1. Introduction

Following the development of p-type GaN and the demonstration of the blue light emitting diode (LED) [1,2], III-nitride light emitting devices have been aggressively pursued for commercial applications. Most commercially available III-nitride devices are grown on the c-plane of GaN and are affected by the presence of inherent spontaneous and piezoelectric polarization. This has led to active research on nonpolar and semipolar orientations of GaN to reduce the effects of polarization [3–6]. Reducing the polarization leads to a reduction in the quantum confined stark effect, higher radiative recombination rates, and increased differential gain [7].

The top cladding of a conventional III-nitride laser diode (LD) consists of thick p-(Al,Ga,In)N layers which are grown at temperatures much higher than the quantum wells (QWs) in the active region (AR). This can lead to thermal damage to the QWs, resulting in material degradation and reduced performance [8–11]. To mitigate this issue, p-GaN can be grown at colder temperatures of around 850-950 °C instead of the optimal 1000-1100 °C. One approach is to use non-epitaxial layers to replace or reduce the thickness of III-nitride p-cladding. Violet and blue c-plane LDs that omit the top cladding and rely on the exceptionally low refractive index of a Ag or APC (Ag-Pd-Cu) p-contact have been demonstrated [12–14]. However, this approach is limited by the loss in the metallic layers. Another approach to solving this design limitation is to replace a part of the III-nitride top cladding with a relatively low loss transparent conducting oxide (TCO). This design was demonstrated using indium-tin-oxide (ITO) as the top TCO layer in ($20\overline{2}1$) III-nitride LD, with threshold current density (J_th) of 6.2 kA/cm² and threshold voltage of 8.9 V [15].

As an alternative to ITO, ZnO can be used as the TCO layer for III-nitride LDs. ZnO is a wurtzite semiconductor, with a direct bandgap of 3.3 eV at room temperature [16]. The c-plane lattice mismatch of ZnO to GaN is only ~2%, which facilitates high quality single-crystal growth of ZnO on GaN. ZnO is an attractive alternative to ITO because its optical absorption coefficient [17], is more than one order of magnitude lower than ITO in the blue and green regions of the spectrum [15]. Recent studies found that external quantum efficiency (EQE), luminous efficacy, and series resistance were improved in LEDs with ZnO current spreading layers relative to ITO, indicating that ZnO layers had both lower absorption and lower contact resistance to p-GaN [18]. Introducing dopants has also been used to demonstrate ZnO ohmic contacts to p-GaN [19]. In this letter, we use numerical modeling to investigate the impact of using a lower loss cladding design with ZnO as the TCO layer instead of ITO and demonstrate a blue $20\overline{2}1$) III-nitride LD with a ZnO top cladding design.

2. Modeling

One benefit of using ZnO as a TCO layer is that it avoids the use of thick p-AlGaN layers and allows the use of thinner p-GaN cladding layers. It is important to understand how the thickness of the p-GaN layers and the choice of TCO affects device performance. To investigate the effect of changing the thickness of the p-GaN layers, 2D optical modes of the proposed LD structure were simulated using Fimmwave, a vectorial mode solver for 2D waveguides (WG)s [20]. GaN, InGaN and AlGaN refractive indices were taken from Goldhahn et. al. [21], and the absorption coefficients were taken from Kioupakis et. al. [22]. The ITO refractive index and absorption coefficient were taken from Hardy et. al. [15]. The ZnO refractive index and absorption coefficient were taken from Reading et. al. [17]. The confinement factor (Γ) and internal loss (α_i) were calculated using the Fimmwave real film mode matching method (FMM) and then used to calculate threshold material gain (g_th) and differential efficiency (η_d) assuming uncoated laser facets and injection efficiency (η_i) of 65% [23].

