Rising temperatures are curbing ocean’s capacity to store carbon

If there is anywhere for carbon dioxide to disappear in large quantities from the atmosphere, it is into the Earth’s oceans. There, huge populations of plankton can soak up carbon dioxide from surface waters and gobble it up as a part of photosynthesis, generating energy for their livelihood. When plankton die, they sink thousands of feet, taking with them the carbon that was once in the atmosphere, and stashing it in the deep ocean.

The oceans, therefore, have served as a natural sponge in removing greenhouse gases from the atmosphere, helping to offset the effects of climate change.

But now MIT climate scientists have found that the ocean’s export efficiency, or the fraction of total plankton growth that is sinking to its depths, is decreasing, due mainly to rising global temperatures.

In a new study published in the journal Limnology and Oceanography Letters, the scientists calculate that, over the past 30 years, as temperatures have risen worldwide, the amount of carbon that has been removed and stored in the deep ocean has decreased by 1.5 percent.

To put this number in perspective, each year, about 50 billion tons of new plankton flourish in the surface ocean each year, while about 6 billion tons of dead plankton sink to deeper waters. A 1.5 percent decline in export efficiency would mean that about 100 million tons of extra plankton have remained near the surface each year.

“We figured the amount of carbon that is not sinking out as a result of global temperature change is similar to the total amount of carbon emissions that the United Kingdom pumps into the atmosphere each year,” says first author B.B. Cael, a graduate student in MIT’s Department of Earth, Atmospheric and Planetary Sciences (EAPS). “If carbon is just standing in the surface ocean, it’s easier for it to end up back in the atmosphere.”

Cael’s co-authors on the paper are Kelsey Bisson of the University of California at Santa Barbara and Mick Follows, an associate professor in EAPS.

Photosynthesizers versus respirers

In 2016, the team first started looking into whether sea surface temperature has an effect on the ocean’s export efficiency. The group’s main research focus is on marine microbes, including interactions between communities, and their effects on and responses to climate change.

In studying export efficiency, the researchers identified two processes in surface ocean microbes that affect the rate at which carbon is drawn down to the deep ocean: Photosynthesizing organisms such as plankton absorb carbon dioxide from surface waters, fixing carbon into their systems; respiring organisms such as bacteria and krill take in oxygen and emit carbon dioxide into the surrounding waters.

Based on the chemistry of photosynthesis and respiration, the researchers realized that the two processes respond differently depending on temperature. Photosynthesizers grow and die relatively faster in colder environments, while respirers are relatively more active in warmer temperatures.

In 2016, the researchers developed a simple model to predict the ocean’s rate of drawing down carbon at given sea surface temperatures. Their results matched with recorded observations of the amount of carbon exported to the deep ocean.

“We had a simple way to describe how we think temperature influences export efficiency, based on this fundamental metabolic theory,” Cael says. “Now, can we use that to see how export efficiency has changed over the time period where we have good temperature records? That’s how we can estimate whether export efficiency is changing as a result of climate change.”

Out to sea

For this new paper, the researchers used the model to estimate the ocean’s export efficiency over the last three decades. Since 1982, satellites, ships, and buoys have made measurements of sea surface temperatures around the world, which scientists have averaged for each measured location and aggregated into publicly available databases.

For this study, the team used temperature measurements from three different databases, taken every month from 1982 to 2014, for locations around the world. The group used the temperature to estimate export efficiency across the global ocean for each month, based on their simple model. They traced the change in export efficiencies across the globe, over the 33-year period during which measurements were available.

They found that, worldwide, the rate at which the ocean draws down carbon has declined by 1 to 2 percent since 1982. Sea surface temperatures have increased during this period.

“People probably expected a decline in export efficiency, but the thing I find interesting is, we have a nice way to try and quantify it,” Cael says. “We’re able to estimate that over last 30 years, export efficiency has declined by 1 or 2 percent, so 1 to 2 percent less of total plankton productivity is making it out of the surface ocean, which is actually a pretty big number.”

