Site recommendation: Inside Science

A brief post in a busy time:  If you like well-written, even-handed journalistic discussions of science, I strongly recommend checking out Inside Science, an editorially independent, non-profit science news service affiliated with the American Institute of Physics.   The writing is engaging and of consistently high quality, and I'm glad it's supported by underwriters so that it's not dependent on clicks/ad revenue.

What is Intel's Optane memory?

Intel has developed a new product, dubbed Optane, that is a memory technology that supposedly combined the speed of conventional DRAM with the nonvolatility of flash memory.   It would appear that the devices function as a form of "3d crosspoint" memory, where the functionality is all in a blob of material at the crossing point between two wires (a bit line and a word line).  Depending on some particular voltage pulse applied to the junction, the blob of material either has a high electrical resistance or a low electrical resistance, corresponding to the two different states of a binary bit.   The upsides here are that information is stored in a material property rather than as charge (making it non-volatile, probably radiation hard, probably uninfluenced by magnetic fields, etc.), and the devices can be packed very densely, since they need fewer transistors, etc. than conventional DRAM.

There are multiple different mechanisms to achieve that kind of response from the mysterious blob of material.  For example, you can have a material that changes structural phase under current flow, between two different structures with different electrical properties.  You can have a material where some redox chemistry takes place, switching on or off a conductive filament.  You can have a material where other redox chemistry takes place, along with the migration of oxygen vacancies, to create or destroy a conductive filament (as in the HP implementation of memristors, which I've written about before).  You could use magnetic data storage of some sort, with spin transfer torque driving switching of some giant magnetoresistive or tunneling magnetoresistive device.

Breathless articles like this one this week make some pretty bold claims for Optane's performance.  However, no one seems to know what it is.  There has been speculation.  Intel's CEO says it's based on actual bulk changes in the electrical properties of the mysterious material.  Intel categorically denies that it is based on phase changes, "memristor" approaches, or spin transfer torque.  

Well, now that they are actually shipping chips, it's only a very short matter of time before someone cuts one open and reverse-engineers what is actually in there.  So, we have only a little while to speculate wildly or place bets.  Please go for it in the comments, or chime in if you have an informed perspective!  Personally, I suspect it really is some form of bias-driven chemical alteration of material, whether this is called "memristor" in the HP sense of the word or not.  (Note that something rather analogous happened back when IBM and Intel switched to using "high-k" dielectrics in transistors.  They wouldn't say what material they'd come up with, and in the end it turned out to be (most commonly) hafnium oxynitrides.)

Hysteresis in science and engineering policy

I have tried hard to avoid political tracts on this blog, because I don't think that's why people necessarily want to read here.  Political flamewars in the comments or loss of readers over differences of opinion are not outcomes I want.  The recent proposed budget from the White House, however, inspires some observations.  (I know the President's suggested budget is only the very beginning of the budgetary process, but it does tell you something about the administration priorities.)

The second law of thermodynamics tell us that some macroscopic processes tend to run only one direction.  It's easier to disperse a drop of ink in a glass of water than to somehow reconstitute the drop of ink once the glass has been stirred.  

In general, the response of a system to some input (say the response of a ferromagnet to an applied magnetic field, or the deformation of a blob of silly putty in response to an applied stress) can depend on the history of the material.  Taking the input from A to B and back to A doesn't necessarily return the system to its original state.  Cycling the input and ending up with a looping trajectory of the system in response because of that history dependence is called hysteresis.  This happens because there is some inherent time scale for the system to respond to inputs, and if it can't keep up, there is lag.

The proposed budget would make sweeping changes to programs and efforts that, in some cases, took decades to put in place.   Drastically reducing the size and scope of federal agencies is not something that can simply be undone by the next Congress or the next President.  Cutting 20% of NIH or 17% of DOE Office of Science would have ripple effects for many years, and anyone who has worked in a large institution knows that big cuts are almost never restored.   Expertise at EPA and NOAA can't just be rebuilt once eliminated.  

