Vertical-cavity surface-emitting lasers (VCSELs) have demonstrated significant potential for all-optical signal processing, especially in the realization of optical flip-flop devices. However, the theoretical relationship between the switching time and the intrinsic laser parameters remains largely unexplored. Our research conducts a numerical emulation utilizing the spin-flip model to analyze polarization switching behavior under polarized optical injection. The dependence of flip-flop switching time on laser internal parameters is systematically analyzed. Subsequent parameter optimization achieves minimized switching time, permitting high-speed operation at repetition frequencies as high as 10 GHz. Further improvements in repetition frequency are attainable by implementing injection detuning and carefully chosen injection intensity.
Optical logic gates offer a promising route toward high-speed, energy-efficient photonic computing. Still, conventional intensity-encoded architectures often require a reference or bias light to implement nontrivial Boolean functions such as NAND and NOR. Here, we propose an inverse-designed optical logic gate that encodes binary information in guided spatial modes rather than optical intensity. In the proposed scheme, the logical states “0” and “1” are represented by the TE$_{00}$ and TE$_{10}$ modes of a multimode silicon waveguide, enabling both logic states to carry optical power and eliminating the need for an external reference beam. Using full three-dimensional adjoint-based topology optimization, we design six two-input, single-output Boolean gates on a silicon-on-insulator platform, including AND, OR, XOR, NAND, NOR, and NXOR, within $4 \times 4~\mu \rm {m}^{2}$ footprint. The optimized devices realize the target truth-table behavior by selectively transmitting the desired optical modes while suppressing unwanted mode crosstalk, ensuring that each input combination yields the correct logical output (0,1) for the corresponding optical modes (TE$_{00}$, TE$_{10}$). The XOR gate maintains a signal-to-noise ratio(SNR) above 5 dB across the 1.5–1.6 $\mu$m wavelength band despite being optimized only at 1.55 $\mu$m, owing to distributed multimode interference rather than narrowband resonance. We also show that the same physical structure can switch between XOR and NXOR operations by tuning only the relative phase between the two input signals. These results demonstrate compact, broadband, and reconfigurable mode-encoded Boolean logic elements while identifying the power and phase requirements that must be addressed for their extension toward multistage photonic circuits.
Coherently coupled ring-arrays of vertical cavity surface emitting lasers (VCSELs) have been investigated with the goal of increasing brightness. The in-phase supermode has a narrow divergent axis intensity lobe created by 0-degree phase shift between the elements of the coherently coupled array. We demonstrate a method for predicting the operating bias conditions for the in-phase supermode and show an approach for reduction of the number of independent biases needed. Finally, we demonstrate the optimized beam profile characteristics that are obtained using the predicted array bias conditions.
Next-generation intensity-modulation (IM) and direct-detection (DD) systems used in data centers are expected to operate at 400 Gb/s/lane and beyond. Achieving such data rates requires either increasing the system bandwidth or employing denser modulation formats, both of which demand higher signal-to-noise ratio (SNR). Such SNR requirements are typically met by increasing the transmitted optical power. However, operating at high optical powers enhances the impact of relative intensity noise (RIN), a signal-dependent impairment inherent to the transmitter laser, which ultimately limits the performance of the system leading to an error floor in the bit error rate (BER). Six-ary pulse amplitude modulation (PAM-6) is a promising candidate for the modulation format of next-generation systems. In this paper, we study the optimal PAM-6 constellation design and use current forward error correction (FEC) standards to predict the performance of 600 and 800 Gb/s/lane systems. We propose a new PAM-6 modulation that improves the SNR and BER for both a standalone KP4 and a concatenated (KP4+Hamming) FEC schemes. Next, the geometry of the constellation is optimized to further improve the system performance. Our results show that a geometrically-shaped PAM-6 decreases the optical power requirement at the pre-FEC KP4 threshold and lowers the BER error floor by up to three orders of magnitude.
