1. Introduction

Since the development of p-type GaN [1] and the demonstration of III-nitride light emitting diodes (LEDs) [2], III-nitride light emitting devices have become prolific in commercial applications such as solid-state lighting and display backlighting. More recently, III-nitride laser diodes (LDs) are finding use in applications such as projection [3], laser-based white lighting [4–6], and visible light communications [7,8]. State-of-the-art III-nitride LEDs and LDs, which are grown on the c-plane of GaN, are affected by its inherent spontaneous and piezoelectric polarizations [9]. Quantum well (QW) LEDs and LDs grown on semipolar planes benefit from reduced effects of polarization relative to the basal plane, which lead to better wave function overlap, higher radiative recombination rates, and increased differential gain [10–12].

The top cladding of conventional III-nitride LDs is comprised of p-type III-nitride layers that require high growth times and temperatures [13] and have high electrical resistivity compared to n-type GaN [14] or non-epitaxial cladding layers [15]. Growing thick top cladding layers at high temperature can result in thermal damage to the InGaN QW active regions [16], and high resistivity layers can increase device operating voltages. P-cladding can be grown at colder than optimal temperatures to reduce thermal damage, but this results in higher electrical resistivity [17]. Non-epitaxial cladding layers, such as metals [18,19], and transparent conducting oxides (TCOs) [20,21], facilitate an elegant mitigation of conventional top cladding limitations because they require much lower growth temperatures and reduce p-cladding bulk resistance. Replacing part of the III-nitride top cladding with a TCO layer is a particularly attractive top cladding design because TCOs provide the necessary index contrast for high mode confinement without significantly contributing to device loss. Indium-tin-oxide (ITO), which is widely used for III-nitride LEDs, has also been used to demonstrate TCO top cladding for both AlGaN-cladding-free semipolar LDs [20] and semipolar LDs with AlGaN cladding [21].

The composition and thickness of the bottom cladding of semipolar III-nitride LDs is limited by high lattice mismatch between GaN and III-nitride ternary alloys. Ternary alloy n-cladding layers, such as AlGaN, are beneficial because they provide a higher refractive index contrast for the LD waveguide (WG), which leads to higher confinement for the optical mode. Stress relaxation can be mitigated by decreasing the density and run length of misfit dislocations (MDs) [22,23], which are formed via threading dislocation (TD) glide at epitaxial growth interfaces. TD glide can be impe ded by breaking the planarity of the growth surface, which is achieved through pre-growth patterning, a technique known as limited area epitaxy (LAE) [24–26]. LAE allows for coherent growth of AlGaN (or InGaN) past the predicted Matthews-Blakeslee critical thickness, and our group has demonstrated n-AlGaN bottom clad semipolar III-nitride LDs with both conventional p-AlGaN top cladding layers [27] and with ITO-based top cladding layers [21].

The simultaneous incorporation of LAE-enabled n-AlGaN bottom cladding and a TCO-based top cladding enables thinner p-GaN in the top cladding and a higher index contrast in the LD WG, which respectively reduce thermal damage of the InGaN QWs and increase the confinement factor (Γ) of the lasing mode. Our first demonstration of this design, which we refer to as the LAE-TCO design, used ITO as the TCO layer in the top cladding. However, ITO can make a significant contribution to the LD internal loss due to higher residual absorption in the visible spectrum. ZnO is a viable alternative to ITO as a top cladding layer because the optical absorption coefficient of ZnO [28] is more than an order of magnitude lower than the optical absorption coefficient of ITO in the blue and green regions of the spectrum [29], all with a comparable resistivity and contact resistance contribution to the top cladding [30,31]. ZnO top cladding was previously demonstrated with a 1.6 μm wide by 1200 μm long ridge waveguide AlGaN-cladding-free semipolar III-nitride LD, with a threshold current density (J_th) of 8.6 kA/cm² and a threshold voltage (V_th) of 10.3 V [31]. In this letter, we present a semipolar III-nitride LD design that simultaneously incorporates thin p-GaN and a ZnO-based top cladding with an LAE-enabled n-AlGaN bottom cladding for improved confinement of the lasing mode.

2. Modeling

The effects of replacing ITO with ZnO top cladding on the LAE-TCO design were evaluated with Fimmwave, a vectorial mode solver for 2D WGs [32]. GaN, InGaN, and AlGaN refractive indices were taken from Goldhahn et. al. [33], and the absorption coefficients were taken from Kioupakis et. al. [34]. The ITO refractive index and absorption coefficient were taken from Hardy et. al. [20], and ZnO refractive index and absorption coefficient were taken from Reading et. al [28]. Fimmwave was used to calculate the Γ and internal mode loss (α_i) of a laser structure, which were then used to calculate threshold material gain (g_th) and differential efficiency (η_d), assuming uncoated laser facets and an injection efficiency (η_i) of 65% [35].

