word Theory To Mean A Well-Tested Model That's Almost A Fact Because That's Not What The Damn Thing Means So Just Get Over It Already.
Admittedly, I do keep flip-flopping on whether "theory" means, "incredibly well-tested" model or something else. This post should definitively clear that up by making it abundantly clear than everything is much more complicated than that.
Clearly there are some theories which do extraordinarily well - sometimes so well that theory and fact are indistinguishable. It might be fair to start to describe these as laws, not theories - the law of gravity, the law of evolution. Both of these things are established factual processes. Yet even these are like Russian dolls : within them we find detailed theoretical models of how they occur, and within those we find competing hypotheses as to how particular aspects proceed and even rivals to the theory - but not the laws. Gravity is a thing. Evolution happens. It's the mechanisms by which these things occur that's open to debate (at least a little bit), not their very existence.
Even if we were to insist that hypothesis only means, "explanation with little or no testing" (which it does) and theory only means, "well-tested explanation" (which it doesn't), then it wouldn't be easy to distinguish between the two. No strict criteria of what "well tested" means exists. It's probably impossible anyway, given the incredibly diverse nature of theories. You can't equate cat emotions with the distortion of spacetime around a black hole, or at least you shouldn't.
The reality is, though, that the vast majority of theories fall somewhere between these two extremes. They aren't just speculations based on limited data, and they aren't so convincing that no other explanations are plausible. They've had some testing and they generally work, but they have room for improvement. Some of them might turn out to be completely wrong, others just need tweaking.
I'm all for rigorous definitions wherever that's possible and appropriate. But in the case of "theory" I think that neither is the case. The simple truth of the matter is that science isn't always purely objective. It's a murky, messy business of turning facts into models, testing those models, rejecting some while provisionally tolerating others. Pretending that it's more objective than it actually is won't work, because it simply isn't true. Would it be nice if it was ? Sure ! But that's not what it's like, and that murkiness is sometimes what makes it fun.
No definition will stop the most ardent from bullshitting about science, because these people simply do not care - and you can't argue with someone who doesn't care, you can only have shouting matches. But for the rest, let's not set ourselves up for disaster by pretending we know things we do not. Simply admit the plain truth of it - that we know hardly anything for certain, but we're far, far more confident about some things than others. If this leaves people feeling lost and insecure, then that would be a good start. Perhaps (and I say this cautiously, knowing how damaging bullshit and stupidity can be) then they'd stop the chest-thumping for a moment, begin to realise that not everything can be quantified, and actually learn how to think.
https://astrorhysy.blogspot.com/2017/06/in-theory.html https://astrorhysy.blogspot.com/2017/06/in-theory.html
Admittedly, I do keep flip-flopping on whether "theory" means, "incredibly well-tested" model or something else. This post should definitively clear that up by making it abundantly clear than everything is much more complicated than that.
Clearly there are some theories which do extraordinarily well - sometimes so well that theory and fact are indistinguishable. It might be fair to start to describe these as laws, not theories - the law of gravity, the law of evolution. Both of these things are established factual processes. Yet even these are like Russian dolls : within them we find detailed theoretical models of how they occur, and within those we find competing hypotheses as to how particular aspects proceed and even rivals to the theory - but not the laws. Gravity is a thing. Evolution happens. It's the mechanisms by which these things occur that's open to debate (at least a little bit), not their very existence.
Even if we were to insist that hypothesis only means, "explanation with little or no testing" (which it does) and theory only means, "well-tested explanation" (which it doesn't), then it wouldn't be easy to distinguish between the two. No strict criteria of what "well tested" means exists. It's probably impossible anyway, given the incredibly diverse nature of theories. You can't equate cat emotions with the distortion of spacetime around a black hole, or at least you shouldn't.
The reality is, though, that the vast majority of theories fall somewhere between these two extremes. They aren't just speculations based on limited data, and they aren't so convincing that no other explanations are plausible. They've had some testing and they generally work, but they have room for improvement. Some of them might turn out to be completely wrong, others just need tweaking.
