WELCOME TO MY TECHNICAL UNIVERSE the physics of cognitive systems

I used to have a description of each of my papers on this page, but it got very boring to read as the numbers grew, so I moved most of it to here. After graduate work on the role of atomic and molecular chemistry in cosmic reionization, I have mainly focused my research on issues related to constraining cosmological models. A suite of papers developed methods for analyzing cosmological data sets and applied them to various CMB experiments and galaxy redshift surveys, often in collaboration with the experimentalists who had taken the data. Another series of papers tackled various “dirty laundry” issues such as microwave foregrounds and mass-to-light bias. Other papers like this one develop and apply techniques for clarifying the big picture in cosmology: comparing and combining diverse cosmological probes, cross-checking for consistency and constraining cosmological models and their free parameters. (The difference between cosmology and ice hockey is that I don’t get penalized for cross-checking…) My main current research interest is cosmology theory and phenomenology. I’m particularly enthusiastic about the prospects of comparing and combining current and upcoming data on CMB, LSS, galaxy clusters, lensing, LyA forest clustering, SN 1, 21 cm tomography, etc. to raise the ambition level beyond the current cosmological parameter game, testing rather than assuming the underlying physics. This paper contains my battle cry. I also retain a strong interest in low-level nuts-and-bolts analysis and interpretation of data, firmly believing that the devil is in the details, and am actively working on neutral hydrogen tomography theory, experiment and data analysis for our Omniscope project, which you can read all about here.

OTHER RESEARCH: SIDE INTERESTS Early galaxy formation and the end of the cosmic dark ages One of the main challenges in modern cosmology is to quantify how small density fluctuations at the recombination epoch at redshift around z=1000 evolved into the galaxies and the large-scale structure we observe in the universe today. My Ph.D. thesis with Joe Silk focused on ways of probing the interesting intermediate epoch. The emphasis was on the role played by non-linear feedback, where a small fraction of matter forming luminous objects such as stars or QSO’s can inject enough energy into their surrounding to radically alter subsequent events. We know that the intergalactic medium (IGM) was reionized at some point, but the details of when and how this occurred remain open. The absence of a Gunn-Peterson trough in the spectra of high-redshift quasars suggests that it happened before z=5, which could be achieved through supernova driven winds from early galaxies. Photoionization was thought to be able to partially reionize the IGM much earlier, perhaps early enough to affect the cosmic microwave background (CMB) fluctuations, especially in an open universe. However, extremely early reionization is ruled out by the COBE FIRAS constraints on the Compton y-distortion. To make predictions for when the first objects formed and how big they were, you need to worry about something I hate: molecules. Although I was so fed up with rate discrepancies in the molecule literature that I verged on making myself a Ghostbuster-style T-shirt reading “MOLECULES – JUST SAY NO”, the irony is that my molecule paper that I hated so much ended up being one of my most cited ones. Whereas others that I had lots of fun with went largely unnoticed…

Math problemsI’m also interested in physics-related mathematics problems in general. For instance, if you don’t believe that part of a constrained elliptic metal sheet may bend towards you if you try to push it away, you are making the same mistake that the famous mathematician Hadamard once did.

