Precise Waveguide Loss Extraction using Wafer-Level OFDR for Optical Process Control Monitoring

Introduction

In the manufacturing of silicon-based photonic integrated circuits (Si-PICs), monitoring fabrication variations that affect the performance of optical devices is crucial. One key parameter is the propagation loss of silicon waveguides, which is typically evaluated using the optical process control monitor (PCM) designed for the cutback method. However, the cutback method requires a large PCM footprint on wafers for multiple long waveguides of varying lengths. Optical frequency domain reflectometry (OFDR) offers an alternative approach to extract waveguide loss with a smaller footprint and reduced measurement time.

OFDR for Waveguide Loss Extraction

OFDR works by measuring the reflected intensity from a waveguide as a function of distance from the input. The propagation loss is defined as the rate of decrease in the reflected intensity with increasing distance of the reflection point. An OFDR system, consisting of a broadband tunable laser source and an optical module for frequency domain interference, was integrated into a wafer prober with an automatic alignment module [Fig. 1(a)].

Fig. 1. (a) Setup for wafer-level optical frequency domain reflectometry (OFDR). (b) Waveguide element for evaluating waveguide loss using OFDR. (c) Spectra showing coupling efficiency between the probing fiber and grating coupler. (d) Typical reflection versus distance profile from OFDR measurement. (e) Contour map of propagation loss for 64 waveguides, each 480 nm wide, derived from OFDR.

Silicon wire waveguides with a path length of 6.45 cm were fabricated on a single silicon-on-insulator (SOI) wafer using the 300-mm AIST SCR photonics platform technology [Fig. 1(b)]. The waveguide height was 200 nm, and the width was varied between 420, 440, 460, and 480 nm. Grating couplers for TE mode coupling with fiber were formed at the waveguide ends, and one was used for OFDR measurement.

The wavelength sweep range for OFDR measurements was 1520 to 1540 nm, matching the 1-dB-bandwidth for optical coupling between the grating coupler and fiber [Fig. 1(c)]. Random reflections occurred from the entire waveguide, and the reflection intensity attenuated with increasing distance [Fig. 1(d)]. The propagation loss was extracted from the reflection vs. distance profile using Equation 1.

Where α is the propagation loss, R1 and R2 are the attenuating reflections at distances L1 and L2 from the light source, and ng is the group refractive index for the propagation mode. Figure 1(e) shows a contour map for propagation loss extracted by OFDR for waveguides with a width of 480 nm on a 300-mm SOI wafer.

Comparison with Cutback Method

To validate the OFDR method, the propagation losses for 256 waveguides with widths of 420, 440, 460, and 480 nm were extracted using both OFDR and the cutback method. For the cutback method, transmission spectra for four waveguide lengths (6.45, 5.25, 1.44, and 0.24 cm) were measured, and the average propagation loss for 1520-1540 nm was used for comparison.

Fig. 2 Correlation between the propagation losses extracted by the OFDR method and cut-back method.

Figure 2 shows the correlation between the propagation losses extracted by the two methods. The OFDR and cutback methods were found to provide almost the same loss, with a correlation coefficient of 0.96. This high correlation validates the OFDR method for waveguide loss extraction.

In the cutback method, variations in coupling efficiency due to fabrication variations in grating coupler formation and fiber-to-grating alignment errors are the main sources of measurement error. For OFDR, reflection peaks from grating couplers and random reflections from the waveguide can affect the extraction validity. While it is difficult to determine which method is inherently better, their complementary nature allows for more accurate propagation loss extraction when used together.

Advantages of Wafer-Level OFDR

From the perspective of manufacturing monitoring, the small optical PCM footprint required for OFDR is a significant advantage over the cutback method. The OFDR method eliminates the need for multiple waveguides of varying lengths, reducing the area required for PCMs on the wafer. Additionally, OFDR measurements can be performed more quickly than the cutback method, which requires measuring transmission spectra for multiple waveguide lengths.

Conclusion

The propagation loss of silicon waveguides extracted from the reflection vs. distance profile obtained by the OFDR method was verified by comparing it with the cutback method. Using an OFDR system integrated into a wafer prober with automatic alignment, it was confirmed that the OFDR method provided almost the same propagation loss values as the cutback method for 256 waveguides. The high correlation between the two methods demonstrates the applicability of wafer-level OFDR for extracting propagation loss in manufacturing monitoring of Si-PICs. The small footprint and reduced measurement time of OFDR make it an attractive option for optical process control monitoring during device fabrication.

Reference

[1]   T. Horikawa, A. Kitamura, M. Yatani, N. Nishiyama, "Precise Waveguide Loss Extraction Using 300-mm Wafer-Level OFDR for Optical Process Control Monitoring," Dept. of Electrical and Electronic Engineering, Tokyo Institute of Technology, Tokyo, Japan; Santec LIS Corporation, Komaki, Japan; Institute of Innovative Research, Tokyo Institute of Technology, Tokyo, Japan; Photonics Electronics Technology Research Association (PETRA), Tokyo, Japan, 2024, pp. 1-6, doi: 979-8-3503-9404-7/24/$31.00 ©2024 IEEE.