[Apologies — this is long, technical, and there are too few examples. I am putting it out for commentary more than anything else…]
In some recent articles and blog posts (including one in response to astronomer David Hogg), Columbia University statistician Andrew Gelman has outlined the philosophical position that he and some of his colleagues and co-authors hold. While starting from a resolutely Bayesian perspective on using statistical methods to measure the parameters of a model, he and they depart from the usual story when evaluating models and comparing them to one another. Rather than using the techniques of Bayesian model comparison, they eschew them in preference to a set of techniques they describe as ‘model checking’. Let me apologize in advance if I misconstrue or caricature their views in any way in the following.
In the formalism of model comparison, the statistician or scientist needs to fully specify her model: what are the numbers needed to describe the model, how does the data depend upon them (the likelihood), as well as a reasonable guess for what those numbers night be in the absence of data (the prior). Given these ingredients, one can first combine them to form the posterior distribution to estimate the parameters but then go beyond this to actually determine the probability of the fully-specified model itself.
The first part of the method, estimating the parameters, is usually robust to the choice of a prior distribution for the parameters. In many cases, one can throw the possibilities wide open (an approximation to some sort of ‘complete ignorance’) and get a meaningful measurement of the parameters. In mathematical language, we take the limit of the posterior distribution as we make the prior distribution arbitrarily wide, and this limit often exists.
The problem, noticed by most statisticians and scientists who try to apply these methods is that the next step, comparing models, is almost always sensitive to the details of the choice of prior: as the prior distribution gets wider and wider, the probability for the model gets lower and lower without limit; a model with an infinitely wide prior has zero probability compared to one with a finite width.
In some situations, where we do not wish to model some sort of ignorance, this is fine. But in others, even if we know it is unreasonable to accept an arbitrarily large value for some parameter, we really cannot reasonably choose between, say, an upper limit of 10100 and 1050, which may have vastly different consequences.
The other problem with model comparison is that, as the name says, it involves comparing models: it is impossible to merely reject a model tout court. But there are certainly cases when we would be wise to do so: the data have a noticeable, significant curve, but our model is a straight line. Or, more realistically (but also much more unusually in the history of science): we know about the advance of the perihelion of Mercury, but Einstein hasn’t yet come along to invent General Relativity; or Planck has written down the black body law but quantum mechanics hasn’t yet been formulated.
These observations lead Gelman and company to reject Bayesian model comparison entirely in favor of what they call ‘model checking’. Having made inferences about the parameters of a model, you next create simulated data from the posterior distribution and compare those simulations to the actual data. This latter step is done using some of the techniques of orthodox ‘frequentist’ methods: choosing a statistic, calculating p-values, and worrying about whether your observation is unusual because it lies in the tail of a distribution.
Having suggested these techniques, they go on to advocate a broader philosophical position on the use of probability in science: it is ‘hypothetico-deductive’, rather than ‘inductive’; Popperian rather than Kuhnian. (For another, even more critical, view of Kuhn’s philosophy of science, I recommend filmmaker Errol Morris’ excellent series of blog posts in the New York Times recounting his time as a graduate student in philosophy with Kuhn.)
At this point, I am sympathetic with their position, but worried about the details. A p-value is well-determined, but remains a kind of meaningless number: the probability of finding the value of your statistic as measured or worse. But you didn’t get a worse value, so it’s not clear why this number is meaningful. On the other hand, it is clearly an indication of something: if it is unlikely to have got a worse value then your data must, in some perhaps ill-determined sense, be itself unlikely. Indeed I think it is worries like this that lead them very often to prefer purely graphical methods — the simulations ‘don’t look like’ the data.
The fact is, however, these methods work. They draw attention to data that do not fit the model and, with well-chosen statistics or graphs, lead the scientist to understand what might be wrong with the model. So perhaps we can get away without mathematically meaningful probabilities as long as we are “just” using them to guide our intuition rather than make precise statements about truth or falsehood.
Having suggested these techniques, they go on to make a rather strange leap: deciding amongst any discrete set of parameters falls into the category of model comparison, against their rules. I’m not sure this restriction is necessary: if the posterior distribution for the discrete parameters makes sense, I don’t see why we should reject the inferences made from it.
In these articles they also discuss what it means for a model to be true or false, and what implications that has for the meaning of probability. As they argue, all models are in fact known to be false, certainly in the social sciences that most concerns Gelman, and for the most part in the physical sciences as well, in the sense that they are not completely true in every detail. Newton was wrong, because Einstein was more right, and Einstein is most likely wrong because there is likely to be an even better theory of quantum gravity. Hence, they say, the subjective view of probability is wrong, since no scientist really believes in the truth of the model she is checking. I agree, but I think this is a caricature of the subjective view of probability: it misconstrues the meaning of ‘subjectivity’. If I had to use probabilities only to reflect what I truly believe, I wouldn’t be able to do science, since the only thing that I am sure about my belief system is that it is incoherent:
Do I contradict myself?
Very well then I contradict myself,
(I am large, I contain multitudes.)
— Walt Whitman, Song of Myself
Subjective probability, at least the way it is actually used by practicing scientists, is a sort of “as-if” subjectivity — how would an agent reason if her beliefs were reflected in a certain set of probability distributions? This is why when I discuss probability I try to make the pedantic point that all probabilities are conditional, at least on some background prior information or context. So we shouldn’t really ever write a probability that statement “A” is true as P(A), but rather as P(A|I) for some background information, “I”. If I change the background information to “J”, it shouldn’t surprise me that P(A|I)≠P(A|J). The whole point of doing science is to reason from assumptions and data; it is perfectly plausible for an actual scientist to restrict the context to a choice between two alternatives that she knows to be false. This view of probability owes a lot to Ed Jaynes (as also elucidated by Keith van Horn and others) and would probably be held by most working scientists if you made them elucidate their views in a consistent way.