Cael says the team’s model could potentially be applied to predict the ocean’s future as a carbon sink, though uncertainty in temperature projections makes this a much more complicated goal.

“How carbon moves around on Earth is fundamental to understanding both Earth’s biosphere and climate, and requires understanding how carbon moves through the ocean,” Cael says. “This [model] is something you could potentially apply to temperature projections, to guess how carbon will move through the Earth in the future.”

This research was supported, in part, by the National Science Foundation, the Simons Foundation, and the Gordon and Betty Moore Foundation.

de MIT News http://ift.tt/2trOr97

Researchers use Kinect to scan T. rex skull

Last year, a team of forensic dentists got authorization to perform a 3-D scan of the prized Tyrannosaurus rex skull at the Field Museum of Natural History in Chicago, in an effort to try to explain some strange holes in the jawbone.

Upon discovering that their high-resolution dental scanners couldn’t handle a jaw as big as a tyrannosaur’s, they contacted the Camera Culture group at MIT’s Media Lab, which had recently made headlines with a prototype system for producing high-resolution 3-D scans.

The prototype wasn’t ready for a job that big, however, so Camera Culture researchers used $150 in hardware and some free software to rig up a system that has since produced a 3-D scan of the entire five-foot-long T. rex skull, which a team of researchers — including dentists, anthropologists, veterinarians, and paleontologists — is using to analyze the holes.

The Media Lab researchers report their results in the latest issue of the journal PLOS One.

“A lot of people will be able to start using this,” says Anshuman Das, a research scientist at the Camera Culture group and first author on the paper. “That’s the message I want to send out to people who would generally be cut off from using technology — for example, paleontologists or museums that are on a very tight budget. There are so many other fields that could benefit from this.”

Das is joined on the paper by Ramesh Raskar, a professor of media arts and science at MIT, who directs the Camera Culture group, and by Denise Murmann and Kenneth Cohrn, the forensic dentists who launched the project.

The system uses a Microsoft Kinect, a depth-sensing camera designed for video gaming. The Kinect’s built-in software produces a “point cloud,” a 3-D map of points in a visual scene from which short bursts of infrared light have been reflected back to a sensor. Free software called MeshLab analyzes the point cloud and infers the shape of the surfaces that produced it.

A high-end commercial 3-D scanner costs tens of thousands of dollars and has a depth resolution of about 50 to 100 micrometers. The Kinect’s resolution is only about 500 micrometers, but it costs roughly $100. And 500 micrometers appears to be good enough to shed some light on the question of the mysterious holes in the jaw of the T. rex skull.

Cretaceous conundrum

Discovered in 1990, the Field Museum’s T. rex skeleton, known as Sue, is the largest and most complete yet found. For years, it was widely assumed that the holes in the jaw were teeth marks, probably from an attack by another tyrannosaur. Ridges of growth around the edges of the holes show that Sue survived whatever caused them.

But the spacing between the holes is irregular, which is inconsistent with bite patterns. In 2009, a group of paleontologists from the University of Wisconsin suggested that the holes could have been caused by a protozoal infection, contracted from eating infected prey, that penetrated Sue’s jaw from the inside out.

The 3-D scan produced by the MIT researchers and their collaborators sheds doubt on both these hypotheses. It shows that the angles at which the holes bore through the jaw are inconsistent enough that they almost certainly weren’t caused by a single bite. But it also shows that the holes taper from the outside in, which undermines the hypothesis of a mouth infection.

One of the great advantages of 3-D scans is that they can be shared remotely. The Field Museum limits the time that researchers can spend with Sue’s skull, so the Wisconsin paleontologists’ analysis was largely based on photographs. But photographs don’t permit the comparison of the holes’ diameters at the inner and outer surfaces.

And if researchers working with a scan needed to examine a particular feature in close detail, they could use a 3-D printer to build a replica. To demonstrate this capacity, Das and his colleagues used their scan of Sue’s skull to produce a few models of it, at one-eighth the actual size.

