Optically Pumped GaN-based Vertical Cavity Surface Emitting Lasers: Technology and Characteristics

Transcription

1 Japanese Journal of Applied Physics Vol. 46, No. 8B, 27, pp #27 The Japan Society of Applied Physics Review Paper Optically Pumped GaN-based Vertical Cavity Surface Emitting Lasers: Technology and Characteristics Shing-Chung WANG, Tien-Chang LU, Chih-Chiang KAO, Jong-Tang CHU, Gen-Sheng HUANG, Hao-Chung KUO, Shih-Wei CHEN, Tsung-Ting KAO, Jun-Rong CHEN, and Li-Fan LIN Department of Photonics and Institute of Electro-Optical Engineering, National Chiao Tung University, 11 Ta Hsueh Road, Hsinch 31, Taiwan (Received February 19, 27; accepted May 1, 27; published online August 23, 27) We review the fabrication technology and performance characteristics of optically pumped GaN-based vertical cavity surface emitting lasers (VCSELs). Two types of VCSELs with different microcavity structures are described. First type of VCSEL has a hybrid microcavity structure that consists of an epitaxially grown AlN/GaN distributed Bragg reflector (DBR), a GaN active layer with InGaN/GaN multiple quantum wells (MQWs), and a Ta 2 O 5 /SiO 2 dielectric DBR. Second type of VCSEL has a dielectric DBR microcavity structure that has a similar InGaN/GaN MQWs active layer sandwiched in two dielectric DBRs formed by Ta 2 O 5 /SiO 2 and TiO 2 /SiO 2. Both types of VCSELs achieved laser action under optical pumping at room temperature with emission wavelength of 448 and 414 nm for hybrid DBR VCSEL and dielectric DBR VCSEL, respectively. Both lasers showed narrow emission linewidth with high degree of polarization and large spontaneous emission coupling factors of about 1 2. In addition, a high characteristic temperature of over 24 K was measured, and a distinct spatially inhomogeneous emission pattern was observed. [DOI: /JJAP ] KEYWORDS: GaN, vertical cavity surface emitting laser, DBR, laser lift-off 1. Introduction GaN-based wide band gap optoelectronic devices such as light emitting devices and lasers have become important devices for practical applications such as display, storage, and illumination. Vertical cavity surface emitting laser (VCSEL) structure, in particular with a small optical mode volume can emit a single longitudinal mode with a symmetrically circular beam and a small beam divergence that are superior than the edge emitting lasers and desirable for many practical applications in high density optical storage, laser printing, etc. In addition to the practical applications, the strong cavity field in the VCSEL microcavity can facilitate the investigation of the cavity quantum electrodynamics effect such as single photon and polariton emission, controlled spontaneous emission, and low threshold or thresholdless lasing. 1 3) In the GaN-based VCSEL structure, a micro cavity with a few in the optical thickness and a pair of high reflectivity (above 99%) distributed Bragg reflectors (DBRs) are necessary for reducing the lasing threshold. The requirement of high reflectivity and high quality DBRs using Al x Ga 1 x N and GaN materials is quite formidable since these two materials have large lattice mismatch and difference in thermal expansion coefficients that tends to form cracks in the epitaxially grown DBR structure. These cracks in DBR could result in the reduction of reflectivity and increase in scattering loss. Recently, several groups have reported optically pumped GaN-based VCSELs mainly using three different kinds of vertical resonant cavity structure forms: (1) monolithic grown vertical resonant cavity consisting of epitaxially grown III nitride top and bottom DBRs (all epitaxial DBR VCSEL), (2) vertical resonant cavity consisting of dielectric top and bottom DBR (dielectric DBR VCSEL), (3) vertical resonant cavity consists of an epitaxially grown III nitride top DBR and a dielectric DBR (hybrid DBR VCSEL). In 1996, Redwing et al. address: scwang@mail.nctu.edu.tw 5397 proposed an all epitaxial DBR VCSEL structure consisting of a 1-mm GaN microcavity embedded by two epitaxially grown 3-pair Al :12 Ga :88 N/Al :4 Ga :6 N DBRs. 4) The reflectivity of top and bottom DBR is 93 and 84%, respectively. Although the reflectivity of the DBR was not very high, a stimulated emission with a wavelength of 363 nm was observed due to the thick gain layer (1-mm GaN). In 1998, Arakawa et al. grew an InGaN multiple quantum wells (MQWs) on 35-pair Al :34 Ga :66 N/GaN DBR and deposited a 6-pair SiO 2 /TiO 2 on the grown structure forming the hybrid DBR VCSEL structure. 