Still, these philosophical points do not take away from Gelman’s more practical ones, which to me seem distinct from those loftier questions and from each other: first, that the formalism of model comparison is often too sensitive to prior information; second, that we should be able to do some sort of alternative-free model checking in order to falsify a model even if we don’t have any well-motivated substitute. Indeed, I suspect that most scientists, even hardcore Bayesians, work this way even if they (we) don’t always admit it.
Like my friend and colleague Peter Coles, I am just returned from the fine wine-soaked dinner for the workshop “Cosmology and Astroparticle physics from the LHC to PLANCK” held at the Niels Bohr Institute in Copenhagen. It is an honor to visit the place where so many discoveries of 20th Century physics were made, and an even greater honor to be able to speak in the same auditorium as many of the best physicists of the last hundred years. (You can see the conference photo here; apparently, the trumpet is just the latest in a long series.)
I talked about the most recent results from the Planck Satellite, gave an overview of the state of the art of (pre-Planck) measurements of the Cosmic Microwave Background, and found myself in what feels like the unlikely role as a mouthpiece for a large (and therefore conservative) community, basically putting forward the standard model of cosmology: a hot big bang with dark matter, dark energy, and inflation — a model that requires not one, not two, but (at least) three separate additions to particles and fields that we know from terrestrial observations: one to make up the bulk of the mass of today’s Universe, another to make the Universe accelerate in its expansion today, and another to make it accelerate at early times. It would sound absurd if it weren’t so well supported by observations.
What are blogs for, if not self-publicity? In that vein, I’ll be appearing at the Spacetacular! night on April 12, in honor of Yuri’s night: the 50th anniversary of Yuri Gagarin’s first-ever manned space flight.
The evening is organized by Londonist editor Matt Brown along with comedian and presenter Helen Keen, hosting a line-up of comedians and scientists. I promise not to be funny so you can tell which I am — I’ll be talking for ten minutes or so about my adventures in space (well, working on a big space-based project, the Planck Surveyor Satellite).
We scientists often, and correctly, make the point that manned space flight has almost nothing to do with science. But I certainly wouldn’t be the scientist I am if it weren’t a morning long ago in Hooks Lane Nursery School watching one of those early moon launches, thinking I wanted to have something, anything, to do with that. So let us know if you want to come celebrate [the Facebook event link is currently broken, but this one is still up.] this amazing human achievement with comedy and science (and spacey costumes) at the Camden Head Pub in London next week.
Many of my colleagues in the EBEX experiment have just lit out for the west. Specifically, the team is heading off to Palestine (pronounced “Palesteen”), Texas, to get the telescope and instrument ready for its big Antarctic long-duration balloon flight at the end of the year, when we hope to gather our first real scientific data and observe the temperature and polarization of the cosmic microwave background (CMB) radiation. Unlike the Planck Satellite, which has a few dozen detectors changed little from those that flew on MAXIMA and BOOMEReNG in the 1990s, EBEX can use more modern technology, and will fly with thousands of detectors, allowing us to achieve far greater sensitivity to the smallest variations in the CMB.
Asad, one of the EBEX postdocs, involved in the experiment for several years, will be writing on the EBEX in Flight blog about the experiences down in Texas and, we hope, the future path of the team and telescope down to Antarctica. Follow along as the team drives across the country (at least twice), assembles and tests the instrument, breaks and fixes things, sleeps too little, works too hard, and, we hope, builds the most sensitive CMB experiment yet deployed. (And of course, eats cheeseburgers.)
And if you want a change from cosmology, you can instead follow along with another friend, Marc, who is trying to see if he can come to grips with writing on an iPad in the supposedly post-PC world, over at typelesswriter.
One of the perks (perqs?) of academia is that occasionally I get an excuse to escape the damp grey of London Winters. The Planck Satellite is an international collaboration and, although largely backed by the European Space Agency, it has a large contribution from US scientists, who built the CMB detectors for Planck’s HFI instrument, as well as being significantly involved in the analysis of Planck data. Much of this work is centred at NASA’s famous Jet Propulsion Lab in Pasadena, and I was happy to rearrange my schedule to allow a February trip to sunny Southern California (I hope my undergraduate students enjoyed the two guest lectures during my absence).
Visiting California, I was compelled to take advantage of the local culture, which mostly seemed to involve meals. I ate as much Mexican food as I could manage, from fantastic $1.25 tacos from the El Taquito Mexicano Truck to somewhat higher-end fare at Tinga in LA proper. And I finally got to taste bánh mì, French-influenced Vietnamese sandwiches (which have arrived in London but I somehow haven’t tried them here yet). And I got to take in the view from the heights of Griffith Park:
as well as down at street level:
And even better, I got to share these meals and views with old and new friends.
Of course I was mainly in LA to do science, but even at JPL we managed to escape our windowless meeting room and check out the clean-room where NASA is assembling the Mars Science Lab:
The white pod-like structure is the spacecraft itself, which will parachute into Mars’ atmosphere in a few years, and from it will descend the circular “sky crane” currently parked behind it which will itself deploy the car-sized Curiosity Rover to do the real work of Martian geology, chemistry, climatology and (who knows?) biology.
But my own work was for the semi-annual meeting of the Planck CTP working group (I’ve never been sure if it was intentional, but the name always seemed to me a sort of science pun, obliquely referring to the famous “CPT” symmetry of fundamental physics). In Planck, “CTP” refers to Cℓ from Temperature and Polarization: the calculation of the famous CMB power spectrum which contains much of the cosmological information in the maps that Planck will produce. The spectrum allows us to compress the millions of pixels in a map of the CMB sky, such as this one from the WMAP experiment (the colors give the temperature or intensity of the radiation, the lines its polarization), into just a few thousand numbers we can plot on a graph.
OK, this is not a publishable figure. Instead, it marks the tenth anniversary of the first CTP working group telecon in February 2001 (somewhat before I was involved in the group, actually). But given that we won’t be publishing Planck cosmology data for another couple of years, sugary spectra will have to do instead of the real ones in the meantime.
