Improvement of current crowding effect in VCSEL arrays with non-uniform oxidation aperture design

Yuanyuan Liu, Dongyue Jin, Xin Lei, Yuxin Zhou, Wanrong Zhang, Hongytm Xie, Weicong Na, Jlngfing Dai, Zhiyong WangMATEC 2024 · DOI

TL;DR

Researchers have developed a way to boost the performance of Vertical Cavity Surface Emitting Laser (VCSEL) arrays by intentionally making them non-uniform. By shrinking the oxidation apertures of central cells relative to the outer ones, they improved current and temperature uniformity by over 36%, leading to an 8.91% increase in peak optical power.

What Problem Does This Paper Address?

VCSELs are the workhorses of modern data centers and optical interconnects because they are efficient and consume little power. However, when you need high power, a single VCSEL isn't enough; you have to pack them into two-dimensional arrays. This introduces two major headaches:

  1. Thermal Rollover: The cells in the center of an array get much hotter than those on the edges because they are surrounded by other heat-generating neighbors. This "self-heating" eventually causes the laser's power to drop off as you increase the current.
  2. Current Crowding: Within each individual laser cell, the electrical current doesn't like to spread out evenly. Instead, it tends to "crowd" around the edges of the oxide aperture. This leads to wasted energy, more heat, and reduced light output.

The challenge is finding a way to make the current flow and the heat dissipate more evenly across the entire array to push the "thermal rollover" point further back.

Methods

The authors tackled this using a two-pronged approach: simulation and fabrication.

  • The Model: They established a compact electro-opto-thermal model. This isn't just a simple circuit diagram; it’s a mathematical framework that accounts for how electrical current turns into light and heat, and how that heat, in turn, changes the electrical properties of the laser.
  • Non-Uniform Design: Most arrays use identical cells. The researchers proposed "Array-2," where the central cell has a smaller oxidation aperture (3.3 µm) compared to the surrounding cells (5.3 µm).
  • Fabrication: To prove the theory, they fabricated 3×3 VCSEL arrays of both types (uniform and non-uniform) and measured their Light-Current-Voltage (L-I-V) characteristics to see which performed better in the real world.

Datasets

This paper does not use external machine learning datasets. Instead, the "data" consists of:

  • Simulated results generated from their custom electro-opto-thermal model.
  • Experimental measurements taken from the 3×3 VCSEL arrays they fabricated in their lab.

Key Insights and Ideas

  • Smaller is Sometimes Better: Counter-intuitively, reducing the aperture of the central cell improves the whole array. A smaller aperture reduces the time and distance current needs to travel to reach the center of the laser, which effectively "fights" the current crowding effect.
  • Symmetry Matters: The authors grouped the 3×3 array into three "groups" based on their distance from the center. This allowed them to analyze the non-uniformity of the array as a whole rather than just looking at individual cells in isolation.
  • The Feedback Loop: The paper highlights that current crowding and thermal hotspots are a vicious cycle. By fixing the current distribution (improving it by ~42%), they saw a massive knock-on benefit in temperature uniformity (~36%).

Key Results

The results show that "Array-2" (the non-uniform design) outperformed the standard "Array-1" across the board:

  • Power Boost: The peak output power reached 1.83 mW, an 8.91% improvement over the uniform design.
  • Thermal Stability: The peak junction temperature was lowered by 6.73 K, and the overall temperature distribution became 36.52% more uniform.
  • Current Efficiency: The uniformity of the current distribution across the array improved by 42.08%.
  • Consistency: Perhaps most importantly, the total optical power of the non-uniform array was superior to the uniform one across a wide range of bias currents, not just at the peak.

Limitations and Future Work

While the 3×3 array provided an excellent proof-of-concept, the paper leaves a few stones unturned:

  • Scaling Up: A 3×3 array is relatively small. In industrial applications, arrays can be much larger. It remains to be seen how complex the "aperture gradient" would need to be for a 10×10 or 20×20 array.
  • Optimization Algorithms: The researchers chose specific sizes (3.3 µm and 5.3 µm). Future work could involve using optimization algorithms to find the perfect aperture size for every single cell in an array to maximize performance.
  • Long-term Reliability: The paper focuses on immediate performance metrics (L-I-V curves). Long-term aging tests would be necessary to ensure that non-uniform apertures don't introduce new degradation mechanisms over thousands of hours of operation.

Summary updated

Abstract

A compact electro-opto-thermal model of 2-D vertical cavity surface emitting laser (VCSEL) arrays considering the current crowding effect in each array cell is established to study the impact of oxidation aperture on the device performance. Simulated results shows that increasing oxidation aperture of array cell is helpful to improve the uniformity of current density distribution. With careful design of non-uniform oxidation aperture layout, both the uniformity of the temperature distribution and the current distribution is improved by 36.52% and 42.08%, respectively. Furthermore, 3×3 VCSEL arrays with uniform oxidation aperture (array-1) and non-uniform oxidation aperture (array-2) are fabricated and the L-I-V curves of two types of VCSEL arrays at different biases are also measured. The peak output optical power of array-2 is enhanced to 1.83mW with an improvement of 8.91% when compared with that of array-1. Moreover, the total optical output power of array-2 is always superior to that of array-1 over a wide bias current range.
BibTeX
@inproceedings{liu2024improvement,
  author = {Yuanyuan Liu and Dongyue Jin and Xin Lei and Yuxin Zhou and Wanrong Zhang and Hongytm Xie and Weicong Na and Jlngfing Dai and Zhiyong Wang},
  title = {{Improvement of current crowding effect in VCSEL arrays with non-uniform oxidation aperture design}},
  booktitle = {MATEC},
  year = {2024},
  doi = {10.1051/matecconf/202439501002},
}
Figure