In recent years, AlGaN-based deep-ultraviolet light-emitting diodes (DUV-LEDs) have been considered to be the most promising DUV light sources because of their wide applications in water purification, UV curing, environmental sensing, plant growth lighting, phototherapy and so on [1]. Nevertheless, the development of AlGaN-based DUV-LEDs is still limited by the relatively low external quantum efficiency (EQE), whose maximum has been reported to be 20.3% up to now [2]. Quite a number of issues, including the poor hole injection efficiency and severe electron leakage are the key factors hindering the improvement of the device performance. Therefore, it is essential to probe device structures to promote the hole injection, and meanwhile effectively suppress the electron leakage.

To improve the hole injection efficiency, Kozodoy et al. [3] adopted polarization-induced p-type AlGaN/GaN superlattices, in which alternate polarization modulation played a critical role to enhance the Mg doping. Apart from increasing the hole concentration in the p-type region, Zhang et al. put forward some special structure designs to improve the hole transport. For example, inserting a lower Al composition AlGaN thin layer in the p-type electron blocking layer (EBL) to enhance the tunneling probability for holes [4], or adopting an electric-field reservoir (EFR) consisting of a p-Al_xGa_1-xN/p-GaN heterojunction which can provide energy continuously for holes thus facilitates the hole injection efficiency [5]. Besides, Zhang et al. [6–8] suggested an interband tunneling contact instead of direct p-type contact by taking advantage of the polarization properties of III-nitride material, which should increase the hole injection efficiency and meanwhile minimize the internal light absorption caused by the p-GaN layer. On the other hand, many approaches were proposed to suppress the electron leakage. Guo et al. [9] recommended single spike barriers in the active region to slow down the velocity of electrons, thus increase the electron capture efficiency of multiple quantum wells (MQWs) in the AlGaN-based LEDs. Li et al. [10] reported that the AlGaN EBL with graded Al composition worked well in suppressing the electron leakage because the polarization induced charges would modulate the energy band. Hirayama et al. [11] introduced a multiple quantum barrier (MQB) structure as an EBL in an AlGaN-based LED, which increased the effective barrier height for electrons. However, problem comes there that the potential barrier for hole injection is always increased accompanying with the effective suppression for electron leakage. Therefore, seeking some structures that can balance electron leakage and hole injection is of special significance.

In this work, we put forward a MQB-EBL structure with Al composition and thickness graded in every barrier layer. Polarization induced bulk charges are introduced to increase the potential barrier for electrons and meanwhile decrease the one for holes, thus improve the vertical transport of carriers. As a result, the internal quantum efficiency (IQE) and light output power (LOP) are experimentally proved to be higher and numerically compared to the traditional structure.

In order to illustrate the device performance and inherent physical mechanism of the DUV-LEDs with the proposed MQB-EBL, two samples are grown by metal-organic chemical vapor deposition (MOCVD) for comparison. As shown in Fig. 1(a), both structures were grown on (0001) c-plane sapphire substrates, initiated with a 1-μm-thick AlN buffer layer, which were treated by high temperature annealing, and the detail can be found in our previous work [12,13]. Then a 0.5-μm-thick AlN/AlGaN superlattice layers and a 1.5-μm-thick Si-doped n-Al_0.55Ga_0.45N layer (n~2.0 × 10¹⁸ cm⁻³) were set in sequence. Afterwards, the active region with five pairs of 2.3-nm-thick Al_0.37Ga_0.63N quantum well (QW) and 10-nm-thick Si-doped Al_0.5Ga_0.5N barrier (n~1.0 × 10¹⁸ cm⁻³) except for the last 20 nm thick undoped barrier were grown at 1080 °C. A 27-nm-thick Mg-doped (~1.0 × 10¹⁹ cm⁻³) p-type EBL and 90-nm-thick Mg-doped p-AlGaN/AlGaN superlattice layer (p~1.0 × 10¹⁸ cm⁻³) were then grown on top of the active region. Finally, a 10-nm-thick Mg-doped (p~1.0 × 10¹⁸ cm⁻³) p-GaN was deposited as the Ohmic contact layer. As the reference sample (structure A), the traditional bulk EBL with Al composition of 65% is shown in Fig. 1(b), while the proposed MQB-EBL structure(B)is shown in Fig. 1(c). In this proposed structure, the thickness and Al composition for every well layer of the EBL are constantly 2 nm and 60%, respectively. While for barriers, the thickness varies from 3 to 5 nm, and then from 5 to 3 nm,along [0001] direction, and the Al composition in every barrier layer is graded from 75% to 65% along [0001] direction (structure B). After finishing growth of these two LED structures, a standard micro-fabrication technique was adopted to fabricate the DUV-LEDs with the chip size of 899 × 899 μm². The electrodes with interdigital shapes were adopted to guarantee current spreading. Ti/Al/Ni/Au metals were deposited on the mesa surface as the n-type electrode, and then were rapidly thermally annealed under N₂ atmosphere at 850 °C to form Ohmic contact. The p-type electrode was composed of Ni/Au metals, which were then annealed in O₂ atmosphere at 550 °C. The chips were obtained by adopting flip-chip configuration and the DUV light was collected from the sapphire side.

