Selective MOVPE of InGaN-based LED structures on non-planar Si (1 1 1) facets of patterned Si (1 0 0) substrates

1. Introduction

Light emitting diodes (LED) play a key role in modern energy-saving applications for general lighting, multimedia and automotive industry. Especially LED in the UV and blue to green color range can be fabricated by using group III-nitrides . These LED structures use active InGaN/GaN multiple quantum wells (MQW) and are mainly grown on sapphire in the commercial sector. Ongoing research more and more focuses on Si as an alternative substrate for III-nitride growth in order to reduce costs. The growth technique on Si (1 1 1) is quite challenging. In order to achieve high-quality crack-free epitaxial layers, the high lattice mismatch and large differences in thermal expansion coefficients need to be taken into account and be compensated, but different concepts have recently been demonstrated with good control of these factors . The first major concern, which has to be considered, is that GaN is not suited for direct epitaxial growth on Si, due to a strong chemical reaction of Ga and Si, also known as meltback etching . One is constrained to deposit AlN as a nucleation layer on Si to protect the surface before GaN growth is applied . Studies on the AlN/Si (1 1 1) interface show that hexagonal AlN can be grown on Si (1 1 1) and that the transition is very smooth . Due to the large lattice mismatch of about 19%, strain can be released by a specific atomic arrangement with a periodic array of misfit dislocations.

The major methodology for growth on planar Si is to control the stress in the layer stack, monitored by in-situ measurements of the curvature of the wafer during epitaxy. Different approaches of strain engineering can be found in literature in order to achieve a crack-free layer after cool-down. Most of them use initial AlGaN buffer layers with high Al contents and subsequent step-wise or gradual reduction of the Al content . Another way is the insertion of several low-temperature AlN interlayers in a GaN layer series with subsequently increasing GaN thickness after each interlayer . These AlN interlayers decouple the thick GaN layers, which then can induce compressive strain in order to counterbalance the tensile thermal stress. A more sophisticated approach is the growth on pre-patterned substrates. Here, mainly two methods have to be mentioned: The first one is a combination of masking the Si substrate and applying selective lateral overgrowth, which takes the advantage of strain release via a free semi-polar facet and reduced dislocation density due to lateral growth . Different Si substrate orientations, in which the Si (1 1 1) facet is provided by etching, were tested so far in order to establish a semipolar surface for the active LED. But the problem of either misaligned nitride crystals, stemming from different (1 1 1) facet orientations, and the challenge of defect formation at the point, at which nitride crystals coalesce, has not been solved yet. The second approach is selective area growth directly on the Si (1 1 1) plane, which uses mask-patterned separated openings in order to reduce the coherent area of the epitaxial layers . Direct growth on exactly orientedSi (1 0 0) substrates is challenging due to the difficulty to grow mono-crystalline III-nitrides on these substrates. The four-fold symmetry of Si (1 0 0) allows two preferred crystalline orientations of hexagonal GaN . At boundaries of the differently aligned islands, the coalescence is impeded. Advances has been achieved by using misoriented Si (1 0 0) substrates with offcuts in the range of a few degrees. On these substrates, one crystallite orientation has a higher growth rate. Finally, a coalesced GaN film with reasonable crystal quality for device operation can be achieved . Still, these layers exhibit higher dislocations densities than GaN on Si (1 1 1).An alternative method of growing GaN on the Si (1 0 0) plane was demonstrated by using a compliant silicon-on-insulator (SOI) substrate, in which strain between the epitaxial layer and the substrate is released by partial accommodation in the compliant overlay .

In this work, a technology is proposed which combines several approaches mentioned before. By providing Si (1 1 1) facets with quite long facet distances of ∼20 µm in patterned Si (1 0 0) substrates, LED structures can be directly fabricated on these facets. It is shown that c-plane AlN/AlGaN/GaN buffer layers and c-plane InGaN/GaN MQW can be grown selectively on these Si (1 1 1) facets by metal organic vapour phase epitaxy (MOVPE). This approach also reduces the required effort for strain-engineering during epitaxy and overcomes problems related to coalescence because the epitaxial layers grow only in the openings of the mask and are hence separated from each other, similar to selective area growth mentioned above.

