Global Sources
EE Times-Asia
Stay in touch with EE Times Asia
EE Times-Asia > Memory/Storage

Memory alternatives wait for flash to flame out

Posted: 16 Dec 2004 ?? ?Print Version ?Bookmark and Share

Keywords:non-volatile memory? mram? carbon nanotube? spansion llc? flash memory?

Memory makers are waiting for flash to die down in the hope to provide new alternatives in the future.

The fight to become the top non-volatile memory of the future is shaping up as a donnybrook. While emerging technologies such as MRAM and carbon nanotubes hover over the market like vultures waiting for their dinner to expire, providers of today's dominant flash memories are showing a burst of new optimism. Their most recent lab results suggest that flash may be scalable into at least 2007 or 2008, which may be beyond the lifetime of some emerging technologies.

Last month, Spansion LLC, the flash memory subsidiary of advanced micro devices Inc. and fujitsu Ltd, announced a road map stretching through 2007 that uses the company's MirrorBit NOR technology to create 8Gb chips. Toshiba Corp. and other flash suppliers have similarly optimistic plans for NAND.

"It's fair to say that existing flash technology will run out of steam, but it's not likely to happen until toward the end of the decade," said Stefan Lai, VP of technology and manufacturing group at NOR flash provider Intel Corp.

"Conventional materials and technologies don't just suddenly stop; they just get harder to do with each succeeding generation," Lai said. "At some point, scaling the existing technology becomes hard enough that a replacement becomes viable. But even then, you need to understand the economics of the replacement so you can orchestrate a successful transition. New technologies don't just burst in and take over."

Lai is optimistic about conventional floating-gate flash cells for the time beingThe cell is working fine, even for multilevel storage, he said. In the near future, Lai added, Intel is investigating changes in materials to achieve the high voltage necessary to induce tunneling to the floating gate.

"Scaling gets to be a real issue," Lai explained. "At the 65nm node, for instance, the control gate dielectric is only 1nm or maybe 1.2nm thick, but you have to make the dielectric under the floating gate 8nm thick. And the devices being proposed to help with transistor scaling, such as the trigate transistor, aren't applicable at the high voltages we need."

New materials may be the way out of the dilemma, Lai said. "Intel is looking at things like oxide-nitride-oxide sandwiches and other composite materials for the tunneling dielectric. Research has shown that a composite material can reduce the voltage necessary for tunneling and increase charge retention." The company is also exploring "nanocrystal structures for the floating gate itself," Lai said. "And further out, we are looking at ovonic tubes. These structures have a memory characteristic and appear to be scalable."

As flash vendors push forward, boosters of non-conventional technologies smell blood. "We're going to need a new embeddable non-volatile technology to be ready by late 2006 or early 2007," said Michael Kozicki, co-founder and CTO of startup Axon Technologies Corp. "If it's not ready by then, we are doomed."

Kozicki and other startup executives pitched their ideas at Semico Research Corp.'s Impact conference last month. All presented views of a future in which SRAM, embedded DRAM and flash are ripe for the picking.

Saied Tehrani, director of MRAM technology at Freescale Semiconductor Inc., claimed that no single established technology could meet all the needs of embedded memory in future SoC devices: high density, high speed, indefinite reliable lifetime, non-volatility, low power and low process cost. Further, he maintained that all established technologies will face serious scaling issues, ranging from the shrinking capacitor area for DRAM to the inability to reduce programming voltage for flash.

Tehrani said MRAM could address each of those issues. He said Freescale has demonstrated symmetric 25ns read and write times on a device built in 180nm technology. He expects Freescale to sample a 1Mb MRAM chip in 2005 and to have the device qualified for production by the end of next year. "We believe the cell will be scalable," he said. "The issue is scalability of the magnetoresistive switching element."

Avo Kanadjian, VP of strategic marketing for intellectual-property startup Nantero Inc., spoke in favor of carbon nanotubes and described the non-volatile RAM array Nantero is researching. The storage element in an NRAM, Kanadjian said, is a little cluster of nanotubes suspended between upper and lower electrodes in a micro vacuum chamber. Applying a charge to the electrodes causes the nanotubes to sag and make contact with the lower electrode. The tubes remain in this state until a reverse electric field makes them bend upward, breaking the connection.

Kanadjian claimed that in the laboratory, Nantero has observed switching times and cell densities similar to those of current SRAM cells. Durability, he said, was on the order of 1,015 cycles. He further claimed that the process requires the introduction of no new materialsexcept carbonand no new equipment to a CMOS fab. There are no scaling limits, he said.

Axon, meanwhile, is working on what it calls the programmable metallization cell, wherein a self-assembling nanoscale electrolyte is grown on selected sites in a memory array. "This material consists of ions embedded in a glassy matrix," Kozicki said. Applying an electric field to metal electrodes on either side of the matrix causes metal ions to be electrodeposited through the matrix, forming a conductive bridge in a matter of nanoseconds. Reversing the field causes the metal ions to reverse their migration.

Kozicki admitted it wouldn't be easy displacing the established powers. "I believe any memory technology must be proven as a discrete, high-volume product before it will be accepted for embedded use," he said. He hinted that Axon's PMC technology will appear in a DRAM-alternative chip in early 2007.

- Ron Wilson

EE Times

Article Comments - Memory alternatives wait for flash t...
*? You can enter [0] more charecters.
*Verify code:


Visit Asia Webinars to learn about the latest in technology and get practical design tips.

Back to Top