1. Introduction

Nitride nanorod (NR) light-emitting diode (LED) is attractive because of the advantages of dislocation-free, high-crystal-quality growth, a larger emission area (emission from NR sidewalls), and non-polar or semi-polar quantum well (QW) formation based on sidewall or slant-facet growth. Randomly-distributed self-organized GaN NRs have been widely grown on sapphire and silicon substrates with molecular beam epitaxy (MBE) [1–5] and metalorganic chemical vapor deposition (MOCVD) [6–13]. Selected or patterned growths of GaN NRs have also been widely implemented with MBE [14, 15] and MOCVD [16–23]. The patterning methods include electron-beam writing [14], focused ion beam [15], holography [16], and nano-imprint lithography [20, 23]. In regularly patterned growth, vertically oriented NRs of uniform size and height distributions can be obtained. With MOCVD growth, the pulsed growth mode, in which the gallium and nitrogen sources are switched on and off alternatively, has been used for implementing the self-catalytic vapor-liquid-solid growth mode [16, 17, 20–22]. For fabricating NR LEDs, InGaN/GaN QW p-i-n NR structures have been widely grown. On a c-axis-oriented GaN NR, QWs can be grown on the flat top face of an NR to form a disc-like c-plane (polar) QW structure. Such a QW structure can usually be formed with MBE [3–5, 15], but can also be formed with MOCVD [10]. Also, on the slant {1-101} facets of top pyramid geometry in an NR, semi-polar QWs can be grown with MBE [24] and MOCVD [25]. Meanwhile, on the sidewalls of an NR, a core-shell m-plane (non-polar) QW structure can be formed on either Ga-face [20–22, 26–28] or N-face NRs [12, 13, 29, 30]. With MOCVD, core-shell QWs and hetero-structures can also be grown on a-axis-oriented GaN NRs [31–33].

With the techniques of nano-imprint lithography and MOCVD pulsed growth, regularly-patterned, c-axis nitride NR arrays of uniform geometry with simultaneous depositions of c-plane disc-like and m-plane core-shell InGaN/GaN QW structures have been fabricated [20]. Also, the dependencies of the cross-sectional sizes of GaN and QW NRs with different heights and different hexagon orientations on the hole diameter, pitch, and crystal orientation of the patterned template have been studied [21]. It was found that the cross-sectional size of the GaN NR was mainly controlled by the patterned hole diameter and was weakly dependent on the NR height, pitch, and pattern orientation. Also, the thickness of the sidewall QW structure was mainly determined by the NR height and was weakly affected by the patterned hole diameter, pitch, and pattern orientation. Based on the QW NR growth technique, a core-shell p-i-n structure was grown for fabricating a blue-emitting NR LED array by depositing a conformal layer of transparent conductor, GaZnO, with MBE on the NRs [22].

It has been speculated that indium incorporation in growing an InGaN/GaN QW on a GaN NR, with either a disk-like or a core-shell structure, can be enhanced for elongating the emission wavelength due to the strain relaxation in such a nanostructure [34, 35]. If this speculation can be proved to be true, it is useful for implementing a non-polar or semi-polar LED with an NR structure in the green-yellow range since indium incorporation in growing a planar non-polar or semi-polar QW is usually lower than that in a polar QW [36]. Due to the zero and reduced quantum-confined Stark effect (QCSE) in a non-polar and a semi-polar QWs, respectively, their light emission efficiencies can be enhanced. However, so far, the emission spectral peaks of all the m-plane and {1-101}-plane QWs grown on the sidewalls and slant-facets, respectively, of the regularly-patterned c-axis-oriented GaN NR arrays with MOCVD reported in literature are shorter than 500 nm [21, 23, 27, 35, 37–39]. Although a long wavelength of 598 nm was achieved by lowering the QW growth temperature to 550 °C in a randomly distributed NR structure with MOCVD growth [40], such a low temperature is not usually used for high-quality QW growth. The emission wavelengths from other randomly distributed QW NR structures reported in literature are all significantly shorter than 500 nm [29, 41, 42]. To implement an efficient long-wavelength LED, it is important to understand the conditions for achieving an emission wavelength longer than 500 nm based on an m-plane, QCSE-free QW structure on an NR. For this purpose, we need to systematically test various reasonable QW growth conditions by varying three commonly used QW growth parameters, including growth temperature (controlling indium incorporation), growth time (controlling QW width), and indium supply rate (also controlling indium incorporation). In particular, the growth scenarios of InGaN/GaN QWs on an NR are quite different from those in planar growth. Also, because the emission wavelength of such a QW is related to its indium composition and width, it is crucial to identify the relations between the growth conditions and the QW parameters. Such relations can also help us in exploring the growth mechanisms of the QWs on an NR. In this paper, we report the growth results of regularly patterned InGaN/GaN QW NR arrays of similar cross-sectional sizes and heights on the same patterned templates, but under different conditions of growth temperature, growth duration, and TMIn flow rate in QW growth and compare the QW emission wavelengths among different NR samples and with that of a planar structure. To understand the QW structures of those samples, including well widths and indium contents, strain state analyses (SSA) based on two-beam transmission electron microscopy (TEM) are undertaken for comparing with the variation trends of QW emission wavelength. In section 2 of this paper, the sample structures, growth conditions, and characterization methods are presented. The results of scanning electron microscopy (SEM) and TEM observations are shown in section 3. Then, in section 4, the basic photoluminescence (PL) and cathodoluminescence (CL) measurement results are reported. The comparisons of PL and CL results among different samples are given in section 5. The SSA results are presented in section 6. Next, discussions are given in section 7. Finally, conclusions are drawn in section 8.