By using two equivalent Si (1 1 1) facets in trenches, tilted by 54.7° against the Si (1 0 0) surface, the active emission area can be increased theoretically by a factor of 1.7 in comparison to the planar area. With a conservative implementation with spaces between trenches, still an active area enlargement compared to the planar area can be expected, which might be an economical benefit for saving resources in addition to increased available wafer sizes of Si (1 0 0).

2. Experimental details

As starting material, 500 µm thick 2-inch n-doped Si (1 0 0) wafers are used. The patterning is realized by applying anisotropic etching with aqueous potassium hydroxide (KOH) on masked wafers, as shown in Fig. 1. Mask material for etching is a 250 nm thick silicon nitride (SiN) layer (Fig. 1a). The naturally appearing planes on the Si (1 0 0) wafer are the Si {1 1 1} facets, due to an anisotropic etch rate in [1 0 0] and equivalent [1 1 1] directions (Fig. 1b) . Epitaxial growth takes place on the Si {1 1 1} facets, tilted by 54.7° (Fig. 1c). A schematic of the layer stack of the investigated LED structure is shown in Fig. 1d). All heterostructures were grown by metal organic vapor phase epitaxy (MOVPE) in an AIXTRON reactor with standard precursors trimethylaluminum (TMAl), trimethylindium (TMIn), trimethylgallium (TMGa), triethylgallium (TEGa) and ammonia. Hydrogen (H₂) is used as carrier gas, except for MQW growth, in which nitrogen (N₂) is employed. Full spectroscopic in-situ measurements of the reflectance at different wavelengths between 276 nm and 775 nm by a measurement tool from LayTec with additional true-temperature pyrometer module enables monitoring and controlling growth surface temperature and film formation during epitaxy.

Fig. 1. a) Si (1 0 0) wafer with SiN mask on top, b) Trenches are etched by KOH with SiN mask, c) MOVPE growth on Si (1 1 1) facet, d) LED layer stack used here.

MOVPE growth is initiated with a desorption step under H₂ to clean the surface, followed by a 10 nm AlN nucleation layer grown at 1000 °C and 100 nm high-temperature AlN, grown at 1130 °C. To reduce the cracking tendency, strain is controlled by a 180 nm thick AlGaN layer . The Al content is gradually reduced from 50% to 20% by adjusting a temperature gradient from 1210 °C to 960 °C during growth. A thin AlN interlayer grown at 900 °C in between decouples these buffer layers . The whole buffer is doped with Si in order to provide a conductivity through these layers for a provisional backside contact via the Si substrate. On top of this buffer structure, a conventional GaN-based diode is grown at standard growth parameters with 1 µm Si-doped n-GaN and 100 nm Mg-doped p-GaN. In between, an InGaN/GaN MQW is used as active recombination zone.

The SiN mask layout, which is used for trench etching and MOVPE growth, is shown in Fig. 2. The pattern was divided into 142 fields of 3 mm×3 mm. Each field contains 110 µm spacing at every side, 20 bond pads and 652 opening trenches, in which the Si (1 1 1) facet appears while etching. Each trench has a 20 µm×260 µm opening size and a 20 µm wide separating bar to the next trench. This results in 41% planar wafer area, which is covered with trenches, and 59% wafer area, covered with SiN on the Si (1 0 0) surface.

Fig. 2. a) 2” Si (1 0 0) wafer with 146 times a 3 mm·3 mm pattern (red rectangles) and etch alignment fields (blue rectangles), b) Each pattern field contains 692 trenches with 20 µm spacing, c) SEM image of the 20 µm×260 µm trenches in birds-eye view.