Remote research

Das envisions that Kinect scans could prove as useful in other fields, such as archaeology and anthropology, as they could in paleontology. An archaeologist who unearths a large, fragile, artifact in a remote corner of the world could scan it and immediately share the scan with colleagues around the world.

“It’s that critical size,” Das says. “If it’s something really small, you can use a 3-D scanner. But if you have something stationary that’s difficult to move, you just put on the [Kinect] rig and walk around.”

Indeed, when Das scanned Sue’s skull, he mounted the Kinect in a modified camera harness and wore it on his chest. The space in which he performed the scan was irregularly shaped and presented various immovable obstacles, so it took him some time to find a route that would permit him to maintain a fixed distance from the skull as he walked around. But once he identified the route, the scan itself took about two minutes.

In ongoing work, Das, Murmann, Cohrn, Raskar, and a team of collaborators including the Wisconsin paleontologists, are looking at fragmentation patterns at the edges of the holes and at the holes’ depths and diameters, to see if they can infer anything about the shape, hardness, and velocity of whatever object might have caused them.

“Three-dimensional scanning has really revolutionized paleontology,” says Peter Mackovicky, associate chair of paleontology at the Field Museum. “We’re able to ask and answer a lot of quantitative questions. But in general we are a pretty underfunded field, and for a lot of folks, off-the-shelf scanning systems are still out of the usual reach of a research budget. Having something that is very cheap, versatile, and relatively fast is certainly useful. And one nice thing about [the new] system is that your results are immediate. You can see in real-time whether you’re capturing the data you need, which is a great benefit.”

“For me, one of the challenges is working in other countries,” Mackovicky adds. “I worked with local paleontologists in Argentina, and we have some material that’s been dug up, and we have a nice new species of very large meat-eating dinosaur. I tried some years ago to do some scanning with an entry-level tabletop laser-scanner system, and it was fairly good for some parts but extremely time consuming — literally hours for a single bone. [This] system is much, much faster. For a project like that, for me to have copies of these bones — which are far too large to ship here for study and also would require a very extensive permit application process to leave their country of origin — this is a really useful tool.”

de MIT News http://ift.tt/2sqhUiF

New 3-D chip combines computing and data storage

As embedded intelligence is finding its way into ever more areas of our lives, fields ranging from autonomous driving to personalized medicine are generating huge amounts of data. But just as the flood of data is reaching massive proportions, the ability of computer chips to process it into useful information is stalling.

Now, researchers at Stanford University and MIT have built a new chip to overcome this hurdle. The results are published today in the journal Nature, by lead author Max Shulaker, an assistant professor of electrical engineering and computer science at MIT. Shulaker began the work as a PhD student alongside H.-S. Philip Wong and his advisor Subhasish Mitra, professors of electrical engineering and computer science at Stanford. The team also included professors Roger Howe and Krishna Saraswat, also from Stanford.

Computers today comprise different chips cobbled together. There is a chip for computing and a separate chip for data storage, and the connections between the two are limited. As applications analyze increasingly massive volumes of data, the limited rate at which data can be moved between different chips is creating a critical communication “bottleneck.” And with limited real estate on the chip, there is not enough room to place them side-by-side, even as they have been miniaturized (a phenomenon known as Moore’s Law).

To make matters worse, the underlying devices, transistors made from silicon, are no longer improving at the historic rate that they have for decades.

The new prototype chip is a radical change from today’s chips. It uses multiple nanotechnologies, together with a new computer architecture, to reverse both of these trends.

Instead of relying on silicon-based devices, the chip uses carbon nanotubes, which are sheets of 2-D graphene formed into nanocylinders, and resistive random-access memory (RRAM) cells, a type of nonvolatile memory that operates by changing the resistance of a solid dielectric material. The researchers integrated over 1 million RRAM cells and 2 million carbon nanotube field-effect transistors, making the most complex nanoelectronic system ever made with emerging nanotechnologies.