Incorporating LAE-enabled AlGaN bottom cladding allows for thinner p-GaN in the TCO-based top cladding and higher index contrast in the laser waveguide [21]. The index contrast is affected by the AlN fraction in the bottom cladding, so the p-GaN thickness in the top cladding was varied in the Fimmwave simulations to find the optimal thickness for lateral mode confinement for different AlN fractions. The LD structure used in the optical modeling consisted of a 1 μm Si-doped GaN template, a 1 μm Si-doped AlGaN n-cladding with Al composition ranging from 0% to 6%, an n-In_0.08Ga_0.92N WG layer with a thickness between 0 and 100 nm, an undoped active region consisting of three 4.8 nm In_0.2Ga_0.8N quantum wells (QWs) and four 7.6 nm GaN quantum barriers (QBs), a 10 nm p-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 20 nm to 550 nm, a 10 nm p + GaN contact layer, and 250 nm of ITO or ZnO. The doping concentrations of the Al_0.05Ga_0.95N n-cladding, In_0.08Ga_0.92N n-waveguide, In_0.08Ga_0.92N p-waveguide, and GaN p-cladding measured by SIMS were 2 x 10¹⁸ cm⁻³, 2.9 x 10¹⁸ cm⁻³, 1 x 10¹⁹ cm⁻³, and 1.3 x 10¹⁹ cm⁻³, respectively. For the optical simulations, we assumed a QW carrier density of 2.45 x 10¹⁹ cm⁻³, which was based on previously reported gaindata [35]. The total thickness of the combined n- and p-In_0.08Ga_0.92N WGs 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.

Figure 1(a) shows the dependence of confinement factor on p-GaN thickness. The TCO material does not affect Γ for a given Al composition value because ZnO and ITO have almost identical refractive indices (~2.0) in the blue region of the spectrum [20,28]. For this reason, only Γ values of the ZnO top clad LD are shown in Fig. 1(a). For all p-GaN thicknesses, Γ increases with increasing Al composition in the n-AlGaN bottom cladding because the index contrast of the waveguide increases with increasing Al composition. Γ decreases significantly for p-GaN thicknesses less than 100 nm because the optical mode has poor lateral confinement due to the shallow etched ridge [21,31]. For p-GaN thicknesses between 100 nm and 400 nm, the mode spreads out into the thicker p-GaN layers and Γ shows little dependence on p-GaN thickness for all Al compositions.

 

Fig. 1 Dependence of (a) confinement factor, (b) internal loss, and (c) differential efficiency on p-GaN thickness for a structure with 250 nm of ITO or ZnO in the top cladding and different Al concentrations in the n-AlGaN bottom cladding.

Download Full Size | PDF

The dependence of the internal mode loss on p-GaN thickness for both ZnO and ITO and for the various Al compositions is shown in Fig. 1(b). Unlike its effect on Γ, the choice of TCO in the top cladding layer has a pronounced effect on α_i of the LAE-TCO LD design. For all AlN fractions, ZnO top cladding results in lower α_i values than ITO top cladding. These results arereasonable, considering that the absorption coefficient values used for the FIMMWAVE simulations for ZnO and ITO were 100 cm⁻¹ [28], and 2000 cm⁻¹ [20], respectively. The difference between the ZnO and ITO contributions to the overall α_i is most noticeable for p-GaN thicknesses between 100 nm and 350 nm. For these thicknesses, the difference between the ZnO and ITO contributions to the overall α_i is significant and it increases with increasing Al composition in the bottom cladding. As the p-GaN thickness increases beyond 350 nm, the mode overlap with the TCO is minimal and the choice of TCO has negligible effect on α_i.

The dependence of η_d on p-GaN thickness for both TCOs is shown in Fig. 1(c). For these calculations, the cavity length and mean reflection coefficient, R [36], were assumed to be 1200 μm and 0.18, respectively, corresponding to a mirror loss of 14.3 cm⁻¹. For p-GaN thicknesses greater than 350 nm, the TCO choice has 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. As shown in Fig. 1(c), LD structures with ITO top cladding have lower η_d than LD structures with ZnO top cladding. For p-GaN thicknesses between 100 nm and 350 nm, the η_d of ITO top clad devices decreases rapidly for decreasing p-GaN thicknesses because the α_i of ITO top clad devices increases rapidly for decreasing p-GaN thicknesses. In contrast, the α_i and therefore η_d of the devices with ZnO top cladding remains relatively constant with decreasing p-GaN thickness. Thus, replacing ITO with ZnO in the top cladding has a considerable effect on η_d and should have an even higher impact for LD designs that have lower internal mode loss [37].

