0.75% Δ GeO₂-doped silica channel waveguide (6×6 µm) ↔ SMF-28 at 1550 nm — a mode-expanding segmented SSC with all mask features ≥ 0.6 µm. Rigorous local-mode EME device insertion loss 0.082 dB (input coupling 0.055 dB), within the 0.1 dB target; 3D-BPM shown for the field physics.
SMF-28 MFD is ≈ 10.4 µm at 1550 nm. The bare 6 µm core (0.75% Δ) guides a ≈ 7.4 µm mode, so the SSC only gently expands it up to ≈ 10.4 µm to match the fiber. The last field is the mode-field diameter the SSC facet presents to the SMF (the design target).
Analytic Gaussian-overlap estimate for a quick look; the authoritative numbers below are the mode-overlap integral between the true 2-D waveguide/facet mode (from a mode solver) and the SMF-28 field.
Mode field cross-section — blue: SMF Gaussian, red dashed: bare WG mode, green: expanded SSC-tip mode. Larger overlap with the SMF means lower loss.
Butt-coupling loss is the mode-overlap integral between the fiber mode and the waveguide (facet) mode — the robust, standard definition:
Three independent calculations, in increasing rigour:
E_wg and the SMF-28 field E_smf (MFD 10.4 µm), both from an imaginary-distance mode solver, give the butt-coupling mismatch at the chip edge = 0.055 dB. This is the input coupling only; it assumes the on-chip taper is perfectly adiabatic.T = |<E_wg|E_smf>|² · Π|<ψ_i|ψ_{i+1}>|². This adds the real taper non-adiabaticity (including the solid → first-segment junction) and is free of any propagation artefact: 0.082 dB (0.055 input + 0.025 junction + 0.002 taper).Why not use the raw BPM number for the loss? The expanded facet mode is weakly guided (near cutoff), where the scalar paraxial BPM has a known mode-beating artefact. Propagating the exact eigenmode for the device length gives a self-overlap of 99.8 % for the well-guided chip mode but only ~87 % for the near-cutoff facet mode — a spurious ~0.58 dB. So the raw propagate-and-overlap reads ≈ 0.42 dB even though the residual power stays at 0.985 (essentially nothing radiates). The EME (0.082 dB) is free of that artefact; the BPM field plot is shown for the physics, not the loss number.
Chip waveguide 6 µm × 6 µm solid; a 50 µm solid lead-in feeds the segmented taper, which ramps the duty 0.80 → 0.48 (cosine) and the width 6 → 7 µm at pitch 3.2 µm over ≈ 434 µm, expanding the mode from 7.4 µm to ≈ 10.4 µm at the SMF facet. A gap-then-tooth placement keeps every gap ≥ 0.64 µm while the duty starts high (0.80) for a gentle, low-loss start.
| Quantity | Value |
|---|---|
| Input coupling — SSC facet ↔ SMF-28 (mode-overlap) | 0.055 dB (η = 98.75 %) |
| Device insertion loss — local-mode EME (artefact-free) | 0.082 dB (η = 98.14 %) |
| Coupling loss — bare 6 µm chip (no SSC) | 0.49 dB (η = 89.2 %) |
| Chip mode field diameter (D4σ) | 7.4 × 7.4 µm (n_eff = 1.44984) |
| SSC facet mode field diameter (D4σ) | 10.6 × 10.1 µm (n_eff = 1.44585) |
| Facet effective width / duty | 7.0 µm / 0.48 (pitch 3.2 µm, duty 0.80 → 0.48) |
| Min tooth / min gap (DRC, real geometry) | 1.54 µm / 0.64 µm (≥ 0.6 µm ✓) |
| 3D-BPM propagate-and-overlap (see note) | 0.42 dB (residual 0.985; near-cutoff artefact, not the loss) |
Expanding the mode to a round D4σ ≈ 10.6 × 10.1 µm brings it right onto the SMF-28 field (10.4 µm): the input coupling is 0.055 dB. Adding the taper non-adiabaticity — dominated by the gentle solid → first-segment junction (0.025 dB) — the rigorous local-mode EME device insertion loss is 0.082 dB, within the 0.1 dB target. The taper obeys the 0.6 µm minimum linewidth/gap rule (min tooth 1.54 µm, min gap 0.64 µm, verified on the as-drawn GDS), so it is directly mask-writable. The facet mode is weakly guided (near cutoff), so a thick cladding is assumed; the raw 3D-BPM propagate number (0.42 dB) is inflated by the paraxial mode-beating artefact and is not the coupling loss.
The SSC built from this result as a ready-to-place, DRC-clean mask cell. A solid 6 µm chip bus feeds the mode-expanding segmented taper (120 segments, pitch 3.2 µm, duty 0.80 → 0.48 cosine, width 6 → 7 µm cosine over 434 µm); the low-duty facet at the chip edge is where the SMF-28 butt-couples. A gap-then-tooth placement makes every gap a full period, so in the as-drawn geometry every tooth ≥ 1.54 µm and every gap ≥ 0.64 µm (min feature rule 0.6 µm); the solid bus and lead-in are merged into one polygon.
ssc_device.gds (core layer 1/0) — solid 6 µm input bus → segmented mode-expanding taper → 7 µm-wide, 0.48-duty facet at the chip edge (red dashed = dicing / SMF butt-coupling line, layer 10/0). Zoom confirms the DRC min tooth 1.54 µm / min gap 0.64 µm.SSC_0P75PCT_6UM_1550): 150 µm solid input bus + the segmented taper on core layer 1/0, a facet/dicing line on layer 10/0, and a design label on layer 63/0. All features ≥ 0.6 µm; centred on y = 0, +z propagation.numpy, gdstk, matplotlib, bpm3d.py).Layer map — (1,0) waveguide core (0.75% Δ GeO₂-silica), (10,0) facet / dicing reference, (63,0) text label. The taper geometry matches ssc_optimized.gds below; this cell adds the routing bus, facet marker, and label for direct placement.
Everything needed to reproduce the result above:
Run with python3 run_bpm_ssc_075_6um_1550.py --out . (needs numpy, gdstk, matplotlib, and bpm3d.py). It prints the DRC check, the mode-overlap coupling loss (0.055 dB), and the 3D-BPM propagate value, and writes the GDS + figures.
coupling_loss.py mfd_coupling.py Coupling loss tool → MFD & coupling loss tool →