2. Sample structures, growth conditions, and characterization methods

For this study, we prepare eight NR-array samples based on the same triangular hole pattern on GaN templates through the nano-imprint lithography technique. The hole diameter and pitch (the center-to-center distance between two nearest neighboring holes) of the hole pattern are 250 and 700 nm, respectively. The un-doped GaN NRs are grown using the pulsed growth technique [16]. Different QW growth conditions, including the TMIn flow rate, growth temperature, and growth duration, for the InGaN well layers, are used for preparing the eight NR array samples (samples B-I), as listed in Table 1For comparison, a planar QW sample (sample A) is also prepared under the same QW growth conditions as those for NR sample B.

Table 1. Growth conditions and the peak wavelengths (λ) of the PL and CL spectra of the nine samples under various measurement conditions. Sample A has a planar QW structure. Samples B-I have NR structures.

View Table | View all tables in this article

The growth of an un-doped GaN NR array is patterned with a 40-nm-thick SiO₂ mask on a 2-μm-thick GaN template, which is deposited with MOCVD on c-plane sapphire substrate. The growth of GaN NRs starts with a hole-filling process under the MOCVD conditions of 100 torr in chamber pressure, 1500 rpm in turbo disc speed, 1050 °C in substrate temperature, 50 sccm in TMGa flow rate, 1500 sccm in NH₃ flow rate (1100 in V/III ratio), and 6 sec in growth duration. After the hole filling process, the pulsed MOCVD growth mode is used at 1050 °C with V/III ratio at 550 for forming essentially flat-top GaN NRs of a hexagonal cross-sectional shape. In this process, TMGa (50 sccm) and NH₃ (500 sccm) flows are switched on and off alternatively with the flow durations of TMGa and NH₃ at 20 and 30 sec, respectively. A pause of 0.5 sec in duration is applied after each TMGa flow half-cycle. After the pulsed growth of 37 cycles (~31 min in duration), regularly patterned GaN NRs of uniform geometry with a height of ~1000 nm are obtained. On the GaN NRs, three periods of InGaN/GaN QWs are deposited with the two-dimensional MOCVD growth mode under the conditions of 200 torr in chamber pressure, 750 rpm in turbo disc speed, 32 sccm in TEGa flow rate, and 2700 sccm in NH₃ flow rate. The growth temperature and duration for a GaN barrier layer are fixed at 870 °C and 11 min, respectively. For samples B-E, the QW growth temperature and growth duration are fixed at 700 °C and 50 sec, respectively, while the TMIn flow rate is varied from 240 through 420 sccm. Then, for samples D, F, and G, the TMIn flow rate and QW growth duration are fixed at 360 sccm and 50 sec, respectively, while the QW growth temperature is decreased from 700 to 670 °C. Next, for samples G-I, the TMIn flow rate and QW growth temperature are fixed at 360 sccm and 670 °C, respectively, while the QW growth duration is increased from 50 to 100 sec.

The SEM measurement is performed with a JEOL JSM-7001F system, in which a Gatan MonoCL4 module is installed for CL measurement. The SEM (CL) results are obtained with the operation conditions of 10 kV (5 kV) in acceleration voltage and 0.3 nA (7 nA) in probe current. The spatial resolution of SEM is ~3 nm and that of CL is ~10 nm. A local CL spectral measurement covers a spatial domain of 100 nm x 75 nm in dimension. The spectral range in a monochromatic CL mapping is 2 nm. The TEM investigation is performed using a Philips Tecnai F30 field-emission electron microscope with an accelerating voltage of 300 kV and a probe forming lens of Cs = 1.2 mm. The high-resolution micrographs for SSA are taken with two-electron-beam interference [43–45]. The current density focused onto the sample is estimated to be lower than 16 A/cm². In the SSA, the indium composition analysis is based on the measurement of local lattice parameter, which is the distance between the neighboring local intensity maximum positions in a two-beam TEM image. The local composition is obtained with the reference of GaN in the barrier region in the same image. The average QW width in an SSA image is obtained by first taking the average of the cross-well indium composition profiles and then evaluating the full-width at half-maximum (FWHM) of the average profile. The average indium content of a QW is obtained by integrating the indium composition in an SSA image covering a section of a QW and then being divided by the effective QW area, which is equal to the product of the average QW width and the section length of the QW.