 

Fig. 1 (a) Schematic structure of AlGaN-based LED and the EBL region of (b) structure A and (c) structure B.

Download Full Size | PDF

In addition, to further investigate the difference on energy band and carrier transport behavior between the traditional structure and the proposed one, we conducted simulations by the Advance Physical Model of Semiconductor Devices (APSYS) simulation program [14]. In the simulations, the Shockley-Read-Hall (SRH) lifetime and Auger recombination coefficient within the active region were chosen to be 10 ns and 1.0 × 10⁻³⁰ cm⁶/s, respectively. Besides, the offset ratio $\Delta E_c/\Delta E_g$ between conductive band and total bandgap for AlGaN/AlGaN interface was chosen to be 0.7. And since there was no special treatment on surface structures for light extraction, we supposed the light extraction efficiency (LEE) to be 5% in the simulations. The other parameters used in the APSYS simulation process can be found in our previous work [9,15].

Figure 2(a) shows the electroluminescence (EL) spectra at the current of 100 mA. The single peak emission with the wavelength of 273 nm could be observed from both structures A and B, and the peak width at half height (FWHM) of EL spectra for structures A and B are severally 9.0 nm and 9.8 nm. Further, we also perform the integrated LOP and EQE measurement with different injection currents, as shown in Fig. 2(b). The maximums of LOP and EQE over the whole current range (0-250 mA) for structure B are severally 9.6 mW and 1.03%, being increased by 405% and 249% respectively compared to structure A. Besides, the efficiency droop at 250 mA is dramatically reduced from 42.2% to 16.6% when adopting the proposed EBL structure. That indicates that structure B with the Al composition and thickness graded MQB-EBL has a remarkable improvement in device performance. Figure. 2(c) further presents the simulated dependences of the LOP and EQE on current for these two structures. It is obvious that the simulation results agree well with the experimental results from the comparison of Figs. 2(b) and 2(c).

 

Fig. 2 (a) EL characteristics of structures A and B at 100mA with single peak emission at 273 nm; (b) The experimental results and (c) the simulation results of EQE and LOP at different injection currents.

Download Full Size | PDF

In order to figure out the underlying mechanism for the above mentioned improvement, we put forward a physical model as Fig. 3 to reveal the carrier transport properties within the proposed p-EBL region. For structure A with conventional EBL as shown in Fig. 3(a), holes (represented by red hollow circles) should overcome a large potential barrier between p-superlattice and p-EBL, and then be injected into the MQWs region. While for structure B with Al composition and thickness graded MQB-EBL [Fig. 3(b)], there are polarization induced negative bulk charges that are distributed in every barrier layer due to the gradient of Al composition along [0001] direction [16]. When holes are injected from one well to the next one along $\left[000\overline{1}\right]$ direction, some holes (represented by blue hollow circles) will be captured because of the attraction of those negative bulk charges, and thus the hole concentration within the barrier layers will be increased. It is well known that the relationship between the potential barrier height (${\Phi }_b$) and concentration for holes at the interface which composed the barrier ($p_{\mathrm{int}er}$) can be expressed by [17]:

 

Fig. 3 Schematic models of (a) the conventional p-EBL and (b) the Al composition and thickness graded MQB-EBL.

Download Full Size | PDF

In terms of electron leakage, for structure A, the electron leakage would happen when electrons (represented by pink solid circles) cross over the potential barrier between the last barrier and p-EBL. For structure B, however, before electrons get into the p-superlattice region, they must cross over five barriers. As the composed p-EBL in Fig. 3(b) owns the negative polarization induced bulk charges in the barrier regions, part of electrons would be repulsed out of the barrier. Similarly, the potential barrier height for electrons (${\Phi }_b^{\text{'}}$) is determined by [17]:

