RICHLAND, Wash. A new technique for fabricating magnetic tunnel junctions would enable magnetoresistive random-access memories (MRAMs) to be economically manufactured, according to researchers here at the Pacific Northwest National Laboratory. Acting on theoretical evidence uncovered by Sandia National Laboratory solid-state theorist Dwight Jennison, PNNL chief scientist Scott Chambers formed atomically flat crystalline films of metal measuring only a few atoms thick on sapphire (aluminum oxide).
"We believe that the ability to fabricate such thin layers of single-crystal metal will enable much more economical magnetic tunnel junctions in MRAMs ones that will require lower currents to switch the direction of magnetic fields," said Chambers. Postdoctoral fellow Tim Droubay assisted Chambers in producing an atomically flat film of cobalt on hydroxylated sapphire.
Semiconductor makers currently deal with the tendency of metal atoms to cluster into islands during oxide deposition by making thicker films. The downside is that the thicker the metals, the higher the power required to achieve ferromagnetism. In the process, of course, extra heat is generated.
"With our new method, you can achieve crystallinity with only a few atomic layers, and the inherent structural strength should produce greater durability in electronic devices," said Chambers.
The discovery could realize considerable cost saving in the production of catalysts as well as benefiting semiconductor design. According to Chambers, the reactive metals would only have to be one atomic layer thick (held by an oxide support). Catalysts are involved in approximately two-thirds of the gross domestic product of the United States, particularly in processing oil into products. The national lab has filed several patents related to the finding's diverse applicability.
Magnetic tunnel junctions function by using an oxide to separate a layer of metal with a fixed magnetic orientation from one that can be magnetically flipped. The two distinct magnetic orientations of the second layer produce different values of resistance to the flow of current through the junction. Measuring the resistance value then becomes a way to detect the orientation of the variable magnetic layer.
Another approach, employed in giant-magnetoresistive heads (GMRs) in disk drives, does not feature electron tunneling but uses a similar scheme with a fixed magnetized layer sandwiched with a variable magnetic layer. High resistance results when the magnetic layers are oppositely oriented; low resistance results when the magnetic moments are aligned.
If MRAMs are indeed enabled by Chambers' discovery in the manner he predicts, they would be nonvolatile and theoretically would store data in much more dense arrays than existing nonvolatile technologies. Chambers predicts a day when the information that used to be stored on a boot disk will stay resident in an MRAM for "instant on" computers.
'Put down anchors'
It was Sandia's Jennison who came to Chambers with the suggestion that now promises to enable highly dense MRAMs. Jennison explained that the tendency of metals to clump into islands during deposition on oxides could be mitigated with hydroxyls (the OH part of water, without the second H). His theory was that those water fragments, at a very low pressure, would enable metal atoms to "put down anchors" on the oxide, resulting in the formation of flat films on sapphire rather than three-dimensional islands.
Chambers verified that insight in the lab with Droubay by producing an atomically flat film of cobalt on hydoxylated sapphire. As Jennison had predicted, the low-pressure hydroxlys permitted the cobalt to self-assemble automatically into flat layers. Chambers predicts that the technique will also work for other metals, such as copper, iron and nickel, and for such catalysts as ruthenium and rhodium.
Chambers explains the unusual behavior whereby the metal lays down perfect, atomically precise layers as a process in which a prepared sapphire surface yields a hydrogen atom from a surface bond in exchange for a hydrogen from the cobalt.
"We have been able to lay down layers, one by one, up to about 10 so far, and the process works very well," he said. "The sapphire link with the cobalt is actually more stable than it was with the original hydrogen. The cobalt becomes oxidized, which forms a very strong bond to the surface."
Chambers estimates that the strong oxidation bonds are put down about 10 atoms apart. The rest of the layer is filled in with free cobalt atoms, which form a complete, atomically thin crystalline metallic layer. Subsequent cobalt layers could be put down as desired. All processing was performed using standard deposition equipment at room temperature.
To uncover just what was happening on the surface of the oxide to make it bond so strongly with the cobalt, Chambers collaborated at Sandia with Thomas Mattsson, who used Sandia's supercomputer to perform diffusion and reaction studies to compute the critical reaction barriers. The collaboration enabled theoretical first principles to be validated with experimental results.
According to the researchers' calculations, in the presence of low-pressure hydroxyl, the higher-temperature cobalt atoms would be able to bond to the sapphire in a tenth of a picosecond long before they would begin clumping (which only occurred if the cobalt was left in the diffusion chamber long enough to cool down to the same temperature as the substrate).
Mattsson provided calculations that quantified the energy barrier to the desired reaction and the time needed for completion vs. the time it takes cobalt atoms to lose temperature and begin to clump.