ECTC 2026 | Qnity Electronics | Multi-Mode Polymer Waveguides for Co-Packaged Optics

2 days ago

5 min read

Qnity Electronics (spun off from DuPont's electronic materials business) used ECTC 2026 to push an "unglamorous but manufacturable" path: multimode polymer waveguides made from BCB (benzocyclobutene, CYCLOTENE) dry film within a standard PCB process. The whole flow uses the lamination, panel-level exposure, development and thermal cure that PCB shops already have, so the adoption barrier is low, and propagation loss is as low as 0.07–0.11 dB/cm. Reliability is solid too: no statistically significant change in insertion loss after 10/20 reflow cycles, and only ~0.02 dB/cm added after 1,000 hours at 85°C/85%RH. It also breaks a common assumption, that "well-spaced multimode waveguides don't crosstalk": measurements show that even at spacings of tens of microns, some supermodes still couple over long enough distances (about 1% of power), and whether that matters depends on the loss budget. In one sentence: build optical waveguides into the PCB process and trade the lowest adoption barrier for datacom-grade optical interconnect.

1. Background: building optical waveguides into the PCB process

This paper comes from Qnity Electronics (Wilmington, USA and other sites, spun off from DuPont's electronics business) and was presented at the 2026 IEEE 76th ECTC. Throughput demands from HPC and next-generation data centers are pushing the performance and energy efficiency of traditional electrical interconnect to the limit, and polymer optical waveguides are a candidate for integrating high-bandwidth, low-loss optical connections "directly into the printed circuit board (PCB)."

Qnity uses CYCLOTENE dry film based on BCB chemistry (CWG2100 as cladding, CWG2000 as core); BCB is known in the semiconductor world for its thermal and chemical stability. The key is the process: it can be made with PCB shops' existing lamination, panel-level exposure, alkaline development and thermal cure, so the adoption barrier is low; propagation loss is about 0.1 dB/cm, a strong alternative to copper interconnect.

For context on polymer waveguides as optical interconnect, see Paper analysis | Polymer waveguides withstand +20 dBm for six hours; glass as an optical substrate is another route, see Glass substrates are no longer slideware: the TGV race. Qnity's paper bets on the manufacturing angle of "PCB process compatibility."

Fig. 1: Dry-film waveguide process: laminate the lower cladding, laminate the core and expose/develop it, then laminate the upper cladding to complete the optical stack. Source: Qnity Electronics, ECTC 2026 - Fig. 1

2. The core question: the whole paper in one sentence

What this paper sets out to prove: can multimode polymer waveguides made from PCB-process-compatible BCB dry film simultaneously achieve low loss, acceptable bending and crosstalk, and reliability through reflow and high temperature/high humidity?

The answer is broadly yes, and along the way it corrects the common misconception that "widely spaced multimode waveguides don't crosstalk."

3. Key figures, one by one

3.1 This figure shows "PCB processes can do it, at 0.07–0.11 dB/cm loss"

These figures (Fig. 1, Fig. 3) cover the process and loss. Dry-film lamination + photopatterning (photomask or laser direct imaging, LDI) + TMAH development, fully compatible with PCB workflows. Cut-back measurements: the glass-substrate sample shows propagation loss of 0.07 dB/cm and double-sided coupling loss of 0.09 dB; the dry-film sample on a PCB substrate shows 0.11 dB/cm and coupling loss of 0.26 dB. The PCB sample's higher coupling loss comes from the mode-field mismatch between the 40×40 µm square waveguide and 50 µm round multimode fiber. The difference in propagation loss between the two is still under study (interaction between higher-order modes and sidewall roughness, PCB surface roughness, scattering centers from the dry-film process, and so on).

Key point: reaching ~0.1 dB/cm with existing PCB equipment is the core selling point of the "low adoption barrier."

