Semiconductor Laser Parameter Inverse Design Method Based on Artificial Neural Network and Particle Swarm Optimization

Fig. 7. Comparison between inverse design power spectra and target power spectrum. (a) Comparison among two sets of inverse design power spectra, numerical inverse power spectrum, and target power spectrum; (b) deviation of two sets of inverse design power spectra and numerical inverse power spectrum compared with target power spectrum

Download full size

Table 1. Comparison of training error between different network structures
View table
Table 1. Comparison of training error between different network structures
n 10 15 20 25 30 35 40
MSE /mW 3.20 1.81 0.55 0.49 0.34 0.28 0.17

Table 2. Comparison of partial output data between TWM simulation and neural network for sample data
View table
Table 2. Comparison of partial output data between TWM simulation and neural network for sample data
Paremeter Value
I /mA 28 48 68 88 108 128
W_O /mW 6.81 12.13 17.11 21.35 23.73 21.32
W_N /mW 7.23 12.55 16.99 21.13 23.90 21.53
E /% 6.0 3.4 0.7 1.0 0.7 0.9

Table 3. Comparison of partial output data between TWM simulation and neural network for test data
View table
Table 3. Comparison of partial output data between TWM simulation and neural network for test data
Paremeter Value
I /mA 28 48 68 88 108 128
W_T /mW 6.61 12.45 17.67 21.46 22.01 15.91
W_N /mW 6.57 12.78 18.33 22.06 21.78 16.35
E /% 0.6 2.7 3.7 2.8 1.0 2.8

Table 4. Comparison of partial data between target power spectrum and inverse design power spectra
View table
Table 4. Comparison of partial data between target power spectrum and inverse design power spectra
Parameter Value
I /mA 28 48 68 88 108 128
W_T /mW 6.01 10.49 15.34 18.44 18.72 11.24
W_S /mW 6.63 12.20 17.00 19.83 19.24 13.08
W₁ /mW 6.61 11.06 15.22 18.41 18.55 12.32
W₂ /mW 6.04 11.13 15.33 18.49 18.61 12.13

Table 5. Comparison of parameters between target power spectrum and two sets of inverse design power spectra
View table
Table 5. Comparison of parameters between target power spectrum and two sets of inverse design power spectra
Parameter η_effc R_t /(K·J^-1) R_s /Ω K_e /K K_g /K K_n /K l /cm^-1
S_T 0.59 6.69×10⁸ 15.19 304.58 105.91 18.25 23.80
S₁ 0.80 7.68×10⁸ 15.69 317.85 94.08 69.98 46.22
S₂ 0.78 7.08×10⁸ 20.00 325.61 96.49 108.73 45.31

Tools

Get Citation

Copy Citation Text

Pei Feng, Yu Li. Semiconductor Laser Parameter Inverse Design Method Based on Artificial Neural Network and Particle Swarm Optimization[J]. Chinese Journal of Lasers, 2019, 46(7): 0701001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: laser devices and laser physics

Received: Dec. 25, 2018

Accepted: Mar. 11, 2019

Published Online: Jul. 11, 2019

The Author Email: Li Yu (li.yu@sdu.edu.cn)

DOI:10.3788/CJL201946.0701001

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Comparison of training error between different network structures

Table 1. Comparison of training error between different network structures

Table 2. Comparison of partial output data between TWM simulation and neural network for sample data

Table 2. Comparison of partial output data between TWM simulation and neural network for sample data

Table 3. Comparison of partial output data between TWM simulation and neural network for test data

Table 3. Comparison of partial output data between TWM simulation and neural network for test data

Table 4. Comparison of partial data between target power spectrum and inverse design power spectra

Table 4. Comparison of partial data between target power spectrum and inverse design power spectra

Table 5. Comparison of parameters between target power spectrum and two sets of inverse design power spectra

Table 5. Comparison of parameters between target power spectrum and two sets of inverse design power spectra