Slashdot Mirror
Perv-y Material Heralds Move From Silicon
RalphTWaP writes "The Register has posted this report about the successful use of perovskite oxides as a replacement for silicon oxide in chip manufacture. As the Reg reports, the journal Science contains the original article. Best of luck getting at it though. Perhaps that kind of thing is what this other article was talking about."
As someone who works in this field I thought I might throw in my two cents.
:-). Suffice it to say that there are zillions of other interesting properties you can get from various perovskites. The famous 1-2-3 superconductor YBa2Cu3O(7-x) is a modified perovskite structure; colossal magnetoresistive materials like (La,Sr)MnO3 have been studied for use in hard drive read heads; (Pb,Zr)TiO3 is a so-called piezoelectric material-- moves when you put a voltage on it-- so you could use it for nanomachines... etc. etc. Silicon, of course, is silicon. So it's a cool thing to put them together.
[For the nosy kind of people who like qualifications: I'm a collaborator of Dr. McKee. See for example Lin et. al., "Epitaxial Growth of Pb(0.2)Zr(0.8)TiO3 on Si and etc.", Appl. Phys. Lett. 78, 2034 (ref. 27 in the Science article) on which I am an author.]
First, this technology is not enormously new. It's gotten into Science now, which makes it more high-profile, but it dates back at least to a publication in 1998 (McKee, Walker, and Chisholm, "Crystalline Oxides on Silicon: The First Five Monolayers", Phys. Rev. Lett. 81, 3014).
People have been trying to put perovskite oxides on silicon for quite some time (see, e.g., Refs. 12-15 in the PRL paper), for various reasons. [By the way, the prototypical perovskite oxide has the form ABO3, where A is a 2+ cation and B is a 4+ cation; the structure is cubic with a unit cell consisting of A at the vertices, O centered on the faces, and B in the center of the O tetrahedron. They're named after the original "perovskite" mineral, CaTiO3. It has nothing even remotely perverted about it. So get your mind out of the gutter.] One of these reasons is, as the article mentions, gate dielectrics (insulators) for field-effect transistors. A field-effect transistor consists of a conducting channel, usually made out of a semiconductor. By applying an electric field to the semiconductor you can enormously change the conductivity of the channel. However, you don't want the electrode you use to apply this field to short to the channel, so you have to put an insulator in between. There is a lower limit on the thickness of the gate insulator imposed by the desire to limit the leakage current between the gate electrode and the channel. By switching insulators, you can extend this limit.
Another use is for capacitors in DRAM. A DRAM cell works by either storing or not storing charge in a capacitor. The capacitor has to store a certain charge (~10 000 electrons) to be reliably read as on or off. If you recall the definition of capacitance Q = C V and the parallel-plate capacitor equation C = epsilon A / d (epsilon is the dielectric constant and depends on what material you put in the capacitor; A is the area; d is the spacing between plates; V is the voltage across the capacitor), you realize that there are four ways to increase Q: increase V, increase epsilon, increase A, decrease d. V is set by the operating voltages in the device, and there's a lower limit on d, again, due to leakage. You want to decrease A as much as possible to squeeze more cells on the chip. So one attractive thing to do is to switch to a material with greater epsilon. Perovskite oxides give you that opportunity.
Perhaps the most exciting use of COS at the moment is to integrate ferroelectric materials with silicon. Ferroelectrics (e.g. BaTiO3, (Pb,Zr)TiO3, both of which are perovskites) are materials with a permanent electric polarization, just like ferromagnets have a permanent magnetic polarization. You can use this permanent field to store data by switching its direction (hence the notion of a 15-year state in a processsor). One particularly attractive way to do this is to replace the gate dielectric in a field-effect transistor with a ferroelectric. Since the ferroelectric has its own built-in field, the transistor remains on or off without your putting in a gate voltage, and even when you power down the device. Hence you can set the state of a processor built from these things, walk away, and come back years later and it will still be there. These materials are also candidates for use in devices to replace Flash memory and that ilk.
I'd go on but I'm supposed to get back to work
-------------
If the internal fields remain 'for 15 years' then the power consumption of the chip should hopefully be lower, since the cache shouldn't need power refresh.
:).
Caches already don't need (much) power to refresh. They're made of CMOS SRAM, which means that they only dissipate a lot of power when their state changes (ideally they'd dissipate none when the state isn't changing, but there's leakage current across gates and across junctions to the substrate).
Caches are power-hungry because any access has to a) do a tag lookup and b) propagate itself across the entire cache. There are tricks you can use to reduce the power cost of this, but this will still dominate by far over static power requirements.
Heat generation within both caches and the chip core is mainly caused by changing states and shuffling information around, not by maintaining an existing state. I've been studying this for a few years now
"We show that the physical and electrical structure and hence the inversion charge for crystalline oxides on semiconductors can be understood and systematically manipulated at the atomic level. Heterojunction band offset and alignment are adjusted by atomic-level structural and chemical changes, resulting in the demonstration of an electrical interface between a polar oxide and a semiconductor free of interface charge. In a broader sense, we take the metal oxide semiconductor device to a new and prominent position in the solid-state electronics timeline. It can now be extensively developed using an entirely new physical system: the crystalline oxides-on-semiconductors interface."
URLs Abstract and article (subscription may be required...)
My summary:
Oak Ridge National Lab scientsts demonstrate "crystalline oxide semiconductors", that are a combination of Ba-SrO and SrTiO3 on Silicon or BaTiO3 on Germanium. The cool thing is it looks like this will enable germanium field effect transistors that could switch faster than the 210 GHz Si-Ge transistors that IBM can now produce.