Long considered a promising next-generation memory thanks to its nonvolatility and fast access times, the MRAM has to date been limited to megabit densities. With its advance, the AIST-Anelva team said it will be possible to fabricate gigabit-level MRAMs using a conventional silicon wafer production process.

MRAMs generally use aluminum oxide as the tunneling barrier for their magnetic-tunnel junctions. Aluminum oxide is amorphous and its disordered atoms disperse electrons, which limits the magnetoresistance ratio (MR) to around 70 percent. But AIST (www.aist.go.jp) has been making progress with magnesium oxide as the tunneling barrier — the insulator layer between ferromagnetic electrodes. Shinji Yuasa, chief researcher at AIST, announced in March the creation of an MgO-based tunneling magnetoresistance (TMR) device that achieved an 88 percent MR at room temperature — the highest performance reported up to that time.

That device was fabricated on a single-crystal MgO substrate by a molecular-beam epitaxy (MBE) system. Neither the process nor the substrate is suitable for mass production, AIST said.

The MR achievable with an aluminum oxide tunneling barrier makes it difficult to obtain an output voltage for signal readout higher than 200 millivolts, which is about half a DRAM's signal readout. That limitation has kept MRAM density at about 64 Mbits. An output voltage of about 400 mV is considered necessary to realize a 1-Gbit MRAM.

In the single-crystal MgO substrate, electrons can move straight between ferromagnetic electrodes without suffering dispersion. With this latest advance, said Yuasa, the researchers achieved an output voltage of 380 mV, nearly double the highest level reported thus far.

Rather than an MBE system, the AIST and Anelva team employed a silicon wafer and a sputtering system, a combination widely used in mass manufacturing today.

The researchers' work with MgO was called a significant advance by one engineer who until recently managed a major MRAM development project, and who asked not to be identified.

"There is a lot more experience with aluminum oxide, and people like myself who have worked with aluminum oxide probably have a conservative streak," he said. "It takes time to bring on a new material like magnesium oxide, and companies probably would have to redesign their MRAMs with shorter bit lines and other changes. That all takes a lot of time. But if a company wants to begin with a new material set, then magnesium oxide has some wonderful advantages that could impact the costs and densities of MRAM."