5) The stimulated emission with 381 nm in wavelength and linewidth smaller than.1 nm were observed as the pumping power was above the threshold condition at 77 K. Cavity quality factor Q of the resonant cavity was estimated to be 165. Thereafter, in 1999, Song et al. demonstrated a dielectric DBR VCSEL structure consisting of InGaN MQWs and 1-pair HfO 2 / SiO 2 top and bottom DBR using laser left-off technology. 6) The reflectivity of top and bottom DBRs were 99.5 and 99.9%, respectively. The emission wavelength and linewidth of the resonant cavity tested at room temperature were 437 and.7 nm, respectively. Because of the high reflectivity DBRs, the cavity had a high Q factor of 6. In 25, Feltin et al. showed that the Al 1 y In y N/GaN material system could be well suited to the growth of vertical cavity structures 7) because Al 1 y In y N(y ¼ :17) was lattice-matched to GaN thus avoiding the subsequent appearance of additional structural degradation (new dislocations and/or cracks) while presenting a refractive index contrast around 7 8%, where the DBR requires over 4 pairs to reach a reflectivity of 99%. Their VCSEL structure consisted of In x Ga 1 x N/ GaN (x ¼ :15) MQWs and 28 (bottom)/23 (top) AlInN/ GaN DBRs and exhibited a Q factor of 8. However there is so far few detailed report of the performance characteristics of optically pumped GaN VCSELs. In this paper, we first describe the design considerations and fabrication technology of GaN-based VCSELs with two types of VCSELs with different microcavity structures for optical pumping. First type has a hybrid microcavity

2 structure that consists of an epitaxially grown AlN/GaN DBR, a GaN active layer with InGaN/GaN MQWs, and a Ta 2 O 5 /SiO 2 dielectric DBR. Here, in order to reduce the optical absorption of the dielectric material for pumping laser light, Ta 2 O 5 /SiO 2 DBR was chosen. Second type has a dielectric DBR microcavity structure that has a similar InGaN/GaN MQWs active layer sandwiched in two dielectric DBRs formed by Ta 2 O 5 /SiO 2 and TiO 2 /SiO 2. Then we describe and discuss the detailed laser performance characteristics of these two types of VCSELs under optical pumping at room temperature conditions. 2. Key Issues and Basic Design Considerations of GaN VCSEL Since the main requirements for the GaN VCSELs are high reflectivity DBRs and a micro-cavity with only a few in the optical thickness, the previously mentioned three types VCSEL structures have their pros and cons in terms of laser design and process difficulties. For the first type structure, all epitaxial DBR VCSELs, the main advantage is its monolithically grown process. However, the epitaxially grown DBRs are still difficult to fabricate. In addition, the relatively thick epitaxially grown DBRs could have problems of high dislocation density and cracks resulting from the mismatch of lattice constants and thermal expansion coefficients between the DBR materials. In addition, p-type conductive GaN-based DBR is still difficult to realize, thus limiting the future application in electric pumped VCSELs. For the second type structure, dielectric DBR VCSEL, the dielectric DBRs can have high reflectivity and are relatively easy to fabricate. Since the deposition of dielectric DBR is relatively simple in comparison to the epitaxially grown DBRs. In addition, the whole VCSEL structure could incorporate with co-planar contacts and a patterned dielectric DBR mesa for the future electrically pumped VCSELs. However, one of the dielectric DBR has to be deposited on one side of the GaN microcavity with the InGaN/GaN MQWs after the GaN thin films were detached from the sapphire substrate by the laser lift-off technique. 