Friday, July 24, 2015

Is the Fermi gamma-ray excess evidence for dark matter or due to milli-second pulsars?

Fermi Satellite, 3d model.
Image Source: NASA

The Large Area Telescope on board the Fermi spacecraft looks out for the most extreme events in the cosmos. Launched 2008, it scans the whole sky for gamma rays in the very high energy range, from 20 MeV to about 300 GeV; a full scan takes about 3 hours. One of Fermi’s most interesting findings is an unexpectedly large amount of gamma-rays, ordinary light but at enormous energy, stemming from the center of our galaxy.

The Fermi gamma-ray excess has proved difficult to explain with standard astrophysical processes. The spectral distribution of the observed gamma-rays over energies bulges at about 2 GeV and it is hard to come up with a mechanism that produces particles at such high energies that prefer this particular spectral feature. The other puzzle is that whatever the sources of the gamma-rays, they seem homogeneously distributed in the galactic center, out to distances of more than ten degrees, which is about 5,000 light years – a huge range to span.

The most exciting proposal to solve the riddle is that the gamma-rays are produced by dark matter annihilation. Annihilation spectra often bulge at energies that depend on both the mass of the particles and their velocity. And the dark matter distribution is known to be denser towards centers of galaxies, so one would indeed expect more emission from there. While dark matter isn’t entirely homogeneous but has substructures, the changes in its density are small, which would result in an overall smooth emission. All of this fits very well with the observations.

For this reason, many particle physicists have taken their dark matter models to see whether they can fit the Fermi data, and it is possible indeed without too much difficulty. If the Fermi gamma-ray excess was due to dark matter annihilation, it would speak for heavy dark matter particles with masses of about 30 to 100 GeV, which might also show up at the LHC, though nothing has been seen so far.

Astrophysicists meanwhile haven’t been lazy and have tried to come up with other explanations for the gamma-ray excess. One of the earliest and still most compelling proposals is a population of millisecond pulsars. Such objects are thought to be created in some binary systems, where two stars orbit around a common center. When a neutron star succeeds in accreting mass from a companion star, it spins up enormously and starts to emit large amounts of particles including gamma rays reaching up to highest energies. This emission goes into one particular direction due to the rapid rotation of the system, and since we only observe it when it points at our telescopes the source seems to turn on and off in regular intervals: A pulsar has been created.

Much remains to be understood about millisecond pulsars, including details of their formation and exact radiation characteristics, but several thousands of them have been observed so far and their properties are observationally well documented. Most of the millisecond pulsars on record are the ones in our galactic neighborhood. From this we know that they tend to occur in globular clusters where particularly many stars are close together. And the observations also tell us the millisecond pulsar spectrum peaks at about 2 GeV, which makes them ideal candidates to explain the Fermi gamma-ray excess. The only problem is that no such pulsars have been seen in the center of the galaxy where the excess seems to originate, at least so far.

There are plausible reasons for this lack of observation. Millisecond pulsars tend to be detected in the radio range, at low energies. Only after astronomers have an idea of where exactly to look, they can aim precisely with telescopes to confirm the pulsation in the higher energy range. But such detections are difficult if not impossible in the center of the galaxy because observations are shrouded by electron gas. So it is quite plausible that the millisecond pulsars are in the center, they just haven’t been seen. Indeed, model-based estimates indicate that millisecond pulsars should also be present in the galactic center, as laid out for example in this recent paper. It would seem odd indeed if they weren’t there. On the other hand it has been argued that if millisecond pulsars were the source of the gamma-ray excess, then Fermi should also have been able to pinpoint a few of the pulsars in the galactic center already, which has not been the case. So now what?

The relevant distinction between the both scenarios to explain the gamma ray excess, dark matter annihilation or millisecond pulsars, is whether the emission comes from point sources or whether the sources are indeed homogeneously distributed. This isn’t an easy question to answer because Fermi basically counts single photons and their distribution is noisy and inevitably sometimes peaks here or there just by coincidence. Making estimates based on such measurements is difficult and requires sophisticated analysis.

In two recent papers now, researchers have taken a closer look at the existing Fermi data to see whether it gives an indication for point-like sources that have so far remained below the detection threshold beyond which they would be identified as stellar objects. For this they have to take the distribution of the measured signals, extract peaks ordered by magnitude, and test this measured distribution against a random distribution.
    Strong support for the millisecond pulsar origin of the Galactic center GeV excess
    Richard Bartels, Suraj Krishnamurthy, Christoph Weniger
    arXiv:1506.05104 [astro-ph.HE]

    Evidence for Unresolved Gamma-Ray Point Sources in the Inner Galaxy
    Samuel K. Lee, Mariangela Lisanti, Benjamin R. Safdi, Tracy R. Slatyer, Wei Xue
    arXiv:1506.05124 [astro-ph.HE]
The difference between these papers is the method they used to identify the point sources. The first paper by Bartels et al uses a wavelet analysis on the data, that is somewhat like a Fourier transform with a local profile, to pick up potential sources with low statistical significance. The Lee et al paper tries to generate a pattern close to the observed one by using various mixtures of noise and point sources of particular spectra. In both papers the researchers find that the data indicates the origin of the gamma-rays is point sources and not entirely smoothly distributed. In the first paper, the authors moreover extract the necessary density of pulsars in the galactic center to explain the observations, and demonstrate that it is possible the pulsars give rise to the observed excess and might so far have stayed just below the detection threshold for point sources.

Taken together, it looks like the evidence has now shifted in favor of millisecond pulsars. As Christopher Weniger from the University of Amsterdam put it “[The pulsars] are there, we know they are there, and they have the right spectrum. We first have to rule out that this isn’t what we see.” Rather than ruling out astrophysical sources as origin of the gamma-ray excess however, the researchers are now well on the way to confirm it’s the pulsars that cause the signal.

Finding definite evidence that the Fermi gamma-ray excess is due to millisecond pulsars is difficult but not impossible. What is needed is more statistics that will allow resolving the point sources better, and more time will bring more statistics. The puzzle isn’t solved yet, but chances are good it will be solved within the next years. What constitutes dark matter however at least for now remains a mystery.

Monday, July 20, 2015

The number-crunchers. How we learned to stop worrying and love to code.

My grandmother was a calculator, and I don’t mean to say I’m the newest from Texas Instruments. I mean my grandmother did calculations for a living, with pencil on paper, using a slide rule and logarithmic tables. She calculated the positions of stars on the night sky, for five minute intervals, day by day, digit by digit.

Today you can download one of a dozen free apps to display the night sky for any position on Earth, any time, any day. Not that you actually need to know stellar constellations to find True North. Using satellite signals, your phones can now tell your position to within a few meters, and so can 2 million hackers in Russia.

My daughters meanwhile are thoroughly confused as to what a phone is, since we use the phone to take photos but make calls on the computer. For my four-year old a “phone” is pretty much anything that beeps, including the microwave, which for all I know by next year might start taking photos of dinner and upload them to facebook. And the landline. Now that you say it. Somebody called in March and left a voicemail.

Jack Myers dubbed us the “gap generation,” the last generation to remember the time before the internet. Myers is a self-described “media ecologist” which makes you think he’d have heard of search engine optimization. Unfortunately, when queried “gap generation” it takes Google 0.31 seconds to helpfully bring up 268,000,000 hits for “generation gap.” But it’s okay. I too recall life without Google, when “viral” meant getting a thermometer stuffed between your lips rather than being on everybody’s lips.

I wrote my first email in 1995 with a shell script called “mail” when the internet was chats and animated gifs. Back then, searching a journal article meant finding a ladder and blowing dust off thick volumes with yellowish pages. There were no keyword tags or trackbacks; I looked for articles by randomly browsing through journals. If I had an integral to calculate, there were Gradshteyn and Ryzhik’s tables, or Abramovitz and Stegun's Handbook of Special Functions, and else, good luck.

Our first computer software for mathematical calculations, one of the early Maple versions, left me skeptical. It had an infamous error in one of the binomial equations that didn’t exactly instill trust. The program was slow and stalled the machine for which everybody hated me because my desk computer was also the institute’s main server (which I didn’t know until I turned it off, but then learned very quickly). I taught myself fortran and perl and java script and later some c++, and complained it wasn’t how I had imagined being a theoretical physicist. I had envisioned myself thinking deep thoughts about the fundamental structure of reality, not chasing after missing parentheses.

It turned out much of my masters thesis came down to calculating a nasty integral that wasn’t tractable numerically, by computer software, because it was divergent. And while I was juggling generalized hypergeometric functions and Hermite polynomials, I became increasingly philosophic about what exactly it meant to “solve an integral.”

