Portland, Ore. -- Though the International Semiconductor Roadmap predicted that 65-nanometer chips would require high-k dielectrics, some chip makers, Intel Corp. among them, put off switching until the 45-nm node, where researchers widely agree the high-k dielectrics will have to be used. Now, a researcher at the University of Utah reports, cobalt-doped high-k dielectrics can double as filters for "spintronic" transistors at the 45-nm node and beyond.
"If you dope cerium-oxide with cobalt, it becomes a ferromagnetic insulator," said professor Ashutosh Tiwari, a materials scientist who led a team at the university's Nanostructured Materials Research Laboratory in Salt Lake City. "It becomes a multifunctional material."
Ferromagnetic insulators polarize the electrons tunneling through them, en-abling spintronic devices to be built.
Researchers worldwide are experimenting with high-k dielectrics of the lanthanoid family, from lanthanum and cerium to gadolinium and dysprosium. Instead of heeding the International Semiconductor Roadmap's forecast that high-k dielectrics would serve as subnanometer insulators for the gates of complementary metal-oxide semiconductor (CMOS) transistors below the 65-nm node, chip makers instead have increased channel strain and enhanced carrier mobilities, thereby avoiding the leakage and reliability challenges involved with scaling down gate oxide thicknesses below 1 nm. But to reach the 45-nm node and beyond, high-k dielectrics are largely seen as essential so that gate oxides can be thinned down to 7 angstroms or less.
"Cerium oxide, for instance, has a very high dielectric constant, of 26, giving it the potential to replace silicon dioxide [which has a dielectric constant of only 3.9] below 1-nm thicknesses," said Tiwari.
Now Tiwari's team offers a second incentive to adopt lanthanoid oxides as high-k dielectrics: harnessing the spin of electrons instead of just storing a charge.
Today's chips ignore the spin polarization of electrons--usually specified as "up" or "down"--and instead just store randomly polarized charge carriers. Up to the transistor dimensions of today, logic values are represented by the bulk flow of large numbers of electrons, which are randomly polarized. But as devices continue to shrink, the absolute number of electrons they channel will also decrease. By harnessing the spin of those electrons, Tiwari argues, the injection and detection of smaller numbers of polarized electrons will enable transistor sizes to shrink further, faster, by using spintronics. "Our material has application in spintronic devices that use the magnetic spin of electrons," said Tiwari. "It can act as a spin filter."
Harnessing magnetic spin requires electrons to be injected with a known polarization. For years, researchers have randomly polarized electrons into streams where the spin is known. But the efficiency of current approaches has never been good enough for devices of commercial quality.
"Creating sources of spin-polarized electrons has been a holy grail of spintronics," said Tiwari. "Other groups have reported efficiencies of only about 3 percent for their spin filters, but our technique has over 80 percent efficiency--meaning that 80 percent of the electrons tunneling through our material will be polarized in the same direction."
By using these high-k dielectric formulations doped with cobalt to make them ferromagnetic, future chip makers could not only reach beyond 65 nm with subnanometer ferromagnetic oxides but could simultaneously enable spintronic devices that harness polarized electrons.
Others have experimentally demonstrated that very low-temperature spin polarization filters are possible, but Tiwari's doping method makes even-better-working room-temperature devices, he said.
"The main problem with others' work is that their spin-polarized emitters and filters only work at very low temperatures, whereas our approach works at room temperature and even at high temperatures--all the way up to 875 Kelvin."
Tiwari's group has yet to create real transistors, but it has proven that its technique can work when lanthanoid-based materials are used with silicon. In real devices, a ferromagnetic metal would be deposited on top of the cobalt-doped ferromagnetic insulator. In tests, Tiwari's group proved the doped cerium oxide polarized 80 percent of the electrons that tunneled through the ferromagnetic oxide, potentially enabling spin polarization devices to be built.
"Our next step will be to make real devices out of our material," said Tiwari.
The team reports the largest magnetic moment ever observed in a cobalt oxide, Tiwari said, accomplished by depositing 3 percent cobalt-doped cerium-oxide films on lanthanum aluminum-oxide substrates with a pulsed laser. The resulting transparent films had a very high Curie temperature and a giant magnetic moment that was constant up to 500 K and still usable to 875 K. Transmission electron micro-scopy revealed a face-centered cubic system with a crystalline lattice that closely matches silicon's, Tiwari said.