Figure 1(a) shows the forward current density–voltage (J–V) characteristic for all chips. The turn-on voltage of µ-LEDs from large to small chips is 3.05, 3.06, 3.04, 3.13, and 2.99 V, respectively, for an injected current density of approximately 10 A/cm2 (corresponding 1 mA for 10×10 µm2). At 4 V, the current density of µ-LEDs from large to small chips is 92.7, 99.8, 131, 154, and 210 A/cm2, respectively. Clearly, the current density increases as the size decreases. When the µ-LED light-emitting area is reduced, the volume and resistivity also reduce. Under the same current density, the forward voltage decreases as the chip size is reduced. Moreover, under the same bias voltage, a small-sized µ-LED will have a larger current density.
The main issue is the leakage current for µ-LEDs. The leakage behavior for µ-LEDs with different sizes is shown in Fig. 1(b). It was previously reported that the leakage current is related to the sidewall [5]. The relationship between the sidewall surface area and the total surface area of the µ-LED is shown in Table 1. First, the same number of defects per unit sidewall surface area is assumed. As the size is reduced, the sidewall surface area and the number of sidewall defects also decreases; nevertheless, the ratio of sidewall surface area to the total surface area increases. This suggests that the number of defects per unit total area increases and the effect of sidewall defects also increases for small-sized µ-LEDs.
It is found that the leakage current density increases as the size decreases under a reverse voltage less than 3 V. The leakage density of µ-LEDs with the smallest dimensions was the highest. This indicates that the leakage current could be caused by sidewall defects. Moreover, it could also be caused the largest electric field from the electrode to the edge (shortest distance = 3.5 µm for µ-LED with 10×10 µm2) as the reverse voltage less than 3 V. The electric field (@ −3 V) from the electrode to the edge was 2.3⊆103, 4.0⊆103, 6.0⊆103, 4.8⊆103, and 8.5⊆103 V/cm of µ-LEDs from large to small chip size. It is worthy to mention that the leakage current density of µ-LEDs from large to small chip size is 2.14 × 10−2, 1.96×10−3, 1.09×10−2, 4.59×10−3, and 7.2 ×10−3 A/cm2 for a reverse voltage of 5 V. As the reverse voltage increases, the leakage increases more obviously for larger chip size, especial for the µ-LED with 100×100 µm2. This suggests that not only is the leakage current relative to the sidewall passivation and the lateral electric field, the defects distributed on the surface could also affect the leakage current density. The large-sized LEDs cover a large area. It causes the leakage current density of large-sized µ-LEDs (100×100, 75×75, and 50×50 µm2) to increase and to be close to, and even higher than, those of small-sized µ-LEDs (20×20 and 10×10 µm2). Because there exist many defects in the InGaN epilayers [12, 15], this could also cover more defects. This causes leakage as the reverse bias increases.
Table 1
Comparison of total surface area and side wall surface area of different sizes of µ-LEDs
µ-LED dimensions (µm2)
|
Mesa area (µm2)
|
Sidewall area (µm2) *1
|
Sidewall surface ratio *2
|
10 × 10
|
100
|
20.6
|
0.171
|
25 × 25
|
625
|
51.5
|
0.076
|
50 × 50
|
2500
|
103
|
0.040
|
75 × 75
|
5625
|
154.5
|
0.027
|
100 × 100
|
10000
|
206
|
0.020
|
*1 Mesa depth = 515 nm |
*2 Sidewall surface ratio = sidewall area/total surface area |
|
Figure 2 (a) shows the optical output power density as a function of the injection current density. As the injected current increases, the optical output power increases for all the chips. It was found that there were two groups. One group is for the larger chips with areas of 100×100 and 75×75 µm2. The other group is for chip sizes with areas 50×50, 25×25, and 10×10 µm2. If the output power density is affected by the sidewall, the smallest chip µ-LEDs should show the lowest power density. As shown in Table 1, the µ-LED with an area of 10×10 µm2 presents the largest sidewall surface ratio. If leakage is the main issue affecting the output power, the short distance between the p-electrode and the edge could be an important parameter. As shown in Fig. 5 (b), the short distance between the p-electrode and the edge is 12.5, 7.5, 5, 6.3, and 3.5 µm for all chips and sizes in order. Moreover, at the same injection current density, the distance from the electrode to the edge is only 3.5 µm for the smallest size chip. Considering the sidewall surface ratio and the distance to the edge, the smallest dimension µ-LEDs would obtain the smallest output power density. However, for the 10×10 µm2 µ-LEDs shown in Figure 2(a), this does not produce the smallest output power density. Evidently, the passivation layer provides the sidewall protection from leakage. This causes the output power density to be very similar for smaller µ-LEDs. This is different from the behavior of AlGaInP µ-LEDs. Obviously, the sidewall defect is more easily repaired by passivation. This is similar to the behavior of blue µ-LEDs. Larger chips (100×100 and 75×75 µm2) show higher output power density than smaller chips (50×50, 25×25 and 10×10 µm2). This may be so because larger chips contribute to light reflection from the bottom.