We say an integral is solved if we can write it down as a composition of known functions. But this selection of functions, even the polynomials, are arbitrary choices. Why not take the supposedly unsolvable integral, use it to define a function and be done with it? Why are some functions solutions and others aren’t? We prefer particular functions because their behaviors are well understood. But that again is a matter of how much they are used and studied. Isn’t it in the end all a matter of habit and convention?

After two years I managed to renormalize the damned integral and was left with an expression containing incomplete Gamma functions, which are themselves defined by yet other integrals. The best thing I knew to do with this was to derive some asymptotic limits and then plot the full expression. Had there been any way to do this calculation numerically all along, I’d happily have done it, saved two years of time, and gotten the exact same result and insight. Or would I? I doubt the paper would even have gotten published.

Twenty years ago, I like most physicists considered numerical results inferior to analytical, pen-on-paper, derivations. But this attitude has changed, changed so slowly I almost didn’t notice it changing. Today numerical studies are still often considered suspicious, fighting a prejudice of undocumented error. But it has become accepted practice to publish results merely in forms of graphs, figures, and tables, videos even, for (systems of) differential equations that aren’t analytically tractable. Especially in General Relativity, where differential equations tend to be coupled, non-linear, and with coordinate-dependent coefficients – ie as nasty as it gets – analytic solutions are the exception not the norm.

Numerical results are still less convincing, but not so much because of a romantic yearning for deep insights. They are less convincing primarily because we lack shared standards for coding, whereas we all know the standards of analytical calculation. We use the same functions and the same symbols (well, mostly), whereas deciphering somebody else’s code requires as much psychoanalysis as patience. For now. But imagine you could check code with the same ease you check algebraic manipulation. Would you ditch analytical calculations over purely numerical ones, given the broader applicability of the latter? How would insights obtained by one method be of any less value than those obtained by the other?

The increase of computing power has generated entirely new fields of physics by allowing calculations that previously just weren’t feasible. Turbulence in plasma, supernovae explosion, heavy ion collisions, neutron star mergers, or lattice qcd to study the strong nuclear interaction, these are all examples of investigations that have flourished only with the increase in processing speed and memory. Such disciplines tend to develop their own, unique and very specialized nomenclature and procedures that are difficult if not impossible to evaluate for outsiders.
Lattice QCD. Artist’s impression.

Then there is big data that needs to be handled. May that be LHC collisions or temperature fluctuations in the CMB or global fits of neutrino experiments, this isn’t data any human can deal with by pen on paper. In these areas too, subdisciplines have sprung up, dedicated to data manipulation and -handling. Postdocs specialized in numerical methods are high in demand. But even though essential to physics, they face the prejudice of somehow being “merely calculators.”

The maybe best example is miniscule corrections to probabilities of scattering events, like those taking place at the LHC. Calculating these next-to-next-to-next-to-leading-order contributions is an art as much as a science; it is a small subfield of high energy physics that requires summing up thousands or millions of Feynman diagrams. While there are many software packages available, few physicists know all the intricacies and command all the techniques; those who do often develop software along with their research. They are perfecting calculations, aiming for the tiniest increase in precision much like pole jumpers perfect their every motion aiming after the tiniest increase in height. It is a highly specialized skill, presently at the edge of theoretical physics. But while we admire the relentless perfection of professional athletes, we disregard the single-minded focus of the number-crunchers. What can we learn from it? What insight can be gained from moving the bar an inch higher?

What insight do you gain from calculating the positions of stars on the night sky, you could have asked my grandmother. She was the youngest of seven siblings, her father died in the first world war. Her only brother and husband were drafted for the second world war and feeding the family was left to the sisters. To avoid manufacturing weapons for a regime she detested, she took on a position in an observatory, calculating the positions of stars. This appointment came to a sudden stop when her husband was badly injured and she was called to his side at the war front to watch him die, or so she assumed. Against all expectations, my grandfather recovered from his skull fractures. He didn’t have to return to the front and my grandma didn’t return to her job. It was only when the war was over that she learned her calculations were to help the soldiers target bombs, knowledge that would haunt her still 60 years later.

What insight do we gain from this? Precision is the hallmark of science, and for much of our society science is an end for other means. But can mere calculation ever lead to true progress? Surely not with the computer codes we use today, which execute operations but do not look for simplified underlying principles, which is necessary to advance understanding. It is this lacking search for new theories that leaves physicists cynical about the value of computation. And yet some time in the future we might have computer programs doing exactly this, looking for underlying mathematical laws better suited than existing ones to match observation. Will physicists one day be replaced by software? Can natural law be extracted by computers from data? If you handed all the LHC output to an artificial intelligence, could it spit out the standard model?

In an influential 2008 essay “The End of Theory,” Chris Anderson argued that indeed computers will one day make human contributions to theoretical explanations unnecessary:
“The reason physics has drifted into theoretical speculation about n-dimensional grand unified models over the past few decades (the "beautiful story" phase of a discipline starved of data) is that we don't know how to run the experiments that would falsify the hypotheses — the energies are too high, the accelerators too expensive, and so on[...]

The new availability of huge amounts of data, along with the statistical tools to crunch these numbers, offers a whole new way of understanding the world. Correlation supersedes causation, and science can advance even without coherent models, unified theories, or really any mechanistic explanation at all.”
His future vision was widely criticized by physicists, me included, but I’ve had a change of mind. Much of the criticism Anderson took was due to vanity. We like to believe the world will fall into pieces without our genius and don’t want to be replaced by software. But I don’t think there’s anything special about the human brain that an artificial intelligence couldn’t do, in principle. And I don’t care very much who or what delivers insights, as long as they come by. In the end it comes down to trust.

If a computer came up with just the right string theory vacuum to explain the standard model and offered you the explanation that the world is made of strings to within an exactly quantified precision, what difference would it make whether the headline was made by a machine rather than a human? Wouldn’t you gain the exact same insight? Yes, we would still need humans to initiate the search, someone to write a code that will deliver to our purposes. And chances are we would celebrate the human, rather than the machine. But the rest is overcoming prejudice against “number crunching,” which has to be addressed by setting up reliable procedures that ensure a computer’s results are sound science. I’ll be happy if your AI delivers a theory of quantum gravity; bring it on.

My grandmother outlived her husband who died after multiple strokes. Well in her 90s she still had a habit of checking all the numbers on her receipts, bills, and account statements. Despite my conviction that artificial intelligences could replace physicists, I don’t think it’s likely to happen. The human brain is remarkable not so much for its sheer computing power, but for its efficiency, resilience, and durability. You show me any man-made machine that will still run after 90 years in permanent use.

Monday, July 13, 2015

Book review: “Eureka” by Chad Orzel.

Eureka: Discovering Your Inner Scientist
By Chad Orzel
Basic Books (December 9, 2014)

It’s a good book, really. The problem isn’t the book, the problem is me.

Chad Orzel’s new book “Eureka” lays out how the scientific enterprise reflects in every-day activities. In four parts, “Look, Think, Test, Tell,” Chad connects examples from sports, cooking, collecting, crossword puzzles, mystery novels, and more, to the methodology used in scientific discovery. Along the way, he covers a lot of ground in astrophysics, cosmology, geology, and atomic physics. And the extinction of the dinosaurs.

It’s a well-written book, with relevant but not excessive references and footnotes; the scientific explanations are accurate yet non-technical; the anecdotes from the history of science and contemporary academia are nicely woven together with the books theme, that every one of us has an “inner scientist” who is waiting to be discovered.

To my big relief in this recent book Chad doesn’t talk to his dog. Because I’m really not a dog person. I’m the kind of person who feeds the neighbor’s cat. Sometimes. I’m also the kind of person who likes baking, running, house music, and doesn’t watch TV. I order frozen food and get it delivered to the door. Chad cooks, believes baking is black magic, plays basketball, likes rock music, and his writing draws a lot on contemporary US TV shows or movies. He might be a good science popularizer, but his sports popularization is miserable. It doesn’t seem to have occurred to him that somebody could read his book who, like me, doesn’t know anything about baseball, or bridge, or basketball, and therefore much of his explanations are entirely lost on me.

I don’t think I’ve ever read a book that made me feel so distinctly that I’m not the intended audience. Of course the whole idea of “Eureka” is totally backwards for me. You don’t have to convince me I’m capable of scientific reasoning. I have even proved capable of convincing others I’m capable of scientific reasoning. Sometimes. But I do not have the slightest idea why somebody would want to spend hours trying to throw a ball into a basket, or, even more bizarre, watch other people trying to throw balls into baskets. So some stretches of the book stretched indeed. Which is why it’s taken me so long to get through with it, since I had an advance proof more than a year ago.