3. Nanorod morphologies and nanostructures

Figures 1(a) and 1(b) [1(c) and 1(d)] show plan-view (cross-sectional) SEM images of sample B with a smaller and a larger scale, respectively. From Fig. 1(a), one can see that the cross-sectional dimension (the distance between two parallel lateral sides of the hexagonal cross section) of the NR is about 425 nm. From Fig. 1(c), one can see that the NR height is about 1 μm. The NR top shows a truncated pyramidal geometry. Figure 2(a) shows a cross-sectional TEM image of an NR in sample B. Here, three periods of c-plane QW can be seen on the top face of the NR. However, the sidewall QWs are unclear in this TEM image. The HAADF image of a neighboring NR is shown in Fig. 2(b). In this HAADF image, both the top-face c-plane and sidewall m-plane QWs can be clearly seen. The magnified TEM images of the circled portions of the top-face QWs, the sidewall QWs near the top, near the middle height, and near the bottom of the NR in Fig. 2(a) are shown in Figs. 2(c)-2(f), respectively. Here, one can see that the thicknesses of the well and barrier layers of the top-face c-plane QWs are significantly larger than those of the sidewall m-plane QWs.

 

Fig. 1 (a) and (b) [(c) and (d)]: Plan-view (cross-sectional) SEM images of sample B with a small and a large scale, respectively.

Download Full Size | PDF

 

Fig. 2 (a): A cross-sectional TEM image of an NR in sample B. (b): A HAADF image of a neighboring NR in sample B. The magnified TEM images of the circled portions at the top-face QWs, the sidewall QWs near the top, near the middle height, and near the bottom of the NR in part (a) are shown in parts (c)-(f), respectively.

Download Full Size | PDF

4. Cathodoluminescence and photoluminescence measurements

Figures 3(a)-3(c) show a plan-view SEM image, the panchromatic CL image, and the CL image at 428 nm, respectively, at the same location of the NR array in sample B. The strong emission at the rim of each NR in Fig. 3(b) originates from the sidewall QWs. The relatively weaker emission around the NR center shown in Fig. 2(b) comes from the top-face c-plane QWs. It is weaker because of the smaller emission volume (QW volume) per unit planar area in the penetration depth (~150 nm) of the accelerated electrons (5 kV) in the CL measurement, when compared with that of the sidewall m-plane QW at the rim of a plan-view NR. Although the wavelength of 428 nm corresponds to the CL emission spectral peak of the top-face QWs, as to be discussed later, in the monochromatic CL image at this wavelength, as shown in Fig. 2(c), the emission intensity of the sidewall QWs at the NR rim is still higher than that of the top-face QW near the NR center. Figures 3(d)-3(f) show a cross-sectional SEM image, the panchromatic CL image, and the CL image at 406 nm, respectively, at the same location of the NR array in sample B. Here, in Figs. 3(e) and 3(f), one can see that the CL emission intensity in the upper one-half of the NRs is significantly stronger than that of the lower one-half. The wavelength of 406 nm corresponds to the spectral peak in the cross-sectional CL measurement. The CL emission intensity difference between the upper and lower one-halves is not caused by the difference of emission spectrum, as to be discussed later.

 

Fig. 3 (a)-(c) [(d)-(f)]: Plan-view (Cross-sectional) SEM image, the panchromatic CL image, and the CL image at 428 (406) nm, respectively, at the same location of the NR array in sample B.

Download Full Size | PDF

Figure 4 shows the CL spectra of sample B under different measurement conditions, including the large-scale plan-view (PV) and cross-section (CS) measurements, and the plan-view local measurement at the NR center (PV-c), the cross-sectional local measurement at the top for the top-face QWs (CS-top), the cross-sectional local measurements for the sidewall QWs near the top (SW-t), near the middle height (SW-m), and near the bottom (SW-b). Here, one can see that the PV spectrum has a peak at 428 nm and a shoulder feature at 406 nm. On the other hand, the CS spectrum has a peak at 406 nm and a shoulder feature at 428 nm. Therefore, the peak at 428 (406) nm originates from the emission of the top-face (sidewall) QWs. The peak at 365 nm in each curve comes from the emission of GaN. The spectral portion longer than 500 nm is caused by defect emission, which can usually be observed in CL measurement. Although the emission peaks of curves PV-c and CS-top are supposed to originate from the top-face QWs and are dominated by the feature at 428 nm, the contributions of the 406 nm feature are quite strong due to the coverage of the electron footprint (100 nm x 75 nm) onto the top portion of the sidewall QWs in the local CL measurement. Similarly, the spectral peak of curve SW-t has a shoulder feature at 428 nm, indicating that the coverage of the electron footprint onto the top-face QWs. The spectral peaks of curves SW-t, SW-m, and SW-b are slightly different at 410, 406, and 402 nm, respectively. The similar CL spectral results from other samples are also obtained. Although their spectral peak wavelengths are different, the variation trends are the same. The overall plan-view and cross-sectional CL spectral peaks of all the eight NR samples are listed in columns 6 and 7, respectively, of Table 1.