8) As a result, the VCSEL fabrication process would become complicated and the precise control of the GaN microcavity thickness would be difficult. Furthermore, the thickness of the GaN microcavity should keep as thick as possible to avoid the damage of the InGaN/GaN MQWs during the laser lift-off process. Such the thick cavity length could increase the threshold condition and reduce the microcavity effect. Finally, the third type VCSEL structure combining an epitaxially grown DBR and a dielectric type DBR compromises the advantages and disadvantages of the above two VCSEL structures. The hybrid DBR VCSEL structure can eliminate the complex process of laser lift-off while maintaining feasibility of coplanar contacts with dielectric DBR mesas for the future electrically pumped VCSEL applications. One of the critical requirements for the hybrid DBR VCSEL is to fabricate high reflectivity and crack-free epitaxially grown DBRs. Various attempts to grow crack free, highly reflective, III nitride DBRs has been reported using a better lattice matched composition of Al x Ga 1 x N/GaN DBR. 9) However due to the small refractive index contrast between Al x Ga 1 x N and GaN, 5 pairs of Al :2 Ga :8 N/GaN DBRs were needed to achieve a peak reflectivity of 99% at 38 nm Reflectance(%) n/n=2% AlN n/n=7.5% AlInN.1 n/n=6% AlGaN Number of Pairs Fig. 1. Reflectivity and penetration depth vs DBR pairs for three DBR combinations (AlN/GaN, n ¼ 2%; AlInN/GaN, n ¼ 7:5%, AlGaN/ GaN, n ¼ 6%) with different index contrasts. The black and blue curves are calculated reflectivity and penetration depth, respectively. The reflectivity and penetration depth curves are calculated by R DBR ¼ f½1 ðn 1 =n 2 Þ 2m Š=½1 þðn 1 =n 2 Þ 2m Šg 2 and L pen ¼½ðL 1 þ L 2 Þ=ð4rÞŠ tanhð2mrþ, where n 1 and n 2 are refractive indices for both layers in each DBR pair, L 1 and L 2 are thickness for both layers in each DBR pair, r is the reflectivity for one DBR pair, and m is the number of DBR pairs. as shown in Fig. 1. Another 4-pair Al :82 In :18 N/GaN DBRs was also reported to achieve a crack free surface with a peak reflectivity of 99.4% at 45 nm. 1) However, all these DBRs grown on sapphire would require a large number of pairs (4) to reach high reflectivity and also have longer penetration depth, which could result in higher possibility of absorption and higher threshold condition for VCSEL operation. As shown in Fig. 1, due to the relatively high contrast of the refractive index between GaN and AlN in comparison to the GaN/AlGaN DBRs, a small number of GaN/AlN DBR pairs are required to achieve a high reflectivity with a broad stopband width. However, the large lattice mismatch between GaN and AlN induces a lot of cracks in the DBR when the number of the epitaxial pairs increases. These cracks tend to grow into V-shaped grooves, which seriously affect the reflectivity of the DBR due to scattering, diffraction and absorption of the propagated light. Therefore, it is necessary to suppress crack generation to achieve a smooth surface and a high reflectivity. Ng et al. reported a molecular beam epitaxy grown GaN/AlN DBR with a reflectivity of 99%. 11) However, these samples exhibited extensive cracking surfaces. Many research groups have studied and reported the approaches of using GaN/AlGaN or AlN/AlGaN superlattice (SL) insertion layers to reduce the biaxial tensile strain and successfully suppress crack generation while growing high Al-contained structures ) We have reported the growth of the crackfree 2-pair GaN/AlN DBRs with insertion of three sets of 5.5 periods of GaN/AlN SL during the growth and the achievement of high reflectivity with wide stopband width. 17) We apply this technique to fabricate the crackfree, high reflectivity AlN/GaN DBR for the hybrid DBR VCSEL. Another consideration of the VCSEL design is the thickness and position of the InGaN/GaN MQWs inside the GaN microcavity. Typically, the cavity length of Penetration Depth(µm)