Besides this, I have a general issue with the well-meant message that we were born to think scientifically, as I elaborated on in this recent post. Chad’s argument is that every one of us brings the curiosity and skills to be a scientist, and that we use many of these skills intuitively. I agree on this. Sometimes. I wish though that he had spent a few more words pointing out that being a scientist is a profession after all, and one that requires adequate education for a reason. While we do some things right intuitively, intuition can also mislead us. I understand that Chad addresses an existing cultural misconception, which is that science is only for a gifted few rather than a general way to understand the world. However, I’d rather not swap this misconception for another misconception, which is that scientific understanding comes with little guidance and effort.

In summary, it’s a great book to give to someone who is interested in sports but not in science, to help them discover their inner scientist. Chad does an excellent job pointing out how much scientific thought there is in daily life, and he gets across a lot of physics along with that. He tells the reader that approaching problems scientifically is not just helpful for researchers, but for every one to understand the world. Sometimes. Now somebody please explain me the infield fly rule.

Wednesday, July 08, 2015

The loneliness of my notepad

Few myths have been busted as often and as cheerfully as that of the lone genius. “Future discoveries are more likely to be made by scientists sharing ideas than a lone genius,” declares Athene Donald in the Guardian, and Joshua Wolf Shenk opines in the NYT that “the lone genius is a myth that has outlived its usefulness.” Thinking on your own is so yesterday; today is collaboration. “Fortunately, a more truthful model is emerging: the creative network,” Shenk goes on. It sounds scary.

There is little doubt that with the advance of information technology, collaborations have been flourishing. As Mott Greene has observed, single authors are an endangered species: “Any issue of Nature today has nearly the same number of Articles and Letters as one from 1950, but about four times as many authors. The lone author has all but disappeared. In most fields outside mathematics, fewer and fewer people know enough to work and write alone.”

Science Watch keeps tracks of the data. The average number of authors per paper has risen from 2.5 in the early 1980s to more than five today. At the same time, the fraction of single authored papers has declined from more than 30% to about 10%.



Part of the reason for this trend is that the combination of expertise achieved by collaboration opens new possibilities that are being exploited. This would suggest that the increase is temporary and will eventually stagnate or start declining again once the potential in these connections has been fully put to use.

But I don’t think the popularity of larger collaborations is going to decline anytime soon, because for some purposes one paper with five authors counts as five papers. If I list a paper on our institutional preprint list, nobody cares how many coauthors it has – it counts as one more paper, and my coauthors’ institutions can be equally happy about adding a paper to their count. If you work on the average with 5 coauthors and divide up the work fairly, your publication list will end up being five times as long as if you’d be working alone. The logical goal of this accounting can only be that we all coauthor every paper that gets published.

So, have the numbers spoken and demonstrated the lone genius is no more?

Well, most scientists aren’t geniuses and nobody really agrees what that means anyway, so let us ask instead what happened to the lone scientists.

The “lone scientist” isn’t so much a myth than an oxymoron. Science is ultimately a community enterprise – an idea not communicated will never become accepted part of science. But the lone scientist has always existed and certainly still exists as a mode of operation, as a step on the path to an idea worth developing. Declaring lonely work a myth is deeply ironic for the graduate student stuck with an assigned project nobody seems to care about. In theoretical physics research often means making yourself world expert on whatever topic, whether you picked it for yourself or whether somebody else thought it was a good idea. And loneliness is the flipside of this specialization.

That researchers may sometimes be lonely in their quest doesn’t imply they are alone or they don’t talk to each other. But when you are the midst of working out an idea that isn’t yet fully developed, there is really nobody who will understand what you are trying to do. Possibly not even you yourself understand.

We use and abuse our colleagues as sounding boards, because attempting to explain yourself to somebody else can work wonders to clarify your own thoughts, even though the person doesn’t understand a word. I have been both at the giving and receiving end of this process. A colleague, who shall remain unnamed, on occasion simply falls asleep while I am trying to make sense. My husband deserves credit for enduring my ramblings about failed calculations even though he doesn’t have a clue what I mean. And I’ve learned to listen rather than just declaring that I don’t know a thing about whatever.

So the historians pointing out that Einstein didn’t work in isolation, and that he met frequently with other researchers to discuss, do not give honor to the frustrating and often painful process of having to work through a calculation one doesn’t know how to do in the first place. It is evident from Einstein’s biography and publications that he spent years trying to find the right equations, working with a mathematical apparatus that was unfamiliar to most researchers back then. There was nobody who could have helped him trying to find the successful description that his intuition looked for.

Not all physicists are Einstein of course, and many of us work on topics where the methodology is well developed and widely shared. But it is very common to get stuck while trying to find a mathematically accurate framework for an idea that, without having found what one is looking for, remains difficult or impossible to communicate. This leaves us alone with that potentially brilliant idea and the task of having to sort out our messy thoughts well enough to even be able to make sense to our colleagues.

Robert Nemiroff, Professor of Physics at Michigan Tech, noted down his process of working on a new paper aptly in a long string of partly circular and confused attempts intersected by doubts, insights and new starts:
“writing short bits of my budding new manuscript on a word processor; realizing I don't know what I am talking about; thinking through another key point; coming to a conclusion; coming to another conclusion in contradiction to the last conclusion; realizing that I have framed the paper in a naive fashion and delete sections (but saving all drafts just in case); starting to write a new section; realizing I don't know what I am talking about; worrying that this whole project is going nowhere new; being grouchy because if this is going nowhere then I am wasting my time; comforting myself with the thought that at least now I better understand something that I should have better understood earlier; starring at my whiteboard for several minute stretches occasionally sketching a small diagram or writing a key equation; thinking how cool this is and wondering if anyone else understands this mini- sub- topic in this much detail; realizing a new potential search term and doing a literature search for it in ADS and Google; finding my own work feeling reassured that perhaps I am actually a reasonable scientist; finding key references that address some part of this idea that I didn't know about and feeling like an idiot; reading those papers and thinking how brilliant those authors are and how I could never write papers this good...”
At least for now we’re all alone in our heads. And as long as that remains so, the scientist struggling to make sense alone will remain reality.

Thursday, July 02, 2015

Analog Duality

I have recently begun working on a research topic entirely new for me. The idea is to combine analog gravity with the applications of the gauge-gravity duality to condensed matter systems. Two ingredients are necessary for this: First, there is the AdS/CFT duality that identifies a gravitational system in asymptotic anti-de Sitter (AdS) space-time with a conformal field theory on the boundary of that space. And second, there is analog gravity, which is the property of some weakly coupled condensed matter systems to give rise to an effective metric. Combine both, and you obtain a new relation, “analog duality,” that connects a strongly coupled with a weakly coupled condensed matter theory, where the metric in AdS space plays the role of an intermediary.

I have now given several seminars about this, and the most common reaction I get is “Why hasn’t anybody done this before?” Well, I don’t know! It might sound obvious, but it isn’t all that trivial once you look at the details. For this junction of two different research directions to work, you have to show that it is indeed possible to obtain the metrics used in the gauge-gravity applications as effective metrics in analog gravity.
The gauge-gravity duality (left) identifies some solutions in General Relativity with strongly coupled condensed matter systems. Analog gravity (right) identifies some solutions in General Relativity with effective metrics of certain types of weakly coupled condensed matter systems.

Provided one can show that the set of metrics used in the gauge-gravity duality overlaps with the set of metrics used in analog gravity, one obtains a new relation, “analog duality,” between strongly and a weakly coupled condensed matter systems.



Concretely, for the applications of the gauge-gravity duality to condensed matter systems people do not use the full mathematical apparatus of this correspondence. To begin with they only work in the limit where in the AdS space one has only classical gravity, and some fields propagating in it, but no string effects. And they are not using a system that is known to correspond to any well-understood theory on the boundary. These are phenomenological models, which is why I find this research area so appealing. Mathematical beauty is nice, but for me describing reality scores higher. For the phenomenological approach one takes the general idea to use a gravitational theory to describe a condensed matter system, and rather than trying to derive this from first principles, sees whether it fits with observations.

In my paper, what I have done is to look at the metrics that are being used to describe high temperature superconductors, and have shown that they can indeed be obtained as effective metrics in certain analog gravity systems. To prove this, one first has to convert the metric into a specific form, which amounts to a certain gauge condition, and then extract the degrees of freedom of the fluid. After this, one has to show that these degrees of freedom fulfil the equations of motion, which will not in general be the case even if you can bring a metric into the necessary form.

Amazingly enough, this turned out to work very well, better indeed than I expected. The coordinate systems commonly used in the AdS/CFT duality have the metric already in almost the right form, and all one has to do is to apply a little shear to it. If one then extracts the degrees of freedom of the fluid, they quite mysteriously fulfil the equations of motion automatically. The challenge in that wasn’t so much doing the calculation but to find out how to do it in the first place.

