The Most-Patented Steps Are Not the Hardest: Inside the 7 Process Steps of a CW DFB Laser and 13 Vendors' Patent Portfolios
Co-packaged optics has suddenly made one component critical: the 1310 nm continuous-wave DFB laser. Silicon does not emit light, so every optical lane on a CPO switch is fed by an external laser module (ELSFP), and the laser accounts for roughly 65–70% of that module's bill of materials.
The market currently compares these lasers by output power. 70 mW and 100 mW parts are in volume production, 200 mW and 400 mW parts are sampling, and Lumentum showed a sample above 1 W at 25 °C at OFC 2026. But output power is only the final report card; behind it is the combined result of seven process steps.
1. The Seven Process Steps of a CW DFB Laser
Epitaxy: grow the quantum wells and grating layer on an n-InP substrate, which sets how the optical field is distributed.
Grating: etch a grating with a period of about 200 nm, which makes the laser emit a single wavelength.
Regrowth: bury the grating and grow the p-type cladding and contact layers.
Front-end process: form a ridge waveguide (RWG) or buried heterostructure (BH) that confines light and current to a channel 1–2 µm wide.
Back-end process: thinning, backside metal, cleaving into bars, AR/HR facet coating, bar testing, and singulation into chips.
Chip bonding: bond the chip p-side down with AuSn eutectic onto an AlN submount to carry the heat away.
Burn-in and test: screen out early failures under current and heat, then measure LIV, side-mode suppression ratio and high-temperature performance.
▲ Seven process steps of a CW DFB laser, from epitaxy to shipment (Image source: Simple Tech Trend original)
2. Three Metrics, Each Set by Different Steps
Power: epitaxy, front-end, chip bonding.
Single-mode yield: grating, regrowth, test.
Long-term reliability: regrowth, back-end, burn-in screening.
We mapped 123 public patents from 13 vendors, including Lumentum, Sumitomo, Coherent, Broadcom, MACOM and Mitsubishi Electric, onto these seven steps, and found a contrast.
STT's view: the most heavily patented steps are not the hardest ones. Front-end structures and high-power DFB-plus-optical-amplifier architectures are crowded with patents from the major vendors; the hardest parts, grating fabrication, regrowth and screening data, have almost no patents.
▲ Patent density and know-how difficulty sit in different steps (Image source: Simple Tech Trend original; data from 123 representative patents on Google Patents)
Related context: high-power CW lasers mean fewer lasers per epitaxial wafer; see LandMark Optoelectronics 2026 Q3 earnings call highlights. For InP substrate supply, see AXT FY2026 Q3 earnings call highlights.
Inside the Premium Section
A station-by-station breakdown: structure, why it matters, technical challenges, and what each vendor's patents protect
The RWG vs BH trade-off, and which weakness Lumentum, Sumitomo, Sivers and AAOI each address
High power is not just a bigger DFB: three cavity architectures and Sumitomo's 5 DFB+SOA patent applications
Published power-versus-temperature data from Sumitomo and Lumentum
A patent heatmap of 13 vendors × 10 process steps, and four portfolio camps
What each of the 13 vendors protects, and where its gaps are
7 counterarguments and 4 signals to watch
──────────────────────────────
That's the summary of this article.
STT's full analysis—a station-by-station breakdown of the seven steps, what each of the 13 vendors' patents protects, and how high-power architecture patents are distributed—is in the premium section, with 21 original charts.
👉 Subscribe to STT Premium to read the full analysis on vocus (in Chinese):
──────────────────────────────
Comments