Machine-learning inverse design of nonlinear plasmonic metasurfaces is limited by the cost of full-wave simulation and by the scarcity of labeled nonlinear optical data. We demonstrate a personal-computer-scale workflow for periodic Au nanorod metasurfaces in which normalized linear finite-difference time-domain (FDTD) fields are extracted on near-surface shell planes and converted into relative second-harmonic-generation (SHG) proxy targets. A 35-sample Latin-hypercube dataset over nanorod length, width, height, lattice period, and pump polarization was generated using MEEP and used to train a radial-basis-function kernel-ridge-regression surrogate. Five-fold cross-validation gives a KRR mean absolute error of 0.493$\pm$0.123 log$_{10}$ units across three SHG-proxy targets, compared with 0.557$\pm$0.150 log$_{10}$ units for degree-1 ridge regression. Direct surrogate maximization overpredicted sparse regions, motivating local active refinement, an offline Bayesian-optimization baseline, and explicit convergence testing. The original sharp-block surface proxy did not pass mesh convergence. A smooth ellipsoid, mesh-locked convergence branch reduced the successive-resolution change from 0.305 log$_{10}$ at 80–100 px/$\mu$m to 0.072 log$_{10}$ at 120–140 px/$\mu$m, with the center-wavelength target reaching the 0.05 log$_{10}$ criterion but the full three-frequency target still failing. The robust conclusion is therefore a convergence-limited, resolution-aware workflow rather than a mesh-converged optimized-device claim.
Scalable on-chip quantum photonics requires efficient quantum interconnects between deterministic single-photon sources and low-loss waveguides. The discovery of intrinsic, nonstoichiometric nitrogen-rich silicon nitride (SiN) emitters provides a promising route for monolithic integration, enabling both enhanced emission rates and high extraction efficiencies. Here, we present a topology-optimized hybrid photonic-plasmonic quantum interconnect that functions as a cavity and coupler; it simultaneously enhances spontaneous emission and directs photons into a waveguide. By combining an Au nanodisk for Purcell enhancement with an inverse-designed SiN routing structure, we increase the emission rate while maintaining a coupling efficiency of $\sim$79% (corresponding to a coupling loss of $\sim$1.02 dB). These co-optimized plasmonic and dielectric components play complementary roles, enabling superior total collected photon power compared to dielectric-only designs. Native integration of single-photon sources in our design enables a robust quantum interconnect with the critical material uniformity across different functional elements and demonstrates a fabrication-aware, scalable approach to addressing key efficiency bottlenecks in large-scale quantum architectures.
Miniaturizing photonic components without compromising performance remains a fundamental challenge, especially when leveraging advanced 3D nanofabrication techniques where fabrication time scales with component volume. Addressing this challenge requires a departure from conventional design methods, with inverse design emerging as a promising strategy to realize compact, high-performance photonic components. To evaluate the effectiveness of such an approach, here we consider optical waveguide tapers, which are among the most basic yet fundamental building blocks in 3D integrated photonics, playing a critical role in efficiently interfacing optical fibers with on-chip components as well 3D printed polymer devices. Deviating from conventional linear taper designs, we first investigate curvature-parameterized tapers defined by a single analytical expression and conduct a classical parameter optimization study. Our results reveal a non-intuitive shape of adiabatic tapering, where shorter tapers outperform longer ones, achieving transmission efficiencies of 99% for designs of 50 μm length. We then proceed with an inverse design approach by considering segmented tapers, exploring novel geometries through evolutionary particle swarm optimization. By independently optimizing the curvature, length, and width of each segment, we achieve transmission coefficients exceeding 90% for tapers as short as 13.9 μm. To validate our design approach, we experimentally study both types of tapers by fabricating them on planar substrates, enabling transmission and coupling efficiency characterization using a fiber-to-fiber configuration. Following this, we fabricate the segmented taper directly on an optical fiber tip. These initial experimental results underscore the potential of optimization algorithms, when combined with accurate numerical simulations and 3D nanofabrication technology, to uncover high performance, non-intuitive designs for compact and efficient optical waveguide components.
This article presents an asterisk shaped ultra-wideband absorber designed with nickel (Ni) as conducting layers and polyimide substrate to form a three-layered structure. Its unit cell volume is 45×45×9nm3 (0.00000267 λ3L at 158.174THz) while operating across the frequency spectrum of 158.174THz (1896.645nm)-≥ 4500THz (≤66.67nm) which covers infrared, visible, and ultraviolet spectra for absorption (A) ≥ 90% with an average (Aavg) of 96.35%. It has absorption >95% across 191.31-388.98THz and 773.3-3876.75THz while simulating using CST Microwave Studio Suite (v. 2021). Symmetricity of asterisk's geometry leads to polarization insensitivity while having Aavg>90% for 0°≤incident angle(θ)≤60°. The solar absorber efficiency (ƞA) achieved for the proposed absorber is 94.66% with thermal emission efficiency (ƞE) of 91.69%, 93.8%, 94.51%, 94.61% and 94.53% at 1500K, 2000K, 2500K, 3000K, and 3500K, respectively. The range of the proposed absorber, with all its operating frequency bands, makes it a suitable model for uses like solar energy harvesting, solar detectors, sensors, and so on.
