Authors:

Yu Dian LimHong Yu LiSimon Chun Kiat GohXiangyu WangPeng ZhaoChuan Seng Tan

Paper:

https://arxiv.org/abs/2408.10287

Introduction

Background

Over the past decade, integrated silicon photonics (SiPh) gratings have been extensively developed for optical addressing in ion trap quantum computing. These gratings are crucial for guiding laser beams to trapped ion qubits, a fundamental component in quantum computing. However, accurately determining the heights at which beam profiles are located when viewed through infrared (IR) cameras remains a challenge.

Problem Statement

The difficulty in determining the corresponding heights of beam profiles from SiPh gratings when viewed through IR cameras poses a significant challenge. This issue is exacerbated by alignment errors in multiple beams, which can affect the performance of complex operations such as ion shuttling and multi-qubit operations. To address this, the study proposes using transformer models to recognize the height categories of beam profiles from SiPh gratings.

Related Work

Silicon Photonics in Quantum Computing

SiPh devices have been increasingly adopted in ion trap quantum computing due to their potential for wafer-scale integration and low fabrication costs. Various SiPh devices, such as optical switches, micro-ring resonators, and on-chip lasers, have been developed and integrated into ion trap setups. However, alignment errors in multiple beams present a significant challenge that affects performance.

Transformer Models

Transformer models, initially introduced by Google for natural language processing (NLP), have shown promise in various applications due to their self-attention mechanism. Vision Transformer (ViT) models, an extension of transformer models for image recognition, have been used in photonics applications such as 3D holography and optical communication. Despite their success, the usage of ViT models in silicon photonics has been limited.

Research Methodology

Objective

The objective of this study is to develop a transformer model capable of recognizing the height categories of beam profiles from SiPh gratings captured by an IR camera.

Data Acquisition

The IR camera is fixed at a position above the silicon photonics chip (z = 0 µm) and elevated periodically to capture beam profiles at every 5 µm increment in elevation, ranging from z = 0 to 805 µm. The beam profiles are then categorized into four regions (A, B, C, and D) based on their z-position.

Model Training

Two techniques are used to train the transformer model:
1. Input Patches: The beam profiles are divided into patches and used as input for the model.
2. Input Sequence: The beam profiles are converted into sequences of intensity values along the x-axis and used as input for the model.

Experimental Design

Fabrication of Gratings

The gratings are fabricated using a conventional full-wafer CMOS process on a 12-inch silicon substrate. The sample chip comprises an input grating connected to 16 output gratings via waveguides and multi-mode interferometers (MMIs). The pitch of the input grating is optimized to maximize the coupling efficiency of fiber-to-chip coupling of light with a 1,092 nm wavelength.

Beam Profile Acquisition

Beam profiles are acquired using an IR camera attached to a microscopic lens. The z-position of the microscope lens is adjusted to obtain the sharpest possible beam profiles. The beam profiles are captured at every 5 µm increment in elevation, ranging from z = 0 to 805 µm.

Data Preparation

The acquired beam profiles are cropped and reshaped for further data analysis and model training. The beam profiles are categorized into four regions (A, B, C, and D) based on their z-position.

Results and Analysis

Beam Analysis and Categorization

The beam profiles from gratings with various pitches show different intensity distributions. The beam profiles are categorized into four regions based on their z-position:

  • Region A: Multiple beam spots with curvy morphologies.
  • Region B: Single brightest beam spot with high focusing.
  • Region C: Slightly distorted beam profiles with reduced intensity.
  • Region D: Highly distorted beam profiles with the lowest intensity.

Model Development and Evaluation (Patches Input)

The transformer model trained using input patches achieved a recognition accuracy of 0.938. The model correctly recognized almost all testing beam profiles, with mistakes occurring mainly at the edges of each region.