WELCOME TO MY TECHNICAL UNIVERSE I love working on projects that involve cool questions, great state-of-the-art data and powerful physical/mathematical/computational tools. During my first quarter-century as a physics researcher, this criterion has lead me to work mainly on cosmology and quantum information. Although I’m continuing my cosmology work with the HERA collaboration, the main focus of my current research is on the physics of cognitive systems: using physics-based techniques to understand how brains works and to build better AI (artificial intelligence) systems. If you’re interested in working with me on these topics, please let me know, as I’m potentially looking for new students and postdocs (see requirements). I’m fortunate to have collaborators who generously share amazing neuroscience data with my group, including Ed Boyden, Emery Brown and Tomaso Poggio at MIT and Gabriel Kreimann at Harvard, and to have such inspiring colleagues here in our MIT Physics Department in our new division studying the physics of living systems. I’ve been pleasantly surprised by how many data analysis techniques I’ve developed for cosmology can be adapted to neuroscience data as well. There’s clearly no shortage of fascinating questions surrounding the physics of intelligence, and there’s no shortage of powerful theoretical tools either, ranging from neural network physics and non-equilibrium statistical mechanics to information theory, the renormalization group and deep learning. Intriguingly and surprisingly, there’s a duality between the last two. I recently helped organize conferences on the physics of information and artificial intelligence. I’m very interested in the question of how to model an observer in physics, and if simple necessary conditions for a physical system being a conscious observer can help explain how the familiar object hierarchy of the classical world emerges from the raw mathematical formalism of quantum mechanics. Here’s a taxonomy of proposed consciousness measures. Here’s a TEDx-talk of mine about the physics of consciousness. Here’s an intriguing connection between critical behavior in magnets, language, music and DNA. In older work of mine on the physics of the brain, I showed that neuron decoherence is way too fast for the brain to be a quantum computer. However, it’s nonetheless interesting to study our brains as quantum systems, to better understand why they perceives the sort of classical world that they do. For example, why do we feel that we live in real space rather than Fourier space, even though both are equally valid quantum descriptions related by a unitary transformation?

Quantum information My work on the physics of cognitive systems is a natural outgrowth of my long-standing interest in quantum information, both for enabling new technologies such as quantum computing and for shedding new light on how the world fundamentally works. For example, I’m interested in how the second law of thermodynamics can be generalized to explain how the entropy of a system typically decreases while you observe a system and increases while you don’t, and how this can help explain how inflation causes the emergence of an arrow of time. When you don’t observe an interacting system, you can get decoherence, which I had the joy of rediscovering as a grad student – if you’d like to know more about what this is, check out my article in with John Archibald Wheeler in Scientific American here. I’m interested in decoherence both for its quantitative implications for quantum computing etc and for its philosophical implications for the interpretation of quantum mechanics. For much more on this wackier side of mine, click the banana icon above. Since macroscopic systems are virtually impossible to isolate from their surroundings, a number of quantitative predictions can be made for how their wavefunction will appear to collapse, in good agreement with what we in fact observe. Similar quantitative predictions can be made for models of heat baths, showing how the effects of the environment cause the familiar entropy increase and apparent directionality of time. Intriguingly, decoherence can also be shown to produce generalized coherent states, indicating that these are not merely a useful approximation, but indeed a type of quantum states that we should expect nature to be full of. All these changes in the quantum density matrix can in principle be measured experimentally, with phases and all.

Cosmology My cosmology research has been focused on precision cosmology, e.g., combining theoretical work with new measurements to place sharp constraints on cosmological models and their free parameters. (Skip to here if you already know all this.) Spectacular new measurements are providing powerful tools for this:

So far, I’ve worked mainly on CMB, LSS and 21 cm tomography, with some papers involving lensing, SN Ia and LyAF as well. Why do I find cosmology exciting?(Even if you don’t find cosmology exciting, there are good reasons why you should support physics research.)

  1. There are some very basic questions that still haven’t been answered. For instance,
    • Is really only 5% of our universe made of atoms? So it seems, but what precisely is the weird “dark matter” and “dark energy” that make up the rest?
    • Will the Universe expand forever or end in a cataclysmic crunch or big rip? The smart money is now on the first option, but the jury is still out.
    • How did it all begin, or did it? This is linked to particle physics and unifying gravity with quantum theory.
    • Are there infinitely many other stars, or does space connect back on itself? Most of my colleagues assume it is infinite and the data supports this, but we don’t know yet.
  2. Thanks to an avalanche of great new data, driven by advances in satellite, detector and computer technology, we may be only years away from answering some of these questions.