3.2 This figure shows "the real limits of bending and crosstalk"

These figures (Fig. 5, Fig. 6) cover bending and crosstalk. Bending: S-bend simulations match measurements, and at a 5 mm bend radius, transmitted power drops to about 46%: multimode waveguides are sensitive to bend radius. Crosstalk: insertion loss at 40 µm and 250 µm spacing differs by only 0.01 dB (seemingly no crosstalk either way); but modal analysis reveals a counterintuitive conclusion: even at spacings of tens of microns, some supermodes still couple over long enough distances. Of about 1,200 modes, only about 15 supermode pairs couple significantly; with a balanced mode distribution, about 1% of power takes part in coupling, which only becomes noticeable over long distances. Whether it matters depends on the loss budget and channel isolation requirements.

This is the most valuable section: it debunks the common assumption that "well-spaced multimode waveguides are immune to crosstalk" and gives designers an honest reminder.

Fig. 5: Normalized average transmission vs. bend radius: it drops to about 46% at a 5 mm RoC, showing multimode waveguides are sensitive to bend radius. Source: Qnity Electronics, ECTC 2026 - Fig. 5

3.3 This figure shows "reliability holds through reflow and high temperature/high humidity"

These figures (Fig. 7, Fig. 8) cover reliability. Reflow: no statistically significant change in insertion loss after 10 and 20 thermal cycles, showing BCB's excellent thermal stability. High temperature/high humidity (HTHH, 85°C/85%RH): insertion loss rose only ~0.02 dB/cm after 1,000 hours, so moisture-driven degradation is minimal. The paper also honestly notes that HTHH is a worst case (the optical layer fully exposed rather than sealed inside a PCB stack), and structures fully sandwiched in the stack still await validation.

4. Technical highlights

The first highlight is PCB process compatibility + a low adoption barrier: BCB dry film can be made with PCB shops' existing lamination, panel exposure, development and thermal cure, with propagation loss of 0.07–0.11 dB/cm. For anyone wanting to "lay optical interconnect into the board," this is the path with the fewest process changes.

The second highlight is an honest quantification of multimode crosstalk: supermode analysis debunks the assumption that "enough spacing means no crosstalk," shows that about 1% of power still couples over long distances, and puts that back in the context of the loss budget. This willingness to spell out limitations is more useful as a reference than reporting only good numbers.

5. Industry links: how far from volume production, and who benefits?

Distance: maturity sits at "materials/process + reliability validation." Loss, bending, crosstalk, reflow and HTHH have all been validated, but structures fully sealed inside the PCB stack and system integration with transceivers/connectors still have a way to go. Qnity notes that loss in the dry-film process is still being optimized.

Beneficiaries: most directly, PCB makers and the board-level optical interconnect ecosystem, who can enter optical waveguides with existing equipment; this is home turf for Qnity with its DuPont electronic materials lineage. Next come system makers pursuing on-board optics. A note of caution: this is a multimode route, a different battlefield from single-mode high-speed long-reach. Multimode suits short-reach, board-level, cost-sensitive links, and its bend-radius sensitivity (down to 46% at 5 mm) also constrains routing. Its sweet spot is "low cost, PCB-compatible, short-reach board level," not chasing the highest data rate or the longest reach.

6. Conclusion

The one sentence to remember from this paper: build optical waveguides into the PCB process and trade the lowest adoption barrier for datacom-grade optical interconnect. Qnity's BCB dry film achieves 0.07–0.11 dB/cm, survives reflow and HTHH, and honestly flags long-distance crosstalk in multimode waveguides. For those tracking board-level optical interconnect, the thing to watch: multimode polymer waveguides compete on "PCB compatibility × loss × bend tolerance × crosstalk management." Qnity wins on a low process barrier, and its battlefield is short-reach board level, not single-mode high speed.

References

Ross Johnson, Yaming Jiang, Rui Zhang, James Ryley et al., "Multi-Mode Polymer Waveguides for Co-Packaged Optics," 2026 IEEE 76th ECTC. Qnity Electronics, Inc., Wilmington, DE, USA.