The RRAM and carbon nanotubes are built vertically over one another, making a new, dense 3-D computer architecture with interleaving layers of logic and memory. By inserting ultradense wires between these layers, this 3-D architecture promises to address the communication bottleneck.

However, such an architecture is not possible with existing silicon-based technology, according to the paper’s lead author, Max Shulaker, who is a core member of MIT’s Microsystems Technology Laboratories. “Circuits today are 2-D, since building conventional silicon transistors involves extremely high temperatures of over 1,000 degrees Celsius,” says Shulaker. “If you then build a second layer of silicon circuits on top, that high temperature will damage the bottom layer of circuits.”

The key in this work is that carbon nanotube circuits and RRAM memory can be fabricated at much lower temperatures, below 200 C. “This means they can be built up in layers without harming the circuits beneath,” Shulaker says.

This provides several simultaneous benefits for future computing systems. “The devices are better: Logic made from carbon nanotubes can be an order of magnitude more energy-efficient compared to today’s logic made from silicon, and similarly, RRAM can be denser, faster, and more energy-efficient compared to DRAM,” Wong says, referring to a conventional memory known as dynamic random-access memory.

“In addition to improved devices, 3-D integration can address another key consideration in systems: the interconnects within and between chips,” Saraswat adds.

“The new 3-D computer architecture provides dense and fine-grained integration of computating and data storage, drastically overcoming the bottleneck from moving data between chips,” Mitra says. “As a result, the chip is able to store massive amounts of data and perform on-chip processing to transform a data deluge into useful information.”

To demonstrate the potential of the technology, the researchers took advantage of the ability of carbon nanotubes to also act as sensors. On the top layer of the chip they placed over 1 million carbon nanotube-based sensors, which they used to detect and classify ambient gases.

Due to the layering of sensing, data storage, and computing, the chip was able to measure each of the sensors in parallel, and then write directly into its memory, generating huge bandwidth, Shulaker says.

Three-dimensional integration is the most promising approach to continue the technology scaling path set forth by Moore’s laws, allowing an increasing number of devices to be integrated per unit volume, according to Jan Rabaey, a professor of electrical engineering and computer science at the University of California at Berkeley, who was not involved in the research.

“It leads to a fundamentally different perspective on computing architectures, enabling an intimate interweaving of memory and logic,” Rabaey says. “These structures may be particularly suited for alternative learning-based computational paradigms such as brain-inspired systems and deep neural nets, and the approach presented by the authors is definitely a great first step in that direction.”

“One big advantage of our demonstration is that it is compatible with today’s silicon infrastructure, both in terms of fabrication and design,” says Howe.

“The fact that this strategy is both CMOS [complementary metal-oxide-semiconductor] compatible and viable for a variety of applications suggests that it is a significant step in the continued advancement of Moore’s Law,” says Ken Hansen, president and CEO of the Semiconductor Research Corporation, which supported the research. “To sustain the promise of Moore’s Law economics, innovative heterogeneous approaches are required as dimensional scaling is no longer sufficient. This pioneering work embodies that philosophy.”

The team is working to improve the underlying nanotechnologies, while exploring the new 3-D computer architecture. For Shulaker, the next step is working with Massachusetts-based semiconductor company Analog Devices to develop new versions of the system that take advantage of its ability to carry out sensing and data processing on the same chip.

So, for example, the devices could be used to detect signs of disease by sensing particular compounds in a patient’s breath, says Shulaker.

“The technology could not only improve traditional computing, but it also opens up a whole new range of applications that we can target,” he says. “My students are now investigating how we can produce chips that do more than just computing.”

“This demonstration of the 3-D integration of sensors, memory, and logic is an exceptionally innovative development that leverages current CMOS technology with the new capabilities of carbon nanotube field–effect transistors,” says Sam Fuller, CTO emeritus of Analog Devices, who was not involved in the research. “This has the potential to be the platform for many revolutionary applications in the future.” 

