1. Introduction

Semipolar ($20\overline{2}1$) plane In_xGa_1-xN/GaN quantum wells (QWs) are an attractive alternative to the commonly used polar QWs, especially in the green spectral region because of the good In uptake and the small intrinsic electric fields [1]. Green ($20\overline{2}1$) plane light emitters are typically based on single or a few narrow QWs [2–8]. Using a larger number of QWs would enhance the volume of the active region, increase the output power and reduce the droop effect. However, increasing the number of wells might have several drawbacks. First, carrier, especially hole, distribution over the QWs might be nonuniform. The issue might find its solution via band gap engineering [9]. Second, increasing the cumulative QW width over the critical thickness [10] might induce misfit stress relaxation via dislocation formation [11], which could affect the emission efficiency and broaden the spectra. The emission broadening is particularly critical for lasers because it affects the modal gain. However, even without the strain relaxation multiple QW (MQW) emission might have a larger linewidth because of different properties of individual QWs. These properties include well width, alloy composition and strain variations, and abruptness of QW interfaces.

The spectral linewidth of InGaN QWs with a high In content has a large contribution from the inhomogeneous broadening [12], which can be assigned to 1-10 nm scale band potential fluctuations [13]. For MQWs, the band potential fluctuations of individual QWs might be correlated by strain fields, similarly to multiple layer quantum dot structures [14], or might occur at random. Different band fluctuation mechanisms might have a different impact on the MQW linewidth and its spatial distribution compared to the single QW (SQW) case. These issues are investigated in this work by studying PL spectra from semipolar ($20\overline{2}1$) plane InGaN SQW and MQW structures emitting in the green spectral region. The PL is measured in the near-field regime, which is a proper way to study the spectral broadening, as it has been proven for InGaN QWs of different crystallographic orientations [12–17].

2. Experimental details

The QW structures were grown by metal organic chemical vapor deposition on low (~5 × 10⁶ cm⁻²) dislocation density bulk ($20\overline{2}1$) plane GaN substrates, provided by Mitsubishi Chemical Corporation. The structures consisted of a 1 µm thick undoped GaN template layer, a SQW or a MQW structure consisting of 4 nm In_0.27Ga_0.73N wells and 8 nm GaN barriers, and a 10 nm GaN cap layer. The barriers were wide enough to prevent QW coupling and formation of superlattice type common energy levels. The In content in the QWs was evaluated by the X-ray analysis. The number of QWs in the samples was 1, 5 and 10. In detail, the growth procedure is described in [12].

Near-field PL measurements were carried out at room temperature with a scanning near-field optical microscope (SNOM). The PL was excited and collected through a 100 nm aperture Al-coated probe. For excitation, 200 fs pulses from a frequency doubled Ti:sapphire laser were used. Near-field PL spectra were measured in the time-integrated mode with a spectrometer and a cooled CCD detector. Data presented below are for a 355 nm excitation wavelength and a 0.4 mW average power coupled into the fiber. Changing the excitation wavelength to 420 nm or the power to 40 µW did not markedly change the results. This shows that the carrier transfer from the barriers to the wells does not affect the spatial distribution of the QW PL, and that 0.4 mW average excitation power does not induce band filling [19]. The feedback signal of the SNOM tuning fork was used to obtain surface morphology maps.

3. Results and discussion

Figure 1 shows normalized far-field spectra of the studied QW structures. The spectra have slightly asymmetric Gaussian shapes modulated by weak Fabry-Perot oscillations. PL around the peak wavelengths is generated by optical transitions between the band edges. The long wavelength tails of the spectra can be assigned to transitions at localized states [12,17].

 

Fig. 1 Normalized far-field PL spectra of 1, 5 and 10 QW structures.

Download Full Size | PDF

Figure 2 displays maps of the surface morphology, the peak PL wavelength and the peak PL intensity for the SQW structure. The surface morphology map has a root mean square (rms) value of 1.37 nm and displays undulations characteristic for the ($20\overline{2}1$) plane surface [20,21]. In ($20\overline{2}1$) QWs, the PL peak corresponds to transitions between band edges in the conduction and valence bands [12]. Moreover, room temperature carrier lifetimes in semipolar InGaN QWs are governed by the nonradiative recombination [23]. Thus, spatial variations of the peak wavelength can be assigned to fluctuations of the InGaN alloy composition and/or the QW width. The PL intensity map indirectly reflects distribution of the nonradiative recombination centers. A full width at half maximum (FWHM) map (not shown) reveals spatial variations of the localized state contribution to the emission spectra.