3 8 pairs Ta 2 O 5 /SiO 2 DBR 1. Light Field Index 2.4 p-gan 1 pairs In.2 Ga.8 N/GaN (2.5nm/7.5nm n-gan Light Field Intensity (arb. unit) Index (arb. unit) 29 pairs AlN/GaN DBR Position (µm) (a) 6 pairs TiO 2 /SiO 2 InGaN MQW 3. Index of refraction Squared electric field 8 pairs Ta 2 O 5 /SiO 2 GaN 24.5 λ Index of refraction Distance (um) Squared electric field (a.u.) (b) Fig. 2. (a) Layer structure and the simulated standing wave patterns inside the cavity for the hybrid DBR VCSEL structure. (b) Layer structure and the simulated standing wave patterns inside the cavity for the dielectric DBR VCSEL. VCSELs is on the order of few half operating wavelengths. In such a short cavity device, the electromagnetic waves would form standing wave patterns with nodes (electromagnetic wave intensity minima) and anti-nodes (electromagnetic wave intensity maxima) within the GaN microcavity. The location of the InGaN/GaN MQWs with respect to the anti-modes can significantly affect the coupling of laser mode with the cavity field. The proper alignment of the MQWs region with the anti-nodes of the cavity standing wave field patterns will enhance the coupling and reduce laser threshold condition. As a result, the precise layer thickness control in the VCSEL fabrication is important. We used ten pairs InGaN/GaN MQWs to form a 1=2 optical thickness to fully overlap with one standing wave pattern in order to have a more thickness tolerance during the fabrication and to have a higher longitudinal confinement factor with respect to the total cavity length. Figure 2(a) shows the layer structure and the simulated standing wave patterns inside the cavity for the hybrid DBR VCSEL structure. Since the bottom half cavity layers of the hybrid DBR VCSEL are epitaxially grown, the precise layer thickness can be controlled by the in-situ monitoring setup inside the reactor for the layer growth. The total thickness of the GaN microcavity is 5 in optical thickness. The n-type GaN layer is designed for the co-planar n-contact in the electrically pumped VCSEL structures. The p-type GaN layer with a 5=4 optical thickness is designed for the top p- contact. The relatively thinner thickness of the p-type GaN 5399 layer can reduce the free carrier absorption in the GaN microcavity. Figure 2(b) shows the layer structure and the simulated standing wave patterns inside the cavity for the dielectric DBR VCSEL structure. The layer thickness of the p-type GaN and the InGaN/GaN MQWs are the same with the hybrid DBR VCSEL. The thick n-type GaN layer can prevent the damage on the InGaN/GaN MQWs during the laser lift-off process. Since the total cavity thickness can not be precisely controlled after the laser lift-off process, the MQW region with 1=2 optical thickness can compensate the misalignment error between the anti-nodes of the standing wave pattern and the active region position. 3. Growth and Fabrication Process of GaN VCSEL 3.1 Hybrid DBR VCSEL The hybrid DBR VCSEL samples were first grown epitaxially by a low pressure metal organic chemical vapor deposition (MOCVD) system (EMCORE D75) on two-inch diameter (1)-oriented sapphire substrates using trimethylgallium and trimethylaluminum as group III source materials and ammonia as the group V source material. Then, the growth process was as follows. The substrate was thermally cleaned in hydrogen ambient for 5 min at 11 C, and then a 3-nm-thick GaN nucleation layer was grown at 5 C. The growth temperature was raised up to 11 C for the growth of 2-mm GaN buffer layer. Then the 29 pairs of AlN/GaN DBR with 6 AlN/GaN SL insertion layers were grown under the fixed chamber pressure of 1 Torr similar

4 SL 1 µm (1) GaN Fig. 3. Cross-sectional TEM image of 29 pairs of AlN/GaN DBR incorporated with six superlattice insertion layers. to the previous reported growth conditions. 