People can have legitimate discussions and differences of opinion about the role of the government and what it should be funding.  However, everyone should recognize that these are serious decisions, many of which are irreversible in practical terms.   Acting otherwise is irresponsible and foolish.

APS March Meeting 2017 Day 3 - updated w/ guest post!

Hello readers - I have travel plans such that I have to leave the APS meeting after lunch today.  That means I will miss the big Kavli Symposium session.  If someone out there would like to offer to write up a bit about those talks, please email me or comment below, and I'd be happy to give someone a guest post on this.

Update:  One of my readers was able to attend the first two talks of the Kavli Symposium, by Duncan Haldane and Michael Kosterlitz, two of this year's Nobel laureates.  Here are his comments.  If anyone has observations about the remaining talks in the symposium, please feel free to email me or post in the comments below.

I basically ran from the Buckley Prize talk by Alexei Kitaev down the big hall where Duncan Haldane was preparing to talk.  When I got there it was packed full but I managed to squeeze into a seat in the middle section.  I sighted my postdoc near the back of the first section; he later told me he’d arrived 35 minutes early to get that seat.

I felt Haldane’s talk was remarkably clear and simple given the rarified nature of the physics behind it.  He pointed out that condensed matter physics really changed starting in the 1980’s, and conceptually now is much different than the conventional picture  presented in books like Ashcroft and Mermin’s Solid State Physics that many of us learned from as students.  One prevailing idea leading up to that time was that changes in the ground state must always be related to changes in symmetry.  Haldane’s paper on antiferromagnetic Heisenberg spin chains showed that the ground state properties of the chains were drastically different depending on whether  the spin at each site is integer (S=1,2,3,…) or half-integer (S=1/2, 3/2, 5/2 …) , despite the fact that the Hamiltonian has the same spherical symmetry for any value of S.  This we now understand on the basis of the topological classifications of the systems.  Many of these topological classifications were later systematically worked out by Xiao-Gang Wen who shared this year’s Buckley prize with Alexei Kitaev. Haldane flashed a link to his original manuscript on spin chains which he has posted on arXiv.org as https://arxiv.org/abs/1612.00076 , and which he noted was “rejected by many journals”.  He was also amused or bemused or maybe both by the fact that people referred to his ideas as “Haldane’s conjecture” rather than recognizing that he’d solved the problem.  He noted that once one understands that the topological classification determines many of the important properties it is obvious that simplified “toy models” can give deep insight into the underlying physics of all systems in the same class.  In this regard he singled out the AKLT model, which revealed how finite chains of spin S=1 have effective S=1/2 degrees of freedom associated with each end.  These are entangled with each other no matter how long the finite chain – a remarkable demonstration of quantum entanglement over a long distance.  This also is a simple example of the special nature of surface states or excitations in topological systems. 

Kosterlitz began by pointing out that the Nobel prize was effectively awarded for work on two distinct aspects of topology in condensed matter, and both of these involved David Thouless which led to his being awarded one-half of the prize, with the other half shared by Kosterlitz and Haldane.  He then relayed a bit about his own story: he started as a high energy physicist, and apparently did not get offered the position he wanted at CERN so he ended up at Birmingham, which turned out to be remarkably fortuitous.  There he teamed with Thouless and gradually switched his interests to condensed matter physics.  They wanted to understand data suggesting that quasi-two-dimensional films of liquid helium seemed to show a phase transition despite the expectation that this should not be possible.  He then gave a very professorial exposition of the Kosterlitz-Thouless (K-T) transition, starting with the physics of vortices, and how their mutual interactions involve a potential that depends on the logarithm of the distance.  The results point to a non-zero temperature above which the free energy favors free vortices and below which vortex-anti vortex pairs are bound. He then pointed out how this is relevant to a wide variety of two dimensional systems, including xy magnets, and also the melting of two-dimensional crystals in which two K-T transitions occur corresponding respectively to the unbinding of dislocations and disclinations.  

I greatly enjoyed both of these talks, especially since I have experimentally researched both spin chains and two-dimensional melting at different times in my career. 