 

Fig. 2 Maps of the surface morphology$20\overline{2}1$ (a), the peak PL wavelength (b) and the peak PL intensity (c) for the SQW structure. Part (d) shows near-field PL spectra measured at points A and B indicated on the peak wavelength map.

Download Full Size | PDF

Near-field spectra, measured at various points of the scan, except for small variations in intensity, are very similar, Fig. 2(d). FWHM values are about 190 meV (Table 1), not much different from the previous data for () single QWs (SQWs) emitting in the green spectral region [12,15]. About 90 meV of the FWHM can be assigned to the homogeneous broadening and the random alloy distribution; the rest originates from the inhomogeneous broadening [12]. The inhomogeneous broadening in a thin SQW is determined by 1-10 nm scale band gap fluctuations that cannot be explicitly revealed to to the limited spatial resolution (100 nm) of the SNOM. Recent atom probe tomography measurements showed that ($20\overline{2}1$) plane QW interfaces may be faceted with a few nm period [20]. The faceting causes well width variations, which can be a likely reason for the inhomogeneous broadening of the PL.

Table 1. Statistical parameters of the near-field PL scans for the studied QW structures

View Table

Despite the very uniform spectra, the near-field scans reveal sub-μm to µm scale islands of similar values. In the peak intensity and the FWHM maps, these islands appear to be random. On the other hand, features in the peak wavelength map are aligned along the [$\overline{1}2\overline{1}0$] direction and are related to the surface morphology. Possibly, the aligned islands of similar peak wavelengths originate from well width variations occurring on a 100 nm to 1 μm scale [20,21]. A similar correlation between the surface morphology and the peak wavelength has been observed in earlier studies of semipolar InGaN QWs as well [15,18].

Variations of the PL parameters are very small, which results in remarkably small values of standard deviations of the spectral parameters (Table 1). The ratio of the peak wavelength deviation to the average value is of the order of 10⁻⁴. If the whole peak wavelength variation were related to fluctuations of alloy composition, the standard deviation of the x value would be only ~3 × 10⁻⁴. Ratios of the peak intensity deviations and the average values are just a few percent indicating a uniform distribution of the nonradiative recombination centers. The variations of the PL parameters are considerably smaller than previously reported values for semipolar ($20\overline{2}1$) and ($20\overline{2}\overline{1}$) plane InGaN QWs [12,15,18]. The increased spatial uniformity of the PL may be attributed to optimization of the growth conditions. For instance, a previous study of ($20\overline{2}\overline{1}$) InGaN QWs has revealed that growth temperature has a strong influence on the band potential fluctuations and the uniformity of the PL [18]. Correlations between the spectral parameters are very weak with the values of Pearson product-moment correlation coefficient close to zero (Table 1). This confirms that spatial distribution of the carrier lifetimes (reflected by the PL intensity map) and localized states (the FWHM map) are neither correlated between themselves nor with the composition and/or well width fluctuations, reflected in the peak wavelength map.

For the MQW structures, the surface undulation pattern is preserved, Fig. 3(a), even though the surface becomes smoother ($\text{rms}=0.95\text{nm}$). However, this pattern is less pronounced in the peak wavelength distribution, Fig. 3(b). Features in the peak intensity and the FWHM maps are not correlated with the surface morphology, similarly to the SQW.

 

Fig. 3 Maps of the surface morphology (a), the peak PL wavelength (b), the peak PL intensity (c) and the FWHM for the 10 QW structure.