I found this connection really surprising. The metrics are derived within classical General Relativity. They are solutions to a set of equations that know absolutely nothing about hydrodynamics; it’s basically Maxwell’s equations in a curved background with a cosmological constant added. But once you have learned to look at this metric in the right way, you find that it contains the degrees of freedom of a fluid! And it is not the fluid on the boundary of the space-time that belongs to the gauge-theory, but a different one. This adds further evidence to the connection between gravity and hydrodynamics that has been mounting for decades.

One then has two condensed matter systems describing aspects of the same gravitational theory, which gives rise to relations between the two condensed matter systems. It turns out for example that the temperature of the one system (on the boundary of AdS) is related to the speed of sound of the other system. There must be more relations like this, but I haven’t yet had the time to look into this in more detail.

As always, there is some fineprint that I must mention. It turns out for example that both the analog gravity system and the boundary of the AdS space must have dimension 3+1. This isn’t something I got to pick; the identification only works in this case. This means that the analog gravity system describes a slice of AdS space, but unlike the boundary it’s a slice perpendicular to the horizon. The more relevant limitation of what I have so far is that in general the equations of motion for the background of the fluid in the analog gravity system will not be identical to Einstein’s field equations. This means that at least for now I cannot include backreaction, and the identification only works on the level of quantum field theory in a fixed background.

I have also in the present paper worked only in a non-relativistic limit, which is a consistent approximation and fine to use, but if the new relation is to make sense physically there must be a fully relativistic treatment. The relativistic extension is what I am working on right now, and while the calculation isn’t finished yet, I can vaguely say it looks good.

So I’m still several steps away from actually proving that a new duality exists, but I have shown that several necessary conditions required for its existence are fulfilled, which is promising. Now suppose I’d manage to prove in generality that there is a new duality that can be obtained by combining the gauge-gravity duality with analog gravity, what would it be good for?

To begin with, I find this interesting for purely theoretical reasons. But beyond this, the advantage of this duality over the gauge-gravity duality is that both systems (the strongly and the weakly coupled condensed matter system) can be realized in the laboratory. (At least in principle. I have a Lagrangian, but not sure yet exactly what it describes.) This means one could experimentally test the validity of this relation by comparing measurements on both systems. Since the gauge-gravity applications to condensed matter systems were indirectly used to obtain the new duality, this would then serve as an implicit experimental test for the validity of the AdS/CFT duality in these applications.

A year ago, I knew very little about any of the physics involved in this project. I’ve never before worked on AdS/CFT, analog gravity, magnetohydrodynamics, or superconductivity, and this is all still very new for me. It’s been a steep learning curve, and I am nowhere near having the same overview on this research than I have on quantum gravity phenomenology – there I’ve reached a level where it feels like I’ve seen it all before. I’m not good with planning ahead my research because I tend to go where my interest takes me, but I think I’ll stick around with this topic and see how far I can push this idea of a new duality. I definitely want to see if I can make the case with backreaction work, at least in the simplest scenarios.

For more details, the paper is here:
and I have slides of my talk here.

Sunday, June 28, 2015

I wasn’t born a scientist. And you weren’t either.

There’s a photo which keeps cropping up in my facebook feed and it bothers me. It shows a white girl, maybe three years, kissing a black boy the same age. The caption says “No one is born racist.” It’s adorable. It’s inspirational. But the problem is, it’s not true.

Children aren’t saints. We’re born mistrusting people who look different from us, and we treat those who look like us better. Toddlers already have this “in-group bias” research says. Though I have to admit that, as a physicist, I am generally not impressed by what psychologists consider statistically significant, and I acknowledge it is generally hard to distinguish nature from nurture. But that a preference for people of similar appearance should be a result of evolution isn’t so surprising. We are more supportive to who we share genes with, family ahead of all, and looks are a giveaway.

As we grow up, we should become aware that our bias is both unnecessary and unfair, and take measures to prevent it from being institutionalized. But since we are born being extra suspicious about anybody not from our own clan, it takes conscious educational effort to act against the preference we give to people “like us.” Racist thoughts are not going away by themselves, though one can work to address them – or at least I hope so. But it starts with recognizing one is biased to begin with. And that’s why this photo bothers me. Denying a problem rarely helps solving it.

On the same romantic reasoning I often read that infants are all little scientists, and it’s only our terrible school education that kills curiosity and prevents adults from still thinking scientifically. That is wrong too. Yes, we are born being curious, and as children we learn a lot by trial and error. Ask my daughter who recently learned to make rainbows with the water sprinkler, mostly without soaking herself. But our brains didn’t develop to serve science, they developed to serve ourselves in the first place.

My daughters for example haven’t yet learned to question authority. What mommy speaks is true, period. When the girls were beginning to walk I told them to never, ever, touch the stove when I’m in the kitchen because it’s hot and it hurts and don’t, just don’t. They took this so seriously that for years they were afraid to come anywhere near the stove at any time. Yes, good for them. But if I had told them rainbows are made by garden fairies they’d have believed this too. And to be honest, the stove isn’t hot all that often in our household. Still today much of my daughters’ reasoning begins with “mommy says.” Sooner or later they will move beyond M-theory, or so I hope, but trust in authorities is a cognitive bias that remains with us through adulthood. I have it. You have it. It doesn’t go away by denying it.

Let me be clear that human cognitive biases aren’t generally a bad thing. Most of them developed because they are, or at least have been, of advantage to us. We are for example more likely to put forward opinions that we believe will be well-received by others. This “social desirability bias” is a side-effect of our need to fit into a group for survival. You don’t tell the tribal chief his tent stinks if you have a dozen fellows with spears in the back. How smart of you. While opportunism might benefit our survival, it rarely benefits knowledge discovery though.

It is because of our cognitive shortcomings that scientists have put into place many checks and methods designed to prevent us from lying to ourselves. Experimental groups for example go to lengths preventing bias in data analysis. If your experimental data are questionnaire replies then that’s that, but in physics data aren’t normally very self-revealing. They have to be processed suitably and be analyzed with numerical tools to arrive at useful results. Data has to be binned, cuts have to be made, background has to be subtracted.

There are usually many different ways to process the data, and the more ways you try the more likely you are to find one that delivers an interesting result, just by coincidence. It is pretty much impossible to account for trying different methods because one doesn’t know how much these methods are correlated. So to prevent themselves from inadvertently running multiple searches for a signal that isn’t there, many experimental collaborations agree on a method for data analysis before the data is in, then proceed according to plan.

(Of course if the data are made public this won’t prevent other people to reanalyze the same numbers over and over again. And every once in a while they’ll find some signal whose statistical significance they overestimate because they’re not accounting, can’t account, for all the failed trials. Thus all the CMB anomalies.)

In science as in everyday life the major problems though are the biases we do not account for. Confirmation bias is the probably most prevalent one. If you search the literature for support of your argument, there it is. If you try to avoid that person who asked a nasty question during your seminar, there it is. If you just know you’re right, there it is.

Even though it often isn’t explicitly taught to students, everyone who succeeded making a career in research has learned to work against their own confirmation bias. Failing to list contradicting evidence or shortcomings of one’s own ideas is the easiest way to tell a pseudoscientist. A scientist’s best friend is their inner voice saying: “You are wrong. You are wrong, wrong, W.R.O.N.G.” Try to prove yourself wrong. Then try it again. Try to find someone willing to tell you why you are wrong. Listen. Learn. Look for literature that explains why you are wrong. Then go back to your idea. That’s the way science operates. It’s not the way humans normally operate.

(And lest you want to go meta on me, the title of this post is of course also wrong. We are scientists in some regards but not in others. We like to construct new theories, but we don’t like being proved wrong.)

But there are other cognitive and social biases that affect science which are not as well-known and accounted for as confirmation bias. “Motivated cognition” (aka “wishful thinking”) is one of them. It makes you believe positive outcomes are more likely than they really are. Do you recall them saying the LHC would find evidence for physics beyond the standard model. Oh, they are still saying it will?

Then there is the “sunk cost fallacy”: The more time and effort you’ve spent on SUSY, the less likely you are to call it quits, even though the odds look worse and worse. I had a case of that when I refused to sign up for the Scandinavian Airline frequent flyer program after I realized that I'd be a gold member now had I done this 6 years ago.

I already mentioned the social desirability bias that discourages us from speaking unwelcome truths, but there are other social biases that you can see in action in science.

The “false consensus effect” is one of them. We tend to overestimate how much and how many other people agree with us. Certainly nobody can disagree that string theory is the correct theory of quantum gravity. Right. Or, as Joseph Lykken and Maria Spiropulu put it:
“It is not an exaggeration to say that most of the world’s particle physicists believe that supersymmetry must be true.” (Their emphasis.)
The “halo effect” is the reason we pay more attention to literally every piece of crap a Nobelprize winner utters. The above mentioned “in-group bias” is what makes us think researchers in our own field are more intelligent than others. It’s the way people end up studying psychology because they were too stupid for physics. The “shared information bias” is the one in which we discuss the same “known problems” over and over and over again and fail to pay attention to new information held only by a few people.