Satellites Rock! Since our atmosphere messes up most electromagnetic waves coming from space (the main exceptions being radio waves and visible light), the advent of satellites has revolutionized our ability to photograph the Universe in microwaves, infrared light, ultraviolet light, X-rays and gamma rays. New low-temperature detectors have greatly improved what can be done from the ground as well, and the the computer revolution has enabled us to gather and process huge data quantities, doing research that would have been unthinkable twenty years ago. This data avalanche has transformed cosmology from being a mainly theoretical field, occasionally ridiculed as speculative and flaky, into a data-driven quantitative field where competing theories can be tested with ever-increasing precision. I find CMB, LSS, lensing, SN Ia, LyAF, clusters and BBN to be very exciting areas, since they are all being transformed by new high-precision measurements as described below. Since each of them measures different but related aspects of the Universe, they both complement each other and allow lots of cross-checks. What are these cosmological parameters?Cosmic matter budget In our standard cosmological model, the Universe was once in an extremely dense and hot state, where things were essentially the same everywhere in space, with only tiny fluctuations (at the level of 0.00001) in the density. As the Universe expanded and cooled, gravitational instability caused these these fluctuations to grow into the galaxies and the large-scale structure that we observe in the Universe today. To calculate the details of this, we need to know about a dozen numbers, so-called cosmological parameters. Most of these parameters specify the cosmic matter budget, i.e., what the density of the Universe is made up of – the amounts of the following ingredients:

  • Baryons – the kind of particles that you and I and all the chemical elements we learned about in school are madeof : protons & neutrons. Baryons appear to make up only about 5% of all stuff in the Universe.
  • Photons – the particles that make uplight. Their density is the best measured one on this list.
  • Massive neutrinos – neutrinos are very shy particles. They are known to exist, and now at least two of the three or more kinds are known to have mass.
  • Cold dark matter – unseen mystery particles widely believed to exist. There seems to be about five times more of this strange stuff than baryons, making us a minority in the Universe.
  • Curvature – if the total density differs from a certain critical value, space will be curved. Sufficiently high density would make space be finite, curving back on itself like the 3D surface of a 4D hypersphere.
  • Dark energy – little more than a fancy name our ignorance of what seems to make up abouttwo thirdsof the matter budget. One popular candidates is a “Cosmological constant”, a.k.a. Lambda, which Einstein invented and then later called his greatest blunder. Other candidates are more complicated modifications toEinsteinstheory of Gravity as well as energy fields known as “quintessence”. Dark energy causes gravitational repulsion in place of attraction. Einstein invented it and called it his greatest mistake, but combining new SN Ia and CMB data indicates that we might be living with Lambda after all.

Then there are a few parameters describing those tiny fluctuations in the early Universe; exactly how tiny they were, the ratio of fluctuations on small and large scales, the relative phase of fluctuations in the different types of matter, etc. Accurately measuring these parameters would test the most popular theory for the origin of these wiggles, known as inflation, and teach us about physics at much higher energies than are accessible with particle accelerator experiments. Finally, there are a some parameters that Dick Bond, would refer to as “gastrophysics”, since they involve gas and other ghastly stuff. One example is the extent to which feedback from the first galaxies have affected the CMB fluctuations via reionization. Another example is bias, the relation between fluctuations in the matter density and the number of galaxies.One of my main current interests is using the avalanche of new data to raise the ambition level beyond cosmological parameters, testing rather than assuming the underlying physics. My battle cry is published here with nuts and bolts details here and here. The cosmic toolboxHere is a brief summary of some key cosmological observables and what they can teach us about cosmological parameters.

Photos of the cosmic microwave background (CMB) radiation like the one to the left show us the most distant object we can see: a hot, opaque wall of glowing hydrogen plasma about 14 billion light years away. Why is it there? Well, as we look further away, we’re seeing things that happened longer ago, since it’s taken the light a long time to get here. We see the Sun as it was eight minutes ago, the Andromeda galaxy the way it was a few million years ago and this glowing surface as it was just 400,000 years after the Big Bang. We can see that far back since the hydrogen gas that fills intergalactic space is transparent, but we can’t see further, since earlier the hydrogen was so hot that it was an ionized plasma, opaque to light, looking like a hot glowing wall just like the surface of the Sun. The detailed patterns of hotter and colder spots on this wall constitute a goldmine of information about the cosmological parameters mentioned above. If you are a newcomer and want an introduction to CMB fluctuations and what we can learn from them, I’ve written a review here. If you don’t have a physics background, I recommend the on-line tutorials by Wayne Hu and Ned Wright. Two new promising CMB fronts are opening up — CMB polarization and arcminute scale CMB, and are likely to keep the CMB field lively for at leastr another decade. Hydrogen tomography Mapping our universe in 3D by imaging the redshifted 21 cm line from neutral hydrogen has the potential to overtake the cosmic microwave background as our most powerful cosmological probe, because it can map a much larger volume of our Universe, shedding new light on the epoch of reionization, inflation, dark matter, dark energy, and neutrino masses. For this reason, my group built MITEoR, a pathfinder low-frequency radio interferometer whose goal was to test technologies that greatly reduce the cost of such 3D mapping for a given sensitivity. MITEoR accomplished this by using massive baseline redundancy both to enable automated precision calibration and to cut the correlator cost scaling from N2 to N log N, where N is the number of antennas. The success of MITEoR with its 64 dual-polarization elements bodes well for the more ambitious HERA project, which incorporates many of the technologies MITEoR tested using dramatically larger collecting area