 

Fig. 4 CL spectra measured under different conditions of sample B, including the large-scale plan-view (PV) and cross-sectional (CS) measurements, the plan-view local measurement at the NR center (PV-c), the cross-sectional local measurement for the top-face QWs (CS-top), and the cross-sectional local measurements for the sidewall QWs near the top (SW-t), near the middle height (SW-m), and near the bottom (SW-b).

Download Full Size | PDF

As shown in Fig. 3(f) for sample B, the CL emission at the spectral peak (406 nm) in cross-sectional measurement is stronger in the upper one-half of the NRs. Figures 5(a)-5(h) show the cross-sectional CL images at individual spectral peaks of samples B-I, respectively. For samples B-I, the mapping wavelengths are 406, 414, 422, 416, 434, 450, 490, and 500 nm, respectively, as listed in column 7 of Table 1. Figure 5(a) is a duplicate of Fig. 3(f) for easy comparison. Here, one can see that as shown in Figs. 5(a)-5(d) for samples B-E, respectively, the CL emission intensities are significantly stronger in the upper one-third to one-half of the NRs, when compared with the rest portions. Then, in Fig. 5(e) for sample F, the bright region extends toward the bottom of the NRs. Next, in Figs. 5(f)-5(h) for samples G-I, the bright regions cover almost the whole NR height range. These differences in bright region are related to the QW growth temperature. The InGaN well layers in samples B-E, sample F, and samples G-I are grown at 700, 685, and 670 °C, respectively. It is noted that in Figs. 5(g) and 5(h), the bright CL images covering the whole NR sidewalls are partly contributed by defect emission, as to be shown in Fig. 6.

 

Fig. 5 (a)-(h): Cross-sectional CL images at individual spectral peaks at 406, 414, 422, 416, 434, 450, 490, and 500 nm for samples B-I, respectively.

Download Full Size | PDF

 

Fig. 6 Normalized PL spectra of the nine samples.

Download Full Size | PDF

Figure 6 shows the normalized PL spectra of the nine samples. The PL measurement is excited by a HeCd laser at 325 nm. Because the samples are excited from the NR top-side (30 degrees incident angle with respect to the vertical direction) and the 325-nm laser can penetrate into the nitride materials for only ~100 nm, the PL emissions mainly come from the top-face QWs and the top portion of the sidewall QWs. In the PL spectra of samples B-G in Fig. 6, besides the individual major peaks from QW emissions, the small peaks at 365 nm correspond to the emission of GaN and the broad humps with the maxima around 545 nm originate from defect emission. The defect emission intensity is generally stronger as the QW emission wavelength of the sample becomes longer. In the spectrum of sample H, two peaks can be seen. The long-wavelength peak at 545 nm originates from defect emission. The short-wavelength one at 509 nm comes from QW emission in this sample. In sample I, besides the defect emission peaked at 545 nm, the shoulder feature around 518 nm is supposed to come from QW emission. The defect emissions in samples H and I look stronger because of the relatively weaker QW emissions in these two samples. In samples H and I, although the PL emission spectra mix with defect emission spectra, the PL spectral peaks can still be clearly identified. It is noted that in Fig. 6, except sample A, either shoulder features or secondary peaks can be observed on the short-wavelength side of the major peaks in all samples. It is believed that the major PL spectral peaks originate from the emissions of the top-face QWs and the top-portion of the sidewall QWs. The short-wavelength shoulder features or secondary peaks are caused by the emissions from the lower-portions of the sidewall QWs. Because of the shadowing effect of PL excitation, the emission from the lower-portion of the sidewall QWs is weak in the PL measurement. The major peak wavelengths of the PL spectra from the QWs in the nine samples are listed in column 5 of Table 1. Their wavelengths are relatively closer to the spectral peaks of the CL emission from the top-face QWs, when compared with those from the sidewall QWs. It is noted that the Fabry-Perot oscillation in the PL spectrum of sample A shown in Fig. 6 due to the flat sample surface cannot be observed in NR samples.

