Fig. 1. (a) The proposed 8-wavelength 4x4 LED array chip image, the colors of each dash line box represent the color of the LED unit. (b) The scanning electron microscope (SEM) image of the V-pit structure. (c) The vertical profile of the V-pit structure. (d) Layers of the red LED units (660 nm and 620 nm). (e) Layers of the GaN-based LED units (wavelengths from 570 nm to 450 nm).

Fig. 2. The current density distribution in simulation (a) and its detailed view (b). The h+ carriers are attracted by the V-pit structure and enter the sidewall quantum well area. Next, they are horizontally transported into the wells in the flat quantum wells and increases the injection into these wells.

Fig. 3. Comparison of the maximum recombination rate (SRH, Radiative, and Auger) between the sample with and without V-pits in each quantum well in 30 mA (a) and 350 mA (b) cases. (c) The hole current density simulation for samples with or without V-pits. Notice that the sample with V-pit presents a significant gain in current density in each quantum well.

Fig. 4. (a) The proposed equivalent circuit for fitting both LEDs with and without V-pits (with tiny V-pits). The branch in the dash yellow box is dedicated to representing the extra current introduced by the V-pit area. And the other branch in the intrinsic LED part represents the flat quantum well region. The fitting result using the proposed equivalent circuit for (b) the sample without V-pits and (c) the sample with V-pits.

Fig. 5. (a) The comparison between the I-V curves of the LED sample with or without the V-pit structure. (b) The LI (left y-axis) and illumination efficacy (right y-axis) versus the current density of the LED sample with or without the V-pit structure.

Fig. 6. (a, b) The electroluminescence (EL) spectrum of the LED samples with and without V-pits. (c) The full-width-half-maximum (FWHM) of EL spectrum shown in (a) and (b).

Fig. 7. (a, b) The pulse response of the LED samples with and without V-pits in 300 mV Vpp. (c, d) The pulse response of LED samples with and without V-pits in 1 V Vpp.

Fig. 8. The setup diagram and the DSP process in the TX and RX sides.The red blocks represent the bit-power loading process which modifies the modulation order and power of each DMT sub-carrier and helps the system achieve higher spectrum efficiency and communication capacity.

Fig. 9. The contour diagrams of each LED unit in searching for the best working point of each channel.

Fig. 10. The spectra of the original, received and NN equalized signals in the channel estimation stage of the communication experiment for each channel.

Fig. 11. The bit-power loading result in the final communication test.The first row of each subplot shows the modulation order and SNR on each subcarrier, and the second row is the allocated power ratio, demonstrating the modification to the power of subcarriers.

Fig. 12. Summary of the communication rate.(a) The spectrum efficiency (SE) and modulation bandwidth of the proposed 8-wavelength WDM system. (b) The BER of each channel, all the BERs are lower than the 7% HD-FEC threshold. (c) Comparison in data rate with the original design. The total data rate of the proposed size-improved design device is 31.38 Gb/s.