APS March Meeting 2017 Day 2

Some highlights from day 2 (though I spent quite a bit of time talking with colleagues and collaborators):

Harold Hwang of Stanford gave a very nice talk about oxide materials, with two main parts.  First, he spoke about making a hot electron (metal base) transistor  (pdf N Mat 10, 198 (2011)) - this is a transistor device made from STO/LSMO/Nb:STO, where the LSMO layer is a metal, and the idea is to get "hot" electrons to shoot over the Schottky barrier at the STO/LSMO interface, ballistically across the metallic LSMO base, and into the STO drain.  Progress has been interesting since that paper, especially with very thin bases.  In principle such devices can be very fast. 

The second part of his talk was about trying to make free-standing ultrathin oxide layers, reminiscent of what you can see with the van der Waals materials like graphene or MoS2.  To do this, they use a layer of Sr3Al2O6 - that stuff can be grown epitaxially with pulsed laser deposition on nice oxide substrates like STO, and other oxide materials (even YBCO or superlattices) can be grown epitaxially on top of it. Sr3Al2O6 is related to the compound in Portland cement that is hygroscopic, and turns out to be water soluble (!), so that you can dissolve it and lift off the layers above it.  Very impressive.

Bharat Jalan of Minnesota spoke about growing BaSnO3 via molecular beam epitaxy.  This stuff is a semiconductor dominated by the Ba 5s band, with a low effective mass so that it tends to have pretty high mobilities.  This is an increasingly trendy new wide gap oxide semiconductor that could potentially be useful for transparent electronics.  

Ivan Bozovic of Brookhaven (and Yale) gave a very compelling talk about high temperature superconductors, specifically LSCO, based on having grown thousands of extremely high quality (as assessed by the width of the transition in penetration depth measurements) epitaxial films of varying doping concentrations.   Often people assert that the cuprates, when "overdoped", basically become more conventional BCS superconductors with a Fermi liquid normal state.  Bozovic presents very convincing evidence (from pretty much the data alone, without complex models for interpretation) that shows this is not right - that instead these materials are weird even in the overdoped regime, with systematic property variations that don't look much like conventional superconductors at all.  In the second part of his talk, he showed clear transport evidence for electronic anisotropy in the normal state of LSCO over the phase diagram, with preferred axes in the plane that vary with temperature and don't necessarily align with crystallographic axes of the material.  Neat stuff.   

Shang-Jie Yu at Maryland spoke about work on coherent optical manipulation of phonons.  In particular, previous work from this group looked at ensembles of spherical core-shell nanoparticles in solution, and found that they could excite a radial breathing vibrational mode with an optical pulse, and then measure that breathing in a time-resolved way with probe pulses.  Now they can do more complex pulse sequences to control which vibrations get excited - very cute, and it's impressive to me that this works even when working with an ensemble of particles with presumably some variation in geometry.

APS March Meeting 2017 Day 1

Some talks I saw today at the APS March Meeting in New Orleans:

John Martinis spoke about "quantum supremacy".  Quantum supremacy means achieving performance truly superior to classical situation - in Martinis' usage, the idea is to look at cross-correlations between different qubits, and compare with expectations for fully entangled/coherent systems, to assess how well you are able to set, entangle, and preserve the coherence of your quantum bits.

An optical analog:  Coherent light (laser pointer) incident on frosted glass results in a diffuse spot that is, when examined in detail, an incredibly complicated speckle pattern.  The statistics of that speckled light (correlations over different spatial regions) are very different than if you just had a defocused spot.  In his system, he is taking nine (superconducting, tunable transmon) qubits, where they can control both the coupling between neighboring bits and the energy of each bit.  They set the system in an initial state (injecting a known number of microwave photons into particular qubits); set the energies in a known but randomly selected way, turn on and off the neighbor couplings (25 ns timescale) for some number of cycles, and then look where the microwave photons end up, and take the statistics.  They find that they get good agreement with an error rate of 0.3%/qubit/cycle.  That's enough that they could conceivably do something useful.

