Friday, March 03, 2006

MS+S: Bob Clark

Bob Clark represented 150 people, funded by both the Australian and U.S. governments. The overall effort in Australia is major, including theory, semiconductor, and optical. He talked mostly about progress in fabbing their solid-state devices. They are famous for the Kane model, of course, but Bob says they have "moved on" from that somewhat.
In one sense, these guys set themselves a difficult task. Although it's based on standard silicon, the fabrication of their devices is complex, compared to the superconducting teams. The superconducting folks have their own problems, of course, including (for some devices) some issues with aligning structures to a particular axis of the crystalline lattice, but I think overall the superconducting devices are a little easier to fab.
In the Australian system, the idea is to implant phosphorus atoms below the surface of silicon, and the P will localize an electron that can then be used as a qubit. They have to put in the P, layer some more Si over it, then build standard CMOS gates aligned to the P atom. Their fab approach is to cover the Si with a protective layer, etch small wells down to the Si surface, sputter P atoms so that they get one into each well, then strip off the protective layer, put down some more Si, then build the gates.
One part of their fab uses an idea that had never occurred to me before: they build electrical structures on the chip that are then used during later steps of the fab. We're accustomed to passive alignment markers on the wafer to get proper registration between layers, but not active elements. In particular, when a P atom impacts the Si surface, it generates about a hundred electron-hole pairs, and they put a detector near each of the wells so they can know with absolute certainty when an atom has made it into the well. Bob says they have ideas that may scale to allow them to reliably implant a million atoms in precise locations (plus/minus a few nanometers) on the surface of a wafer.
Their first experiments are aimed at creating a charge qubit, and Bob expects that moving from charge to spin will be straightforward. They are currently doing experiments on a two-atom device, with the atoms placed 50nm apart, each coupled to an SET (single electron transistor). For this setup, T1 is 10msec. As the atoms are brought closer together, the lifetime will go down, but if they can make the spin device work, lifetimes will be very long; tens of msec, maybe hundreds of msec.
They have also been doing a lot of work on integrating various important control structures directly into the device, which will reduce the need for signal generators and whatnot outside the dil fridge. To me, this is critical work, and I hope they are successful, and that some of what they learn can be applied to other solid-state technologies.
Bob's talk included a number of fantastic pictures of their data; I wish the talks were online. One paper is this one by Schofield et al.
They don't really have a working qubit yet; hope to in a few months.

0 Comments:

Post a Comment

<< Home