Length-Matching Routing for Programmable Photonic Circuits Using Best-First Strategy

Xiaoke Wang, Dirk Stroobandt · 2025

Read Length-Matching Routing for Programmable Photonic Circuits Using Best-First Strategy by Xiaoke Wang, Dirk Stroobandt (2025) 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-2509-23463.txt.

Introduction

Chips that steer light need wires of an exact length, not the shortest path. Xiaoke Wang and Dirk Stroobandt study that routing task for programmable photonic circuits, where parts such as ring resonators and delay lines depend on precise lengths. They present length-matching strategies built on best-first search, guided by a new scoring rule that never overstates the cost left. A pruning step trims dead ends from the search.

For two-pin routes, the pair prove their methods complete when the scoring rule stays consistent. Multi-pin routes get a pin-ordering trick based on detour room, which lowers the chance of blocking a workable route early. Tests on varied layouts show how run time and search quality trade off.

Catalog record

CREATORXiaoke Wang, Dirk Stroobandt
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE38 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-23463

Browse the full library · About the Tower of Babel

Full text

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

Length-Matching Routing for Programmable Photonic Circuits Using Best-First Strategy Xiaoke Wang, Dirk Stroobandt Abstract In the realm of programmable photonic integrated circuits (PICs), precise wire length control is crucial for the performance of on-chip programmable components such as optical ring resonators, Mach-Zehnder interferometers, and optical true time-delay lines. Unlike conventional routing algorithms that prioritize shortest-path solutions, these photonic components require exact-length routing to maintain the desired optical properties. To address these challenges, this paper presents different length-matching routing strategies to find exact-length paths while balancing search space and runtime efficiently. We propose a novel admissible heuristic estimator and a pruning method, designed to enhance the accuracy and efficiency of the search process. The algorithms are derived from the Best-First search with modified evaluation functions. For two-pin length-matching routing, we formally prove that the proposed algorithms are complete under monotonic heuristics. For multi-pin length-matching challenges, we introduce a pin-ordering mechanism based on detour margins to reduce the likelihood of prematurely blocking feasible routes. Through evaluations on various length-matching benchmarks, we analyze runtime and heuristic performance, demonstrating the effectiveness of the proposed approaches across different layout scenarios. I Introduction Figure 1 : Hexagonal programmable photonic circuit with integrated phase shifters and tunable 2\times 2 couplers (a), showing coupler states (b), three other different architectures of programmable photonic circuits (c–e), and directed routing resources graph of tile and coupler (f, g). Silicon photonics and Photonic Integrated Circuits (PICs) have seen significant advancements in recent decades, offering compact, high-speed, and energy-efficient solutions for optical signal processing. Among these developments, programmable photonics has emerged as a promising approach [ 1 , 2 ] , enabling reconfigurable circuits composed of tunable elements and functional building blocks. Unlike traditional PICs, where connectivity is fixed, programmable architectures allow dynamic configuration, introducing new challenges in circuit routing and resource allocation. Various routing algorithms have been proposed to address these challenges in programmable photonics, such as sequential routing [ 3 ] , Aurora [ 4 ] , and CRoute [ 5 ] . However, existing solutions mainly focus on shortest-length routing while satisfying photonic constraints, with an emphasis on scalability and efficient resource usage. In addition to resource utilization, another key aspect of routing in photonics circuits is the management of optical waveguide lengths. Unlike electrical interconnects, where wire length primarily affects resistance and capacitance, optical paths must be carefully controlled to maintain phase coherence and timing synchronization. The degree of length sensitivity varies depending on the application. In some cases, such as simple interconnects, precise length control is not critical. In contrast, for components like Mach-Zehnder Interferometers (MZIs), waveguide length directly affects phase delay and must be strictly managed. This distinction is particularly crucial in programmable PICs, where waveguide paths must be dynamically adjusted to achieve specific optical behaviour. Length-matching techniques have long been studied in the printed circuit board (PCB) and analogue circuit design [ 6 , 7 ] , where ensuring signal arrival synchronization is essential