SCIENCE

RICHLAND, WA -- A newly patented way to deposit metal atoms on very thin oxide layers may help next-generation computers boot up instantly, making entire memories immediately available for use. The technique also may help fabricate less expensive catalysts for chemical reactions and lead to better nanotechnology devices and ceramic/metal seals. The method, described in the August 2 issue of Science, anchors ultrathin metallic layers to metal oxides by using a chemical reaction discovered at the Pacific Northwest National Laboratory and understood and generalized by theoretical scientists at Sandia National Laboratory in Albuquerque. The inexpensive trick bypasses the hurdle created when metal atoms cluster together into three-dimensional islands when deposited on oxide surfaces. These ultrasmall islands of metal-similar to water beads on a waxed car-produce discontinuous, noncrystalline metal films. The new smooth interfaces achieve crystallinity by only a few atomic layers and should also produce greater durability in electronic devices. The findings may have the most immediate bearing on magnetic tunnel junctions, slated for use in magnetoresistive random access memory, or MRAM. MRAM will allow computers to store information in a nonvolatile fashion, meaning that the information is not lost when the computer is turned off. As a result, MRAM promises a day when computers would boot up instantly once turned on, rather than slowly retrieving information during the boot-up stage. Major corporations have begun developing MRAM modules in hopes of generating robust nonvolatile memory in the next few years. In a magnetic tunnel junction, an ultrathin layer of insulator, typically aluminum oxide with a thickness of less than one nanometer, is sandwiched between thin layers of magnetic metal, such as cobalt or nickel-iron. Current flows through the device and the magnetic orientation of the two metal layers can be switched, resulting in different values of the tunneling current, thereby creating an environment in which "bits" of computer memory can be stored. Yet difficulties in growing an atomically flat, ultrathin film of metal on top of any insulator material have been well documented for years. In order to achieve ferromagnetism, thick layers of the top metal must be made. The new discovery should allow for much thinner layers of metal and lower currents needed to switch the direction of the magnetic field. Catalysts are involved in approximately two-thirds of the gross domestic product of the United States, particularly oil. Their wide applications allow chemists to turn one molecule into another. The new discovery should also enable the production of catalysts where the reactive metal on an oxide support is only one atomic layer thick, thereby saving considerable cost. The new method uses a chemical reaction to embed metal atoms at scattered points within the top layer of the oxide, amounting to about one anchor for every ten oxygen atoms in the top layer. These anchoring atoms are then able to bind other metallic atoms just above the oxide surface. The new method can use equipment already in place in chip manufacturing plants. "Many advanced technologies rely on strong interfaces between metals and oxides," said Scott Chambers, PNNL chief scientist and lead author of the Science paper. "These findings are very exciting because they may provide the molecular insight industry needs to create better materials for microelectronics and sensors." Chambers worked in partnership with PNNL postdoc Tim Droubay, who helped with the experiments, and Dwight Jennison, a well-known solid-state theorist from DOE's Sandia National Laboratories. Said Jennison, "The process Scott tested concerns growing cobalt on aluminum oxide. Cobalt's interaction with the oxide is so weak that it would normally ball up when deposited. However, if the surface of the oxide is first completely hydroxylated, i.e. is terminated by a layer of hydrogen and oxygen atoms bound together, cobalt atoms which hit two hydroxyl groups at once can react to release a hydrogen gas molecule. These cobalt atoms then become oxidized themselves and end up in the top layer of the oxide, surrounded by negative ions to which they bind strongly. These are the anchors." "For industry, a solution may be as simple as exposing the thin aluminum oxide films to a low pressure of water vapor before adding a final cobalt layer," said Chambers. The entire process may be done at room temperature, while it is often important to avoid high temperatures in manufacturing. Jennison, who first found which chemical reactions would be energetically favorable, collaborated at Sandia with Thomas Mattsson, who has long experience in first principle based diffusion and reaction studies and in computing critical reaction barriers. Their theoretical first principles calculations predicted some and validated other experimental results. Some of these calculations required one of the world's fastest computers. Notably, the calculations provided insight into what reaction is taking place, where it occurs, the energy barrier for it to happen, and the time it needs to be completed vs. the time it takes arriving cobalt atoms to lose energy while in contact with the surface. In particular, the latter is important because if the reaction were slow, the rapidly diffusing cobalt atoms could find a growing island first. However, because hydrogen molecules are being made, the reaction can be very fast, of the order of tenths of a picosecond and well before the arriving cobalt atoms can assume the temperature of the substrate. Said Jennison, "Otherwise the experimental result would be impossible to explain. However, here we have a wonderful joining of theory and experiment." While the experiment was conducted using cobalt, Jennison's calculations predict the method also would be effective for iron and nickel, two other metals under consideration for MRAM, as well as metals such as copper, ruthenium, and rhodium, the latter two having catalysis applications. The Division of Material Science within the DOE's Office of Basic Energy Sciences supported this research at PNNL.