Further simulations are performed to verify the above analysis, and the results at the current of 100 mA are chosen for demonstration. Figure 4 shows a comparison of the conduction band near the p-EBL and corresponding electron concentration for both two structures, in which the parts with grey, yellow and pink colors represent the LBAR, p-EBL and p-superlattice layers, respectively. As shown in Fig. 4(a) for structure A, the conduction band (CB) of the LBAR bends down and forms a sharp corner (highlighted by the red oval) because of the existence of the positive polarization induced charges at the interface between the LBAR and p-EBL. This sharp corner with rather low energy would result in electron accumulation [highlighted by the dashed box in Fig. 4(c)], which will aggravates electron leakage. In the contrast, for structure B, due to the negative polarization induced bulk charges in the first barrier (closest to the quantum wells) of p-EBL, the positive charges at the interface between LBAR and p-EBL are screened partially. Since the polarization degree is inversely proportional to the thickness of the graded layer [18], we increase polarization induced bulk charges by thinning the barrier thickness of p-EBL near the LBAR to screen more positive charges at the interface between the LBAR and p-EBL. It can be observed in Fig. 4(b) that the band bending is effectively eliminated in structure B, and thus the electron accumulation in the LBAR disappears in structure B [Fig. 4(c)]. Besides, by comparing Figs. 4(a) and 4(b), it is obvious that the highest potential energy barrier height for electrons is increased from 640 meV to 1069 meV, and thus the electron concentration within p-EBL including the $n_{\mathrm{int}er}$ [highlighted by the blue arrows in Fig. 4(c)] for the structure B is decreased remarkably compared to structure A, as shown in Fig. 4(c). That is in complete agreement with the analysis with Fig. 2 (b). And since the change of EBL structure does not influence the electron injection from the n-Al_0.5Ga_0.5N layer to active region, there is no difference on the electron distribution within the active region between structures A and B expect the LBAR [shown in Fig. 4(c)], which has been discussed above.

 

Fig. 4 Conduction band near the p-EBL region of (a) structure A and (b) structure B as well as (c) electron concentration in the LBAR and p-EBL of structures A and B under 100 mA.

Download Full Size | PDF

Figure 5 shows the valence band (VB) as well as hole concentration distribution in the active region and p-EBL at 100mA for structures A and B. The potential barrier height for holes is 493 meV in structure A, while that for structure B is 536 meV and 258 meV for the first barrier (closest to n-side) and the other four barriers in the p-EBL respectively, as shown in Figs. 5(a) and 5(b). Although the barrier height of the first barrier in p-EBL for structure B is higher than that for structure A, the potential barrier height of the other four barriers for structure B decreases dramatically compared to structure A. In addition, the thinner thickness of the first barrier makes it possible for holes to tunnel through the first barrier and be injected into the active region. Also, the barriers near the p-superlattice side is set thinner to enhance the polarization which would elevate the valence band and thus allow more holes climb over the barriers before they tunnel through the higher barrier. Simulation results show that the hole concentration, both in the active region and p-EBL [including $p_{\mathrm{int}er}$ as highlighted by violet arrows in Fig. 5(d)], in structure B increases obviously compared to structure A [Figs. 5(c) and 5(d)], which indicates that the hole injection efficiency is improved significantly when adopting the proposed p-EBL structures.

 

Fig. 5 The conduction bands within the p-EBL region of (a) structure A and (b) structure B as well as the hole concentration within (c) the active region and (d) p-EBL region of structures A and B under 100 mA.

Download Full Size | PDF

Finally, the radiative recombination rates (RRRs) are calculated for these two structures within the active region under 100 mA as presented in Fig. 6. The average RRR of the five quantum wells for structure B is ~1.98 times greater than that of structure A, showing a significant improvement compared to that of structure A. And since there is no difference in the structures of active region between structures A and B, the improved RRR of structure B can basically be ascribed to the increased concentration of holes within the active region as shown in Fig. 5(c). The enhancement of RRR for structure B is also consistent with the improved LOP and EQE as presented in Fig. 2.

 

Fig. 6 Radiative recombination rate within the active region of structures A and B under a current of 100 mA.

Download Full Size | PDF

In conclusion, DUV-LED with Al composition and thickness graded MQB-EBL has been investigated. The maximum of LOP and EQE for the specially designed structure with the pumping current up to 250 mA reach 9.6 mW and 1.03%, which are increased by 405% and 249%, respectively, compared to the traditional LED structure, meanwhile, the efficiency droop at 250 mA is also dramatically reduced from 42.2% to 16.6%. Further simulation analysis indicates that this graded MQB EBL introduces negative polarization bulk charges in every barrier of MQB-EBL, hence both the conduction band and valance band are elevated. As such, the potential barrier height for electrons is increased, while that for holes is decreased, which effectively suppresses the electron leakage, and meanwhile improves the hole injection efficiency. Obviously, the change of the energy band helps to enhance the carrier transport within the p-EBL region, and finally significantly improves the device performance in terms of LOP and EQE.