Structural characterization is performed by X-ray diffraction (XRD), scanning electron microscopy (SEM) and transmission electron microscopy (TEM). Optoelectronic properties are determined by electroluminescence measurements (EL). For quick functional testing, a provisional concept is used at first instead of a full LED process. A semi-transparent 5 nm Ni/20 nm Au contact is deposited on the whole surface of a sample, constituting the p-contact on the Mg-doped GaN. Here a conformal deposition of Ni/Au on both the SiN and the Si (1 1 1) facets could be achieved by using an angle of 30° and rotation during vacuum metal evaporation. In order to realize the n-backside-contact, the Si (1 0 0) wafer was contacted with a paste of Ag at a cleaved edge of the wafer. By using a silane doped buffer structure with moderate doping levels, a limited n-conductivity through the Si (1 0 0) and the AlN/AlGaN to the n-GaN could be established. The SiN mask works as a vertical current aperture in this design.

3. Results and discussion

3.1. Silane doping of the AlN/AlGaN buffer

The first experimental series is represented by four samples with different Si doping levels in the buffer structure. As shown in Table 1, the silane molar flow is reduced from 49.3 nmol/min to 4.5 nmol/min from sample A to C. These molar flow values are applied through the whole buffer structure. An additional sample D was grown, in which the doping level was adjusted separately in each layer of the buffer, starting with no doping in AlN, then a doping level in AlGaN similar to sample C and doping in GaN slightly higher than for sample C. A benchmark for crystal quality is provided by measurements of the GaN rocking curves.In order to detect the signal from layers on the Si (1 1 1) facet, a semi-skew geometry was used. Samples were tilted in χby the inclination angle of the Si (1 1 1) facet of 54.7°. By adjusting the sample that way, four peaks arise in ϕscans of 360° belonging to the four groove side facets. The two strongest peaks belong to the long side facets. With this geometry, symmetrical (0 0 0 2) and asymmetrical (10–12) ω-scans can be measured (0 0 0 2). In Fig. 3, the full width at half maximum (FWHM) values of the (0 0 0 2) and (10–12) ω-scans are plotted as a function of the silane molar flow in the AlN layer. Obvious in this plot are the very high FWHM values for samples A and B, which both were grown at very high doping levels. A reduction of the molar flow to about 10% of the doping of sample A for sample C leads to significantly reduced FWHM values of 1141 arcsec for (0 0 0 2) and 1461 arcsec for (10–12). These values are in a reasonable range for growth on Si (1 1 1) facets, but still about two times higher than state-of-the-art values for GaN on planar Si (1 1 1). Sample D exhibits slightly increased FWHM, which can be explained by the higher doping level in the GaN layer for this sample. The much improved crystal quality at lower silane doping levels is also shown by SEM images, which are displayed in Fig. 4. Huge hexagonal pits, forming inverted pyramids, can be observed on the surface of both samples A and B. It seems that these V-pits appear very early during growth in the AlN buffer layer, which explains the large size of those pits propagating through the whole layer stack. With lowered Si doping for sample C, the V-pit size is strongly reduced. It is difficult to quantify the V-pit density, which might be still in the same order as for sample A and B. The last sample D without doping in AlN shows a clear reduction of the V-pit density, revealing that V-pit formation is most likely starting in the AlN layer due to the silane doping. It has been reported that hexagonal V-pits with (1-101) sidewalls in AlGaN are associated with threading dislocations at their bottoms. Further, in compressively strained AlGaN, as it is the case in our structure, increased stress relaxation is directly correlated to higher Si doping levels . The reason has been found in an inclination of threading dislocations, which we assume might act as a nucleation point for V-pit formation in our structures. However, this might not explain the dependence of V-pit formation in tensile strained AlN on the Si doping level we observe. Hence, we further consider that Si has the tendency to attach to dangling bonds at dislocations leading to SiN formation and dislocation bending. SiN leads to a local pinning of screw type dislocations, which might be the origin of crystal deformation. Theoretical simulations of pit formation at highly tensile strained SiN/Si(1 1 1) show that due to the elastic deformation, the vertical displacement of the atoms is most pronounced under the amouphous SiN in proximity to a dislocation, because of weaker bonds . This model might be transferred to our local tensile strained SiN/AlN interface at a dislocation. Finally, this compressive deformation in c-direction can nucleate a pit in order to release in-plane tensile strain in the AlN.