One of the most problematic distortions in science is that we consider a fact more likely the more often we have heard of it, called the “attentional bias” or the “mere exposure effect”. Oh, and then there is the mother of all biases, the “bias blind spot,” the insistence that we certainly are not biased.

Cognitive biases we’ve always had of course. Science has progressed regardless, so why should we start paying attention now? (Btw, it’s called the “status-quo-bias”.) We should pay attention now because shortcomings in argumentation become more relevant the more we rely on logical reasoning detached from experimental guidance. This is a problem which affects some areas of theoretical physics more than any other field of science.

The more prevalent problem though is the social biases whose effects become more pronounced the larger the groups are, the tighter they are connected, and the more information is shared. This is why these biases are so much more relevant today than a century, even two decades ago.

You can see these problems in pretty much all areas of science. Everybody seems to be thinking and talking about the same things. We’re not able to leave behind research directions that turn out fruitless, we’re bad at integrating new information, we don’t criticize our colleagues’ ideas because we are afraid of becoming “socially undesirable” when we mention the tent’s stink. We disregard ideas off the mainstream because these come from people “not like us.” And we insist our behavior is good scientific conduct, purely based on our unbiased judgement, because we cannot possibly be influenced by social and psychological effects, no matter how well established.

These are behaviors we have developed not because they are stupid, but because they are beneficial in some situations. But in some situations they can become a hurdle to progress. We weren’t born to be objective and rational. Being a good scientist requires constant self-monitoring and learning about the ways we fool ourselves. Denying the problem doesn’t solve it.

What I really wanted to say is that I’ve finally signed up for the SAS frequent flyer program.

Wednesday, June 24, 2015

Does faster-than-light travel lead to a grandfather paradox?

Whatever you do, don’t f*ck with mom.
Fast track to wisdom: Not necessarily.

I stopped going to church around the same time I started reading science fiction. Because who really needs god if you can instead believe in alien civilizations, wormholes, and cell rejuvenation? Oh, yes, I wanted to leave behind this planet for a better place. But my space travel enthusiasm suffered significantly once I moved from the library’s fiction aisle to popular science, and learned that the speed of light is the absolute limit. For all we know. And ever since I have of course wondered just how well we know this.

Fact is we’ve never seen anything move faster than the speed of light (except for illusions of motion), and it is both theoretically understood and experimentally confirmed that we cannot accelerate anything to become faster than light. That doesn’t sound good for what our chances of visiting the aliens are concerned, but it isn’t the main problem. It could just be that we haven’t looked in the right places or not tried hard enough. No, the main problem is that it is very hard to make sense of faster-than-light travel at all within the context of our existing theories. And if you can’t make sense of it, how can you build it?

Special relativity doesn’t forbid motion faster than light. It just tells you that you’d need an infinite amount of energy to accelerate something which is slower than light (“subluminal”) to become faster than light (“superluminal”). Ok, the infinite energy need won’t fly with the environmentalists, I know. But if you have a particle that always moves faster than light, its existence isn’t prohibited in principle. These particles are called “tachyons,” have never been observed, and are believed to not exist for two reasons. First, they have the awkward property of accelerating when they lose energy, which lets them induce instabilities that have to be fixed somehow. (In quantum field theory one can deal with tachyonic fields, and they play an important role, but they don’t actually transmit any information faster than light. So these are not so relevant to our purposes.) Second, tachyons seem to lead to causality problems.

The causality problems with superluminal travel come about as follows. Special relativity is based on the axiom that all observers have the same laws of physics, and these are converted from one observer to another by a well-defined procedure called Lorentz-transformation. This transformation from one observer to the other maintains lightcones, because the speed of light doesn’t change. The locations of objects relative to an observer can change when the observer changes velocity. But two observers at the same location with different velocities who look at an object inside the lightcone will agree on whether it is in the past or in the the future.

Not so however with objects outside the lightcone. For these, what is in the future for one observer can be in the past of another observer. This means then that a particle that for one observer moves faster than light – ie to a point outside the lightcone – actually moves backwards in time for another observer! And since in special relativity all observers have equal rights, neither of them is wrong. So once you accept superluminal travel, you are forced to also accept travel back in time.

At least that’s what the popular science books said. It’s nonsense of course because what does it mean for a particle to move backwards in time anyway? Nothing really. If you’d see a particle move faster than light to the left, you could as well say it moved backwards in time to the right. The particle doesn’t move in any particular direction on a curve in space-time because the particles’ curves have no orientation. Superluminal particle travel is logically perfectly possible as long as it leads to a consistent story that unfolds in time, and there is nothing preventing such a story.

Take as an example the below image showing the worldline of a particle that is produced, scatters twice to change direction, travels superluminally, and goes back in time to meet itself. You could interpret the very same arrangement as saying you have produced a pair of particles, one of which scatters and then annihilates again.

No, there is no problem with the travel of superluminal particles in principle. The problems start once we think of macroscopic objects, like spaceships. We attach to their curves an arrow of time, pointing into the direction in which the travelers age. And it’s here where the trouble starts. Now special relativity indeed tells you that somebody who travels faster than light will move backwards in time for another observer, because a change of reference frame will not reverse the travelers’ arrow of time. This is what creates the grandfather paradox, in which you can travel back in time to kill your own grandfather, resulting in you never be born. Here, requiring consistency would necessitate that it is somehow impossible for you to kill your grandfather, and it is hard to see how this would be insured by the laws of physics.

While it’s hard to see what conspiracy would prevent you from killing your grandpa, it is fairly easy to see that closing the loop backwards in time is prevented by the known laws of physics. We age because entropy increases. It increases in some direction that we can, for lack of a better word, call “forward” in time. This entropy increase is ultimately correlated with decoherence and thus probably also with the restframe of the microwave background, but for our purposes it doesn’t matter so much exactly in which direction it increases, just that it increases in some direction.

Now whenever you have a closed curve that is oriented in the direction in which the traveler presumably experience the passage of time, then the arrow of time on the curve must necessarily run against the increase of entropy somewhere. Any propulsion system able to do this would have to decrease entropy against the universe’s thrust of increasing it. And that’s what ultimately prevents time-travel. In the image below I have drawn the same worldline as above with an intrinsic arrow of time (the direction in which passengers age), and how it is necessarily incompatible with any existing arrow of time along one of the curves, which is thus forbidden.

There is no propulsion system that would be able to produce the necessary finetuning to decrease entropy along the route. But even if such a propulsion existed it would just mean that time in the spaceship now runs backwards. In other words, the passengers wouldn’t actually experience moving backwards in time, but instead moving forwards in time in the opposite direction. This would force us to buy into an instance of a grandfather pair creation, later followed by a grandchild pair annihilation. It doesn’t seem very plausible, and it violates energy conservation, but besides this it’s at least a consistent story.

I briefly elaborated on this in a paper I wrote some years ago as a sidenote (see page 6). But just last month there was a longer paper on the arxiv, by Nemiroff and Russell, that studied the problems with superluminal travel in a very concrete scenario. In their example, a spaceship leaves Earth, visits an exoplanet that moves with some velocity relative to Earth, and then returns. The velocity of the spaceship at the both launches is the same relative to the planet from which the ship launches, which means it’s a different velocity on the return trip.

The authors then calculate explicitly at which velocity the curves start going back in time. They arrive at the conclusion that the necessity of a consistent time evolution for the Earth observer would then require to interpret the closed loop in time as a pair creation event, followed by a later pair annihilation, much like I argued above. Note that singling out the Earth observer as the one demanding consistency with their arrow of time is in this case what introduces a preferred frame relative to which “forward in time” is defined.

The relevant point to take away from this is that superluminal travel in and by itself is not inconsistent. Leaving aside the stability problems with superluminal particles, they do not lead to causal paradoxa. What leads to causal paradoxa is allowing travel against the arrow of time which we, for better or worse, experience. This means that superluminal travel is possible in principle, even though travel backwards in time is not.

That travel faster than light is not prevented by the existing laws of nature doesn’t mean of course that it’s possible. There is also still the minor problem that nobody has the faintest clue how to do it... Maybe it’s easier to wait for the aliens to come visit us.

Thursday, June 18, 2015

No, Gravity hasn’t killed Schrödinger’s cat

There is a paper making the rounds which was just published in Nature Physics, but has been on the arXiv since two years:
    Universal decoherence due to gravitational time dilation
    Igor Pikovski, Magdalena Zych, Fabio Costa, Caslav Brukner
    arXiv:1311.1095 [quant-ph]
According to an article in New Scientist the authors have shown that gravitationally induced decoherence solves the Schrödinger’s cat problem, ie explains why we never observe cats that are both dead and alive. Had they achieved this, that would be remarkable indeed because the problem has been solved half a century ago. New Scientist also quotes the first author as saying that the effect discussed in the paper induces a “kind of observer.”

