Findings

Super Memory
Computer memory has come a long way since the Commodore 64’s cassette-tape drive, but even today’s data-storage devices face a pesky problem.  They come in two flavors: sluggish and stable, or fast and forgetful.  Flash memory—the stuff of iPods and digital camera cards—can survive a sudden power outage, but the business of saving and retrieving is slow going.  The RAM memory that pins a typed character to your computer screen, on the other hand, is lightning-fast but disappears without a trace if you kick out the electrical cord.

A new technique developed by Ritesh Agarwal, an assistant professor in the Department of Materials Science and Engineering, may do away with that tradeoff.  As reported in Nature Nanotechnology in October, Agarwal’s research team has produced a nanowire memory structure that writes and recalls data 1,000 times faster than conventional flash memory, and appears robust enough to withstand 100,000 years of use.

What’s more, they coaxed the tiny wires—which measure 100 atoms in diameter—to form all by themselves.  In fact that was the crucial part.  Conventional memory manufacture relies on lithography, in which strong chemicals and etching tools whittle a material ever smaller.  “But by doing this,” Agarwal says, “all the interesting properties of this material are basically degraded,” impairing its performance.

The Penn team designed a chemical reaction in which a gaseous form of the material, called germanium antimony telluride, crystallizes around nanoscale catalysts.  The result is a high-grade structure that is both speedy and stable—and requires remarkably little power to operate. 

In other words, this may be a kind of Holy Grail of data storage: a “universal memory” for which speed need not be sacrificed for durability.  Commercial availability may take a decade, but Agarwal imagines a world in which laptops take a couple seconds to boot up and the words “data buffering” never interrupt the video streaming on your monitor. 

“If you have this universal memory,” he says, “then you don’t have to have a hard drive, current drive, DRAM, SRAM—you’ll have only one thing.”  And it will be the only thing you need.

 

Surviving the 80-Hour Workweek
Time was, doctors-in-training lived virtually all their waking hours (and plenty of sleeping ones) in the hospital—becoming “residents” in the fullest sense of the word.  Thirty-six-hour shifts routinely bookended 12-hour rest periods, and bleary-eyed workweeks commonly exceeded 100 hours. 

But five years ago the Accreditation Council for Graduate Medical Education mandated an 80-hour week for residents, capping individual shifts at 24 hours.  According to two large studies by researchers in the School of Medicine, that change has been good for patients.  In the September 5 issue of JAMA, Dr. Kevin Volpp, assistant professor of medicine and health care systems, and Dr. Jeffrey Silber, professor of pediatrics, reported that patient mortality rates have either improved or remained stable under the new regulations.

Among 8 million patients in the Medicare system, the new rules have had no discernable impact on patient mortality.  In Veterans Administration hospitals, patient mortality has fallen. 

One of the goals of the new system was to reduce fatigue-related medical errors, yet the tradeoff for shortening shifts was an increase in patient handoffs from one resident to the next, decreasing continuity of care.  The authors noted that in addition to being more teaching-intensive, VA hospitals also have superior patient-information systems, which may make handoffs safer.—T.P.