Table 1.
Silane molar flow levels in the buffer structure for samples A, B, C and D.
Layer	A Silane molar flow [nm/min]	B Silane molar flow [nm/min]	C Silane molar flow [nm/min]	D Silane molar flow [nm/min]
GaN	49.3	14.8	4.5	8.9
AlGaN	49.3	14.8	4.5	4.5
AlN	49.3	14.8	4.5	–

Fig. 3. FWHM values of GaN in (0 0 0 2) and (10–12) rocking curve measurements, taken with an open detector.

Fig. 4. Characteristic top-view SEM images of samples A, B, C and D with different silane doping levels in the buffer structure.

Due to a provisional backside contacting for electroluminescence measurements shown later in this work and hence requiring reasonable conductivity of the AlN, further samples are grown with conditions of sample C making the compromise of the slightly higher pit density.

3.2. SiN mask and growth behaviour at the Si (1 0 0)/Si (1 1 1) vertex

For a better understanding of the growth behaviour at the Si (1 0 0)/Si (1 1 1) vertex, two samples are investigated via TEM images, one sample with SiN mask on the Si (1 0 0) surface and one sample, for which the SiN mask was removed before MOVPE. TEM images of the latter sample without SiN are shown in Fig. 5. Regarding the growth on the Si (1 1 1) facet, visible in the full view on the vertex in inset i), there are two regions of interest: First, the layer growth in c-direction seems to be similar to that on planar Si (1 1 1) showing distinct darker stripes along the [0 0 0 1] growth direction, which can be interpreted as threading dislocations (TD) typically created at interfaces in GaN heteroepitaxy due to relaxation mechanisms . The second interesting region is a dislocation-free epitaxial layer near the facet vertex, which results from epitaxial lateral overgrowth (ELOG) . Generally, a strong overgrowth out of the facet is observed on this sample without SiN mask due to a strong tendency of ELOG. The naturally appearing semipolar plane in the upper ELOG region is identified to be the (1-101) plane with an inclination angle of 61.96° towards the (1-100) plane (m-plane). The nonpolar m-plane perpendicular to the c-plane also arises in that overgrown region, as obvious in inset i).

Fig. 5. TEM image of the Si (1 0 0)/(1 1 1) vertex of a sample with removed SiN before MOVPE. A full view on the vertex is shown in (i). Inset (ii) is a zoomed TEM image with a closer view on the vertex. The MQW also grows on the semi-polar (1-101) plane (Sub-region 1), which is also displayed in zoom in inset (iii). In (iv), a region is shown illustrating MQW thickness inhomogeneities in c-direction (Sub-region 2).

Looking at a zoom TEM image in inset ii), it is clearly visible that MQW grow on all planes, the c-plane, m-plane and semipolar plane. MQW growth on m-plane is not further discussed, as on the other samples discussed later, the m-plane does not emerge. On the (1-101) plane, a 5-fold MQW can be clearly observed in TEM images of the sub-region 1 shown in inset iii). It seems that in semipolar growth direction, the wells appear more and more blurry. Further, there are clear dark stripes in the p-GaN above the MQW, which might be identified as threading dislocations. They arise in the InGaN wells and propagate orthogonally to the c-direction, hence in nonpolar direction. One explanation for crystal quality degradation might be a higher In content in the semipolar MQW, due to enhanced In incorporation efficiencies compared to c-plane MQW . More In leads to larger strain in the MQW, which would increase the probability of dislocation formation. This effect must be verified in further investigations.

