Inverse designed photonic crystal waveguides for pulsed operation: dispersion, losses and controlled light-matter interactions

Dominic Thompson, Stephen Hughes, Nir Rotenberg · 2026

Read Inverse designed photonic crystal waveguides for pulsed operation: dispersion, losses and controlled light-matter interactions by Dominic Thompson, Stephen Hughes, Nir Rotenberg (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-24639.txt.

Introduction

Thompson, Hughes, and Rotenberg inverse design photonic crystal waveguides that handle pulses, not just steady light. These waveguides confine light below the wavelength scale and control its group velocity, enhancing light matter interaction. Applications need a wide band of low loss, constant velocity slow light, though. Earlier dispersion engineering could not deliver that combination. Their 2026 method pairs inverse design with a fast mode solver and physics-based formulas.

Compute time falls by more than a hundredfold. Waveguides with up to an order of magnitude more bandwidth, and several times lower loss than prior state of the art, are unlocked. Trade offs among bandwidth, disorder induced loss, group index, and dispersion are mapped. Two application studies show the reach. They cover broadband, position tolerant Purcell enhancement for quantum light sources, and compact phase shifters for optical communications. Thompson's team offers the workflow as a general design method for integrated photonics.

Catalog record

CREATORDominic Thompson, Stephen Hughes, Nir Rotenberg
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE45 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.24639

