Light as the carrier
A subject the papers are about. The loosest grouping, and the one to reach for last.
What a strand of glass or a waveguide on a die can be made to carry. Collects the optical-communication literature from Kao's 1966 loss argument onward, through the fibre systems that were actually built, to on-die photonic interconnect. The recurring disagreement is whether capacity comes from speed or from parallelism -- more wavelengths, more spatial paths -- and what integration cost each is worth. **Two members added in September 2026 change what the set can argue, and both come from outside optical communication.**
miller1997limit and miller2000rationale supply the quantitative case the set previously lacked. The 1997 paper derives that the bit-rate capacity of an electrical interconnect depends on its **aspect ratio** -- the ratio of length to cross-sectional dimension -- and is therefore **independent of size scale**: shrinking the system does not move the bound, so this is not a limit that Moore's law answers. The 2000 paper carries that forward, notes that splitting one wide line into many narrow parallel ones does not evade it, and draws the comparison in numbers: a single-mode fibre 125 micrometres across and 15 km long carries 100 Mb/s routinely, while an electrical interconnect of the same length and cross-section would manage **less than 0.1 bit/s**.
That is the argument underneath most of this set's engineering. Optical loss is essentially independent of modulation rate; electrical loss is not, and the difference is not marginal but orders of magnitude.
**Miller's own qualification should travel with it**, since it is easy to quote the physics and drop the economics: "**No physical breakthrough is required** to implement dense optical interconnects to silicon chips, though substantial technological work remains. **Cost is a significant barrier** to practical introduction." Written in 2000 by an advocate. The roadmapping and terabit papers at the recent end of this set are, read against that sentence, twenty-five years of the technological work he described rather than the breakthrough he said was unnecessary.
**One member is not about communication at all.** lewis2026finding uses a delta-weakening SMT solver to synthesise photonic circuits that emulate quantum gates, optimising the probability of success. It sits here because the medium is the same and nowhere else in the corpus, and it is the only point of contact between these holdings and the verification sets.