This work was funded by the Defense Advanced Research Projects Agency, the National Science Foundation, Semiconductor Research Corporation, STARnet SONIC, and member companies of the Stanford SystemX Alliance.

de MIT News http://ift.tt/2uqn1AJ

Reverend Kirstin Boswell-Ford to be newest chaplain

MIT will welcome the Reverend Kirstin Boswell-Ford this month as the new chaplain to the Institute and director of religious life. She will succeed Robert Randolph, a member of the MIT community since 1979 who became the first chaplain to the Institute in 2007.

Boswell-Ford comes to MIT from Brown University, where she served as associate university chaplain to the Protestant community. Prior to her service at Brown, Boswell-Ford worked both at Bentley University in Waltham, Massachusetts, and at the International Association of Black Religions and Spiritualities in Chicago. Speaking about working in a university setting, she says the growth and development of individuals is what makes her service so worthwhile.

“You’re looking at students that are going to be our next world leaders,” Boswell-Ford says. “And I love seeing them as they’re just embarking on their careers.”

Senior Associate Dean of Student Support and Wellbeing David Randall, who chaired the search for the new chaplain to the Institute, says the committee “wanted someone who, most importantly, connected with students — someone who could build on the foundation that was set by Bob Randolph, but also create a new vision for the office.”

Boswell-Ford matched the committee’s priorities perfectly. During her time at Brown, she worked closely with a number of smaller groups within the Protestant community, as well as with that university's many interfaith organizations.

“Kirstin has a deep appreciation for diversity and inclusion, and we needed a chaplain who could speak to the whole MIT community,” Randall says. “Kirstin really had a commitment to interfaith work that I think was quite unique.”

While Boswell-Ford is new to MIT, she's no stranger to Cambridge. She has also served as an associate pastor at Union Baptist Church in Central Square, where she worked closely with the congregation’s Women’s Fellowship and provided pastoral care and counseling for community members. She experienced her first interactions with the MIT community when she took some courses at the Institute while studying at Wellesley College (she later transferred to the University of Virginia, where she completed her degree).

“I've always been very impressed with the school’s mission and dedication to the sciences and engineering and technology,” she says. “MIT really is a place where there's a lot of support for students, and that was really impressive to me.”

Given that Boswell-Ford will be only the second chaplain to the Institute in MIT’s history after Randolph, who retired last August, she and the MIT community are looking forward to the strides she can make in the position.

“Bob Randolph really worked hard at establishing the chaplaincy, and I think Kirstin can really take it to the next level,” Randall says. “There are folks in many offices who are very interested in partnering with the chaplaincy, and in making sure that we have a tight web of support for our students. I think she’s going to really have a lot of flexibility in creating her vision for the office, and she’s going to have a lot of eager partners as we think about how to all work together.”

Boswell-Ford says she's “very excited for the challenge.”

“I think that there’s a lot of room for putting my mark on the growth and the implementation of what religious life looks like at MIT, so that’s very exciting to me,” she says.

de MIT News http://ift.tt/2tpOYbK

Modeling innovation

Suppose you help run the R&D unit of a major technology company. To encourage innovation, you might have the unit run a kind of internal race, letting a wide variety of projects unfold on their own. That seems like an enlightened approach, given the difficulty of knowing exactly which ones will pan out — and yet you might actually be discouraging the unit’s overall productivity, according to MIT Associate Professor Alessandro Bonatti.

Bonatti is an economist at the MIT Sloan School of Management whose work models the behavior of firms, employees, and market prices. Across this wide range of topics, Bonatti deploys game theory, the formal study of competition and cooperation, to draw some surprising conclusions.

Consider the R&D scenario. Individual researchers want to press ahead on their own projects, knowing they will benefit by making advances more quickly than their co-workers. But a useful final product for a firm — in computing, biotech, and many other fields — may well be a compromise among several individuals’ initiatives. How should a firm structure its R&D process so that researchers both compete and collaborate?