In contrast to that, the MQW on the c-plane appear to be more defined, especially in the upper wells, as visible in the sub-region 2 in inset iv). But here, some areas of the MQW are disturbed by small V-shaped pits, which most likely stem from high strain in the MQW or TD . Another interesting observation is a kink in the MQW growth. Upwards from the abrubt thickness transition, the MQW suffer from a strongly reduced growth rate.

In order to achieve a more homogeneous film growth at the Si (1 0 0)/Si (1 1 1) vertex, MOVPE growth was investigated on a substrate which was this time SiN masked. Hence, the SiN layer on top of the Si (1 0 0) surface, which is used for etch-patterning the substrates, was not removed before MOVPE. A sample with a 250 nm thick SiN layer is analysed by TEM, shown in Fig. 6. Growth conditions are similar to the aforementioned sample and are given in inset iii). On SiN, the AlN/AlGaN buffer layers are formed with quite good coalescence, which can be explained by the lack of growth selectivity of Al-containing layers over commonly used dielectric masks and substrates . The growth on the Si (1 0 0) surface can also be detected by in-situ reflectivity measurements more or less pronounced depending on growth conditions. No peak can be found in XRD analysis for this layer. A completely different appearance is that of the subsequently grown GaN, which conserves its selectivity and cannot grow in a two-dimensional film on the amorphous AlN/AlGaN on SiN-masked Si (1 0 0) surface. Here crystallites with µm-dimensions can be observed on the surface, also highlighted in the SEM image in Fig. 6, inset i). The poly-crystalline GaN might be etched away in following device processing steps.

Fig. 6. TEM image of the Si (1 0 0)/(1 1 1) vertex of a sample with SiN masking with overhang during MOVPE. Inset (i) shows a SEM image of the GaN crystallites on the SiN surface. Inset (ii) is a TEM image with a closer view on the vertex. The MQW also grows on the semi-polar (1-101) plane. In (iii), the growth parameters of this sample in the GaN buffer and the MQW are given.

The layer growth on the Si (1 1 1) facets features quite similar issues like the sample discussed before: First the c-plane growth with TD in the buffer structure and second the defect-free ELOG region. But some structural details of this sample appear to be quite different. Worth mentioning is the SiN overhang of about 0.5 µm, which originates from the wet-etching step. Under the SiN overhang, the buffer layers experience a reduction in growth rate, most likely due to a lack of local precursor supply under the overhang. GaN growth in [0 0 0 1] direction generates dislocations at that interface propagating in [0 0 0 1]-direction. But due to a dominating ELOG, these dislocations annihilate before the layer is reaching the edge of the overhang. Interesting to see and a real success of the SiN masking with an overhang is a suppression of strong parasitic growth over the facet, if compared to the sample without overhang in Fig. 5. Further, a uniform thickness of MQW on the c-plane all the way down the facet is observed, which results in a very well defined edge of the (0 0 0 1) and (1-101) planes. A parasitic effect here is the MQW growth on the GaN crystallites with divot orientations on the SiN-masked Si (1 0 0) surface, visible in the very left of inset ii).

3.3. MQW growth homogeneity

The thickness homogeneity of an LED structure and especially of the MQW plays a major role for a stable emission wavelength and intensity. As shown in Fig. 7a), the LED structure on the Si (1 1 1) facet exhibits an inhomogeneous thickness down along the facet. While at the top of the facet, the total thickness is about 1.85 µm, the thickness is reduced about four times to only 0.45 µm at the bottom of the facet. This effect might be explained by a weaker supply of precursor material in the gas phase way down the facet. Growth rate variations in trenches for non-planar MOVPE have been reported for III-V compounds before . Due to a limited gas phase diffusion, the precursor concentration is perturbed above a trench, illustrated inFig. 7b). It is obvious from this schematic that the distance Δd between different constant concentrations increases with the trench depth. Hence, in the case of a V-shaped groove, the precursor concentration gradient changes along the facets, which results in a non-uniform precursor supply. Further studies on epitaxial growth parameters, e.g. different pressure and surface temperature regimes, which might improve the precursor supply, are ongoing. Also, by reducing the trench size, this effect might be mitigated.