The simulated LD structure consisted of 1 µm n-GaN cladding layer, an n-In_0.08Ga_0.92N WG layer with a thickness between 0 and 100 nm, an active region consisting of three 4.8 nm In_0.20Ga_0.80N QWs and four 7.6 nm GaN quantum barriers (QBs), a 10 nm Al_0.21Ga_0.79N electron blocking layer (EBL), a p-In_0.08Ga_0.92N WG layer with a thickness between 0 and 100 nm, a p-GaN layer with a thickness between 10 to 900 nm, a 10 nm p + GaN contact layer, and 250 nm of ITO or ZnO. The total thickness of the In_0.08Ga_0.92N WG was fixed at 100 nm and the ridge WG was etched to a depth of 100 nm above the top interface of the last QB.

The use of a TCO in the top cladding creates an asymmetric WG [15,24]. As the p-GaN gets thinner, the modal overlap with the TCO increases and the peak of the mode shifts towards the n-side. For a given p-GaN thickness and a fixed total InGaN WG thickness, the best balance of high Γ and low α_i can be achieved by placing a larger fraction of the InGaN WG on the n-side. Our simulations showed that for a fixed p-GaN thickness both Γ and α_i decrease as the fraction of the InGaN WG on the n-side increases. This happens due to the reduced overlap of the mode with the QWs, the EBL and the p-GaN layers, all of which have relatively high absorption. Placing the QWs symmetrically in the waveguiding layer (i.e. 50/50 nm n-/p-In_0.08Ga_0.92N) results in the highest optical confinement, but results in a relatively high internal loss. However, as a higher fraction of the InGaN waveguiding layer is placed on the n-side, internal loss decreases more rapidly than optical confinement. Therefore, Γ can be minimally sacrificed to achieve considerably lower α_i. The LD design demonstrated in this work had a 325 nm p-GaN layer and a 70/30 nm n-/p-In_0.08Ga_0.92N WG, which resulted in a 3% lower Γ and 12% lower α_i, compared to a LD structure with a symmetric 50/50 n-/p-In_0.08Ga_0.92N WG. The demonstrated LD, whose structure and simulated optical mode profile are shown in Fig. 1, had a Γ and α_i, of 5.2% and 12.8 cm⁻¹, respectively.

 

Fig. 1 (a) Cross sectional schematic of the LD with ZnO cladding. The structure consisted of a 1μm n-GaN cladding layer, a 70 nm n-In_0.08Ga_0.92N waveguide (WG), an active region consisting of three 4.8 nm In_0.2Ga_0.8N quantum wells and four 7.6 nm GaN quantum barriers, a 30 nm p-In_0.08Ga_0.92N WG, a 10 nm Al_0.21Ga_0.79N electron blocking layer (EBL), a 325 nm p-GaN layer, a 10 nm p⁺ GaN contact layer, and 780 nm of ZnO. The thin green layer indicates the location of the p-Al_0.21Ga_0.79N EBL. The ridge WG was etched to a depth of 100 nm above the top interface of the last quantum barrier. (b) Transverse mode profile and refractive index as a function of distance in the growth direction.

Download Full Size | PDF

The dependence of Γ and α_i on p-GaN thickness and TCO composition is shown in Figs. 2(a) and 2(b), respectively. Γ does not depend on the TCO composition, which is expected as ZnO and ITO have almost identical refractive indices in the blue and green regions of the spectrum. For p-GaN thicknesses of less than 100 nm, the optical mode has poor lateral confinement due to the shallow etched ridge WG and Γ decreases significantly [24]. For p-GaN thicknesses of greater than 500 nm, the ZnO and ITO cladding layers have a negligible effect on the optical mode and the Γ and α_i show little dependence on p-GaN thickness.

 

Fig. 2 Dependence of (a) confinement factor and (b) internal mode loss on p-GaN thickness for a structure with 250 nm of ITO or ZnO in the top cladding layer.

Download Full Size | PDF

For p-GaN thicknesses between 100 and 500 nm, Γ increases as the p-GaN thickness decreases because the peak intensity of the mode increases in the vicinity of the QWs as the mode is compressed by the thinner p-GaN. For the device with ITO cladding, α_i increases as the p-GaN thickness decreases because the overlap of the mode with the high loss ITO cladding increases. The mode overlap with the TCO layer is the same for both ITO and ZnO for a given p-GaN thickness. However, since the absorption coefficient for ZnO (100 cm⁻¹) [17], is much lower than the absorption coefficient for ITO (2000 cm⁻¹) [15], the contribution of the ZnO layer to the overall device loss is negligible for devices with ZnO cladding, even for structures with thin p-GaN layers. This means that devices with ZnO cladding layers havelower overall α_i than devices with ITO cladding layers, especially in devices with thin p-GaN layers. This reduction in α_i is a key advantage of replacing ITO with ZnO in a TCO clad LD design.