5. Comparisons of cathodoluminescence and photoluminescence results

Figure 7 shows the variations of PL and CL spectral peak wavelengths under various measurement conditions with TMIn flow rate (including the data of samples B-E with the QW growth temperatures fixed at 700 °C and growth times fixed at 50 sec). The variation trends of all the curves in Fig. 7 are similar. As TMIn flow rate increases for increasing indium supply in growing the QWs, the emission spectral peak first red shifts and then blue shifts beyond 360 sccm. The blue-shift behavior between 360 and 420 sccm is particularly stronger in the emission of the sidewall QWs. The average red shift slope between 240 and 360 sccm of all the curves is 0.15 nm/sccm. Figure 8 shows the variations of PL and CL spectral peak wavelengths under various measurement conditions with QW growth temperature (including the data of samples D, F, and G with the TMIn flow rates fixed at 360 sccm and QW growth times fixed at 50 sec). Here, one can see that the emission spectral peaks in all the curves blue shift with increasing QW growth temperature. Except the steeper variation in the curve for the emission from the sidewall QWs near the top (SW-t), the variation slopes of all other curves are about the same with an average value of −0.9 nm/°C. The slope of curve SW-t is −1.3 nm/°C. Figure 9 shows the variations of PL and CL spectral peak wavelengths under various measurement conditions with QW growth time (including the data of samples G, H, and I with the TMIn flow rates fixed at 360 sccm and QW growth temperatures fixed at 670 °C). Here, the emission spectral peaks red shift with increasing QW growth time or well layer thickness in all curves. Because the collected emissions from the top-face QWs (for curve CS-top) and the sidewall QWs near the top (for curve SW-t) in cross-sectional CL measurement can be mixed, these two curves in either Fig. 8 or 9 are close. However, the emission spectral peak wavelengths of the top-face QWs (CS-top) are always longer than those of the overall emission from the sidewall QWs (CS). Their individual differences (Δλ_c-m) of those NR samples are listed in column 8 of Table 1. One can see that the difference generally decreases with increasing emission wavelength. In Figs. 7–9, one can see the same trend that the emission spectral peaks from the sidewall QWs near the top are the longest, followed by that from the middle height and then by that from the bottom. The differences of emission spectral peaks from the sidewall QWs between those from the top and bottom (Δλ_m) are listed in column 9 of Table 1 for all the NR samples. Here, one can see that this difference generally increases with increasing TMIn flow rate and decreasing QW growth temperature. In other words, it increases with increasing indium incorporation. However, this difference is essentially fixed when QW growth time or well layer thickness is increased.

 

Fig. 7 Variations of PL and CL spectral peak wavelengths under various measurement conditions with TMIn flow rate (including the data of samples B-E with the QW growth temperatures fixed at 700 °C and growth times fixed at 50 sec). The curves of CS-top and CS correspond to the emissions from the top-face QWs and the overall emission from the sidewall QWs, respectively.

Download Full Size | PDF

 

Fig. 8 Variations of PL and CL spectral peak wavelengths under various measurement conditions with QW growth temperature (including the data of samples D, F, and G with the TMIn flow rates fixed at 360 sccm and QW growth times fixed at 50 sec). The curves of CS-top and CS correspond to the emissions from the top-face QWs and the overall emission from the sidewall QWs, respectively.

Download Full Size | PDF

 

Fig. 9 Variations of PL and CL spectral peak wavelengths under various measurement conditions with QW growth time (including the data of samples G, H, and I with the TMIn flow rates fixed at 360 sccm and QW growth temperatures fixed at 670 °C). The curves of CS-top and CS correspond to the emissions from the top-face QWs and the overall emission from the sidewall QWs, respectively.

Download Full Size | PDF

6. Analysis results of quantum well width and indium content

Figure 10(a) shows a typical two-beam TEM image of a section of a top-face QW in sample G. Figure 10(b) shows the corresponding SSA image. Then, Fig. 10(c) shows a cross-well profile of indium composition along the vertical (white) line marked in Fig. 10(b). Figures 11(a)-11(c) demonstrate a two-beam TEM image, the corresponding SSA image, and a cross-well profile of indium composition, respectively, of a section of a sidewall QW near the middle height in sample G, similar to those in Figs. 10(a)-10(c). Based on the SSA image in Fig. 10(b) [Fig. 11(b)], the calibrated average QW width, which is obtained by averaging over all the similar cross-well indium composition profiles in the SSA image, and indium content are 4.52 nm and 15.6% (1.41 nm and 16.8%), respectively. Five similar SSA images at slightly different positions are acquired to evaluate the overall average QW width and average indium content for each designated QW portion, including the top-face QWs, the sidewall QWs near the top, middle height, and bottom. It is noted that the red spots in Figs. 10(b) and 11(b) correspond to indium-rich InGaN nano-clusters in the QWs [46].

 

Fig. 10 (a): Typical two-beam TEM image of a section of a top-face QW in sample G. (b): SSA image corresponding to the TEM image in part (a). (c): A typical cross-well indium composition profile along the vertical (white) line shown in part (b).

Download Full Size | PDF

 

Fig. 11 (a): Typical two-beam TEM image of a section of a sidewall QW near the middle height in sample G. (b): SSA image corresponding to the TEM image in part (a). (c): A typical cross-well indium composition profile along the vertical (white) line shown in part (b).

Download Full Size | PDF

Table 2 shows the evaluated results of average QW widths and average indium contents in various portions of all the nine samples. Here, the numbers before and after the slash represents the values of average QW width (in nm) and average indium content (in %), respectively. Here, one can see that the average QW width of the top-face QWs is significantly larger than those at various sidewall locations of the sidewall QWs in each sample. Also, the average QW width increases with the height on the sidewall. Meanwhile, the average indium content of the top-face QWs is always smaller than those at various sidewall locations of the sidewall QWs in each sample. In addition, on the sidewalls, the average indium content increases with the height of the observation point. Figure 12 shows the average QW widths (the left ordinate) and the average indium contents (the right ordinate) of the top-face QWs (TF) and the sidewall QWs near the top (SW-t), middle height (SW-m), and bottom (SW-b) as functions of TMIn flow rate (for samples B-E) when the QW growth temperature and time are fixed at 700 °C and 50 sec, respectively. Here, one can see that with increasing indium supply, the average QW width is slightly increased. This trend can be due to the stronger process of indium out-diffusion in increasing the total indium content in a QW such that the FWHM of indium distribution becomes larger. Also, one can see that with increasing indium supply, the indium content is generally increased until the TMIn flow rate reaches 360 sccm. The decreasing trend beyond 360 sccm is consistent with the blue-shift trend of QW emission wavelength in Fig. 7. The decreasing trend of indium content when the TMIn flow rate is increased from 360 to 420 sccm can be attributed to the parasitic chemistry effects [47]. The increase range of indium content in increasing TMIn flow rate is quite small, indicating that increasing indium supply is not an effective approach for enhancing indium incorporation and hence elongating emission wavelength under the current NR fabrication conditions.