As Bonatti and co-author Heikki Rantakari PhD ’07, an assistant professor at the University of Rochester, suggest in a recent paper called “The Politics of Compromise,” firms can use rules to orchestrate R&D in these situations, making clear that only projects with certain degrees of compromise will be adopted. A balanced set of rules benefit firms, by helping them develop better products; the same rules give researchers incentives to both push ahead and compromise — since contributing to a successful product is better than having one’s own work be ignored.

“We try to understand how the rules should be written in a way that provides members incentives to do research and work hard, and doesn’t yield disproportionate power to an individual,” Bonatti explains.

So while popular culture fixates on lone innovators, the reality, Bonatti suggests, is usually different. “The goal of a compromise that builds on all core competencies is probably going to produce a better product.”

And while Bonatti’s work is largely theoretical — albeit with many references to real-world cases — he believes that streamlined analysis of a large organization is valuable.

“It’s like taking an X-ray of an organization,” Bonatti says. “You are trying to get at the core mechanism, and the core mechanism can’t be understood if it has 55 moving parts.”  

For his work, including a wide-ranging portfolio of published research, Bonatti was granted tenure last year from MIT — a significant milestone for a scholar who did not gravitate toward his professional field until he reached college.

New direction in Naples

Bonatti, a native Italian, grew up near Naples and studied the classics extensively in school. But as an undergraduate at the University of Naples Federico II, his interest in studying economics took hold — partly because it helped him “understand more of the world,” and partly, Bonatti jokes, “because I didn’t want to be a lawyer, which is what everybody seemed to want to be.”

Yale University accepted Bonatti to its PhD program in economics, and he thrived as a graduate student there, completing in 2009 a thesis on dynamic pricing as deployed by the likes of Netflix. Bonatti’s graduate research led to what he calls a “wave” of published papers early in his career — four completed in 2011 alone, two years after he joined the MIT faculty.

At MIT, Bonatti has added to his portfolio by studying what he calls “the dynamics of incentives” for workers in firms; he still works on pricing as well. This has produced another wave of Bonatti papers, with several more being published in the last three years.

Bonatti’s careful modeling has established rigorous ways of thinking about numerous issues firms encounter. For instance, strict mechanisms that kill off questionable R&D projects, he has found, are important, in order to maintain high standards and save firms money in the long run. Additionally, voting procedures in firms that use a “supermajority” to approve projects can produce better collaboration, by satisfying more interests. Bonatti has also found that, in the R&D context, project deadlines are often more efficient than close supervision.

“I’m interested in different topics, but my method of analysis is always the same,” Bonatti says. “Look at the world, distill it into a framework that can be written out in tractable mathematical form, run the model, do a sanity check with reality — Do I think I’ve captured what matters? — and then, crucially, learn from the model.”

Smart people, faulty systems

One advantage of this approach, Bonatti suggests, is that it can keep firms from incorrectly blaming their problems on particular individuals, and instead may help direct their focus to systemic issues that can help or hinder productivity.

“People’s motives are pretty clear in the R&D world, for instance in pharma,” Bonatti says. “They’re smart. They’re motivated. Their hearts are in the right place, considering they’re trying to cure diseases. It’s unlikely that any failure or productivity problem is a very person-specific issue.”

So systems matter. Still, every topic Bonatti studies is different, meaning that in his work, there is a huge premium on modeling firms and workers carefully, and considering as many aspects of a problem as possible.

In the last few years, Bonatti has also begun engaging with executives and R&D leaders more extensively, something that helps him develop what he calls “a proper conversation between the laboratory and the model.”

After all, the flexibility Bonatti finds useful in a firm’s researchers and managers, as they search for compromises and new solutions, applies to his work as well.

“There isn’t always just one solution to any one problem,” he says.

de MIT News http://ift.tt/2tfCGEp

Summer interns' lab work underway

The Summer Scholars in materials science and engineering have settled on their research projects and lab assignments. The interns, co-sponsored by the Materials Processing Center and the Center for Materials Science and Engineering, faced difficult decisions to choose labs after hearing enticing faculty presentations and taking lab tours.