The dependence of g_th and η_d on p-GaN thickness and TCO composition are shown in Figs. 3(a) and 3(b), respectively. For these calculations, the cavity length and mean mirror reflection coefficient (R) [25], were assumed to be 1200 μm and 0.18, respectively, corresponding to a mirror loss of 14.3 cm⁻¹. For p-GaN thicknesses of greater than 500 nm, the ZnO and ITO cladding layers have a negligible effect on the optical mode, so g_th and η_d show little dependence on p-GaN thickness and are equal for both TCOs. For p-GaN thicknesses between 100 and 500 nm, for the device with ITO cladding, g_th increases slightly and η_d decreases significantly as the p-GaN thickness decreases because Γ only increases slight while α_i increases significantly with decreasing p-GaN thickness. In contrast, for the device with ZnO cladding, both g_th and η_d do not decrease as the p-GaN thickness decreases because Γ increases slightly and α_i actually decreases with decreasing p-GaN thickness. The effects of the cladding layer TCO material on optical power output and η_d are particularlymarked and should be even more apparent for state-of-the-art low-loss high-power LD designs [26].

 

Fig. 3 Dependence of (a) threshold material gain and (b) differential efficiency on p-GaN thickness for a structure with 250 nm of ITO or ZnO in the top cladding layer.

Download Full Size | PDF

3. Experimental

LDs were grown on free standing ($20\overline{2}1$) semipolar GaN substrates provided by Mitsubishi Chemical Corporation using atmospheric pressure metalorganic chemical vapor deposition (MOCVD). The LD structure had a 1 μm n-GaN cladding layer, a 70 nm n-In_0.08Ga_0.92N WG, an AR consisting of three 4.8 nm In_0.2Ga_0.8N QWs and four 7.6 nm GaN QBs, a 30 nm p- In_0.08Ga_0.92N WG, a 10 nm Al_0.21Ga_0.79N EBL, a 325 nm p-GaN layer, and a 10 nm p + -GaN contact layer. A thick n-GaN layer was used for the bottom cladding to avoid the potential relaxation of a strained n-AlGaN cladding layer.

The LDs were fabricated using a self-aligned ridge WG process. Ridge WGs were defined using a BCl₃/Cl₂ reactive ion etch (RIE) and were etched to a depth of 100 nm above the top interface of the last QB. Following the ridge WG etch, SiO₂ was deposited via sputtering on the sidewalls of the ridge WG and in the field for electrical isolation. The samples were then cleaned in a two-step process to remove organic and oxide residue prior to the ZnO deposition. The organic residue was removed with a 10 min RCA standard clean (1:1:5 NH₄OH:H₂O₂) solution heated to 80 °C, which leaves an oxide residue [27]. The oxide residue was removed with a 10 min etch in a 1:1 HCl:H₂O solution. A 780 nm thick undoped ZnO layer was then deposited on the exposed p-GaN surface using aqueous solution deposition, the details of which are discussed elsewhere [28,29]. The first step is the growth of a 200 nm thick partially coalesced ZnO seed layer. The seed layer growth is catalyzed at 90 °C by adding ammonium hydroxide (NH₄OH) to a zinc nitrate hexa-hydrate (Zn(NO₃)₂:6H₂O) and ammonium nitrate (NH₄NO₃) solution. To ensure good electrical contact and adhesion, the sample was then annealed at 500 °C in N₂ for 15 minutes in a rapid thermal anneal (RTA) system. A second 580 nm thick fully-coalesced ZnO layer was deposited using a 90 °C solution of zinc nitrate hexa-hydroxide (Zn(NO₃)₂:6H₂O), trisodium citrate (NaC₆H₅O₇), and ammonium hydrate (NH₄OH). The relatively small chip size (5 mm x 7 mm) ensured uniform concentrations over the deposition area.