The work of the CTP group is exactly concerned with finding the best algorithms for translating CMB maps into these power spectra. They must take into account the complicated noise in the map, coming from our imperfect instruments which observe the sky with finite resolution — that is, a telescope which smooths the sky at a scale from about half down to one-tenth of a degree — and with a limited sensitivity — every measurement has a little bit of unavoidable noise added to it. Moreover, in between the CMB, produced 400,000 years after the Big Bang, and Planck’s instruments, observing today, is the entire rest of the Universe, which contains matter that both absorbs and emits (glows) in the microwaves which Planck observes. So in practice we need to simultaneously deal with all of these effects when reducing our maps down to power spectra. This is a surprisingly difficult problem: the naive, brute-force (Bayesian), solution requires a number of computer operations which scales like the cube of the number of pixels in the CMB map; at Planck’s resolution this is as many as 100 million pixels, and there still are no supercomputers capable of doing the septillion (1024) operations in a reasonable time. If we smooth the map, we can still solve the full problem, but on small scales, we need to come up with useful approximations which take advantage of what we know about the data, usually taking advantage of the very large number of points that contribute, and the so-called asymptotic theorems which say, roughly, that we can learn about the right answer by doing lots of simulations, which are much less computationally expensive.
At the required levels of both accuracy and precision, the results depend on all of the details of the data processing and the algorithm: How do you account for the telescope’s optics and the pixelization of the sky? How do you model the noise in the map? How do you remove those pixels contaminated by astrophysical emission or absorption? All of this is compounded by the necessary (friendly) scientific competition: it is the responsibility of the CTP group to make recommendations for how Planck will actually produce its power spectra for the community and, naturally, each of us wants our own algorithm or computer program to be used — to win. So these meetings are as much about politics as science, but we can hope that the outcome is that all the codes are raised to an appropriate level and we can make the decisions on non-scientific grounds (ease of use, flexibility, speed, etc.) that will produce the high-quality scientific results for which we designed and built Planck — and have worked on it for the last decade or more.
I’ve been meaning to give a shout-out to my colleagues on the ADAMIS team at the APC (AstroParticule et Cosmologie) Lab at the Université Paris 7 for a while: in addition to doing lots of great work on Planck, EBEX, PolarBear and other important CMB and cosmology experiments, they’ve also been running a group blog since the Autumn, Paper(s) of the Week et les autres choses (scientifique) which dissects some of the more interesting work to come out of the cosmology community. In particular, one of my favorite collaborators has written an extremely astute analysis of what, exactly, we on the Planck team released in our lengthy series of papers last month (which I have already discussed in a somewhat more boosterish fashion).
I’ve recently “upgraded” my software which seems to be playing havoc with the format of the blog. The blog is visible, and in many ways nicer than before, but I’ve lost all of my lovely formatting… I hope we’ll be back to normal soon.
In any event, you can probably ignore this and read my post about Planck’s new results instead!
The Satellite now known as the Planck Surveyor was first conceived in the mid-1990s, in the wake of the results from NASA’s COBE Satellite, the first to detect primordial anisotropies in the Cosmic Microwave Background (CMB), light from about 400,000 years after the big bang. (I am a relative latecomer to the project, having only joined in about 2000.)
After all this time, we on the team are very excited to produce our very first scientific results. These take the form of a catalog of sources detected by Planck, along with 25 papers discussing the catalog as well as the more diffuse pattern of radiation on the sky.
Planck is the very first instrument to observe the whole sky with light in nine bands with wavelengths from about 1/3 of a millimeter up to one centimeter, an unprecedented range. In fact this first release of data and papers discusses Planck as a tool for astrophysics — as a telescope observing distant galaxies and clusters of galaxies as well as our own Galaxy, the Milky Way. All of these glow in Planck’s bands (indeed they dominate over the CMB in most of them), and with our high-sensitivity all-sky maps we have the opportunity to do astronomy with Planck, the best microwave telescope ever made. Indeed, to get to this point, we actually have to separate out the CMB from the other sources of emission and, somewhat perversely, actively remove that from the data we are presenting.
Over the last year, then, we on the Planck team have written about 25 papers to support this science; a few of them are about the mission as a whole, the instruments on board Planck, and the data processing pipelines that we have written to produce our data. Then there are a few papers discussing the data we are making available, the Early Release Compact Source Catalog and the various subsets discussing separately objects within our own Milky Way Galaxy as well as more distant galaxies and clusters of galaxies. The remaining papers give our first attempts at analyzing the data and extracting the best science possible.
Most of the highlights in the current papers provide confirmation of things that astronomers have suspected, thanks to Planck’s high sensitivity and wide coverage. It has long been surmised that most stars in the Universe are formed in locations shrouded by dust, and hence not visible to optical telescopes. Rather, the birth of stars heats the dust to temperatures much lower than that of stars, but much higher than the cold dust far from star-forming regions. This warm dust radiates in Planck’s bands, seen at lower and lower frequencies for more and more distant galaxies (due to the redshift of light from these faraway objects). For the first time, Planck has observed this Cosmic Infrared Background (CIB) at frequencies that may correspond to galaxies forming when the Universe was less than 15% of its current age, less than 2 billion years after the big bang. Here is a picture of the CIB at various places around the sky, specifically chosen to be as free as possible of other sources of emission:
Another exciting result has to do with the properties of that dust in our own Milky Way Galaxies. This so-called cosmic dust is known to be made of very tiny grains, from small agglomerations of a few molecules up to those a few tens of micrometers across. Ever since the mid-1990s, there has been some evidence that this dust emits radiation at millimeter wavelengths that the simplest models could not account for. One idea, actually first proposed in the 1950s, is that some of the dust grains are oblong, and receive enough of a kick from their environment that they spin at very high rates, emitting radiation at a frequency related to that rotation. Planck’s observations seem to confirm this prediction quantitatively, seeing its effects in our galaxy. This image of the Rho Ophiuchus molecular cloud shows that the spinning dust emission at 30 GHz traces the same structures as the thermal emission at 857 GHz:
In addition, Planck has found more than twenty new clusters of galaxies, has mapped the dust in gas in the Milky Way in three dimensions, and uncovered cold gas in nearby galaxies. And this is just the beginning of what Planck is capable of. We have not yet begun to discuss the cosmological implications, nor Planck’s abilities to measure not just the intensity of light, but also its polarization.