17) In order to reduce the tensile strain between the AlN and GaN, we inserted one SL into each five DBR periods at first twenty pairs of DBR. Then the SL was inserted into each three DBR periods for the remaining nine pairs of DBR to reduce the tensile strain. The overall AlN/GaN DBRs has 29 pairs with 6 SL insertion layers. Figure 3 shows the TEM images of the DBR. The lighter layers represent AlN layers while the darker layers represent GaN layers. The arrows indicate the SL insertion positions. In Fig. 3, no cracks can be observed in the TEM image. Then the n-type GaN cladding layer and a ten pairs In :2 Ga :8 N/GaN (2.5/1 nm) MQWs and p-type GaN cladding layer were grown to form a 5 cavity in optical thickness for center wavelength of 46 nm. Figure 4 shows the in-situ monitoring signals for the half cavity. Since there are thermal expansion and change in refractive index for a material while the temperature is changed, we have taken the effect of temperature change into consideration during the epitaxial process. By estimation from the empirical rule, we can have an epitaxial layer with correct thickness at room temperature according to the in-situ monitoring signal at high temperature during epitaxial process. As a result, by fixing the monitor wavelength to be 46 nm, we can obtain precise layer thickness control following the reflectance signals during the MOCVD growth. Finally, an eight pairs of Ta 2 O 5 /SiO 2 dielectric mirror were deposited by electronic beam evaporation as the top DBR reflector to form the hybrid microcavity. The whole structure is schematically shown in Fig. 2(a). The hybrid DBR VCSEL samples with the completed layer structure were then sent for the optical characterization. The detailed laser properties will be given in the following sections. One test sample of 29 pairs of AlN/GaN DBR with 6 AlN/GaN SL insertion layers was also grown to measure the reflectivity using an n&k ultraviolet visible spectrometer with normal incidence at room temperature, and to check the surface morphology using atomic force microscope (AFM). Figure 5 shows an AFM image of the 29-pair AlN/GaN DBR test sample. The average roughness measured was about 3.5 nm over a 5 5 mm 2 surface area with crack-free surface morphology. Figure 6 shows the reflectivity spectrum of crack free DBR sample under near normal incidence. The peak reflectivity of 99.4% at 461 nm with flat-topped 54 Reflectivity (arb. unit) Reflectivity (arb. unit) nm GaN SL AlN/GaN DBR SL SL SL nm n-gan stopband of 21 nm was measured, which is in agreement with the designed wavelength. The flat-topped stopband indicates the high crystal quality of the DBR. SL Time (sec) (a) MQW SL p-gan Time (sec) Fig. 4. (a) In-situ normal reflectance measurement during the growth of the AlN/GaN DBR and GaN microcavity with InGaN/GaN MQWs by a fixed measurement wavelength of 45 nm. The AlN/GaN superlattices are inserted into AlN/GaN DBR at the time indicated as SL. (b) Enlarged part of reflectance signals during the growth of the GaN microcavity consisted of n-gan, InGaN/GaN MQWs and p-gan layers. Fig. 5. AFM top-view image of grown 29 pairs of AlN/GaN DBR with six superlattice insertion layers. (b)

5 1 1 Reflectivity (%) Intensity (a.u.) 16.k 8.k.61 nm Reflectivity (%) Fig. 6. Reflectivity spectrum of grown 29 pairs of AlN/GaN DBR with six superlattice insertion layers Fig. 7. Reflectivity and the PL spectra of microcavity at the room temperature. PR InGaN MQW GaN SiO 2 /TiO 2 DBR Sapphire Sapphire SiO 2 /TiO 2 DBR InGaN MQW SiO 2 /TiO 2 DBR SiO 2 /Ta 2 O 5 DBR Silica substrate Silica substrate Sapphire adhesive Fig. 8. Schematic process steps of the dielectric DBR VCSELs incorporating with two dielectric DBRs fabricated by the laser lift-off technique. Figure 7 shows the room temperature reflectivity spectrum of whole microcavity under near normal incidence. The peak reflective is about 97% with a large stopband of 7 nm resulted from the large refractive index contrast between Ta 2 O 5 and SiO 2 layers. Figure 7 also shows the PL emission spectrum of the microcavity at room temperature pumped by the He Cd laser at 325 nm for obtaining the quality factor Q of the fabricated microcavity. The cavity resonance mode at nm with a FWHM of.61 nm is clearly observed. The cavity mode dip is located at reflectivity curve corresponding to the emission peak. This indicates that the InGaN/GaN MQWs emission peak was well aligned with the hybrid microcavity. The cavity Q factor was estimated from the = to be about 76. This Q factor value agrees with the value calculated from the reflectivities of both AlN/GaN and Ta 2 O 5 /SiO 2 DBRs. 3.2 Dielectric