As a demo, they use their qubits to model the Hofstadter butterfly problem - finding the energy levels of a 2d electronic system (on a hexagonal lattice, which maps to a 1d problem that they can implement w. their array of nine qubits).  They can get a nice agreement between theory and experiment.  Very impressive.  He  concluded w/ a warning not to believe all hype from qc investigators, including himself.  In general, the approach is basically brute force up to ~ 45 qubits or more (couple of hundred), to think about optimal control and feedback schemes before worrying about truly huge scaling.  The only downside to the talk was that it was in a room that was far too small for the audience.

Alex MacLeod gave a nice talk about using scanning near-field optical microscopy to study the metal-insulator transition in V2O3, as in this paper.   By performing cryogenic near-field scanning optical microscopy in ultrahigh vacuum (!), they measured scattered light from nanoscale scanning tip, giving local dielectric information (hence distinction between metal and insulator surroundings) with an effective spatial resolution that is basically the radius of curvature of the tip.   There is pattern formation at the metal-insulator transition because the two phases have different crystal structures (metal = corundum; insulator = monoclinic), and therefore the transition is a problem of constrained free energy minimization.  This generically leads to pattern formation in the mixed-phase regime.  They see a clear percolation transition in optical measurements, coinciding w/ long distance transport measurements - they really are seeing metallic domains.  Strangely, they find a temperature offset betw/ the structural transition (as seen through x-ray) vs the MIT.  The structural transition temperature is higher, and coincides with max anisotropy in the imaged patterns.  They also see pieces of persistent metallic state at low T, suggesting that some other frustration is going on to stabilize this.

Anatole von Lilienfeld of Basel gave an interesting talk about using machine learning techniques to get quantum chemistry information about small molecules faster and allegedly with better accuracy than full density functional theory calculations.  Basically you train the software on molecules that have been solved to some high degree of accuracy, parametrizing the molecules by their structure (a "Coulomb matrix" that takes into account the relative coordinates and effective charges of the ions) and/or bonding (a "bag of bonds" that takes into account two-body bonds).  Then the software can do a really good job interpolating quantum properties (HOMO-LUMO gaps, ionization potentials) of related molecules faster than you could calculate them in detail.  Impressive, but it seems like a powerful look-up table rather than providing much physical insight.

Melissa Eblen-Zayas gave a fun talk about trying to upgrade the typical advanced junior lab to include real elements of experimental design.  Best line:  "At times student frustration was palpable."

Dan Ralph gave a very compelling talk about the origins of spin-orbit torques in thin-film heterostructures.  I've written in the past about related work.   This was a particularly clear exposition, and went to new territory.  Traditionally, if you have a thin film of a heavy metal (tantalum, say), and you pass current through that film, at the upper (and lower) film surface you will accumulate spin density oriented in the plane and perpendicular to the charge current.  He made a clear argument that this is required because of the mirror symmetry properties of typical polycrystalline metal films.  However, if instead you work with a thin material with much lower symmetry (WTe2, for example) instead of the heavy metal, you can exert spin torques on adjacent magnetic overlayers as if the accumulated spin was out of the plane (which could be useful for certain device approaches).

APS March Meeting 2017

Once again, it's that time of year when somewhat absurd numbers of condensed matter (and other) physicists gather together.  This time the festivities are in New Orleans.  I'll be at the meeting tomorrow (this will be my first time attending the business meetings as a member-at-large of the Division of Condensed Matter Physics, so that should be new and different)  through Wednesday afternoon.  As in previous years, I will do my best to write up some of the interesting things I learn about.  (If you're at the meeting and you don't already have a copy, now is the perfect time to swing by the Cambridge University Press exhibit at the trade show and pick up my book :-) )

Career guidance and advice - aggregated posts

Similarly, over the years I have written several posts about (academic) career topics.  Google doesn't always pagerank these very highly (that is a form of peer review, I suppose), so here they are in one place.  Again, some should probably be rewritten and updated, but this is a start.

Advice on choosing a graduate school
Advice on choosing/finding a postdoc position
Guide to faculty searches, 2015 edition
Tenure - some advice
How to write a scientific paper
How to write a response to referees
About refereeing
How to carry on a scientific collaboration
About coauthorship
Things no one teaches you as part of your training
Lab habits and data management