Table 2. Average QW widths and average indium contents of the top-face QWs, the sidewall QWs at the top, middle height, and bottom of various samples. The numbers before and after a slash correspond the average QW width (in nm) and average indium content (in %), respectively.

View Table | View all tables in this article

 

Fig. 12 Average QW widths (the left ordinate) and average indium contents (the right ordinate) of the top-face (TF) QWs, the sidewall QWs near the top (SW-t), near the middle height (SW-m), and near the bottom (SW-b) as functions of TMIn flow rate for samples B-E when the QW growth temperature and time are fixed at 700 °C and 50 sec, respectively.

Download Full Size | PDF

Figure 13 shows the average QW widths (the left ordinate) and the average indium contents (the right ordinate) of the top-face QWs (TF) and the sidewall QWs near the top (SW-t), middle height (SW-m), and bottom (SW-b) as functions of QW growth temperature (for samples D, F, and G) when the TMIn flow rate and QW growth time are fixed at 360 sccm and 50 sec, respectively. Here, one can see that with increasing QW growth temperature, the QW width is decreased. This trend is caused by the weaker indium out-diffusion when indium incorporation becomes lower in increasing QW growth temperature such that the FWHM of the cross-well indium profile is decreased. Also, the indium content of the top-face QWs is more sensitive to the variation of QW growth temperature, when compared with the sidewall QWs. Figure 14 shows the average QW widths (the left ordinate) and the average indium contents (the right ordinate) of the top-face QWs (TF) and the sidewall QWs near the top (SW-t), middle height (SW-m), and bottom (SW-b) as functions of QW growth time (for samples G-I) when the TMIn flow rate and QW growth temperature are fixed at 360 sccm and 670 °C, respectively. Here, the variation trend of the average QW width is consistent with that of the QW emission wavelength since a larger QW width generally leads to a longer emission wavelength. Also, with increasing QW growth time, the average indium content is increased. This variation trend is due to the larger increase range of total indium content, when compared with that of QW width, in increasing QW growth time.

 

Fig. 13 Average QW widths (the left ordinate) and average indium contents (the right ordinate) of the top-face (TF) QWs, the sidewall QWs near the top (SW-t), near the middle height (SW-m), and near the bottom (SW-b) as functions of QW growth temperature for samples D, F, and G when the TMIn flow rate and QW growth time are fixed at 360 sccm and 50 sec, respectively.

Download Full Size | PDF

 

Fig. 14 Average QW widths (the left ordinate) and average indium contents (the right ordinate) of the top-face (TF) QWs, the sidewall QWs near the top (SW-t), near the middle height (SW-m), and near the bottom (SW-b) as functions of QW growth time for samples G-I when the TMIn flow rate and QW growth temperature are fixed at 360 sccm and 670 °C, respectively.