Of course the most important thing we have learned so far is how hard it is to work in a team of 400 or so scientists, whom — myself included — like neither managing nor being managed (and are likewise not particularly skilled at either). I’ve been involved in a small way in the editing process, shepherding just a few of those 25 papers to completion, paying attention to the language and presentation as much as the science. Given the difficulties, I am relatively happy with the results — the papers can be downloaded directly from ESA, and will be available on the ArXiV on 12 January 2011, and will eventually be published in the journal Astronomy and Astrophysics. It will be very interesting to see how we manage this in two years when we may have as many as a hundred or so papers at once. Stay tuned.
As a scientist, I am used to my work being read by my peers, and I’ve made it into the occasional magazine or newspaper article, and even the odd TV and radio slot. But last week I travelled to Venice’s Architecture Biennale for the culmination of the first phase of the Architectural Association’s Beyond Entropy art/science project (which I’ve described before). I took a vaporetto to the island of San Giorgio, and next to one of Venice’s more spectacular Palladian churches, I saw the Beyond Entropy banner hanging over the entrance:
(I took these pictures, but there are many much more professional ones taken by the AA’s Valerie Bennett.)
Before arriving, I didn’t know what to expect from the project: small-scale, low-key, amateurish? In this setting, it was clearly big and serious. And inside this lovely building were these, the prototypes for our time machine:
Last year I traveled to South America to witness the launch of our several-hundred million-Euro Planck satellite, surely a big and serious project. But the sight of my own work — our texts, flywheels and gyroscopes — sitting on a plywood plinth, plausibly described as something at least related to the very different creative process of art, was nearly as disconcerting (despite the lack of highly explosive rocket fuel).
I’ll leave any assessment of the overall quality to others, although it became obvious that these pieces really are prototypes for what could become more finished works, but we have a long way to go. Nonetheless, let me explicitly thank my collaborators, Shin Egashira (whom I will also congratulate on his wedding which gave him an excellent reason to not show up in Venice) and Scrap Marshall, a student at the Architectural Association who joined us toward the end of the project and did an enormous amount of practical and creative work getting our pieces together. From speaking to members of some of the other groups, we were lucky to all be based in London, and to eventually come to see our project in similar ways, albeit from different directions; some of the more widely-dispersed groups had to deal with significantly greater practical problems, and the interpersonal ones those ended up causing.
That first day was dedicated to the AA’s visiting school, and the next day was the centrepiece: a marathon symposium of more than thirty talks, dedicated to the themes of “entropy” and “energy”. Remarkably, none of our projects addressed the ecological, societal and political aspects of these topics, while many of the speakers attacked them directly, from Richard Burdett and Reinier de Graaf’s complementary discussions of the bleak picture for energy and climate if we keep to “business as usual” in our habits of consumption and production, to Italian Green Party politician Grazia Francescato’s hopeful discussion of “Green Jobs and Green Economy”. There were a few talks on science per se, from Angelo Merlina’s brief introduction to the LHC at CERN (of which a third talked about cosmology, and a third was pre-recorded), to one of my favourites, biophysicist Tania Saxl’s description of the amazing mechanism behind the motion of rotating bacterial flagella. There was also an inexplicable prerecorded description of “parallel worlds” in film from de Gruyter and Thys, a performance from the Arazzi Laptop ensemble, and contributions from Serpentine Gallery curator Hans-Ulrik Obrist (which was interesting but mostly about himself) and Charles Jencks. Jencks tackled the overlap between science, art and architecture head-on, each as a different metaphorical system for describing and interacting with the world. This culminates in his Scottish Garden of Cosmic Speculation, a hugely symbolic landscape replete with double helixes and grassy knolls in the form of black hole spacetime diagrams (I admit I’ve also found these supposed metaphors a bit too, well, literal for my taste — with insufficient information to be effective teaching tools, but too didactic to be truly beautiful.) I think the most important thing I learned was that, in their own way, the architects are just as nerdy as us scientists, but just better looking dressed.
Also, there was plenty of fine food and free-flowing sparkling wine (which meant that I probably missed about half of the presentations).
Finally, I would like to thank everyone from the AA who made the project happen (and will continue to do so, if further funding is forthcoming): Artemis Doupa, Sylvie Taher, Esther McLaughlin, Aram Mooradian and most especially the ever-enthusiastic project director, Stefano Rabolli Pansera. Thanks also to the AA visiting students, and all of the other participants, especially Ariel Schlesinger and Wilfredo Prieto for giving me a glimpse of the Architecture Biennale through artists’ eyes.
I spent part of this week in Paris (apparently at the same time as a large number of other London-based scientists who were here for other things) discussing whether the European CMB community should rally and respond to ESA’s latest call for proposals for a mission to be launched in the next open slot—which isn’t until around 2022.
As successful as Planck seems to be, and as fun as it is working with the data, I suspect that no one on the Planck team thinks that a 400-scientist, dispersed, international team coming from a dozen countries each with its own politics and funding priorities, is the most efficient way to run such a project. But we’re stuck with it—no single European country can afford the better part of a billion Euros it will cost. Particle physics has been in this mode for the better part of fifty years, and arguably since the Manhattan Project, but it’s a new way of doing things — involving new career structures, new ways of evaluating research, new ways of planning, and a new concentration upon management — that we astrophysicists have to develop to answer our particular kinds of scientific questions.