Model Development and Evaluation (Sequence Input)

The transformer model trained using input sequences achieved a lower accuracy of 0.895. Despite the lower accuracy, the model trained using input sequences showed more consistent outcomes and better fitting to the training and validation datasets.

Repeatability Testing

The model training and evaluation were repeated 150 times to test the robustness of the models. The transformer model trained using input patches showed higher accuracy but less consistency, while the model trained using input sequences showed more consistent performance.

Overall Conclusion

This study successfully developed transformer models to recognize the height categories of beam profiles from SiPh gratings. The model trained using input patches achieved higher accuracy, while the model trained using input sequences showed more consistent performance. The outcomes of this work establish a foundation for recognizing light beam profiles according to their corresponding heights and can be expanded to various applications, including auto-focusing of light beams and auto-adjustment of z-axis stages.

References

  1. Bruzewicz, C. D., Chiaverini, J., McConnell, R., & Sage, J. M. (2019). Trapped-ion quantum computing: Progress and challenges. Applied Physics Reviews, 6(2).
  2. Zhao, P., Lim, Y. D., Li, H. Y., Guidoni, L., & Tan, C. S. (2021). Advanced 3D Integration Technologies in Various Quantum Computing Devices. IEEE Open Journal of Nanotechnology, 2(October), 1–1.
  3. Sia, J. X. B., Li, X., Wang, J., Wang, W., Qiao, Z., Guo, X., Lee, C. W., Sasidharan, A., Gunasagar, S., Littlejohns, C. G., Liu, C., Reed, G. T., Ang, K. S., & Wang, H. (2022). Wafer-Scale Demonstration of Low-Loss (∼0.43 dB/cm), High-Bandwidth (>38 GHz), Silicon Photonics Platform Operating at the C-Band. IEEE Photonics Journal, 14(3), 1–9.
  4. Wang, J., Sia, J. X. B., Li, X., Guo, X., Wang, W., Qiao, Z., Littlejohns, C. G., Liu, C., Reed, G. T., Rusli, & Wang, H. (2023). High-speed 4 × 4 silicon photonic plasma dispersive switch, operating at the 2 µm waveband. Optics Express, 31(20), 33548.
  5. Li, X., Xu, J., Sia, B., Wang, J., Wang, H., Qiao, Z., Xu, J., Sia, B., & Wang, H. (2024). Optoelectronic oscillator based on a silicon microring modulator. Applied Physics Letters, 125, 071104.
  6. Wang, J., Li, X., Guo, X., Loh, T.-H., Ranno, L., Liu, C., Rusli, Wang, H., & Sia, J. X. B. (2024). Scalable single-microring hybrid III-V/Si lasers for emerging narrow-linewidth applications. Optics Express, 32(15), 26751.
  7. Todaro, S. L., Verma, V. B., McCormick, K. C., Allcock, D. T. C., Mirin, R. P., Wineland, D. J., Nam, S. W., Wilson, A. C., Leibfried, D., & Slichter, D. H. (2021). State Readout of a Trapped Ion Qubit Using a Trap-Integrated Superconducting Photon Detector. Physical Review Letters, 126(1), 1–7.
  8. Slichter, D. H., Verma, V. B., Leibfried, D., Mirin, R. P., Nam, S. W., & Wineland, D. J. (2017). UV-sensitive superconducting nanowire single photon detectors for integration in an ion trap. Optics Express, 25(8), 8705.
  9. Mehta, K. K., Bruzewicz, C. D., Mcconnell, R., Ram, R. J., Sage, J. M., & Chiaverini, J. (2016). Integrated optical addressing of an ion qubit. Nature Nanotechnology, 11(12), 1066–1070.