Figure 2(b) shows the EQE as a function of injection current densities. The maximum EQE is 5.11, 4.75, 3.36, 2.81 and 2.91% corresponding to the chip size in order. Note that the maximum EQE occurs at almost the same current density, except the 10⊆10 um2 chip. It is due to the output power is too small to measure as the low injecting current density for the 10⊆10 um2 chip. Moreover, it has been reported that the maximum EQE is shifted to higher current densities as the chip size decreases for the AlGaInP uLEDs. The shift of maximum EQE was related to leakage current and/or an increased Shockley-Read-Hall (SRH) non-radiative recombination at sidewall defects in the smaller geometries [14, 16], although the sidewall had been passivated. However, the phenomena does not be observed in the red InGaN uLEDs. The obtained results were consistent with that in Figure 2(a), i.e., the sidewall defects can be repaired by passivation. On the other hand, the droop behavior was more alleviative for the small size uLEDs. It could be due to the fact that there exist more defects or phase separation for the uLEDs with larger size. These defects could trap the carriers and reduce the EQE. As concerning this point, it has been demonstrate by TEM measurement [12]. Furthermore, it was found that the EQE of the uLEDs with10⊆10 um2 was higher than those of uLED with 25⊆25 um2 and 50⊆50 um2. For smaller mesa size, the contribution of light coupled through the mesa edge compared to the surface becomes larger. This causes an increase of the external quantum efficiency (EQE) by light extraction. It was worthy to mention although the epilayer structure was the same with Refs. [12] and [14], the obtained EQE in this work is higher than those of Refs. [12] and [14]. It could be due to the different device fabrication parameters, geometry of the metallic contacts, mesa outline and chip sizes.
Figure 3 shows the wavelength of µ-LEDs with different sizes as a function of current. The wavelength decreases as the current increases, exhibiting blue shift behavior. It is well known that the blue shift phenomenon in blue and green InGaN-based LEDs grown on c-plane sapphire due to the InGaN QWs is caused by the screening of the piezoelectric field and band filling of the localized state. The same behavior is also exhibited in the InGaN µ-LEDs with high In composition [13–18]. Moreover, the wavelength decreases as the chip size is reduced. This can be attributed to several possible reasons. The first is stress release in small dimensional LEDs. Second, smaller-sized chips have larger current densities, which leads to screening of QCSE and also band filling effects. To verify this, we applied a 1D Poisson and drift-diffusion solver (1D-DDCC) developed by Prof. Wu’s lab in NTU [19, 20]. With the same structure as shown in Fig. 1, we can simulate the band bending, confined energy, and emission spectrum at different densities. Figure 3(a) indicates that at a driving current of 1 mA, the emission wavelengths from large to small are 636.24, 617.55, 612.98, 605.87, and 576.87 nm in order. The current density versus chip size as shown in Figure 3(c) reveals a trend with small variation for different chip sizes. To understand the mechanism for this large range of blue shift, 1D simulation was performed for different current densities. It shows that blue shift from 660 nm (0.1 A/cm2) to 590 nm (1 kA/cm2) is possible. This blue shift is caused by (1) screening of QCSE and (2) band filling effects. Our calculations show that blue shift from 660 nm to 620 nm is mainly due to the screening of QCSE. After 10 A/cm2, the band filling effect becomes stronger and the emission spectrum widens as shown in Figure 3(d). The trend of blue shift matches better with larger chip size. For smaller chip size such as 25 and 10 µm, the trend is slightly different. The discrepancy may be due to several possible reasons: (1) the current crowding effect is different for different chip sizes. Hence, the red shift may not occur for different current densities in the QW as predicted by 1D simulations; (2) for the high indium QW, a large random alloy fluctuation or indium segregation is expected. Hence, the filling of states may not be the same as the ideal cases; (3) although the compressive InGaN is relaxed to 0.38 GPa [12], the top multi-QWs of InGaN still suffer from compressive strain. Hence, the wavelength shift results from a combination of the screening of QCSE, quantum effect, and band filling. This is a complex phenomenon and requires further study. By contrast, there exist extra emission spectra at 380 and 450 nm for the smallest µ-LEDs for high injection current density, as shown in Figure 3(b). Obviously, the overflow current and band filling phenomena occur in the red InGaN µ-LEDs. The inset of Figure 3(b) also shows the emission images of the 10×10 µm2 µ-LED injected at the current from 0.1 to 10 mA. In the simulation, if the leakage path is strong in the epilayer, with a higher density of tail stats in the simulation, we also observe that some holes will reach the blue QWs for current density larger than 103A/cm2 and an emission of blue QWs will be observed. These leakage paths may come from dislocation line and Vpits[12].