But a longer discussion of “big science” is for another time. The next CMB satellite will probably be big, but the coming ESA call is officially for an “M-class” (for “medium”) mission, with a meagre (sic) 600 million euro cap. What will the astrophysical and cosmological community get for all this cash? How will it improve upon Planck?
Well, Planck has been designed to mine the cosmic microwave background for all of the temperature information available, the brightness of the microwave sky in all directions, down to around a few arcminutes at which scale it becomes smooth. But light from the CMB also carries information about the polarisation of light, essentially two more numbers we can measure at every point. Planck will measure some of this polarisation data, but we know that there will be much more to learn. We expect that this as-yet unmeasured polarisation can answer questions about fundamental physics that affects the early universe and describes its content and evolution. What are the details of the early period of inflation that gave the observable Universe its large-scale properties and seeded the formation of structures in it—and did it happen at all? What are the properties of the ubiquitous and light neutrino particles whose presence would have had a small but crucial effect on the evolution of structure?
The importance of these questions is driving us toward a fairly ambitious proposal for the next CMB mission. It will have a resolution comparable to that of Planck, but with many hundreds of individual detectors, compared to Plank’s many dozens—giving us over an order of magnitude increase in sensitivity to polarisation on the sky. Actually, even getting to this point took a good day or two of discussion. Should we instead make a cheaper, more focused proposal that would concentrate only on the question oaf inflation and in particular upon the background of gravitational radiation — observable as so-called “B-modes” in polarisation — that some theories predict? The problem with this proposal is that it is possible, or even likely, that it will produce what is known as a “null result”—that is, it won’t see anything at all. Moreover, a current generation of ground- and balloon-based CMB experiments, including EBEX and Polarbear, which I am lucky enough to be part of, are in progress, and should have results within the next few years, possibly scooping any too-narrowly designed future satellite.
So we will be broadening our case beyond these B-modes, and therefore making our design more ambitious, in order to make these further fundamental measurements. And, like Planck, we will be opening a new window on the sky for astrophysicists of all stripes, giving measurements of magnetic fields, the shapes of dust grains, and likely many more things we haven’t yet though of.
One minor upshot of all this is that our original name, the rather dull “B-Pol”, is no longer appropriate. Any ideas?
(Warning, scattershot blogging echo-chamber post follows.)
Partially because the event was mostly attended by science bloggers themselves, there was a bit of a preaching-to-the-converted sense to the proceedings. (I tried to engage in some good-natured tweaking, pointing out that probably the greatest influence of [supposedly] science blogging has been in absurdly dragged-out climategate saga, but I couldn’t get a rise out of the audience.) But it was heartening to see just how mainstream science blogging has become.
“Only” five years ago (scare-quotes denoting an eternity of internet-time), the academic-blogosphere chattered on about an anonymous article in the Chronicle of Higher Education which contended that bloggers were essentially unsuitable to be hired as faculty members, and a couple of years after that several of my colleagues felt the need to seriously restrict their blogging while searching for permanent positions. I was heartened to see that the question of whether blogging could actually hurt someone’s career seems to be less worrying. Although Petra Boynton said that one of her previous departments were less than enthusiastic about it, most of the panelists have found that, with an increased in impact and communication in general, blogging has taken its position as an effective way to engage with the public.
One of the more novel (to me) things going on at this meeting was the Twitter backchannel: the organizers projected a running stream of tweets marked with the #talkfest tag. It was a decent mix of jokes and apposite comments, especially including erstwhile MP Dr Evan Harris’ provocative comments about whether scientists should be forced to do public engagement at all. It’s certainly good that blogging and communication don’t hurt your career — but should they be requirements for scientific advancement? Not all scientists’ talents lie in that direction, and we shouldn’t expect them to. There was also a twitter discussion of the gender makeup of the panel, which was dishearteningly 1/6 female despite an audience of at least 50% women.
When science blogging started out as its own sub-genre in the middle of the decade, no one was quite sure what it would be for. Would it be used within science as an online lab notebook, or as a substitute or adjunct to papers? That doesn’t seem to have panned out — even in the post-’net open world, the structure of science encourages secrecy, at least until the work can be packaged into what are still more or less old-fashioned papers in what are still more or less old-fashioned journals (albeit with the important twist of pre-publication posting on the arXiv in many fields). Within collaborations, however, wikis, rather than blogs, have become ubiquitous as an easy way to communicate amongst scientists who are already expert — the easy ability to add small chunks of information is exactly what is needed. (Within the Planck Satellite collaboration, we actually use a wiki as a sort of blog — we keep a reverse-chronological list of “posts” discussing our latest results.)
Instead, blogs seem to be used almost exclusively as a window into the life, methods and results of scientists, directed at a knowledgeable but lay public. Indeed, it was suggested at the talkfest that someone could make a very useful living textbook from the scattered blog posts on a given subject. I’m not so sure — one of the advantages of a proper textbook is a single voice and, more prosaically, a single notation starting from scratch— but it’s probably worth trying if someone’s got the wherewithal to do the bit-work involved.
It was especially nice to meet several of my fellow Imperial College bloggers, including biophysicist Stephen Curry (whose own post on the Talkfest also has a list of other reactions to it), whom I was somewhat embarrassed to discover actually works in the same building as I do. As always at these sorts of events, much of the amusement was during the inevitable pub visit afterwards and especially the pre-panel milling about — thanks to the organizers for the excellent combination of cupcakes and beer.
To celebrate, the Planck team have released an image of the full sky. The telescope has detectors which can see the sky with 9 bands at wavelengths ranging from 0.3 millimeters up to nearly a centimeter, out of which we have made this false-color image. The center of the picture is toward the center of the Galaxy, with the rest of the sphere unwrapped into an ellipse so that we can put it onto a computer screen (so the left and right edges are really both the same points).
At the longest and shortest wavelengths, our view is dominated by matter in our own Milky Way galaxy — this is the purple-blue cloud, mostly so-called galactic “cirrus” gas and dust, largely concentrated in a thin band running through the center which is the disk of our galaxy viewed from within.