Fig. 7. Illustration of the non-uniformity of layer deposition: a) Side view of a trench, deposited with the full LED structure, b) Precursor concentration contour lines in the gas phase above a trench, c) TEM image of the MQW at the top of the facet, d) TEM image at 85% distance way down along the facet.

A similar behaviour can be found for the MQW thickness along the facet. Fig. 7c) and d) illustrate the MQW at the top and the MQW at 85% down at the facet. The total thickness starts with about 130 nm and is halved at 50% of the facet distance. At the bottom, MQW growth cannot be observed.The lower thickness of the wells in the active zone leads to an emission color shift along the facet, which was observed on samples with larger trenches and might be an issue on these samples. Thinner wells down the facet result in a blue-shifted emission. A change in the In content along the facet might also contribute to this effect, but this is still under investigation.

3.4. Electroluminescence from LED on Si (1 1 1) facets

EL measurements were carried out on two samples, both grown on SiN masked substrates, which reveal bright electroluminescence from the LED structure on the Si (1 1 1) facets, as shown in Fig. 8. One sample was grown with MQW conditions of 768 °C surface temperature and 400 hPa pressure, further called high-temperature (HT) sample and another sample, grown at lower temperature of 722 °C and lower pressure of 150 hPa, further called low-temperature (LT) sample. Bright blue electroluminescence was observed for the HT-sample, which corresponds to a wavelength of about 480 nm. If we estimate the In content from the bandgap, taking formulas from literature with a bowing parameter of b_InGaN=1.65 eV into account and neglecting quantization effects, this results in about 14% In in the InGaN films . Due to the non-uniform well thickness on the facet, HRXRD measurements deliver insufficient data, hence a precise determination of the In content at that point is not possible. Further analysis of the In content and the uniformity on the facet will be performed. The highest intensity of light emission is observed near the n-contact, most likely due to a non-optimal current distribution in the Si (1 0 0) wafer. Hence, the current density is higher near the n-contact, which results in a brighter emission, clearly visible in Fig. 8.

Fig. 8. Photos, showing electroluminescence of LED on the Si (1 1 1) facet. Growth conditions of the MQW are given in the insets of the images. Photos at the right side were taken through the optical microscope.

The LT-sample shows different emission characteristics. At first, the highest intensity is observed near the n-contact with a color shifted to green likely due to a higher In content compared to the HT sample. A corresponding In content of above 20% can roughly be estimated from the emission wavelength. This is in agreement with the lower growth surface temperature of only 722 °C, which results in an increased incorporation of In in nitride layers . Further, with increasing distance to the n-contact and therefore lower current density, the emission intensity decreases as for the HT-sample, but also the color is red-shifted for the LT-sample. This trend can be partly explained by a strong tilt in the bandstructure in the MQW, due to high internal polarization-induced electric fields . In the regions with higher current densities, this band tilt is reduced because of a screening of the electric field, also known as the quantum confined Stark effect. The polarization-induced red-shift cannot be the only reason for such a strong shift to above 600 nm wavelength, which would correspond to a bandgap belonging to InGaN with an average In content of about 35%.