LAE enables various compositions of Al in the n-AlGaN bottom cladding. In all previously reported LAE III-nitride devices, the bottom cladding thickness has been fixed at ~1 μm [21,23], but reducing the n-AlGaN thickness can reduce the overall stress in the LD structure. The dependence of Γ and α_i on n-AlGaN thickness and TCO material are shown in Fig. 2 for a device with 200 nm of p-GaN, 250 nm of ZnO, and the same WG and AR as the simulations in Fig. 1. Both Γ and α_i are almost unchanged for n-AlGaN thicknesses between 500 nm and 1000 nm. This indicates that the n-AlGaN bottom cladding thickness can be significantly reduced without sacrificing the Γ of the optical mode. The transverse mode profile of a ZnO top-clad LD with 500 nm of n-AlGaN, depicted in Fig. 2(b), shows that there is minimal mode overlap with the n-GaN buffer layer and no mode leakage into the substrate.

 

Fig. 2 (a) Dependence of confinement factor and internal mode loss on the AlGaN bottom cladding for devices with 5% Al composition in the bottom cladding, and 200 nm of p-GaN and 250 nm of ITO in the top cladding. (b) Transverse mode profile and refractive index as a function of distance in the growth direction for an LD with a 500 nm Al_0.05Ga_0.95N bottom cladding.

Download Full Size | PDF

3. Experimental

Taking into account the modeling results discussed above, metalorganic chemical vapor deposition (MOCVD) was used to grow a laser structure on a free-standing semipolar $(20\overline{2}1)$) GaN substrate from Mitsubishi Corporation with a TD density of 5 x 10⁶ cm⁻². LAE mesa stripeswere patterned on the substrate prior to the device growth and were oriented parallel to the $[\overline{1}014]$ direction, the projection of the $[0001]$ axis, to effectively block TD glide resulting from c-plane slip in the semipolar $20\overline{2}1$) plane [38]. The stripes were 14 μm wide and were etched 1 μm into the substrate, using BCl₃/Cl₂ chemistry. The laser structure consisted of a 600 nm n-GaN template, a 186 period Al_0.1Ga_0.9N/GaN modulation-doped short period superlattice (SPSL) n-cladding layer, with a total thickness of ~750 nm and average AlN fraction of 5%, a 60 nm n-In_0.08Ga_0.92N waveguiding layer, an undoped active region consisting of three unintentionally doped (UID) 3.7 nm In_0.2Ga_0.8N QWs and four 7.6 nm GaN QBs, a 10 nm p-Al_0.21Ga_0.79N EBL, a 40 nm p-In_0.08Ga_0.92N waveguiding layer, a 205 nm p-GaN layer, and a 12 nm p⁺ GaN contact layer. This device had a simulated Γ of 4.7% and α_i of 15.1 cm⁻¹. A cross-sectional schematic of the LD is shown in Fig. 3.

 

Fig. 3 Cross-sectional schematic of the device structure reported in this letter, consisting of 750 nm of n-Al_0.1Ga_0.9N/GaN SPSL, 60 nm p-In_0.08Ga_0.92N waveguiding layer, three 3.7 nm In_0.2Ga_0.8N quantum wells and four 7.6 nm quantum barriers, a 10 nm p-Al_0.21Ga_0.79N electron blocking layer (EBL), a 40 nm p-In_0.08Ga_0.92N waveguiding layer, a 200 nm p-GaN layer, and 1.4 μm of ZnO. The thin light-yellow layer indicates the location of the p-Al_0.21Ga_0.79N EBL.

Download Full Size | PDF

The LDs were fabricated using a self-aligned ridge waveguide polished facet process. The ridge waveguides were defined using a BCl₃/Cl₂ RIE etch. Following the waveguide etch, SiO₂ was deposited via sputtering for electrical isolation. After the deposition, the SiO₂ on top of the ridges was lifted off, exposing the top surface of the laser stripes for ZnO deposition. A 1.4 μm 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, 30, 39–41]. The ZnO deposition was followed by the deposition of 30/1000 nm Ti/Au p-contact pads. Dicing and mechanical polishing were then used to form the laser facets. The process was completed with deposition of a common 50/300 nm Al/Au back side n-contact via electron beam evaporation.

