Design of Slow-Light Effects in Truncated-Hole Silicon Photonic Crystal Waveguides (Invited)

Slow light photonic crystal waveguides (PCWs) have shown great potential in integrated optics, particularly for applications requiring enhanced light-matter interaction, optical delay, and compact nonlinear components. However, the trade-off between achieving high group index (ng) and broad bandwidth remains a key challenge in chip-scale implementations. This work aims to address this issue by proposing a dispersion-engineered PCW design based on the W1 line-defect configuration. The approach combines first-row air-hole truncation and second-row cooperative modulation to realize wide flat-band slow-light transmission, while maintaining structural simplicity and fabrication compatibility. Furthermore, a square Fabry-Pérot (FP) cavity is introduced at the waveguide terminal, and is employed in both numerical analysis and experiments to validate the slow-light properties via spectral analysis, enabling accurate extraction of ng variation. The structure is designed on a 220-nm-thick silicon slab corresponding to the device layer of a standard silicon-on-insulator (SOI) platform, with a buried oxide underneath providing vertical index contrast in the fabricated devices. In plane, a triangular lattice with a period of 414 nm and an initial air-hole radius of 0.29a is patterned, and a W1 line defect is formed along the Γ-K direction by removing one row of holes. Different from a conventional W1, the first-row air holes next to the defect are laterally cut to form D-shaped truncated holes, and the remaining part is defined as D′. Reducing D′ increases the local effective index and shifts the guided band, so that a low-dispersion slow-light segment can appear before the intrinsic band edge (Figs. 1(b) and 1(c)). On this basis, a second degree of freedom is introduced by tuning the position or the radius of the second-row holes (Fig. 2): in one case, the holes are shifted toward the line defect; in the other case, their radius is enlarged. Both operations target further dispersion suppression around the designed wavelength. To verify the design on real devices, a pair of square air holes is added at the end of the optimized PCW to form an on-chip FP cavity of 40 μm in length, and the ng is retrieved from the spacing of adjacent resonances and compared with the numerical results (Fig. 3, Fig. 4). Band-structure and ng calculations show that truncating the first-row holes shifts the guided band toward lower normalized frequency, and more importantly, turns the previously steep rise of the ng curve into a segment that becomes flat before the band edge is reached (Fig. 1(c)). When D′ is reduced to 0.6D or 0.5D, a practically usable slow-light window appears in the telecom band, in which the ng variation is limited while the average value stays high. Based on this truncated configuration, shifting the second-row holes toward the defect strengthens the mode-lattice interaction and pushes the band to higher frequency; with a shift of s=0.12a, the sharp ng peak is transformed into a flatter distribution with a low-dispersion bandwidth of 8.25 nm, a relatively high ng of 47.63, and a corresponding normalized delay bandwidth product (NDBP) of 0.249, which is suitable for on-chip delay and nonlinear interaction (Figs. 2(a)-(c)). In the alternative route, enlarging the second-row radius to r′=1.16r produces a red-shifted slow-light window; although the peak ng is slightly lower than that of the shifted case, the flat slow-light bandwidth is broadened to 16.6 nm and the ng is 29.3, resulting in an NDBP of ~0.307 (Fig. 2(d)-(f)). Thus, the two optimizations are complementary: position shift favors higher ng, while radius enlargement favors wider usable bandwidth. To confirm that these effects are not limited to simulations, both optimized PCWs were terminated by square mirrors to form FP cavities, and their transmission spectra were numerically obtained (Fig. 3). For the structure with a second-row position shift (s=0.12a), the simulated resonance spacing decreases toward the target band, and the FSR-extracted ng is 45, which is in good agreement with the bandgap structure results, exhibiting a flat slow-light region with a bandwidth of 8.1 nm (Fig. 3(b)). For the structure with an enlarged second-row radius (r′=1.16r), the same FSR analysis yields a red-shifted and slightly broader flat slow-light window, consistent with the calculated dispersion (Fig. 3(c)). Based on this, fabricated SOI devices were measured. SEM images confirm that the truncated first row, the tuned second row and the square cavity can all be fabricated with good uniformity (Fig. 4(a), Fig. 4(d)). The ng retrieved from the measured FP spectra is 40, with a flat region of 6.1 nm, consistent with the simulation results in Fig. 3 (Fig. 4(b), Fig. 4(c), F ig. 4(e), Fig. 4(f)). This work numerically designed and experimentally demonstrated a slow-light PCW featuring a truncated first row of air holes jointly tuned with a second-row air-hole modification. The proposed structure preserves the simplicity and CMOS compatibility of the W1-type PCW. With the first-row truncation fixed at D′=0.6D, numerical results show that the position-offset design achieves a high group index slow light regime at s=0.12a, delivering =47.63 and Δλ=8.25 nm with a corresponding NDBP of 0.249. By contrast, increasing the second row hole radius to r′=1.16r reduces to 29.26 but expands the low-dispersion bandwidth to 16.59 nm, boosting the NDBP to ~0.307. By integrating a square FP cavity and retrieving ng from the measured FSR, both optimized structures yield a flat slow-light window with =40, where Δλ is 6.1 nm for s=0.12a and 9 nm for r′=1.16r; the slow-light window position and its evolution trend are consistent with the theoretical predictions. The proposed approach provides a feasible structural route for integrating slow-light-enhanced nonlinear optics, optical delay, and ultrafast pulse characterization in photonic integrated circuits.