  10. Niffenegger, R. J., Stuart, J., Sorace-Agaskar, C., Kharas, D., Bramhavar, S., Bruzewicz, C. D., Loh, W., Maxson, R. T., McConnell, R., Reens, D., West, G. N., Sage, J. M., & Chiaverini, J. (2020). Integrated multi-wavelength control of an ion qubit. Nature, 586(7830), 538–542.
  11. Mehta, K. K., Zhang, C., Malinowski, M., Nguyen, T.-L., Stadler, M., & Home, J. P. (2020). Integrated optical multi-ion quantum logic. Nature, 586, 533–537.
  12. Mordini, C., Vasquez, A. R., Motohashi, Y., Müller, M., Malinowski, M., Zhang, C., Mehta, K. K., Kienzler, D., & Home, J. P. (2024). Multi-zone trapped-ion qubit control in an integrated photonics QCCD device. arXiv, 1–15.
  13. Reed, G. T., & Knights, A. P. (2004). Silicon Photonics: An Introduction. John Wiley & Sons, Ltd.
  14. Lim, Y. D., Zhao, P., Guidoni, L., Likforman, J. P., & Tan, C. S. (2023). Development of Mixed Pitch Grating for the Optical Addressing of Trapped Sr+ Ions with Data Analysis Techniques. Optics Express, 31(15), 23801–23812.
  15. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2016). Attention is All You Need. In 31st Conference on Neural Information Processing Systems (NIPS 2017), (Nips), pp. 47–82.
  16. Lim, Y. D., Zhao, P., Guidoni, L., Likforman, J., & Tan, C. S. (2024). Next-Value Prediction of Beam Waists From Mixed Pitch Grating Using Simplified Transformer Model. IEEE Photonics Journal, 16(5), 1–9.
  17. Lim, Y. D., & Tan, C. S. (2024). Modelling and next-value prediction of beam propagation from grating structures using a simplified transformer model. Optics Express, 32(18), 31533–31548.
  18. Jin, Z., Ren, Q., Chen, T., Dai, Z., Shu, F., Fang, B., Hong, Z., Shen, C., & Mei, S. (2024). Vision transformer empowered physics-driven deep learning for omnidirectional three-dimensional holography. Optics Express, 32(8), 14394.
  19. Merabet, B., Liu, B., Li, Z., Tian, J., Guo, K., Shah, S. A. A., & Guo, Z. (2023). Vision transformers motivating superior OAM mode recognition in optical communications. Optics Express, 31(23), 38958.
  20. Cuenat, S., Andréoli, L., André, A. N., Sandoz, P., Laurent, G. J., Couturier, R., & Jacquot, M. (2022). Fast autofocusing using tiny transformer networks for digital holographic microscopy. Optics Express, 30(14), 24730.
  21. Dubé, P., Madej, A. A., Shiner, A., & Jian, B. (2015). Sr 88 + single-ion optical clock with a stability approaching the quantum projection noise limit. Physical Review A – Atomic, Molecular, and Optical Physics, 92(4).
  22. Lim, Y. D., Zhao, P., Guidoni, L., Likforman, J., & Seng, C. (2023). Development of Grating for the Optical Addressing of 88Sr+ Ions with Data Analysis Techniques. IEEE Photonics Journal, PP, 1–8.
  23. Zhao, J., Rong, C., Lin, C., & Dang, X. (2023). Multivariate time series data imputation using attention-based mechanism. Neurocomputing, 542, 126238.
  24. Lim, Y. D. (n.d.). Recognizing Experimentally-Obtained Beam Profiles using Transformer Model: Input Patches. Retrieved from https://raw.githubusercontent.com/yd145763/TransformerExperimentalData/main/ViT.py
  25. Lim, Y. D. (n.d.). Recognizing Experimentally-Obtained Beam Profiles using Transformer Model: Input Sequence. Retrieved from https://raw.githubusercontent.com/yd145763/TransformerExperimentalData/main/Transformer.py
  26. Lim, Y. D. (n.d.). Boxplots for 150 Repetition Cycles. Retrieved from https://raw.githubusercontent.com/yd145763/TransformerExperimentalData/main/boxplot_github.py
Share.

Comments are closed.

Exit mobile version