to prevent timing mismatches. Similar constraints arise in photonics. Unlike electronic circuits, photonic circuits are susceptible to phase errors caused by path length mismatches. While phase shifters offer limited compensation—typically up to a single wavelength—larger mismatches from mesh-based routing cannot be corrected post-fabrication. Thus, precise length matching during routing is essential to ensure phase coherence and maintain the performance of programmable PICs. In traditional photonic circuits, physical design is mostly schematic-driven [ 8 ] , and routing is often performed manually or optimized for specific layouts, and also global route planning [ 9 , 10 ] . However, in programmable PICs, the challenge is different: the interconnect structure is based on a regular tiling architecture, meaning that paths must be computed dynamically while satisfying strict length constraints. This introduces additional complexity, as conventional shortest-path routing is insufficient: Detours must be introduced to ensure that the path length (number of waveguide segments used) exactly meets the required value. This makes length-matching in programmable PICs a unique problem that demands specialized routing algorithms. To the best of our knowledge, the length-matching routing problem in programmable photonics has not been previously addressed. The key distinctions between our approach and existing routing algorithms in this field are summarized in Table I . Table I : Comparison of routing algorithms for programmable photonic integrated circuits (PICs). Router Length Matching Fan-out Support Global / Local Router Sequential Routing [ 3 ] ✗ 2-pin only Global Aurora [ 4 ] ✗ 2-pin only Global C-Route [ 5 ] ✗ ✓ Global This Work (H \& DS-LM) ✓ ✓ Local Component This work focuses on addressing this challenge by introducing an efficient search-based length-matching (LM) routing algorithm. The proposed approach extends best-first search strategies to enforce strict length constraints while exploring feasible detours within the regular tiling architecture of programmable PICs. Furthermore, a dual-stage LM variant is introduced as an extra option, allowing for initial shortest-path exploration followed by targeted detours (close to the target) to achieve precise length matching. The proposed methods provide a practical solution for length-matching in programmable photonic circuits, enabling more efficient and flexible circuit implementations. II Preliminaries II-A Programmable Photonic Circuits We target programmable photonic integrated circuits (PICs) based on a hexagonal mesh of tunable 2\times 2 couplers and integrated phase shifters, as shown in Fig. 1 (a). Each coupler supports three programmable states: bar, cross, and partial coupling (Fig. 1 (b)), allowing flexible routing of light signals. Alternative architectures, such as feed-forward meshes and rectangular or triangular feedback loops, are illustrated in Fig. 1 (c-e). In this paper, we limit our scope to problems based on the hexagonal architecture. For routing purposes, we construct a directed routing resource graph where each coupler and interconnect is represented by multiple vertices and arcs, as shown in Fig. 1 (f-g). This abstraction captures the directionality of light propagation and enables enforcement of physical constraints such as no back-propagation and loop-free paths. Activated, forbidden, and feasible states are encoded in the graph to reflect current routing conditions. II-B Problem Formulation We first clarify the routing requirements for programmable photonic circuits, including physical restrictions. The tunable coupler serves as a unique device within the routing process. Therefore, a customized model needs to be developed, and specific rules must be established to adhere to the physical restrictions. The routing resources graph needs to obey the optical physical restriction, so we summarize the rules into the following two restrictions: 1. Acyclic: Avoiding loops (closed paths) in the circuit to prevent undesirable interference effects, in line with the routing requirements. 2. Avoid Opposite: Each photonic connection or edge in the graph G is inherently directional, allowing light propagation in only one specified direction along the waveguide during routing, as illustrated in Fig. 1 (f-g). A pair of two directional arcs represents each optical waveguide segment. Once an arc (shown in green) is utilized, its opposite arc (shown in red) becomes forbidden. II-C Length Constraints Except for the foundational restrictions introduced above, length-matching routing has an obvious extra constraint: the length. In typical photonic applications, such as optical true time delay lines (OTTDs), Mach-Zehnder interferometers (MZIs), and optical ring resonators (ORRs), the path length directly determines the device’s optical functionality and phase relationship. For interference-based components such as the MZI shown in

Read the complete text