In addition to this so-called diffuse emission, we can also see individual, bright blue-white objects. Some of these are within our galaxy, but many are themselves whole distant galaxies viewed from many thousands or millions of light years distance. Here’s a version of the picture with some objects highlighted:
Even though Planck is largely a cosmology mission, we expect these galactic and extragalactic data to be invaluable to astrophysicists of all stripes. Buried in these pictures we hope to find information on the structure and formation of galaxies, on the evolution of very faint magnetic fields, and on the evolution of the most massive objects in the Universe, clusters of galaxies.
But there is plenty of cosmology to be done: we see the Cosmic Microwave Background (CMB) in the red and yellow splotches at the top and bottom — out of the galactic plane. We on the Planck team will be spending much of the next two years separating the galactic and extragalactic “foreground” emission from the CMB, and characterizing its properties in as much detail as we can. Stay tuned.
I admit that I was somewhat taken aback by the level of interest in these pictures: we haven’t released any data to the community, or written any papers. Indeed, we’ve really said nothing at all about science. Yet we’ve made it onto the front page of the Independent and even the Financial Times, and yours truly was quoted on the BBC’s website. I hope this is just a precursor to the excitement we’ll generate when we can actually talk about science, first early next year when we release a catalog of sources on the sky for the community to observe with other telescopes, and then in a couple of years time when we will finally drop the real CMB cosmology results.
Results from the first major science papers from the Herschel Satellite were released this week at a conference in Holland. Launched almost a year ago on the same rocket as Planck, Herschel is an infrared and sub-millimeter telescope, which lets it see not only the stars that generate the visible light we see with our eyes and ordinary cameras, but also the gas and dust that absorb and re-radiate that light. That gas and dust carries information about both the birth and death of stars: the detritus of exploding stars pollutes the interstellar medium, which eventually condenses out to form new generations of stars. On larger scales, Herschel’s observations let us trace the evolution of entire galaxies, the most important tracers of large-scale structure, formed from seeds laid down somehow in the first instants of the Universe (and, bringing it all back to cosmology, which are viewed by Planck in a much earlier form).
My Imperial colleagues and Herschel scientists Dave Clements and Brian O’Halloran discuss the results in much more detail over on the Herschel mission blog, or you can keep more up to date on twitter. But I’ll just steal some of their bandwidth and show some pretty pictures.
Most of the dots in this picture are one of those distant galaxies, lit up in the infrared due to its once and future stars:
Image courtesy ESA/ATLAS Consortium
Closer to home, this is selection of star-forming regions, turbulent filaments of gas and dust:
Image courtesy ESA/Hi-GAL Consortium
Not coincidentally, Imperial’s Michael Rowan-Robinson, who has been doing infrared astronomy for several decades, appeared on BBC radio 4’s wonderful In Our Time this morning to discuss “The Cool Universe”: covering a century or so of infrared astronomy in forty-five minutes.
We on Planck won’t be coming out with any papers for quite a while. However, many members of the team gathered in Orsay, outside of Paris, this week, to discuss the progress of the observations (and our analyses) and, crucially, to start talking in more detail about the actual papers that we’ll be writing over the next few years. More generally, Planck is doing pretty well. It came out first in NASA’s latest round of evaluations (which is a significant achievement for a mission primarily run by ESA), and which we hope will also give further impetus to keep funds flowing in the UK. This is especially important as the length of the Planck mission is likely to be almost doubled, allowing us to extract even more science than we originally hoped.
I can’t say much more, except that we’ve got a lot of — very exciting — work ahead of us.
I’ve been in Geneva now for a couple of days. We spent yesterday visiting CERN, trying to inspire the artists, architects and scientists alike (I’ve collaborated with people here, but I’ve never visited before).
A mockup of a section of the CERN tunnels. More pictures here.
You can also check out Peter Coles’ blog for his
The second night, after our visit to CERN and a dinner of fondue and swiss music (possibly not the high point of the trip), all of the 24 participants (eight groups each of an architect, artist and a scientist) gave a few-minute presentation on their work and interests. I was, to use the cliché, blown away by the ambition and accomplishment of everyone else involved. In particular, I am lucky enough to be working with Budapest-based artist Attila Csorgo and architect Shin Egashira, who works out of the Architecture Association, the overall initiators and sponsors of the project. Both build amazing machines. Attila’s constructions seem to me to be about the interaction of the machine and the environment, or of the components of the machine itself, whereas Shin’s involve more effort on the part of the viewer/participant (but I am sure I will get to understand their work and their practice better as I spend more time with it and them).
We spent the next day in a lovely old Swiss building, brainstorming our projects — we’re meant to come up with a “prototype” to have in place for this summer’s Architecture Biennale in Venice. Our brief was to explore the concept of “Mechanical Energy”, and we found an area of convergence in the idea of cameras, in the process of taking pictures, areas that both Shin and Attila have explored in their work.
Right now, our first idea is to combine the Planck Surveyor’s method of scanning the sky with a microphone-based sensor and camera, to make sound and light pictures of the volume surrounding the apparatus. We’re looking forward to a weekend retreat into the wilds of Dorset, to Hooke Park, a site run by the AA.
Thanks, finally, to Stefano Rabolli Pansera, the brilliant, optimistic, and enthusiastic mind behind this project, as well as all of the other people from the Architecture Association doing the hard work.
The cosmology community has had a terrible few months.
I am saddened to report the passing of Andrew Lange, a physicist from CalTech and one of the world’s preeminent experimental cosmologists. Among many other accomplishments, Andrew was one of the leaders of the Boomerang experiment, which made the first large-scale map of the Cosmic Microwave Background radiation with a resolution of less than one degree, sufficient to see the opposing action of gravity and pressure in the gas of the early Universe, and to use that to measure the overall density of matter, among many other cosmological properties. He has since been an important leader in a number of other experiments, notably the Planck Surveyor satellite and the Spider balloon-borne telescope, currently being developed to become one of the most sensitive CMB experiments ever built.