After the ZnO deposition, 30 nm of Ti was deposited by electron beam evaporation along the entire length of the laser ridge and then 30/1000 nm Ti/Au was deposited also by electron beam evaporation to form a p-contact pad. The thin Ti layer extended beyond the p-contact pad to ensure uniform current injection. The thin Ti layer also serves to avoid peeling the metal off the ZnO during the facet polishing process. The sample was then diced to form laser bars with a cavity length of 1200 μm. Each facet was polished using diamond polishing films and the grit size was graded from 6.0 μm to 0.1 μm. Facet polishing was completed using 0.05 μm alumina and 0.03 μm silica polishing films, resulting in smooth and featureless facets, as depicted in Fig. 4, which shows a scanning electron microscope (SEM) image of the facet of a fully processed laser, where the epitaxial layers, SiO₂, ZnO bilayer and p-contact pad are clearly visible. The laser fabrication was completed with the deposition of a 50/3000 nm Al/Au common back side contact via electron beam evaporation.

 

Fig. 4 Scanning electron microscope (SEM) image of the polished facet of a fully processed laser, where the epitaxial layers, SiO₂, ZnO bilayer and p-contact pad are clearly visible.

Download Full Size | PDF

4. Results and discussion

Figure 5(a) shows the light-current-voltage (L-I-V) characteristic for a 1.6 μm wide by 1200 μm long blue LD. The device was tested under pulsed electrical injection with a pulse width of 1.0 μs and a duty cycle of 1%. Lasing was achieved at a threshold current (I_th) of 166 mA, a threshold current density (J_th) of 8.6 kA/cm² and threshold voltage (V_th) of 10.3 V. The single facet slope efficiency (η_sl) was 0.37 W/A which corresponds to a single facet η_d of 13.5%. Assuming equal power output from both facets, the experimental η_d for both facets was 27%. This value is close to the 32% η_d predicted by optical modeling for this p-GaN thickness (see Fig. 3(b)). Compared to a similar ($20\overline{2}1$) LD structure with ITO top cladding [15], this LD showed a marked improvement in slope efficiency and output power.

 

Fig. 5 (a) Pulsed L-I-V characteristics, (b) current dependent spectra, and (c) the far field pattern of a 1.6 μm wide by 1200 μm long LD.

Download Full Size | PDF

The differential resistance, R_d, was 9 Ω at currents greater than 166 mA. Assuming that the ZnO resistivity is 4.50 x 10-3 Ω-cm [30], the average resistivity in the p-type (Al,In,Ga)N layers is 1 Ω-cm, the average resistivity in the n-type (Al,In,Ga)N layers is 0.001 Ω-cm, and that the resistance contributions of the n-contact, the substrate, the Ti/ZnO interface, and the heterointerfaces are negligible, we estimate that the specific contact resistivity of the ZnO/p-GaN interface is ~1.35 x 10⁻⁴ Ω-cm². Optimizing the aqueous solution deposition methodology and introducing dopants during the deposition should also improve the electrical properties of the ZnO [19], reduce the specific contact resistivity at the ZnO/p-GaN interface, and lower the overall device operating voltage. The current dependent spectra, which exhibit a peak wavelength of 453 nm above threshold, are depicted in Fig. 5(b). Figure 5(c) shows the far field pattern above threshold.

5. Conclusions

In conclusion, we have investigated using ZnO instead of ITO as the top cladding layer in semipolar ($20\overline{2}1$) III-nitride LDs. Numerical modeling indicated that replacing ITO with ZnO reduces the internal loss and threshold gain and increases the differential efficiency for a TCO clad LD design due to the relatively low optical absorption in ZnO. Lasing was achieved at 453 nm with a threshold current density of 8.6 kA/cm² and a threshold voltage of 10.3 V in a ($20\overline{2}1$) III-nitride LD with ZnO cladding.