The article is very light on the technology. (Quad Junction) GaAs cells have been produced with >40% conversion efficiency. That's more than three times the efficiency of Si or anything else in common use. The GaAs cells have thus far only been used for space-based applications. Somebody should be working on getting production costs down and deploying these things terrestrially.
40% is not, in fact, 3x, or even 2x, current best efficiency of Si or CdTe panels.
And, the important efficiency measure is W/$, where GaAs trails well back. Perovskites may exceed 40% conversion efficiency, and also have good prospects to offer much better W/$. Their endurance has grown encouragingly quickly.
If you looked at the first link I cited, they claim 47% efficiency is possible. That is more than 3x the efficiency of many Si panels on the market today. Unlike perovskites, which to date have only achieved ~25% efficiency, GaAs cells have been produced with efficiencies approaching 70% (under laser illumination). https://www.ise.fraunhofer.de/en/press-media/press-releases/...
W/$ goes down with manufacturing efficiency. Nobody has done large scale manufacturing of multi-junction GaAs cells.
If you looked at the first link I cited, they claim 47% efficiency is possible. That is more than 3x the efficiency of many Si panels on the market today.
Top-end GaAs technology wouldn't compete with bottom-end silicon technology (the below-16% panels that you mention). It would compete with top-end silicon technology, which is already commercially available at module efficiencies above 22%:
GaAs cells reported to operate above this limit are part of multi-junction cells and/or incorporate optical concentration systems to focus sunlight to higher intensities. The experimental record-holder of 47.1% uses a multi-junction cell plus optical concentration. Optical concentration only works with direct normal illumination; light that is scattered through haze or clouds can't be focused, so optical concentration systems are a good match only for sunny areas that have clear skies year-round.
That said, the National Renewable Energy Laboratory is researching ways to make GaAs cells at lower cost:
At low enough cost, GaAs modules could compete directly with premium silicon solar modules. They could conceivably be lighter as well as more efficient than silicon, since thinner layers of GaAs are needed than silicon layers, which in turn reduces the required module rigidity, thickness, and weight.
The fine article was about the efficiency of cells vs. their geometric footprint. If the cost of the footprint is considered, smaller GaAs cells might be more economical than the larger (but cheaper) Si cells. Also, conentrators might be nothing more than a few mirrored/angled surfaces at the edge of the cell, which would increase cell illumination independent of cloud cover. It's not rocket science.
If we are talking about solar installations on land, rather than in the built environment, the W/$ also needs to factor in the cost of the land itself, which is at a premium in many places including across Europe. On that basis there is scope for an increased efficiency to lower the cost delta.
Yes, but is anyone seriously considering multi junction panels for stationary applications? When there is an abundance of free rooftops and parking lots and plenty of low quality rural land for under 20 cents per square metre it doesn't seem to make a lot of sense to pay over double for less than double efficiency.
For the case of perovskite cells, adding layers does not seem to multiply the production cost.
Certain uses place a premium on areal efficiency, notably aerospace.
But anybody who has maxed out their own collection area, whether a roof or reservoir, can get more revenue from using better cells. And, the more of that that is done, the faster their price falls. We may reasonably expect the price of multilayer perovskites to drop below cost of silicon panels. Sooner is better, which is driven by maxed-out demand.
https://arxiv.org/pdf/1905.08024.pdf
https://boeing.mediaroom.com/2010-11-01-Boeings-Spectrolab-P...
https://www.spectrolab.com/photovoltaics.html