Vertical cavity surface emitting lasers (VCSELs) feature layer structures which are among the most complicated ones in compound semiconductor device production. They consist of many layers with stringent requirements on optical thickness to match the device specifications. Especially the optical cavity requires a very precise thickness control going far beyond the requirements of standard edge-emitting lasers. Also, a high number of interfaces needs to be controlled, especially when arsenides and phosphides are combined. Optical in situ characterization is thus helpful, if not indispensable for process development as well as process control in production. Re-establishing growth conditions for a new growth campaign after chamber maintenance can be challenging and time consuming. This work is about how to tackle this challenge by applying in situ optical metrology during growth and processing of GaAs based VCSEL devices as well as post-growth ex-situ wafer mapping. We demonstrate how to efficiently combine in-situ and ex-situ white light reflectance (WLR) measurements and modelling in order to increase the target wavelength accuracy. Additionally, we discuss how to use the data obtained during growth in order to enable automated end pointing for the plasma etching of the VCSEL mesa. Application of reflectometry for GaN VCSELs is briefly discussed and an outlook on the usage of in-situ bow measurements is given.
Reservoir computing (RC) offers an efficient framework for temporal processing by combining nonlinear mapping with short-term memory. However, conventional implementations are often hindered by the high hardware costs of spatial arrays or the speed and integration bottlenecks associated with feedback delay lines and input masking in time delayed RC. In this paper, we propose and experimentally demonstrate a single-node photonic reservoir computing system without physical delay lines and input masking operations. To enhance the representational capability of the single-node system, we introduce a time-concatenated dimensional expansion technique to construct a high-dimensional state space. Furthermore, a hybrid pruning strategy is employed to quantitatively optimize the trade-off between computational resources and forecasting performance. Our results demonstrate a feedback-free photonic reservoir computing framework that combines physical photonic feature generation with digital state construction and optimization for temporal processing tasks with diverse dynamical characteristics.
We present a non-destructive, all-electrical method for extracting the lateral oxidation extent in oxide-confined vertical-cavity surface-emitting lasers (VCSELs) using wafer-scale pulsed I–V measurements of non-lasing test structures. A distributed series-resistance model partitions contact, DBR, and oxide contributions and is fit across arrays of circular and elliptical mesas to recover isotropic and anisotropic oxidation extents, respectively. On commercial 940 nm, 150 mm epitaxial wafers with circular oxide apertures, the method yields a wafer-level mean of 16.10 μm with a standard deviation of 0.45 μm, showing close agreement with measurements from IR reflectivity of mean and standard deviation of 15.97 and 0.44 μm, respectively. Extension of the model to account for anisotropic oxidation extents on 852 nm, 100 mm epitaxial wafers extracts major and minor oxidation extents with values for wafer-level mean and standard deviations of (14.81 ± 0.35) μm and (12.00 ± 0.37) μm, respectively, which are consistent with IR mapping and FIB–SEM cross-sections. The approach is compatible with standard wafer-scale electrical probestations and supports statistically meaningful wafer-scale assessment relevant to VCSEL manufacturing.
A preliminary exploration of symmetry breaking in the triangular lattice photonic crystal surface emitting laser (PCSEL) is outlined. This analysis shows that greater single mode stability and optical efficiencies can be achieved for asymmetric unit cell designs in triangular lattices. In particular, changing the fill factor alone, for void containing or all-semiconductor wedge unit cell devices can provide significant tunability of the radiation constant and gain discrimination. This opens up routes for the use of asymmetric unit cells in triangular lattices to enhance PCSEL performance.
High-speed vertical-cavity surface-emitting lasers (VCSELs) with single-mode (SM) output and strong immunity to optical feedback play a vital role in further increasing the package density of co-packaged optics (CPO) systems for modern AI scale-out and scale-up networks. Compared to the combination of traditional copper cable with digital signal processing (DSP) ICs, the VCSEL based optical solutions offer significant energy savings and provide a much higher data rate per unit area. In this paper, we demonstrate that optimizing the structure of Zn-diffusion apertures inside VCSEL cavities can simultaneously improve the SM output power and speed of both 850 and 1060 nm VCSELs. Compared with the 850 nm VCSEL, the 1060 nm VCSEL exhibits superior SM characteristics, comparable electrical-to-optical (E-O) bandwidth (∼25 GHz), and much less spatial hole burning effect under dynamic operations. Nevertheless, the 850 nm VCSEL exhibits superior transmission performance over the same 1 km length of few mode fibers (FMF) to that of the 1060 nm one due to the stronger bandwidth enhancement effect. The direct-modulation channel bandwidths achieved at 850 and 1060 nm are 39 and 36 GHz, respectively, both with the same optical coupling efficiency as high as around 60% through a single lens on the fiber tip. At the 850 nm wavelength, error-free 56 Gbit/s transmission over 1 km under a low bias current of 4 mA is obtained without using any power-hungry retiming DSP ICs. The transmission distance can be further extended to 2 km FMF at 48 Gbit/sec due to the lower propagation loss for the 1060 nm wavelength window.