I'm all for rigorous definitions wherever that's possible and appropriate. But in the case of "theory" I think that neither is the case. The simple truth of the matter is that science isn't always purely objective. It's a murky, messy business of turning facts into models, testing those models, rejecting some while provisionally tolerating others. Pretending that it's more objective than it actually is won't work, because it simply isn't true. Would it be nice if it was ? Sure ! But that's not what it's like, and that murkiness is sometimes what makes it fun.
No definition will stop the most ardent from bullshitting about science, because these people simply do not care - and you can't argue with someone who doesn't care, you can only have shouting matches. But for the rest, let's not set ourselves up for disaster by pretending we know things we do not. Simply admit the plain truth of it - that we know hardly anything for certain, but we're far, far more confident about some things than others. If this leaves people feeling lost and insecure, then that would be a good start. Perhaps (and I say this cautiously, knowing how damaging bullshit and stupidity can be) then they'd stop the chest-thumping for a moment, begin to realise that not everything can be quantified, and actually learn how to think.
https://astrorhysy.blogspot.com/2017/06/in-theory.html https://astrorhysy.blogspot.com/2017/06/in-theory.html
Saturday, 17 June 2017
Monday, 12 June 2017
Crowdsourced reviewing
Interesting and novel approach. Via Sakari Maaranen.
I am not proposing what is sometimes referred to as crowdsourced reviewing, in which anyone can comment on an openly posted manuscript. I believe that anonymous feedback is more candid, and that confidential submissions give authors space to decide how to revise and publish their work. I envisioned instead a protected platform whereby many expert reviewers could read and comment on submissions, as well as on fellow reviewers’ comments. This, I reasoned, would lead to faster, more-informed editorial decisions.
I recruited just over 100 highly qualified referees, mostly suggested by our editorial board. We worked with an IT start-up company to create a closed online forum and sought authors’ permission to have their submissions assessed in this way. Conventional peer reviewers evaluated the same manuscripts in parallel. After an editorial decision was made, authors received reports both from the crowd discussion and from the conventional reviewers.
This January, we put up two manuscripts simultaneously and gave the crowd 72 hours to respond. Each paper received dozens of comments that our editors considered informative. Taken together, responses from the crowd showed at least as much attention to fine details, including supporting information outside the main article, as did those from conventional reviewers.
So far, we have tried crowd reviewing with ten manuscripts. In all cases, the response was more than enough to enable a fair and rapid editorial decision. Compared with our control experiments, we found that the crowd was much faster (days versus months), and collectively provided more-comprehensive feedback.
https://www.nature.com/news/crowd-based-peer-review-can-be-good-and-fast-1.22072
I am not proposing what is sometimes referred to as crowdsourced reviewing, in which anyone can comment on an openly posted manuscript. I believe that anonymous feedback is more candid, and that confidential submissions give authors space to decide how to revise and publish their work. I envisioned instead a protected platform whereby many expert reviewers could read and comment on submissions, as well as on fellow reviewers’ comments. This, I reasoned, would lead to faster, more-informed editorial decisions.
I recruited just over 100 highly qualified referees, mostly suggested by our editorial board. We worked with an IT start-up company to create a closed online forum and sought authors’ permission to have their submissions assessed in this way. Conventional peer reviewers evaluated the same manuscripts in parallel. After an editorial decision was made, authors received reports both from the crowd discussion and from the conventional reviewers.
This January, we put up two manuscripts simultaneously and gave the crowd 72 hours to respond. Each paper received dozens of comments that our editors considered informative. Taken together, responses from the crowd showed at least as much attention to fine details, including supporting information outside the main article, as did those from conventional reviewers.
So far, we have tried crowd reviewing with ten manuscripts. In all cases, the response was more than enough to enable a fair and rapid editorial decision. Compared with our control experiments, we found that the crowd was much faster (days versus months), and collectively provided more-comprehensive feedback.
https://www.nature.com/news/crowd-based-peer-review-can-be-good-and-fast-1.22072
Subscribe to:
Posts (Atom)
ChatGPT Is A Competent Source Extractor
This builds on my post from two years ago, ChatGPT Is Not A Source Extractor . My, how things change. Last time I wrote : They're still ...