.Galaxy cluster Large-scale structure: 3D mapping of the Universe with galaxy redshift surveys offers another window on dark matter properties, through its gravitational effects on galaxy clustering. This field is currently being transformed by everr larger Galaxy Redshift Survey. I’ve had lots of fun working with my colleagues on the Sloan Digital Sky Survey (SDSS) to carefully analyze the gargantuan galaxy maps and work out what they tell us about our cosmic composition, origins and ultimate fate. The abundance of galaxy clusters, the largest gravitationally bound and equilibrated blobs of stuff in the Universe, is a very sensitive probe of both the cosmic expansion history and the growth of matter clustering. Many powerful cluster finding techniques are contributing to rapid growth in the number of known clusters and our knowledge of their properties: identifying them in 3D galaxy surveys, seeing their hot gas as hot spots in X-ray maps or cold spots in microwave maps (the so-called SZ-effect) or spotting their gravitational effects with gravitational lensing.Gravitational lensing Yet another probe of dark matter is offered by gravitational lensing, whereby its gravitational pull bends light rays and distorts images of distant objects. The first large-scale detections of this effect were reported by four groups (astro-ph/0002500, 0003008, 0003014, 0003338) in the year 2000, and I anticipate making heavy use of such measurements as they continue to improve, partly in collaboration with Bhuvnesh Jain at Penn. Lensing is ultimately as promising as CMB and is free from the murky bias issues plaguing LSS and LyAF measurements, since it probes the matter density directly via its gravitational pull. I’ve also dabbled some in the stronger lensing effects caused by galaxy cores, which offer additional insights into the detailed nature of the dark matter.Supernovae Ia: Supernovae If a white dwarf (the corpse of a burned-out low-mass star like our Sun) orbits another dying star, it may gradually steal its gas and exceed the maximum mass with which it can be stable. This makes it collapse under its own weight and blow up in a cataclysmic explosion called a supernova of type Ia. Since all of these cosmic bombs weigh the same when they go off (about 1.4 solar masses, the so-called Chandrasekhar mass), they all release roughly the same amount of energy – and a more detailed calibration of this energy is possible by measuring how fast it dims, making it the best “standard candle” visible at cosmological distances. The supernova cosmology project and the high z SN search team mapped out how bright SN Ia looked at different redshifts found the first evidence in 1998 that the expansion of the Universe was accelerating. This approach can ultimately provide a direct measurement of the density of the Universe as a function of time, helping unravel the nature of dark energy – I hope the SNAP project or one of its competitores gets funded. The image to the left resulted from a different type of supernova, but I couldn’t resist showing it anyway..

.Lyman Alpha Forest The so-called Lyman Alpha Forest, cosmic gas clouds backlit by quasars, offers yet another new and exciting probe of how dark has clumped ordinary matter together, and is sensitive to an epoch when the Universe was merely 10-20% of its present age. Although relating the measured absorption to the densities of gas and dark matter involves some complications, it completely circumvents the Pandora’s of galaxy biasing. Cosmic observations are rapidly advancing on many other fronts as well, e.g., with direct measurements of the cosmic expansion rate and the cosmic baryon fraction.