New Scientist further tries to make a connection to quantum gravity, even though everyone involved told the journalist it’s got nothing to do with quantum gravity whatsoever. There is also a Nature News article, which is more careful for what the connection to quantum gravity, or absence thereof, is concerned, but still wants you to believe the authors have shown that “completely isolated objects” can “collapse into one state” which would contradict quantum mechanics. If that could happen it would be essentially the same as the information loss problem in black hole evaporation.

So what did they actually do in the paper?

It’s a straight-forward calculation which shows that if you have a composite system in thermal equilibrium and you push it into a gravitational field, then the degrees of freedom of the center of mass (com) get entangled with the remaining degrees of freedom (those of the system’s particles relative to the center of mass). The reason for this is that the energies of the particles become dependent on their position in the gravitational field by the standard redshift effect. This means that if the system’s particles had quantum properties, then these quantum properties mix together with the com position, basically.

Now, decoherence normally works as follows. If you have a system (the cat) that is in a quantum state, and you get it in contact with some environment (a heat bath, the cosmic microwave background, any type of measurement apparatus, etc), then the cat becomes entangled with the environment. Since you don’t know the details of the environment however, you have to remove (“trace out”) its information to see what the cat is doing, which leaves you with a system that has now a classic probabilistic distribution. One says the system has “decohered” because it has lost its quantum properties (or at least some of them, those that are affected by the interaction with the environment).

Three things important to notice about this environmentally induced decoherence. First, the effect happens extremely quickly for macroscopic objects even for the most feeble of interactions with the environment. This is why we never see cats that are both dead and alive, and also why building a functioning quantum computer is so damned hard. Second, while decoherence provides a reason we don’t see quantum superpositions, it doesn’t solve the measurement problem in the sense that it just results in a probability distribution of possible outcomes. It does not result in any one particular outcome. Third, nothing of that requires an actually conscious observer; that’s an entirely superfluous complication of a quite well understood process.

Back to the new paper then. The authors do not deal with environmentally induced decoherence but with an internal decoherence. There is no environment, there is only a linear gravitational potential; it’s a static external field that doesn’t carry any degrees of freedom. What they show is that if you trace out the particle’s degrees of freedom relative to the com, then the com decoheres. The com motion, essentially, becomes classical. It can no longer be in a superposition once decohered. They calculate the time it takes for this to happen, which depends on the number of particles of the system and its extension.

Why is this effect relevant? Well, if you are trying to measure interference it is relevant because this relies on the center of mass moving on two different paths – one going through the left slit, the other through the right one. So the decoherence of the center of mass puts a limit on what you can measure in such interference experiments. Alas, the effect is exceedingly tiny, smaller even than the decoherence induced by the cosmic microwave background. In the paper they estimate the time it takes for 1023 particles to decohere is about 10-3 seconds. But the number of particles in composite systems that can presently be made to interfere is more like 102 or maybe 103. For these systems, the decoherence time is roughly 107 seconds - that’s about a year. If that was the only decoherence effect for quantum systems, experimentalists would be happy!

Besides this, the center of mass isn’t the only quantum property of a system, because there are many ways you can bring a system in superpositions that doesn’t affect the com at all. Any rotation around the com for example would do. In fact there are many more degrees of freedom in the system that remain quantum than that decohere by the effect discussed in the paper. The system itself doesn’t decohere at all, it’s really just this particular degree of freedom that does. The Nature News feature states that
“But even if physicists could completely isolate a large object in a quantum superposition, according to researchers at the University of Vienna, it would still collapse into one state — on Earth's surface, at least.”
This is just wrong. The object could still have many different states, as long as they share the same center of mass variable. A pure state left in isolation will remain in a pure state.

I think the argument in the paper is basically correct, though I am somewhat confused about the assumption that the thermal distribution doesn’t change if the system is pushed into a gravitational field. One would expect that in this case the temperature also depends on the gradient.

So in summary, it is a nice paper that points out an effect of macroscopic quantum systems in gravitational fields that had not previously been studied. This may become relevant for interferometry of large composite objects at some point. But it is an exceedingly weak effect, and I for sure am very skeptical that it can be measured any time in the soon future. This effect doesn’t teach us anything about Schrödinger’s cat or the measurement problem that we didn’t know already, and it for sure has nothing to do with quantum gravity.

Science journalists work in funny ways. Even though I am quoted in the New Scientist article, the journalist didn’t bother sending me a link. Instead I got the link from Igor Pikovski, one of the authors of the paper, who wrote to me to apologize for the garble that he was quoted with. He would like to pass on the following clarification:
“To clarify a few quotes used in the article: The effect we describe is not related to quantum gravity in any way, but it is an effect where both, quantum theory and gravitational time dilation, are relevant. It is thus an effect based on the interplay between the two. But it follows from physics as we know it.

In the context of decoherence, the 'observer' are just other degrees of freedom to which the system becomes correlated, but has of course nothing to do with any conscious being. In the scenario that we consider, the center of mass becomes correlated with all the internal constituents. This takes place due to time dilation, which correlates any dynamics to the position in the gravitational field and results in decoherence of the center of mass of the composite system.

For current experiments this effect is very weak. Once superposition experiments can be done with very large and complex systems, this effect may become more relevant. In the end, the simple prediction is that it only depends on how much proper time difference is acquired by the interfering amplitudes of the system. If it's exactly zero, no decoherence takes place, as for example in a perfectly horizontal setup or in space (neglecting special relativistic time dilation). The latter was used as an example in the article. But of course there are other means to make sure the proper time difference is minimized. How hard or easy that will be depends on the experimental techniques. Maybe an easier route to experimentally probe this effect is to probe the underlying Hamiltonian. This could be done by placing clocks in superposition, which we discussed in a paper in 2011. The important point is that these predictions follow from physics as we know, without any modification to quantum theory or relativity. It is thus 'regular' decoherence that follows from gravitational time dilation.”

Tuesday, June 16, 2015

The plight of the postdocs: Academia and mental health

This is the story of a friend of a friend, a man by name Francis who took his life at age 34. Francis had been struggling with manic depression through most of his years as a postdoc in theoretical physics.

It is not a secret that short-term contracts and frequent moves are the norm in this area of research, but rarely do we spell out the toll it takes one our mental health. In fact, most of my tenured colleagues who profit from cheap and replaceable postdocs praise the virtue of the nomadic lifestyle which, so we are told, is supposed to broaden our horizon. But the truth is that moving is a necessary, though not sufficient, condition to build your network. It isn’t about broadening your horizon, it’s to make the contacts for which you are later being bought in. It’s not optional, it’s a misery you are expected to pretend enjoying.

I didn’t know Francis personally, and I would never have heard of him if it wasn’t for the acknowledgements in Oliver Roston’s recent paper:

“This paper is dedicated to the memory of my friend, Francis Dolan, who died, tragically, in 2011. It is gratifying that I have been able to honour him with work which substantially overlaps with his research interests and also that some of the inspiration came from a long dialogue with his mentor and collaborator, Hugh Osborn. In addition, I am indebted to Hugh for numerous perceptive comments on various drafts of the manuscript and for bringing to my attention gaps in my knowledge and holes in my logic. Following the appearance of the first version on the arXiv, I would like to thank Yu Nakayama for insightful correspondence.

I am firmly of the conviction that the psychological brutality of the post-doctoral system played a strong underlying role in Francis’ death. I would like to take this opportunity, should anyone be listening, to urge those within academia in roles of leadership to do far more to protect members of the community suffering from mental health problems, particularly during the most vulnerable stages of their careers.”
As a postdoc, Francis lived separated from his partner, and had trouble integrating in a new group. Due to difficulties with the health insurance after an international move, he couldn’t continue his therapy. And even though highly gifted, he must have known that no matter how hard he worked, a secure position in the area of research he loved was a matter of luck.

I found myself in a very similar situation after I moved to the US for my first postdoc. I didn’t fully realize just how good the German health insurance system is until I suddenly was on a scholarship without any insurance at all. When I read the fineprint, it became pretty clear that I wouldn’t be able to afford an insurance that covered psychotherapy or medical treatment for mental disorders, certainly not when I disclosed a history of chronic depression and various cycles of previous therapy.

With my move, I had left behind literally everybody I knew, including my boyfriend who I had intended to marry. For several months, the only piece of furniture in my apartment was a mattress because thinking any further was too much. I lost 30 pounds in six months, and sometimes went weeks without talking to a human being, other than myself.

