Anyway, I'm not making the claim that this approach is better than any other approach. It's just different, and these differences may sometimes be useful.
I certainly won't compare this type of antenna-based lens to a zone plate as far as applications go because I don't want to speculate too much. I should say though that while you need to be in the "far field zone" to see the desired beam from a diffraction optics element such as a zone plate (see http://en.wikipedia.org/wiki/Fraunhofer_diffraction), our design forms the desired wave-front almost immediately after the layer of antennas. I'm not sure if it's useful to have a curved wave-front that a lens imparts immediately after the flat wave surface or it's just a curiosity, but it's at the very least interesting. For other applications (for example we've demonstrated vortex plates here with the same technique: http://www.seas.harvard.edu/capasso/wp-content/uploads/publications/Genevet_APL_100_013101_2012.pdf) this may more relevant.
Thank you for your comment. It's not true though that we're making small micro-lenses. A micro-lens is still a regular lens; at the very least that means that it is larger than the wavelength of light. The antennas used here are smaller than the wavelength of light, made of metal, and change the phase of the light due to a resonance behavior. You can see the particular shapes and sizes of the structures used in Fig. 2 of the paper (http://arxiv.org/abs/1207.2194)
As an FYI, many articles that are pay walled can be found on the arxiv pre-print server for free.
Can it be adapted to be *useful* with visible light? Unclear for a variety of reasons. The first is that shifting to the visible will increase metal losses, so more of the light will simply be absorbed instead of focused. Not that the efficiency isn't an issue already: from the article you can see that with the current design, the maximum attainable efficiency is ~10%, with the rest of the light being absorbed (not that much actually) and scattered somewhere else (this is the big one). In fact the presently demonstrated lens has an even lower efficiency, though scaling it up to the 10% figure is fairly trivial. Anyway, in the visible the 10% figure probably drops with the current design, though some design improvements could likely be made. I don't want to give you an upper bound on the efficiency because frankly I'm not sure. Anyway, do you want a lens that only focuses some percentage (say between 10% and 40% just to have some numbers) of the light and throws away the rest? We've gotten so good at making regular old lenses in the visible, that I'm not so sure. On the other hand go to a different frequency range where good lenses are less common, and all of a sudden the present approach may have some value.
There is another issue as well. Look at this diagram of a "beam expander" (or telescope): http://www.cvimellesgriot.com/glossary/imagesDir/BeamExpander.gif
You'll see that while it has two lenses which no doubt have some thickness, there is also some space in the middle. With the approach in the article, the lenses can be very thin. However, to make a telescope there still has to be space in the middle. Can that be overcome to some extent (for example with very high numerical aperture ultra-thin lenses)? That's yet to be determined.
If you wish to make visible or near-IR lenses, you are stuck with things like electron beam lithography and focused ion beam, so it's extremely expensive. Some newer fabrication techniques such as nano-imprint lithography could maybe bring the cost down. .
If you want to scale the entire lens up to, say, terahertz frequencies, you can make the same structures with photolithography and the price goes down tremendously because photolithography is a parallel process (every part of the pattern gets written at once instead of writing each spot point by point).
A computer without COBOL and Fortran is like a piece of chocolate cake without ketchup and mustard.