Luke Soule found all the possible projects interesting, but has honed in on electrochemistry, choosing to work in Department of Materials Science and Engineering Professor Yang Shao-Horn’s Electrochemical Energy Lab (EEL). During a tour of the lab, graduate student Karthik Akkiraju presented several research projects on the role of catalysts in lowering the energy needed to stimulate electrochemical reactions in energy devices. Akkiraju says Shao-Horn looks for students who are excited about the work and encourages students to be independent and to work together as a community. He also emphasizes the family-like atmosphere of the group. “At EEL, you never work alone,” Akkiraju says.

Stephanie Bauman has chosen to work in Assistant Professor Luqiao Liu’s lab, after listening to electrical engineering and computer science graduate student Joseph T. Finley explain how he uses processes such as electron sputtering and ion milling to make magnetic thin films. The lab is developing new magnetically switchable materials for computer memory. “It seems to be mostly focused toward physics which is my major and more so than a lot of the other bio or chem projects,” Bauman says.

Alexandra Oliveira has chosen to work under Fikile R. Brushett, the Raymond A. (1921) and Helen E. St. Laurent Career Development Professor of Chemical Engineering, on redox flow batteries. ‘”Right now I’m working on the permeability of different microstructures for carbon electrodes and I’ll be attempting to electrograft molecules onto the electrodes to change their chemical properties for aqueous and non-aqueous flow batteries,” Oliveira says.

Summer Scholar Grace Noel is working in the lab of Charles and Hilda Roddey Career Development Professor in Chemical Engineering William A. Tisdale, on a project to make and study metal halide perovskite nanoplatelets. These platelets, which are like flat quantum dots, are sometimes just over one-half of a unit cell in thickness, and their color can be adjusted by altering their composition.

Ben Canty is involved in a project to develop a catalyst for breaking down lignins in plant biomass into industrially useful chemicals like benzene, working in the lab of associate professor of chemical engineering Yuriy Román. “I’m mixing in stuff in a tiny little batch reactor, putting it on a heater on a shelf, watching it so it doesn’t explode, centrifuging it, and then running it on gas chromatographs and mass spectrometers,” Canty explains.

NanoStructures Laboratory postdoc Reza Baghdadi impressed Summer Scholar Saleem Iqbal while explaining how Professor Karl Berggren aims to develop superconducting nanowires made of niobium nitride for reducing data processing energy consumption. In the Berggren lab, Iqbal is getting a chance to learn different fabrication skills, such as photolithography and electron beam lithography, and thin film deposition and etching processes, with optical and electrical studies at liquid helium temperatures of about 4.2 kelvins.

AIM Photonics Academy interns were matched separately to their projects. Stuart Daudlin is working on statistical modeling of photonic device variations with Duane Boning, the Clarence J. LeBel Professor of Electrical Engineering. Ryan Kosciolek is working on nonlinear photonic devices with Microphotonics Center Principal Research Scientist Anuradha Agarwal. Summer Scholars attend regular weekly or bi-weekly lab group meetings. Larger groups have dedicated subgroups as well that meet regularly.

The internships are supported in part by the National Science Foundation’s Materials Research Science and Engineering Centers program. Participants will present their results at a poster session the last week of the program, which runs from June 15 to August 5.

de MIT News http://ift.tt/2tENrAD

Practical parallelism

The chips in most modern desktop computers have four “cores,” or processing units, which can run different computational tasks in parallel. But the chips of the future could have dozens or even hundreds of cores, and taking advantage of all that parallelism is a stiff challenge.

Researchers from MIT’s Computer Science and Artificial Intelligence Laboratory have developed a new system that not only makes parallel programs run much more efficiently but also makes them easier to code.

In tests on a set of benchmark algorithms that are standard in the field, the researchers’ new system frequently enabled more than 10-fold speedups over existing systems that adopt the same parallelism strategy, with a maximum of 88-fold.

