Weird that the fab work was done at a new york university, doesn't MIT have facilities for this type of thing?

Direct link to the paper - https://www.nature.com/articles/s41586-018-0028-z

sci-hub link http://sci-hub.tw/https://www.nature.com/articles/s41586-018... (https doesn't work because their CA decided to censor them)

Sci-Hub.tw has been brought down

What single-core frequencies could we hope to achieve if chips were working entirely on optical components? There are no heat problems, or is there?

Signal propagation times still put a size limit, then you have optical attenuation to deal with in tiny fibres, requiring amplification. Edit: apparently heater too for tuning stability. Not especially power saving then.

The main gain is that you can put components further away if latency is not a concern and can use parallel buses due to no interference. So the tech is most suited to replace long range busses like PCIe.

Optical switches are neither small nor efficient compared to a VLSI CMOS transistor. Few things are.

It makes sense two use optical switches for certain specific requirements until they surpass VLSI CMOS in every requirement