I learned about this tragedy on the same day that people are gathering in Berkeley, California, to mourn the passing of another experimental cosmologist, Huan Tran of Berkeley. Huan was an excellent young scientist, most recently deeply involved in the development of PolarBear, another one of the current generation of ultra-sensitive CMB experiments. Huan lead the development of the PolarBear telescope itself, currently being tested in the mountains of California, but to be deployed for real science on the Atacama plane in Chile. We on the PolarBear team are proud to name the PolarBear telescope after Huan Tran, a token of our esteem for him, and a small tribute to his memory.
My thoughts go out to the friends and family of both Huan and Andrew. I, and many others, will miss them both.
Luckily, not all the astrophysics news this week was so bad.
First, and most important, two of our Imperial College Astrophysics postgraduate students, Stuart Sale and Paniez Paykari, passed their PhD viva exams, and so are on their ways to officially being Doctors of Philosophy. Congratulations to both, especially (if I may say so) to Dr Paykari, who I had the pleasure and fortune to supervise and collaborate with. Both are on their way to continue their careers as postdocs in far-flung lands.
Second, the first major results from the Herschel Space Telescope, Planck’s sister satellite, were released. There are impressive pictures dwarf planets in the outer regions of our solar system, of star-forming regions in the Milky Way galaxy, of the vary massive Virgo Cluster of galaxies, and of the so-called “GOODS” (Great Observatory Origins Deep Survey) field, one of the most well-studied areas of sky. All of these open new windows into these areas of astrophysics, with Herschel’s amazing sensitivity.
Finally, tantalisingly, the Cryogenic Dark Matter Search (CDMS) released the results of its latest (and final) effort to search for the Dark Matter that seems to make up most of the matter in the Universe, but doesn’t seem to be the same stuff as the normal atoms that we’re made of. Under some theories, the dark matter would interact weakly with normal matter, and in such a way that it could possibly be distinguished from all the possible sources of background. These experiments are therefore done deep underground — to shield from cosmic rays which stream through us all the time — and with the cleanest and purest possible materials — to avoid contamination with both both naturally-occurring radioactivity and the man-made kind which has plagued us since the late 1940s.
With all of these precautions, CDMS expected to see a background rate of about 0.8 events during the time they were observing. And they saw (wait for it) two events! This is on the one hand more than a factor of two greater than the expected number, but on the other is only one extra count. To put this in perspective, I’ve made a couple of graphs where I try to approximate their results (for aficionados, these are just simple plots of the Poisson distribution). The first shows the expected number of counts from the background alone:
(I should point out a few caveats in my micro-analysis of their data. First, I don’t take into account the uncertainty in their background rate, which they say is really 0.8±0.1±0.2, where the first uncertainty, ±0.1 is “statistical”, because they only had a limited number of background measurements, and the second, ±0.2, is “systematic”, due to the way they collect and analyse their data. Eventually, one could take this into account via Bayesian marginalization, although ideally we’d need some more information about their experimental setup. Second, I’ve only plotted the likelihood above, but true Bayesians will want to apply a prior probability and plot the posterior distribution. The most sensible choice (the so-called Jeffreys prior) for this case would in fact make the probability peak at zero signal. Finally, one would really like to formally compare the no-signal model with a signal-greater-than-zero model, and the best way to do this would be using the tool of Bayesian model comparison.)
Nonetheless, in their paper they go on to interpret these results in the context of particle physics, which can eventually be used to put limits on the parameters of supersymmetric theories which may be tested further at the LHC accelerator over the next couple of years.
I should bring this back to the aforementioned bad news. The UK has its own dark matter direct detection experiments as well. In particular, Imperial leads the ZEPLIN-III experiment which has, at times, had the world’s best limits on dark matter, and is poised to possibly confirm this possible detection — this will be funded for the next couple of years. Unfortunately, STFC has decided that the next generation of dark matter experiments, EURECA and LUX-ZEPLIN, needed to make convincing statements about these results, weren’t possible to fund.
I presume that anyone reading this blog knows that today is the day when the great unwashed masses of UK Astronomers heard about our financial fate from the STFC, the small arm of the UK government responsible for Astrophysics, Particle Physics and Nuclear Physics.
For various reasons, some clear and others manifestly not, STFC is something like £70 million in the red. When all this started about two years ago, one of the main criticisms of the STFC management (beyond wondering how they could have got themselves — and us — into this predicament to begin with) was that they started to impose solutions that seemed to bear little resemblance to what the scientists themselves wanted. Trying to either genuinely ameliorate this, or at least give themselves good cover, they’ve spent much of the last year gathering input from various groups of physicists and astronomers, through a series of reports produced by scientist-led panels. These panels released their results this autumn, and STFC has supposedly used them to make decisions about the next five or so years of funding.
I was selfishly relieved to see that our work with the Planck Surveyor Satellite is rated “alpha 5”, and that our other local grants don’t appear directly affected (i.e., we weren’t drastically cut). However, STFC has “requested” (not sure what that means in this context) that even these projects reduce their costs by 15%. Other programs were not even this lucky — a not-quite-complete list of the cuts is on the STFC site. The cuts (a.k.a. “managed withdrawal”) include the UKIRT telescope, the LOFAR array, future work at the low-background facility at the Boulby mine, and future science exploitation of the XMM and Cassini missions (among many others). Alongside this, there will be a 25% cut in studentships and fellowships, although the details of this have not been revealed.
In his independent response, the Science Minister, Lord Drayson, says “we are investing record amounts into scientific research, but it is absolutely right that it is the scientists themselves, through the Research Councils, that decide how best to spend this money.” Of course we scientists don’t necessarily feel that our voices have been heard. The prioritized list of projects is available from STFC, and although it generally correlates with both the inputs from the various sub-panels and the financial outcome (in particular, many of us were pleased and relieved to see the much-criticised MoonLITE project at the bottom of the heap), there are some striking differences from at least my understanding of the panel recommendations, such as the “alpha 4” grade given to the Aurora human spaceflight program.