For instance, algorithms for solving an important problem called max flow have proven very difficult to parallelize. After decades of research, the best parallel implementation of one common max-flow algorithm achieves only an eightfold speedup when it’s run on 256 parallel processors. With the researchers’ new system, the improvement is 322-fold — and the program required only one-third as much code.

The new system, dubbed Fractal, achieves those speedups through a parallelism strategy known as speculative execution.

“In a conventional parallel program, you need to divide your work into tasks,” says Daniel Sanchez, an assistant professor of electrical engineering and computer science at MIT and senior author on the new paper. “But because these tasks are operating on shared data, you need to introduce some synchronization to ensure that the data dependencies that these tasks have are respected. From the mid-90s to the late 2000s, there were multiple waves of research in what we call speculative architectures. What these systems do is execute these different chunks in parallel, and if they detect a conflict, they abort and roll back one of them.”

Constantly aborting computations before they complete would not be a very efficient parallelization strategy. But for many applications, aborted computations are rare enough that they end up squandering less time than the complicated checks and updates required to synchronize tasks in more conventional parallel schemes. Last year, Sanchez’s group reported a system, called Swarm, that extended speculative parallelism to an important class of computational problems that involve searching data structures known as graphs.

Irreducible atoms

Research on speculative architectures, however, has often run aground on the problem of “atomicity.” Like all parallel architectures, speculative architectures require the programmer to divide programs into tasks that can run simultaneously. But with speculative architectures, each such task is “atomic,” meaning that it should seem to execute as a single whole. Typically, each atomic task is assigned to a separate processing unit, where it effectively runs in isolation.

Atomic tasks are often fairly substantial. The task of booking an airline flight online, for instance, consists of many separate operations, but they have to be treated as an atomic unit. It wouldn’t do, for instance, for the program to offer a plane seat to one customer and then offer it to another because the first customer hasn’t finished paying yet.

With speculative execution, large atomic tasks introduce two inefficiencies. The first is that, if the task has to abort, it might do so only after chewing up a lot of computational cycles. Aborting smaller tasks wastes less time.

The other is that a large atomic task may have internal subroutines that could be parallelized efficiently. But because the task is isolated on its own processing unit, those subroutines have to be executed serially, squandering opportunities for performance improvements.

Fractal — which Sanchez developed together with MIT graduate students Suvinay Subramanian, Mark Jeffrey, Maleen Abeydeera, Hyun Ryong Lee, and Victor A. Ying, and with Joel Emer, a professor of the practice and senior distinguished research scientist at the chip manufacturer NVidia — solves both of these problems. The researchers, who are all with MIT’s Department of Electrical Engineering and Computer Science, describe the system in a paper they presented this week at the International Symposium on Computer Architecture.

With Fractal, a programmer adds a line of code to each subroutine within an atomic task that can be executed in parallel. This will typically increase the length of the serial version of a program by a few percent, whereas an implementation that explicitly synchronizes parallel tasks will often increase it by 300 or 400 percent. Circuits hardwired into the Fractal chip then handle the parallelization.

Time chains

The key to the system is a slight modification of a circuit already found in Swarm, the researchers’ earlier speculative-execution system. Swarm was designed to enforce some notion of sequential order in parallel programs. Every task executed in Swarm receives a time stamp, and if two tasks attempt to access the same memory location, the one with the later time stamp is aborted and re-executed.

Fractal, too assigns each atomic task its own time stamp. But if an atomic task has a parallelizable subroutine, the subroutine’s time stamp includes that of the task that spawned it. And if the subroutine, in turn, has a parallelizable subroutine, the second subroutine’s time stamp includes that of the first, and so on. In this way, the ordering of the subroutines preserves the ordering of the atomic tasks.

As tasks spawn subroutines that spawn subroutines and so on, the concatenated time stamps can become too long for the specialized circuits that store them. In those cases, however, Fractal simply moves the front of the time-stamp train into storage. This means that Fractal is always working only on the lowest-level, finest-grained tasks it has yet identified, avoiding the problem of aborting large, high-level atomic tasks.

de MIT News http://ift.tt/2t9xBMa