The main reason I’m still here is that I’m by nature a loner. When I wasn’t working, I was hiking in the canyons, and that was pretty much all I did for the better part of the first year. Then, when I had just found some sort of equilibrium, I had to move again to take on another position. And then another. And another. It still seems a miracle that somewhere along the line I managed to not only marry the boyfriend I had left behind, but to also produce two wonderful children.

Yes, I was lucky. But Francis wasn’t. And just statistically some of you are in that dark place right now. If so, then you, as I, have heard them talk about people who “managed to get diagnosed” as if depression was a theater performance in which successful actors win a certificate to henceforth stay in bed. You, as I, know damned well that the last thing you want is that anybody who you may have to ask for a letter sees anything but the “hard working” and “very promising” researcher who is “recommended without hesitation.” There isn’t much advice I can give, except that you don’t forget it’s in the nature of the disease to underestimate one’s chances of recovery, and that mental health is worth more than the next paper. Please ask for help if you need it.

Like Oliver, I believe that the conditions under which postdoctoral researchers must presently sell their skills are not conductive to mental health. Postdocs see friends the same age in other professions having families, working independently, getting permanent contracts, pension plans, and houses with tricycles in the yard. Postdoctoral research collects some of the most intelligent and creative people on the planet, but in the present circumstances many are unable to follow their own interests, and get little appreciation for their work, if they get feedback at all. There are lots of reasons why being a postdoc sucks, and most of them we can do little about, like those supervisors who’d rather die then say you did a good job, only once. But what we can do is improve employment conditions and lower the pressure to constantly move.

Even in the richest countries on the planet, like Germany and Sweden, it is very common to park postdocs on scholarships without benefits. These scholarships are tax-free and come, for the employer, at low cost. Since the tax evasion is regulated by law, the scholarships can typically last only one or two years. It’s not that one couldn’t hire postdocs on longer, regular contracts with social and health benefits, it’s just that in current thinking quantity counts more than quality: More postdocs produce more papers, which looks better in the statistic. That’s practiced, among many others, at my own workplace.

There are some fields of research which lend themselves to short projects and in these fields one or two year gigs work just fine. In other fields that isn’t so. What you get from people on short-term contracts is short-term thinking. It isn’t only that this situation is stressful for postdocs, it isn’t good for science either. You might be saving money with these scholarships, but there is always a price to pay.

We will probably never know exactly what Francis went through. But for me just the possibility that the isolation and financial insecurity, which are all too often part of postdoc life, may have contributed to his suffering is sufficient reason to draw attention to this.

The last time I met Francis’ friend Oliver, he was a postdoc too. He now has two children, a beautiful garden, and has left academia for a saner profession. Oliver sends the following message to our readers:
“I think maybe the best thing I can think of is advising never to be ashamed of depression and to make sure you keep talking to your friends and that you get medical help. As for academia, one thing I have discovered is that it is possible to do research as a hobby. It isn't always easy to find the time (and motivation!) but leaving academia needn't be the end of one's research career. So for people wondering whether academia will ultimately take too high a toll on their (mental) health, the decision to leave academia needn't necessarily equate with the decision to stop doing research; it's just that a different balance in one's life has to be found!”

[If you speak German or trust Google translate, the FAZ blogs also wrote about this.]

Friday, June 12, 2015

Where are we on the road to quantum gravity?

Damned if I know! But I got to ask some questions to Lee Smolin which he kindly replied to, and you can read his answers over at Starts with a Bang. If you’re a string theorist you don’t have to read it of course because we already know you’ll hate it.

But I would be acting out of character if not having an answer to the question posed in the title did prevent me from going on and distributing opinions, so here we go. On my postdoctoral path through institutions I’ve passed by string theory and loop quantum gravity, and after some closer inspection stayed at a distance from both because I wanted to do physics and not math. I wanted to describe something in the real world and not spend my days proving convergence theorems or doing stability analyses of imaginary things. I wanted to do something meaningful with my life, and I was – still am – deeply disturbed by how detached quantum gravity is from experiment. So detached in fact one has to wonder if it’s science at all.

That’s why I’ve worked for years on quantum gravity phenomenology. The recent developments in string theory to apply the AdS/CFT duality to the description of strongly coupled systems are another way to make this contact to reality, but then we were talking about quantum gravity.

For me the most interesting theoretical developments in quantum gravity are the ones Lee hasn’t mentioned. There are various emergent gravity scenarios and though I don’t find any of them too convincing, there might be something to the idea that gravity is a statistical effect. And then there is Achim Kempf’s spectral geometry that for all I can see would just fit together very nicely with causal sets. But yeah, there are like two people in the world working on this and they’re flying below the pop sci radar. So you’d probably never have heard of them if it wasn’t for my awesome blog, so listen: Have an eye on Achim Kempf and Raffael Sorkin, they’re both brilliant and their work is totally underappreciated.

Personally, I am not so secretly convinced that the actual reason we haven’t yet figured out which theory of quantum gravity describes our universe is that we haven’t understood quantization. The so-called “problem of time”, the past hypothesis, the measurement problem, the cosmological constant – all this signals to me the problem isn’t gravity, the problem is the quantization prescription itself. And what a strange procedure this is, to take a classical theory and then quantize and second quantize it to obtain something more fundamental. How do we know this procedure isn’t scale dependent? How do we know it works the same at the Planck scale as in our labs? We don’t. Unfortunately, this topic rests at the intersection of quantum gravity and quantum foundations and is dismissed by both sides, unless you count my own small contribution. It’s a research area with only one paper!

Having said that, I found Lee’s answers interesting because I understand better now the optimism behind the quote from his 2001 book, that predicted we’d know the theory of quantum gravity by 2015.

I originally studied mathematics, and it just so happened that the first journal club I ever attended, in '97 or '98, was held by a professor for mathematical physics on the topic of Ashtekar’s variables. I knew some General Relativity and was just taking a class on quantum field theory, and this fit in nicely. It was somewhat over my head but basically the same math and not too difficult to follow. And it all seemed to make much sense! I switched from math to physics and in fact for several years to come I lived under the impression that gravity had been quantized and it wouldn’t take long until somebody calculated exactly what is inside a black hole and how the big bang works. That, however, never happened. And here we are in 2015, still looking to answer the same questions.

I’ll restrain from making a prediction because predicting when we’ll know the theory for quantum gravity is more difficult than finding it in the first place ;o)

Tuesday, June 09, 2015

What is cosmological hysteresis?

Last week there were two new papers on the arXiv discussing an effect dubbed “cosmological hysteresis,” which, so the authors argue, would make cyclic cosmological models viable alternatives to inflation

Hysteresis is an effect more commonly known from solid state materials, when a material doesn’t return to its original state together with an external control parameter. The textbook example is a ferromagnet’s average magnetization whose orientation can be changed by applying an external magnetic field. Turn up the magnetic field and it drags with it the magnetization, but turn back the magnetic field and the magnetization lags behind. So for the same value of the magnetic field you can have two different values of magnetization, depending on whether you were increasing or decreasing the field.

Hysteresis in ferromagnets. Image credit: Hyperphysics.

This hysteresis is accompanied by the loss of energy into the material in form of heat, because one constantly has to work to turn the magnets, and in this cycle entropy increases. In fact I don’t know any example of hysteresis in which entropy does not increase.

What does this have to do with cosmology? Well, nothing really, except that it’s an analogy that the authors of the mentioned papers are drawing upon. They argue that a simple type of cyclic cosmological model with a scalar field has a similar type of hysteresis, but one which is not accompanied by entropy increase, and that this serves to make cyclic cosmology more appealing.

Cyclic cosmological models have been around since the early days of General Relativity. In such a model, each phase of expansion of the universe ends in a turnaround and subsequent contraction, followed by a bounce and a new phase of expansion. These models are periodic, but note that this doesn’t necessarily mean they are time-reversal invariant. (A sine curve is periodic and has a time-reversal invariance around the maxima and minima. A saw-tooth is periodic but not invariant under time-reversal.)

In any case, that the behavior of a system isn’t time-reversal invariant doesn’t mean its time evolution cannot be inverted. It just means it isn’t symmetric under this inversion. To our best present knowledge the time dependence of all existing systems can be inverted – theoretically. Practically this is normally not possible because such an inversion would require an extremely precise choice of initial conditions. It is easy enough to mix flour, sugar, and eggs to make a dough, but you could mix until we run out of oil (and Roberts) and would never see an egg separate from the sugar again.

Statistical mechanics quantifies the improbability in succeeding to reverse a time-dependence by the increase of entropy. A system is likely to develop into a state of higher entropy, but, except for fluctuations that are normally tiny, entropy doesn’t decrease because this is exceedingly unlikely to happen. That’s the second law of thermodynamics.