4. Results and discussion

The LDs were diced to a length of 900 μm and the ridge widths varied from 1.6 μm to 8.0 μm. Figure 4(a) shows the measured pulsed and continuous wave (CW) light-current-voltage (L-I-V) characteristics of a 4.0 μm wide by 900 μm long ridge waveguide LD with uncoated facets. The LDs were tested under pulsed electrical injection with a 2-probe configuration and a pulse width of 1 μs and 1% duty cycle. Pulsed lasing was achieved at a threshold current (I_th) of 236.5 mA, a threshold current density (J_th) of 6.6 kA/cm2, and a threshold voltage (V_th) of 6.9 V. The differential resistance (R_d) measured under pulsed conditions was 4.8 Ω. The peak output power measured from both facets was 236.2 mW and double facet slope efficiency was 1.03 W/A, which corresponded to a double facet η_d of 36.8%. The η_d of this device is significantly higher than the η_d of previously reported AlGaN-free ZnO top clad LDs η_d (27%) [31], and the η_d of previously reported LAE ITO top clad LDs (15.06%) [21]. The η_d of this device is also higher than the calculated η_d from FIMMWAVE simulations (32%), indicating that either the assumption for optical loss in the ZnO is too high, the assumption for injection efficiency is too low, or both. The same device was measured under CW conditions with a 4-probe configuration. The CW V_th was 5.8 V, which was more than 1 V lower than the pulsed V_th. The CW V_th was lower than the pulsed V_th due to better thermal activation of the acceptors in the p-GaN as a result of the self-heating as well as greater accuracy in measurement due to the subtraction of the voltage drop across the probes in 4-probe configuration. The CW measurement showed an increased J_th of 10.4 kA/cm² and a decreased double facet slope efficiency of 0.6 W/A due to self-heating.

 

Fig. 4 (a) Pulsed and CW L-I-V characteristics of a 4.0 μm wide by 900 μm long LD, (b) Current dependent spectra of a 1.6 μm wide by 900 μm long LD, showing emission at 445 nm, and (c) Pulsed temperature dependent L-I-V curves of a different 4.0 μm wide by 900 μm long LD. The inverse of the slope of a linear fit of the ln(I_th) vs. temperature data gives the T_o of the device, which was found to be 104 K.

Download Full Size | PDF

The CW R_d was 1.1 Ω, which was significantly lower than the pulsed R_d of 4.8 Ω, discussed above and the R_d of a previously reported AlGaN-free ZnO top clad LD [31]. This is partly due to the reasons mentioned above related to self-heating and the probing configuration, but it also indicates improved ZnO uniformity and a better physical contact between the ZnO and the p-GaN, compared to previously reported ZnO-top clad LDs [31]. Considering the CW R_d reported above, and assuming that the resistive contribution of all other layers in the LD structure is negligible, we estimate that the specific contact resistivityof the ZnO/p-GaN interface is at most 3.8 x 10⁻⁵ Ω-cm². This value is comparable to state-of-the-art metal alloy contacts to p-GaN and establishes that the specific contact resistivity of the ZnO/p-GaN interface is acceptable for LDs and other devices that operate under high current densities.

Figure 4(b) shows the pulsed current dependent spectra of a 1.6 μm wide by 900 μm long LD. The emission wavelength is 445 nm at 2.4I_th. The pulsed temperature dependent L-I-V curves shown in Fig. 4(c), belonging to a different 4.0 μm wide x 900 μm long LD than the one depicted in Fig. 4(a), were used to calculate the characteristic temperature (T_o) [36], of 103.7 K. Although this T_o (which is for a blue LD) is higher than the T_o for violet m-plane LDs reported by Farrell et.al [35], it is lower than the T_o for green ($20\overline{2}1$) LDs reported by Sizov et.al [42]. This is reasonable because the T_o for this device is expected to be lower than the T_o for the green ($20\overline{2}1$) LDs since the shallower QWs are expected to have more carrier overflow.

5. Conclusions

In conclusion, we have demonstrated blue ($20\overline{2}1$) LDs that utilize LAE n-AlGaN bottom cladding layers and ZnO top cladding. Room temperature CW lasing was achieved at 445 nm with a threshold current density of 10.4 kA/cm², a threshold voltage of 5.8 V and a differential resistance of 1.1 Ω for a 4.0 μm wide by 900 μm long ridge waveguide LD with uncoated facets. The specific contact resistivity of the ZnO/p-GaN interface was estimated to be in the low 10⁻⁵ Ω-cm² range.