A big-picture roadmap for taking photonics from millions → billions of units: on-chip lasers, low-power tuning, smarter packaging/test, and co-design with electronics—what must happen next and why it matters.

We close Season 1 by zooming out to the industry roadmap. This Nature Communications perspective maps silicon photonics from early demos to SSI → MSI → LSI → VLSI eras and asks: what unlocks true scale? Four simple takeaways:

1. Integrate light at the source. Moving from fiber-attached lasers to hybrid/heterogeneous on-chip solutions (and, longer term, monolithic growth) shrinks boxes, lowers loss, and simplifies products.

2. Tune with tiny power. Today’s heaters work but waste energy and create crosstalk. Next-gen low-power phase shifters (e.g., MEMS/NOEMS, Pockels materials) and better modulators (lower α·Vπ·L) are key to large, stable chips.

3. Package like pros. Lower-loss fiber coupling, passive alignment (e.g., photonic wire bonding), and scalable test flows turn great dies into great products.

4. Co-design with electronics. Photonics + CMOS go hand-in-hand: practical systems will mix 2.5D/3D integration, chiplets, and smart control to hit cost, power, and reliability targets.

Where this shows up first: data-center links (IMDD/coherent), photonic switching, LiDAR, sensing, and early photonic computing/quantum—each with a short list of “must-fix” challenges (loss, laser efficiency, driver/receiver noise, reliable tuning). The paper also lays out near-term milestones: wider availability of integrated lasers/SOAs, maturing low-power phase shifters, better fiber I/O, and a stronger design/PDK ecosystem.

Source (with authors):
Sudip Shekhar, Wim Bogaerts, Lukas Chrostowski, John E. Bowers, Michael Hochberg, Richard Soref & Bhavin J. Shastri. Roadmapping the next generation of silicon photonics. Nature Communications (2024).

A programmable photonic chip that measures its own phase and auto-calibrates in ~25 iterations—using an on-chip reference path + Kramers–Kronig to recover the full complex response over hundreds of GHz.This episode unpacks “Self-calibrating programmable photonic integrated circuits” by Xingyuan Xu, Guanghui Ren, Tim Feleppa, Xumeng Liu, Andreas Boes, Arnan Mitchell & Arthur J. Lowery (Nature Photonics, 2022). The problem: big PICs drift and vary—fabrication errors and thermal cross-talk break the link between control voltages and optical function. The fix: add a short-delay on-chip reference path, sweep a laser, and use the Kramers–Kronig relationship to recover phase from intensity, then inverse-FFT to get the complex impulse response and update each MZI/phase shifter—no prior device model needed.

They demo it on a 16-tap FIR with an 8-tap signal-processing core, FSR ≈ 160 GHz (6.25 ps delay steps), dialing diverse functions—complex sinc filters, Hilbert transformer, half-band low/high-pass, differentiator—and converging in ~25 training iterations. Platform: SiN (TriPleX) PIC with ~0.15 dB/cm loss; the reference adds <3 mm² footprint. Result: dial-a-transfer-function across the C-band with fast, stable calibration—a path to real-time reconfigurable comms, neuromorphic optics, and quantum.

Source: Xu et al., “Self-calibrating programmable photonic integrated circuits,” Nature Photonics (2022).