However, Drayson does seem to understand some of the issues: “…there are real tensions in having international science projects, large scientific facilities and UK grant giving roles within a single Research Council. It leads to grants being squeezed by increases in costs of the large international projects which are not solely within their control. I will work urgently with Professor Sterling, the STFC and the wider research community to find a better solution by the end of February 2010.” Not sure what this means, but even if we are grasping at straws, it’s the only promising news of the day.
I’ve got 11 browser tabs open just to get myself up-to-date. Here are some of them:
- An excellent summary of the situation before the announcement is at To the Left of Centre.
- Paul Crowther’s page has become the canonical clearing-house for information on the astronomy side of the “STFC Funding Crisis”.
- Blogs from Andy Lawrence and Peter Coles are both well-wrought and likely feature commentary from the opinionated luminaries of UK astronomy.
- The Institute of Physics and the Royal Astronomical Society respond.
- The BBC talks about the cuts with Lord Drayson.
- Physics World summarizes the situation as “savage cuts”.
- Our US counterparts get in on the commentary at Science magazine.
- There’s a web campaign to help save astronomy in the UK.
This week I was in the truly wonderful city of Bologna, home of possibly the oldest university in Europe. Nowadays, Bologna is also the home of IASF-BO, the Italian Istituto di Astrofisica Spaziale e Fisica Cosmica, and was hosting this year’s Planck Satellite Consortium meeting.
Of course I can’t talk about anything that was actually presented at the meeting — as I’ve mentioned before, there are strong restrictions on what is allowed to be discussed before the data become public in about three years. Indeed, that communication policy was itself the topic of considerable discussion — it turns out that at least a couple of Planck’s “highest ranking” scientists had recently been deemed to be in “non-compliance” with the policy (which may be different from actually violating the policy, but no one is quite sure…).
Luckily, there was plenty to talk about amongst ourselves between the political discussions. I reported on our efforts in London to recover Planck’s “pointing solution” — that is, to figure out where, exactly, each of Planck’s fifty or so detectors are actually looking on the sky at any given moment. This is obviously crucial to getting good science out of Planck — indeed, even though the instrument smears the sky with a resolution of about four arcminutes (about 1/15 of a degree), we want to know the pointing to roughly 10 arcseconds (about 1/360 of a degree)! But there were several hundred scientists at the meeting, so plenty to discuss, besides, over the course of the week, from Planck’s electronics to the eventual scientific results on the earliest instants of the Universe. The first hints of this science, but not much more, are present in the pictures we showed from Planck’s first-light survey. And I should point out that, despite at least one attempt — which I hesitate to even link to — there is really no science to be had in any analysis of what we’ve presented. We’re not taking three years to analyze the data just to be selfish — at least not entirely. It will take that long before we can understand the instrument well enough to interpret the data that comes out of it.
Luckily, Bologna is also known for its food, and aside from the excellent conference snacks and lunches (and a blow-out dinner at a local Palazzo from which I mostly recall the giant parmigiana wheel and the copious grappa), it was pretty easy to find excellent food at pretty much any local Trattoria (like La Montanara and the strangely-named Serghei). So now I am back, fat, happy, and with plenty of Planck work to do in the next few weeks, months and years.
I’m happy to be able to point to ESA’s first post-launch press release from the Planck Surveyor Satellite.
Here is a picture of the area of sky that Planck has observed during its “First Light Survey”, superposed on an optical image of the Milky Way galaxy:
(Image credit: ESA, LFI and HFI Consortia (Planck); Background image: Axel Mellinger. More pictures are available on the UK Planck Site as well as in French.)
The last few months since the launch have been a lot of fun, getting to play with Planck data ourselves. Here at Imperial, our data-processing remit is fairly narrow: we compute and check how well the satellite is pointing where it is supposed to, and calculate the shape of its beam on the sky (i.e., how blurry its vision is). Nonetheless, just being able to work at all with this incredibly high-quality data is satisfying.
Because of the way Planck scans the sky, in individual rings slowly stepping around the sky over the course of about seven months, with a nominal mission of two full observations of the sky, even the two weeks of “First Light Survey” data is remarkably powerful: we have seen a bit more than 5% of the sky with about half of the sensitivity that Planck is meant to eventually have (in fact, we hope to extend the mission beyond the initial 14 months). This is already comparable to the most powerful sub-orbital (i.e., ground and balloon-based) CMB experiments to date.
But a full scientific analysis will have to wait a while: after the 14 month nominal mission, we will have one year to analyze the data, and another year to get science out of it before we need to release the data alongside, we hope, a whole raft of papers. So stay tuned until roughly Autumn of 2012 for the next big Planck splash.
Central London featured two important events this past weekend. First was the annual Gay Pride Parade, a riotous and joyful procession of rainbow flags, pink clothing, and (mostly) ill-fitting dresses on very large people.
Sadly, the only thing that marred the good-natured, family-friendly event were the stupid protesters. But it was wonderful to see that they were just ignored, or occasionally people would point at their sad and pathetic group and just laugh (there was also a much smaller, and yet more pathetic, group of National Front protesters who deserved and received even less attention).
At the same time, the Royal Society, right down the road from Piccadilly Circus, hosted the annual Summer Science Exhibition, and I visited my colleagues (Stuart Lowe and Michael Bridges, here) talking Planck Surveyor science, taking infrared pictures of the visitors and handing out lots of great Planck swag.
In fact, this weekend, Planck has cooled down to just about its final temperature of 100mK (that is 0.1 degrees above absolute zero!) and has made it to its final orbit at the L2 point. So we are starting to get ready to take real data, after we spend the next month or so kicking the tires and checking her out.