GOOD READ Scientists Create First Working Model of a Two-Qubit Electronic Quantum Processo

[David Farber <dave () farber net>]

Begin forwarded message:

From: Rodney Van Meter <rdv () sfc wide ad jp>
Date: July 6, 2009 7:24:14 AM EDT
To: dave () farber net
Cc: "ip" <ip () v2 listbox com>
Subject: Re: [IP] Scientists Create First Working Model of a Two-Qubit  
Electronic Quantum Processor

Dave, for IP, if you wish, and Happy Fourth of July to all.

Re: Yale's "electronic" two-qubit quantum computer:

http://www.nature.com/nature/journal/vaop/ncurrent/abs/nature08121.html
http://www.nsf.gov/news/news_summ.jsp?cntn_id=115089&govDel=USNSF_51

And related links:
http://qulab.eng.yale.edu/transmon_proj.htm
http://arxiv.org/abs/cond-mat/0703002

Let's cut through the press release BS and the non-technical "quantum
phonebook lookup" stuff and get right to the point:

This paper rocks.

With a paper like this on my c.v., I could die a happy man.

This paper has far fewer caveats to it than almost any other
experimental quantum computing paper I have seen.  There are only one
or two issues with it, which we'll get to.

They have actually implemented a two-qubit system and run both
the Grover and Deutsch-Jozsa algorithms, running ten actual quantum
gates (some one-qubit gates, some two-qubit gates) and getting roughly
90% fidelity out.  That's outstanding, and, as the press release says,
the first time that I'm aware of that anyone has done this with a
lithographically fabricated, solid-state system with electronic
control.

Their system uses the "transmon" qubit type, a word you can expect to
hear more often.  The transmon is a relatively new design, and has
been coming on like gangbusters, that group has produced a bunch of
outstanding results in a very short period of time.

The qubit itself is coupled to a long waveguide which serves as a
resonator cavity, with standing wave modes that interact with the
qubit.  It's like an electronic version of cavity QED, in which
photons bounce repeatedly between two mirrors and interact with a
single atom in the cavity, so this is sometimes called a "circuit
cavity QED" architecture.

Most experimental QC papers have at least one of the following flaws:

* no full two-axis control of a qubit (you have to be able rotate the
 vector of a qubit to anywhere on a unit sphere)
* qubits are hard to initialize
* qubits are hard to read out
* gates are slow
* the coupling between two qubits can't really be turned on and off
* the lifetime of the quantum state is so short you can't really do
 anything with it
* qubit-to-qubit variation means it doesn't really work well

A couple of the more famous architectural proposals have the flaw:

* They're almost impossible to fabricate (so no one has succeeded yet)
* or, as in purely tabletop optical systems, everything is hand-built.

This paper has none of these flaws.

The shortcomings their architecture does have:

* the structures are physically large; the supporting structure for a
 single qubit is 300 microns by ~30, and it must be coupled to a long
 waveguide that serves as the cavity -- 1cm long, give or take; that
 length is dictated by the wavelength of the microwave photons used
 for coupling, and (probably) cannot be shortened.
* since the cavity (which they call the quantum bus, no relation to
 the qubus architecture of my collaborators) is shared, it will be
 limited to only a few qubits, and competition for access to the bus
 will limit performance
* they haven't yet talked about how to scale beyond a single cavity
 with a few qubits; switching microwave photons from one cavity to
 another in principle isn't hard, but a lot of engineering will be
 needed as the loss requirements are probably very stringent.
* beyond switching from one cavity to another, they will need the
 ability to couple qubits in separate chips, which will take yet more
 engineering.

In addition, scalability of fabrication will be an issue; they have
cleverly used a partly optical, partly e-beam lithographic method, but
e-beam is low-throughput.  Despite the overall size of the structures,
a few critical components (notably the Josephson junction at the core
of the qubit) require very precise fabrication.  This is a common
shortcoming in proposed solid-state systems (including the one I am
currently working on), but I think it can be solved -- it's "just"
engineering money, and Intel and others have a VERY strong vested
interest in continuing to shrink the size of features they can
reproducibly fabricate at very high speed.

This system, like many solid-state systems, must be run at EXTREMELY
low temperatures; their experiments were done at 13 millikelvin.  Yes,
about 1/80th of one degree.  This is achieved using a dilution
refrigerator.

One strength of bus-based systems that they didn't discuss is their
natural defect tolerance: if one qubit doesn't work, you just ignore
it and use the ones that do.  In some other systems, nearest-neighbor
interactions are required, and one defect can ruin the utility of a
large chain.

Their 90% fidelity is very good for experimental physics, they
demonstrate very clearly that their system works.  But it's still
about two orders of magnitude from what we need to really build
large-scale quantum computers.  Error rates of ~0.7% are currently
considered to be the "threshold" below which using quantum error
correction makes your net error rate lower, rather than higher, but to
be practical you need to be one to two orders of magnitude better than
that.

They say that their bus will support several qubits; that's the
natural next set of experiments.  I would expect to see 3-5 qubits
from them within a year or two, but then system size will likely
stall, and they will go back to working on fidelity.

                --Rod





-------------------------------------------
Archives: https://www.listbox.com/member/archive/247/=now
RSS Feed: https://www.listbox.com/member/archive/rss/247/
Powered by Listbox: http://www.listbox.com