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Inverse designed photonic crystal waveguides for pulsed operation: dispersion, losses and controlled light-matter interactions Dominic Thompson, Stephen Hughes, Nir Rotenberg Abstract Photonic crystal waveguides (PCWs) are a powerful platform for optical technologies because they can spatially confine light on sub-wavelength scales and manipulate the group velocity of propagation modes, both of which enhance light-matter interactions. Many applications in photonics require a large bandwidth of low-loss and constant-velocity slow light, a significant challenge for previous dispersion and Bloch mode engineering techniques. By combining inverse design with an efficient mode solver and physics based formulas, we reduce the computational time of PCW designs by more than 100 100 times, allowing for the realization of PCWs with up to an order of magnitude increase in bandwidth and up to 4 4 times decrease in loss. We then explore the trade-offs between bandwidth, disorder-induce loss, group index, and dispersion. As examples, we apply this approach to two active and practical areas of research for PCWs design: broadband, position-tolerant Purcell enhancement, and compact phase shifters for optical communications. Our results significantly improve state-of-the-art PCW designs and provide a general method to optimize PCWs integrated technologies. I Introduction Photonic crystal waveguides (PCWs) shape light confinement, controlling its flow, spatial structure, and properties at sub-wavelength dimensions 40 ; 61 ; 19 ; 37 . These properties allow PCWs to enhance nanoscale light-matter interactions, for example, to enhance optical nonlinearities for all-optical signal processing 23 ; 3 ; 24 ; 52 , sensing applications 26 ; 66 , and efficiently interface with quantum emitters, providing a platform for light-based quantum circuits 17 ; 30 ; 46 ; 29 ; 27 . Many photonic technologies use short pulses of light, with typical pulse durations ranging from nanoseconds down to 10’s of femtoseconds, and differing frequencies may be used at once to increase the device capacity or processing power. For a typical PCW, these can pose a significant challenge, as illustrated in Fig. 1 (a). Here we observe the calculated profile of three Gaussian pulses [lengths of 3.3 (blue), 6.6 (megenta), and 13.2 ps (orange)] as they propagate through a typical 120 \mu m W1 PCW [top panel in Fig. 1 (a); W1 geometry and mode profile shown in the left panel of Fig. 1 (c)]. Both the shape and relative temporal position of each pulse change as it propagates, as is most evident in the zoom in on the final positions [Fig. 1 (b)] where we show both the initial (dashed grey) and final pulse (dashed color) profiles. In particular, the length of the shortest pulse has increased from 3.3 ps to 5.6 ps, and its amplitude has almost completely attenuated . The pulse reshaping occurs due to two main effects. First, in the typical PCW, different frequency components propagate with different velocities, because of dispersion, which causes the pulse to pull apart and broaden. This can be seen in the dispersion relation of the waveguide, shown in Fig. 1 (d) in the dashed curve, along with the group index n_{g} of the mode. In particular, note that the shortest pulse, whose bandwidth is denoted by the blue shaded region, spans a range of n_{g} from 33 to 130. Second, PCWs can be very lossy because of disorder-induced scattering, particularly for frequencies near the band-edge where light slows down and light-matter interactions are enhanced 8 ; 16 ; 25 ; 43 , as is the case for the shorter pulse. Figure 1: Inverse design of large-bandwidth, slow-light PCWs. (a) Calculated pulse profile of Gaussian pulses with different central frequencies and pulse-widths (13.2 ps in orange, 6.6 ps in magenta, and 3.3 ps in blue) travelling down both the original W1 waveguide (top) and an optimized waveguide (bottom). (b) Zoom in on the pulse profile at the end of the PCW, comparing the initial profile (dashed grey curve) with those of the unoptimized (dashed colored curve) and optimized (solid colored curve) designs. (c) Example intensity distribution of the electric field Bloch mode at 193.5 THz of a unit cell of the original (left) and optimized (right) PCWs in the center of the slab. (d) The corresponding dispersion relations and group index for the two designs. Dashed lines correspond to the original design and solid lines to the optimized design. The orange, magenta, and blue regions correspond to the pulses shown in (a), the gray region to the left is the light line, and the lower gray region is the edge of the band gap for the optimized structure. (e) The number of channels supported by different design bandwidths for different pulse widths. Each channel width is assumed to be three times the transform-limited frequency bandwidth of the Gaussian pulse. Bandwidths for the original design (dashed lines) and best-case improvements (solid lines) are shown for different n_{g} ’s (10 in light gray, 20 in dark gray, and 30 in black). (f) The total device length that can be made given different waveguide losses and total acceptable loss. Best-case improvements to the original device are shown [dashed to solid curves, as in (f)]. Disorder-induced losses are inevitable and arise due to the interaction of the propagating light field with the edges of the holes [cf. Fig. 1 (b)] and scale with n_{g}^{2} 42 ; 32 , as discussed further below. These can significantly limit the practical applications of PCW-based devices, for example in the number of distinct photonic channels that they can support or in their lengths, shown by dashed lines in Figs. 1 (e) and (f), respectively. Dispersion engineering of PCWs has focused on overcoming these effects by increasing the bandwidth across which they support slow light and pulling this region away from the band edge 28 ; 7 . Much of this work has involved a typical brute force approach where the PCW geometry (e.g., hole sizes or positions) is systematically varied, though recent optimization methods such as particle swarm optimization 15 , deep learning approaches 62 ; 4 , and gradient-based inverse design 54 ; 39 ; 38 ; 33 have begun to partly address this issue. In this work, we use the gradient information provided by an efficient mode-solving method, Guided Mode Expansion (GME) 2 ; 64 ; 34 , in conjunction with physics-based formulas and powerful gradient-based inverse design methods 65 to perform multi-dimensional optimization of PCWs rapidly and efficiently. This enables us to solve for 102 million modes across 25,500 separate optimizations targeting different bandwidths, group indexes, and losses while maintaining single-mode operation, for telecom-wavelength PCWs, namely near 1.5 microns (corresponding frequencies of around 200 THz). Our calculations reveal an inherent trade-off between different parameters, which can be optimized for specific applications. As shown by an exemplary optimized waveguide (Fig. 1 (c), right panel), we can find large-bandwidth slow-light modes [here, n_{g}\approx 30 ; solid lines in Fig. 1 (d)] that drastically increase the bandwidth by almost 2-orders of magnitude (Fig. 1 (e), solid lines) with a commensurate increase to the number of supported optical channels. Similarly, the losses of the optimized PCW decrease by about an order-of-magnitude across the n_{g}=30 region, as shown in Fig. 1 (f) (solid lines). Altogether, the drastic improvement in PCW performance can be observed in the pulse propagation, which now

Read the complete text