Download Full Size | PDF

7. Discussions

As shown in Table 1, although the growth conditions for the planar sample (sample A) are the same as those for sample B, in the planar structure, the QW emission wavelength in either PL or CL measurement is significantly longer than that in the NR structure of sample B. From Table 2, one can see that the average QW width of sample A is significantly smaller than that of the top-face QWs in sample B although it is significantly larger than those of the sidewall QWs. Also, the average indium content of sample A is higher than that of the top-face QWs in sample B, but smaller than those of the sidewall QWs in sample B. Compared with the top-face QWs in sample B, the higher indium content and smaller QW width in sample A result in the significantly longer emission wavelength. It is noted that the multiplication result of the average indium content and QW width corresponds to the total indium content per unit area in the QW plane. The multiplication results of samples A and B (for the top-face QWs) are 38.1 and 53.0 nm %, respectively. The larger total indium content in the top-face QWs of sample B implies that the top-face growth on an NR can adsorb the constituent atoms over a planar area larger than the projected area, when compared with the planar growth of sample A. Also, it is noted that the top-face QWs are formed during a tapering growth process. In other words, the top-face area becomes smaller and smaller during this growth process. Therefore, the adsorption of the constituent atoms projected onto a larger top-face area are used to form a c-plane QW on a smaller top-face area since it is difficult to grow a QW on the slant {1-101} facet under our growth conditions. This behavior also explains the larger well and barrier thicknesses of the top-face QWs in NR samples. The larger total indium content in the top-face QWs of sample B, when compared with that in the QWs of sample A, can also be attributed to the higher indium incorporation in sample B due to strain relaxation in such a nanostructure. The growth of sidewall QWs in the NR samples is based on the supply of the constituent atoms, which fall into the gaps between NRs [21]. Part of those atoms is adsorbed onto the sidewalls for forming QWs directly; the rest can fall down to the bottom in the gap regions and then migrate upward along the NR sidewalls for sidewall QW growth. The multiplication result of the average indium content and QW width of the sidewall QWs in sample B ranges from 13.2 through 16.1 nm %, which are significantly smaller than that of sample A and that of the top-face QWs in sample B. This result is partly due to the large sidewall area to share a relatively small amount of supplied constituent atoms. Therefore, the well and barrier thicknesses of the sidewall QWs are smaller than those of the top-face QWs. The lower indium contents in the sidewall QWs can also be attributed to the dependence of indium incorporation on crystal orientation. Usually, indium incorporation in a c-plane QW is higher than that in an m-plane QW [36]. In the sidewall QWs, because of the narrow well layer, the average indium content becomes relatively higher. However, with the stronger quantum confinement in the sidewall QWs, the emission wavelength becomes shorter, when compared with the top-face QWs. The longer emission wavelengths of the top-face QWs, which have lower indium contents, when compared with the sidewall QWs, are attributed not only to their larger QW widths and hence weaker quantum confinements, but also to their stronger QCSEs or larger QW potential tilts, even though the QCSE is reduced due to strain relaxation in such a nanostructure.

The variation trends of QW emission wavelength from PL and CL measurements and those of average QW width and indium content shown in Figs. 7–9, 12–14, Tables 1 and 2 are consistent. In the eighth column of Table 1, except the case of high TMIn flow rate (420 sccm for sample E), the CL emission wavelength difference between the top-face and sidewall QWs (Δλ_c-m) generally decreases with increasing TMIn flow rate and decreasing QW growth temperature. In Fig. 12, one can see that the difference in indium content between the top-face and sidewall QWs generally increases with TMIn flow rate except sample E, leading to the decreasing Δλ_c-m. As shown in the ninth column of Table 1, the increase of indium supply leads to a larger variation range of emission wavelength (Δλ_m) from the sidewall QWs. This trend is essentially consistent with the change of the variation range in indium content on NR sidewall with TMIn flow rate. This trend indicates that as indium supply increases, a higher percentage of indium atoms are adsorbed onto the sidewalls before they can move to the bottom of the NR gaps. As also shown in the ninth column of Table 1, the variation range of the CL emission wavelength from the sidewall QWs (Δλ_m) becomes larger as QW growth temperature decreases. This trend is generally consistent with the larger variation ranges in both QW width and indium content of the sidewall QWs when QW growth temperature is decreased. This behavior indicates that the variation of indium incorporation on NR sidewalls becomes stronger as the general indium incorporation efficiency is increased.

As shown in Fig. 5, the bright range of the cross-sectional CL image at the individual spectral peak extends from top to bottom as QW growth temperature decreases. This trend can be attributed to the more uniform intensity distribution among the emissions from the different heights on the sidewall, which have slightly different spectral peak wavelengths, and the broader spectral widths of local CL emissions as QW growth temperature is decreased. As a result, the spectral peak of the overall sidewall QW emission is not dominated by that near the top as in the case of a higher QW growth temperature [see Figs. 5(a)-5(d)]. At a lower QW growth temperature, the CL emissions at different heights on the sidewall can essentially equally contribute to the emission at the spectral peak of the overall sidewall emission [see Figs. 5(e)-5(h)]. The other attribution for the more uniform emission intensity on the sidewall under the condition of a lower QW growth temperature is that more constituent atoms stay near the bottom of the NR gaps for growing higher-quality QWs in this region on the sidewall.

As shown in Table 1, the CL spectral peak of the sidewall QWs in sample I can reach 500 nm. Because the CL spectrum of a sample is usually blue-shifted from the corresponding PL spectrum, the PL spectral peak of the sidewall QWs in sample I is expected to be longer than 500 nm. This emission wavelength is longer than those from similar NR QW structures (core-shell QW structures in regularly patterned NR arrays grown with MOCVD) reported in literature [21, 23, 27, 35, 37–39]. The reported shorter-wavelength emissions can be due to their relatively higher QW growth temperatures. In one of the previous accomplishments, by reducing the hole diameter on the patterned mask to 143 nm, the m-plane QW can emit a CL signal of 498 nm in spectral peak [27]. In this work, it was shown that the CL spectral peak of the m-plane QW blue-shifted from 498 to 432 nm when the patterned hole diameter was increased from 143 to 236 nm. However, in another previous accomplishment, it was shown that when the patterned hole diameter increased from 400 nm to 2 μm, the CL spectral peak of the m-plane QW was red-shifted from 400 to 420 nm [35]. Therefore, the dependence of m-plane QW emission spectrum on the cross-sectional size of NR is still inconclusive and deserves further investigation. Such an investigation can shed some light on the effect of strain relaxation on indium incorporation in growing a sidewall QW.