Optical microscope investigation of the emissive area, shown in the right side of Fig. 8, reveals that light emission stems only from the facets and not from the SiN-masked Si (1 0 0) surface. However, not every facet emits light, which let us conclude that leakage paths and non-radiative recombination are major concerns in these provisional devices. A more sophisticated device structure has to be established and is in process. In contrast to the HT-sample, which shows good color uniformity from facet to facet, a variation of the emission color is observed for the LT-sample for different facets. With an enhancement in luminescence intensity, the emission wavelength is strongly blue-shifted, which is in agreement with the above mentioned observation. The light emission at longer wavelengths from the LT-sample is likely caused by a non-uniform In content in the well. It was often reported that the luminescence spectra of LED in the green or even red color range is related to the nanostructure of the alloy . The appearance of composition fluctuations with a locally smaller bandgap on the nanoscale leads to improved carrier confinement in band tail states, which dominate the luminescence spectra. It was shown that In-rich nanoclusters also arise in quantum wells and are able to act as local traps for photon-emitting carriers of especially green LED . In clusters with 1–3 nm lateral size with local In content up to 30–40% are observed in a layer with an average composition of 17–20% In. In a recent report, a monolithic two-color LED has been presented with clear degradation of optical properties of the MQW with higher In content . Hence, non-uniformities in InGaN layers are not necessarily correlated to immature technology, it is more a physical issue of critical stress, which can be accumulated in a layer, before relaxation occurs. At a critical thickness, compressively strained nitride layers relax, leading to clustering effects . Hence, the yellow and red colors in the LED presented here are likely caused by In clustering effects, which dominate the emission spectra at low drive currents.

The current-voltage characteristic is shown in Fig. 9. The forward voltages at 20 mA of both devices are in a reasonable range of 3.9 V for the LT-sample and 8.8 V for the HT-sample. The trend to a higher value for the HT-sample is in agreement with the higher bandgap of the InGaN wells, due to a lower In content achieved at higher growth temperatures. However, the threshold voltage difference between both devices is quite high and other parasitic effects stemming from the provisional contact method might play a role here. The high reverse currents are a further hint on leakage paths, either through the SiN, or through non-optimally coated facet bottoms. The electrical performance of the presented structures is likely to be improved by the utilization of appropriate processing techniques used in industry.

Fig. 9. Current-voltage characteristic of the LED on Si (1 1 1) facets. Reasonable forward voltages of 3.9 V for the LT-sample and 8.8 V for the HT-sample at 20 mA are observed. High reverse currents are likely caused by leakage paths through the SiN or defects in the layer structure, especially the bottom of the facet.

4. Conclusions

In this work, MOVPE InGaN/GaN LED growth on Si {1 1 1} facets patterned into Si (1 0 0) substrates and successful LED operation were demonstrated. It was shown that doping the AlN/AlGaN buffer is possible, but affects crystal quality. It was shown that by using a SiN mask with a short overhang, a defined edge at the Si (1 0 0)/Si (1 1 1) vertex with uniform MQW can be achieved and strong parasitic overgrowth of the facet is suppressed. The deposition still suffers from an insufficient precursor supply, which leads to a non-uniform thickness on the facet.

MQW growth is observed on different planes of GaN, especially on c-plane and the semipolar (1-101) plane. An inferior crystal quality is detected on the semipolar (1-101) plane, but by optimizing growth conditions this degradation might be improvable.

Electroluminescence in the blue color range was observed for c-plane MQW grown at a surface temperature of 768 °C resulting in an estimated In content of 14%. A lowered MQW growth surface temperature results in a wavelength shift to green color, due to a higher In content. For low current densities, the emission is shifted to yellow and red, most likely caused by local strain inhomogeneities and In clusters on the nanoscale, which lead to improved carrier confinement in band tail states, dominating the electroluminescence spectra. These preliminary luminescence results are encouraging but the process, however, has to be optimized to improve the LED performance. By using the by 54.7° tilted facets in Si (1 0 0) wafers for LED fabrication, the light emissive area can be theoretically increased by a factor of 1.7. By considering the loss in wafer area due to the spacing between trenches, the active wafer area can easily match or even exceed that achieved on a planar design. Further, these results present a first step towards integration of nitride devices in Si (1 0 0) technology, e.g. a possible implementation of LED on patterned Si (1 0 0) for chip-to-chip communication.

Source:Journal of Crystal Growth

If you need more information about Selective MOVPE of InGaN-based LED structures on non-planar Si (1 1 1) facets of patterned Si (1 0 0) substrates, please visit our website:http://www.qualitymaterial.net, send us email at powerwaymaterial@gmail.com.