This second law of thermodynamics is also the main problem with cyclic cosmologies. Since entropy increases throughout each cycle, the next cycle cannot start from the same initial conditions. Entropy gradually builds up, and this is generally a bad thing if you want conditions in which life can develop because for that you need to maintain some type of order. The major obstacle in making convincing cyclic models is therefore to find a way to indeed reproduce the initial conditions. I don’t really know of a good solution to this. The maybe most appealing idea is that the next cycle isn’t actually initiated by the whole universe but only a small part of it, leading to “baby universe” scenarios. I toyed for some while with the idea to couple two universes that periodically push entropy back and forth, but this ended up in my dead drafts drawer, and ever since I’ve disliked cyclic cosmologies.

In the mentioned papers the authors observe that a cosmology coupled to a scalar field has two different attractors (solutions towards which the field develops) depending on whether the universe is expanding or contracting. In the expanding phase, a scalar field with a potential gets decelerated and slows down, which makes its behavior stable under perturbations because these get damped. In the contracting phase, the field gets accelerated instead, continues to grow, and becomes very sensitive to smallest perturbations because they get blown up. The time-dependence of this system is still reversible in theory, but not in practice for the same reason that you can’t unmix your dough. Since the unstable period tends to be very sensitive to smallest mistakes, you will not be able to reverse it perfectly.

Figure 2 from arXiv:1506.01247. Φ is the scalar field, a dot the time derivative. During expansion of the universe the field evolves along the arrows on the curves in the left figure. Different curves correspond to different initial conditions, but they all converge together. During contraction the field evolves along the curves as shown in the right figure, where the curves diverge with an increasing value of the field.


For the cyclic model this means that basically noise from small fluctuations builds up through each cycle. After the turnaround, the field will not exactly retrace its path but diverge from the time-reversal. That is why they refer to it as hysteresis.

Figure 1 from arXiv:1506.02260. a is the scale factor of the universe - a larger a means a larger universe, and w encodes the equation of state of the scalar field.  In this scenario, the scalar field doesn’t retrace the path it came after turnaround.


It also has the effect that the next cycle starts from a different initial condition (provided there is some mechanism that allows the universe to bounce, which necessitates some quantum gravity theory). In the studies in the paper, the noise is mimicked in a computer simulation by some small random number that is added to the field. More realistically you might think of it as quantum fluctuations.

Now, this all sounds plausible to me. There are two things though I don’t understand about this.

First, I don’t think it’s justified to say that in this case entropy doesn’t increase. The problem of having to finetune initial conditions to reverse the process demonstrates instead that entropy does increase - this is essentially the very definition of entropy increase! Second, and more important, I have no idea why that would makes cyclic cosmological models more interesting because they are just demonstrating exactly that it’s not possible to make these models periodic and one doesn’t return to anywhere close by the initial state.

In summary, cosmological hysteresis seems to exist under quite general circumstances, so there you have another cool word for your next dinner party that will make you sound very sciency. However, I don’t see how that effect makes cyclic cosmologies more appealing. What I learned from the papers though is that this very simple scalar field model already captures the increase in entropy through the cycles, which had not previously been clear to me. In fact this model is so instructive that maybe I should open that drawer with the dead drafts again...

Thursday, June 04, 2015

Social Media for Scientists

I recently gave a seminar on the use of social media for scientists at an internal event at Stockholm University. It was an opportunity to collect my thoughts, and also to summarize what shortcomings the presently available platforms have. It would be odd if I didn’t share this with you, wouldn’t it?

There are many uses of social media, but three of them are particularly important for science: Networking, Communication, and Outreach.

Networking can best be described as the making and maintain of contacts and the exchange of information. It blends into science communication, which is more generally about discussing your own or others’ research, with your community or with the public. And then there is public outreach, which has a broader aim because you may also want to draw attention to your institution or yourself, or to generally get people engaged in science.


Networking is really unavoidable if you want to work in science today – and you are almost certainly doing it already. Communication is essential for research, and so I think using social media to this end is part of being a good scientist. And I strongly encourage you to try if you like public outreach because of its many benefits. I don’t think every scientist must engage in public outreach; in the first line scientists should do science. But it can be very rewarding and helpful to your science too, so if you have both the time and the interest, you should definitely consider it.

A lot of scientists I know shy away from using social media for no good reason and seem to believe twitter and facebook are somehow not intellectual enough. Or maybe they mistrust their own abilities to withstand the temptation of cat videos. To me twitter, facebook, blogger and, to a lesser extent, Google plus and ResearchGate are simply tools that help me to stay up to date, keep in touch with colleagues, discuss science, get feedback and advice, and share my own research.

There are many other reasons to use social media, but they are all driven by the underlying changes in the communities: We are more people in science today than ever before, collaborations are becoming more international, there are more and more papers being published. Social media is a good way to manage this. If you’re not using social media, you are putting yourself at a disadvantage, it’s as simple as that.

I have the slides of my talk online here, where I have some remarks on the social media platforms presently most widely used by scientists: Twitter, Facebook, ResearchGate, LinkedIn and Google+. In physics in particular there are also the PhysicsForums and the Physics Stack Exchage which are well frequented and can be really useful to ask questions, and give or get answers. These services differ somewhat in their aim and use, so if you are new to this you might want to check these out and see what suits you best.

The existing platforms leave me wanting for the following reasons:

  • None of the existing social media sites covers the spectrum from professional to personal contacts.

    Presently we either have pages like LinkedIn and ResearchGate that focus entirely on job experience and skills and, in the case of ReseachGate, publications. Or we have sites like Facebook and Google+ where you don’t have this at all. But I know most of my colleagues personally, and I am also interested to hear what is going on in their life beyond the publication record or changes of affiliation. For me, like for many scientists I know, work life and private life blur together. Maybe it’s the shared experience of losing friends during the postdoc time that creates these ties, be that as it may, it’s a reality of research.

    I mostly use facebook, because there we also talk about the human side of science, the frustration with peer review, the nuisance of writing proposals, the inevitable rejections, the difficulties of balancing work with family, travel stress, conference experiences, and so on. To me this is part of life as a scientist. Every one of us goes through difficult times every one and then, and using social media is a way to both get and give support. So even if I were using a site like ResearchGate, I would still use other sites in addition to this.
  • None of the existing sites integrates a useful archiving function.

    This is something I really don’t understand: Why isn’t there a way to tag posts on either of these platforms with keywords or move them into folders for your own reference? On facebook you can now at least do a keyword search on your timeline, but it is working badly. On twitter too you can search your posts, though you have to use a third party service for that. Still, I and others I talked to, often get frustrated not being able to find a particular post or reference or comment.

    There are of course apps like for example Evernote that allow you to basically archive anything you want in categories of choice with keywords, and to a lesser extent you can do this on Feedly too. But then if you archive a reference, you will not have the discussion about it in the same place.
  • The professional sites are too public.

    Michael Nielsen in his book “Reinventing Discovery” has a charming analogy in which he describes a scientist with an unfinished idea as someone owning only one shoe, looking for a match without wanting to show the shoe to anybody. Nielsen describes how awkward and hesitant scientists can be before they start talking about their lonely shoes, and I find much truth in this analogy. It’s all well and fine if you have a question and ask it at the Stack Exchange or ResearchGate or facebook or wherever. But that really isn’t how it works if you are looking for a collaborator.

    To begin with you might not know exactly what the question is, or what you are looking for isn’t somebody explaining how to do a calculation but somebody interested enough to actually do it in exchange for being coauthor. There is also the prevalent academic paranoia of getting scooped. Especially in fields where competition is high, people don’t normally go around and publicly distribute their half-done research projects. And then, maybe most importantly, both the person asking and the person answering might have some misunderstandings and they might be afraid of their mistakes being publicly documented. Michael in his book lays out a vision for the matching of shoes, and I think what is really important in this is to allow scientists to find others with similar interests and then give them a private space to discuss off the record.
  • None if the existing services has an integration of bibliometric or scientometric data.

    There is plenty of data about coauthor networks within communities and also their evolution over time, mutually quoted references, and various ways to visualize research topics and their relation. I think this is relevant for scientists to know how many other people are working in their area, how it connects to other fields, who is working on making these connections, and how the field develops. Research has shown that many breakthroughs in science originate at the intersection of fields that weren’t previously known to be related, so these maps are interesting from a purely scientific perspective already. But they are also of personal use because they give researchers an idea about how their own research fits into the larger picture.

    Here is for example an interesting paper about pivot points in string theory. I know, the visualization isn’t all that great, but note that the paper is more than a decade old!
I could go on about how I hope that webinars will become better integrated and that conferences will come to have a better online presence since I find it extremely annoying and cumbersome that every institution is using their own registration system, but I can see that for what the software solution is concerned these are difficult to address by any one service.

What social media do you use to discuss science and how has your experience been with that?