To increase the indium content in a sidewall QW for further elongating emission wavelength, the increase of the constituent atom supply in an NR gap region can be an effective approach. In this approach, the pitch of the NR array needs to be increased. However, in this situation, the total sidewall area for emission in a unit planar area is reduced unless NR height is increased. Nevertheless, with a larger NR height, the device fabrication of high yield becomes more difficult. From the insert of Fig. 15, which is a plan-view schematic demonstration of NR arrangement, one can evaluate the gap volume between the three neighboring NRs (the central pink area) to give 3^1/2(d² - a²)h/4 [21]. Here, a is the NR cross-sectional size, d is the NR pitch, and h is the NR height. Also, the total NR sidewall area of the six half-sidewall-faces (contacting the pink area) is 3^1/2ah. Meanwhile, the planar area of a unit cell is 3^1/2d²/4. For understanding the relation between the atom supply level for sidewall QW growth, which is proportional to the gap volume and increases with NR pitch, and the emission area enhancement (from the sidewalls) in an NR structure, which decreases with NR pitch, we define a parameter of atom supply level as the ratio of the gap volume over the total sidewall area sharing the atoms in the gap region for sidewall growth to give (d² - a²)/4a. For more comprehensive discussions, this parameter is further normalized by the sidewall length, a/3^1/2, to give the dimensionless parameter s. For a given a, this parameter is linearly proportional to the atom supply rate to the gap region. Also, by ignoring the emission from the top-face QWs, we define the enhancement ratio of emission area, t, as the ratio of the total sidewall area over the planar area in a unit cell to give t = 4ah/d². Figure 15 shows the dependencies of s (the left ordinate) and t (the right ordinate) on d/a with three h/a values at 2.35, 3.53, and 4.7 for t. In the current experimental study, a = 425 nm, d = 700 nm, and h = 1000 nm. In this situation, d/a is ~1.65, as marked by the vertical dashed line in Fig. 15, and h/a is ~2.35. The operation points for the experimental data in this paper are marked with the two “x” symbols, giving s = 0.74 and t = 3.46. If we intend to double the supplied constituent atoms to the gap region (s = 1.48) for increasing the sidewall QW width and indium content such that the emission wavelength can be elongated, the NR pitch needs to be increased to ~888 nm (d/a = 2.09). In this situation, if the NR height is kept at ~1000 nm (h/a = ~2.35), t is reduced to 2.2. This t value is still large enough to show the advantage of increasing emission area in using the NR structure. It is noted that although not all the supplied atoms in the gap region are adsorbed by the NR sidewalls for QW growth, the discussions above can still provide us with a guideline for designing the parameters of an NR LED array.

 

Fig. 15 Dependencies of parameter s (the left ordinate) and parameter t (the right ordinate) on d/a with three h/a values at 2.35, 3.53, and 4.7 for t. In the current experimental study, d/a is ~1.65, as marked by the vertical dashed line. The operation points under the current experimental conditions are marked with the two “x” symbols.

Download Full Size | PDF

As mentioned in section 5 of this paper, the red-shift slope of emission wavelength ranges from −0.9 to −1.3 nm/°C when QW growth temperature varies between 670 and 700 °C. This slope magnitude is significantly smaller than that in the planar structure. Generally speaking, in varying the emission wavelength of a planar QW, the corresponding slope is −1.5 to −2 nm/°C in the blue range and −2 to −2.5 nm/°C in the green range. Regarding the dependence of QW width on QW growth time, the QW width is almost linearly proportional to growth time in planar growth. However, from Fig. 14, one can see that the sidewall QW width in the NR structure is sub-linearly proportional to growth time. It is noted that under the growth conditions for samples A and B, the planar and NR growths lead to quite different QW emission wavelengths (up to 100 nm in wavelength difference). In other words, the QW growth scenarios on NR sidewalls are quite different from those in the planar structure. Therefore, their comparison does not much help us in the development of an NR LED array.

8. Conclusions

In summary, we have grown regularly patterned InGaN/GaN QW NR arrays under various growth conditions of indium supply rate, QW growth temperature, and QW growth time for comparing their emission wavelength variations of the top-face c-plane and sidewall m-plane QWs based on PL and CL measurements. Although the variation trends of QW emission wavelength by changing those growth conditions in the NR structure were similar to those in the planar structure, the emission wavelength range of the QWs on an NR was significantly shorter than that in a planar structure under the same growth conditions. Nevertheless, under proper growth conditions, an emission wavelength longer than 500 nm from sidewall non-polar QWs could be achieved. Under the growth conditions for a longer NR QW emission wavelength, the difference of emission wavelength between the top-face and sidewall QWs was smaller. Also, the variation range of the emission wavelength from the sidewall QWs over different heights on the sidewall became larger. On the other hand, SSA based on TEM was undertaken to calibrate the average QW widths and average indium contents in the two groups of QW of an NR. The variation trends of the calibrated QW widths and indium contents were consistent with those of the CL emission wavelengths from various portions of NR QWs.