Download Full Size | PDF

The most remarkable feature of the SNOM data is that the average peak wavelength and the FWHM, as well as the deviation values for the SQW and the MQW structures are essentially the same (Table 1). Spans of the peak wavelengths reflecting the range of their variations also remain similar. These observations reveal several important details. First, the growth of all QWs in the MQW structures is identical in terms of well width and alloy composition. Second, no significant relaxation of the QWs occurs, even in the 10 QW structure; or, at least, the relaxation does not markedly affect the emission spectra. Third, the QW width and alloy composition fluctuations in the wells of the MQW structures are similar and uncorrelated. Fourth, the distribution of the carrier lifetimes (and related distribution of the nonradiative recombination centers) in different wells is quite uniform and uncorrelated between the wells. This, for instance, indicates that the main nonradiative recombination centers are point defects and not dislocations, because in the opposite case strong spatial variations of the PL intensity would occur [18,20]. And finally, the reduction of the morphology-related pattern in the peak wavelength maps of MQW structures shows that with increased QW number the influence of the surface undulations on the QW width and/or alloy composition is smeared out. Considering that the peak wavelength variations are very small in the first place, minute changes for individual wells would be sufficient to affect the overall MQW result. In fact, the averaging of the peak wavelength and the FWHM of the individual well PL has a positive effect for the MQW structures in slightly reducing the standard deviations of these parameters.

Intense and relatively narrow MQW PL spectra is an important finding that encourages exploring a possibility of using a large number of QWs in ($20\overline{2}1$) plane InGaN/GaN photonic devices. This is even more remarkable taking into account the large cumulative thickness of the InGaN layers, 20 nm and 40 nm for the 5 and 10 QW structures, respectively. To evaluate whether the thickness of the MQWs structures exceed the critical thickness for the strain relaxation, one may assign the well and the barrier an average In content of 9% (4 nm wells with 27% In and 8 nm barriers with 0% In). The critical thickness for this average alloy composition is ~25 nm [23]. Previous studies of strain relaxation in the ($20\overline{2}1$) plane have shown that the relaxation occurs at thicknesses that are 2 to 3 times larger than the calculated critical thickness [24]. This is close to the thickness of the QW region in the 5 QW sample. For the 10 QW structure, the thickness of the MQW region (10 wells, 9 barriers, 112 nm) is over 4 times larger than the theoretical critical thickness, thus, the structure is expected to be relaxed. However, kinetic factors during growth might delay the onset of the relaxation. Besides, investigations of strain relaxation in ($20\overline{2}1$) plane MQWs with different well and barrier parameters have revealed that one-dimensional relaxation via misfit dislocations do not affect PL spectra [11]. A deterioration of the PL intensity has only been observed after the on-set of two-dimensional relaxation via non-basal plane defects.

In spite of the very similar properties of the SQW and MQW near-field PL, a slight difference in the recombination features can be noticed. It appears in the peak intensity and the FWHM correlations (Fig. 4). For the SQW and 5 QW structures, these parameters have a very weak negative correlation. For the 10 QW structure, the correlation increases.Wider PL spectra have a larger contribution from the localized states. The negative correlation shows that a larger number of localized states reduces the overall PL intensity, or, in other words, that the low energy states are less prone to the nonradiative recombination. These localized states in the 10 QW structure may be related to partially relaxed regions with a higher defect concentration. It should be noted, though, that the partial decrease of the radiative efficiency of the localized states is detected only through the correlations and not by major changes in the PL intensity. Thus, it should not have a marked influence on the device performance.

 

Fig. 4 Correlation between the peak intensity and the FWHM for 1 (a), 5 (b) and 10 (c) QW structures.

Download Full Size | PDF

4. Conclusions

Scanning near-field PL spectroscopy has been applied on ($20\overline{2}1$) plane In_0.27Ga_0.73N/GaN single and multiple QWs to reveal the uniformity of individual QWs of MQW structures. The experimental data show that there is no additional broadening or shift of the PL spectra in the MQWs compared to the SQW structure. The results indicate that the wells in MQW structures are nearly identical and the well width and alloy composition fluctuations are uncorrelated. No marked influence of the possible structural relaxation on the emission properties has been detected, even for 10 QW structure. This suggests that using a large number of QWs in the active region of semipolar light emitters, provided that the uniform carrier distribution could be achieved by the band gap engineering, is a promising way to increase the emission intensity and reduce the droop effect.