Low-energy, low-precision inference is an important enabler for scalable deployment of large language models (LLMs). Optical-electronic computing offers high bandwidth, low latency, and potentially high energy efficiency, making it a promising approach for future LLM deployment. However, existing optical and hybrid optical-electronic inference systems typically use uniform hardware configurations, such as fixed optical power or a fixed electronic-photonic partition across layers. Fine-grained allocation of optical power and optical-electronic computation remains underexplored for photonic LLM inference. Here, we propose three flexible configuration algorithms: Flexible Optical Power Allocation (FOPA), Flexible Weak-Column Allocation (FWCA), and Joint Optical Power and Weak-Column Allocation (JOWA). Using a diagonal Hessian estimator without constructing the full Hessian matrix, these methods jointly optimize optical-electronic resources under predefined energy budgets. Evaluations on Llama 2-7B and DeepSeek-LLM-7B-Base show 40.11%–60.85% higher estimated energy efficiency than FP16 inference on the NVIDIA H100 and 5.34%–38.12% higher estimated energy efficiency than uniform optical-power and weak-column baselines, while maintaining an acceptable perplexity increase. Evaluations using a Tianshu-hardware-calibrated simulator show corresponding improvements of 39.91%–53.67% and 5.02%–26.76%, respectively. These results indicate that fine-grained resource allocation can improve the energy-accuracy trade-off of hybrid optical-electronic systems for LLM inference.
This paper reports on 1.3-μm-wavelength InP-based photonic-crystal surface-emitting lasers (PCSELs) utilizing an in plane hetero photonic crystal (PC) structure. We investigated the impact of introducing the in-plane hetero PC design on lasing characteristics under both pulsed and continuous-wave (CW) operations. Two types of in-plane hetero PC structures were introduced in a 200-μm-diameter device, in which the band-edge frequency was intentionally varied between the central and peripheral regions of the device. The pop-down structure had a lower central band-edge frequency to improve mode stability under pulsed operation, whereas the pop-up structure had a higher central band-edge frequency to enhance optical output power under continuous-wave (CW) operation. Under pulsed operation, the pop-down hetero PC structure exhibited a single lobe circular beam with a narrow divergence angle by suppressing the higher-order mode while maintaining linear light–output characteristics up to 1 W. Under CW operation, the pop-up heterostructure achieved an increase in optical output power of 25% compared with a uniform PC device. Single-mode CW lasing with a maximum output power exceeding 1 W was achieved at 15°C, together with high side-mode suppression ratios exceeding 80 dB. In addition, high wall-plug efficiencies were obtained, reaching 29.1%, 27.1%, and 20% at 15°C, 25°C, and 50°C, respectively.
This study explores the integration of TiO2-based nanofilaments into dye-sensitized solar cells (DSSCs) and perovskite solar cells (PSCs) to enhance their optoelectronic performance. Despite the advancements in one-dimensional lepidocrocite TiO2 nanofilaments, there remains a significant gap in understanding their optical scattering mechanisms and dual applicability in both types of solar cells. We synthesized TiO2 nanofilaments and compared their performance to conventional P25 TiO2 in terms of morphology, electrical, and optical properties. The results indicate that the DSSC utilizing nanofilaments achieved a short-circuit current density of 8.2 mA/cm² and a power conversion efficiency (PCE) of 6.9%, compared to 7.1 mA/cm² and 5.1% for the bare P25 cell. Similarly, the PSCs exhibited a short-circuit current density increase from 16.0 mA/cm² to 19.7 mA/cm² and a PCE improvement from 11.3% to 13.4% when employing nanofilament TiO2. These enhancements are attributed to superior charge transport properties, increased surface area for dye adsorption, and enhanced forward optical scattering, which collectively improve light harvesting and charge collection efficiency.