DBR VCSEL The schematic fabrication steps for dielectric DBR VCSEL are shown in Fig. 8. First, the layers structure of the GaN-based cavity, grown on a (1)-oriented sapphire substrate by MOCVD is described as followed: a 3-nm 541 nucleation layer, a 4-mm GaN bulk layer, MQWs consisting of 1 periods of 5-nm GaN barriers and 3-nm In :1 Ga :9 N wells, and a 2-nm GaN cap layer. The peak emission wavelength of the MQWs for the as-grown sample was obtained to be 416 nm. Then, the dielectric DBR consisting of six pairs of SiO 2 and TiO 2 was evaporated on the top of GaN-based cavity. The stop band center of the DBR was tuned to 45 nm. The reflectivity of the SiO 2 /TiO 2 DBR at 414 nm is obtained to be 99.5%. Next, because the adhesion between the dielectric material and GaN surface is weaker than that between the adhesive and GaN surface, an array of disk-like SiO 2 /TiO 2 DBR mesas with the diameter of 6 mm was formed by standard photolithography and chemical wet etching process. The DBR mesas were formed in order to increase the adhesion area between the adhesive and GaN to prevent the epitaxial structure separating from the silica substrate during the laser lift-off process. The wafer was then mounted onto a silica substrate using the adhesive. A KrF excimer laser radiation at 248 nm was guided into the sample from back side of the sapphire to separate the sapphire from the epitaxial layers. 18) After the laser lift-off process, the sample was dipped into the H 2 SO 4 solution to

6 CCD 1. monochromator polarizer cryostat sample Intensity (a.u.) Pumping energy (nj/pulse) YbO 4 laser Fig. 1. Laser emission intensity as a function of the pumping energy at room temperature conditions for hybrid DBR VCSEL. Fig. 9. Schematic setup of the optical pumping experiment. remove the residual Ga on the exposed GaN buffer layers. In the next step, the sample was lapped and polished by diamond powders to smooth the GaN surface since the laser lift-off process left a roughened surface. The mean surface roughness of the polished GaN surface measured by the AFM is about 1 nm over a scanned area of 2 2 mm 2. However, to prevent the possible damage of the quality in MQWs during the lapping process, the 4-mm GaN bulk layer was preserved. Finally, the second DBR consisting of eight pairs of SiO 2 and Ta 2 O 5 was deposited on the top of the polished GaN surface. The reflectivity of the SiO 2 /Ta 2 O 5 DBR at 414 nm is 97%. The stop band center of the DBR was also tuned to 45 nm. The detailed layer structure of the VCSEL and the simulated electric field distribution in the cavity is shown in Fig. 2(b). The thickness of the whole epitaxial cavity was equivalent to the optical thickness of 24.5 emission wavelength. The optical thickness of the MQWs covered nearly half of the emission wavelength right between two adjacent nodes. Then, the laser emission characteristics of the completed dielectric DBR VCSELs were investigated by the optically pumped scheme. 4. Optical Pumping Setup Figure 9 shows the optical pumping scheme for our hybrid type and dielectric type VCSELs. As shown in Fig. 9, the emission spectrum of the GaN-based VCSEL structure was measured using a microscopy system (WITec, alpha snom) at room temperature. The optical pumping of the samples was performed using a frequency-tripled Nd:YVO nm pulsed laser with a pulse width of :5 ns at a repetition rate of 1 khz. The pumping laser beam with a spot size ranging from 3 to 6 mm was incident normal to the VCSEL sample surface. The light emission from the VCSEL sample was collected using an imaging optic into a spectrometer/charged coupled device (Jobin-Yvon Triax 32 Spectrometer) with a spectral resolution of :1 nm for spectral output measurement. The characteristic temperature of GaN VCSEL sample was measured using the same microscopic optical pumping system with the sample placed in a cryogenics controlled chamber. The temperature of the chamber can be controlled from room temperature of 3 K 542 Intensity (arb. unit) E th laser Fig. 11. The variation of laser emission spectrum with the increasing pumping energy for the hybrid DBR VCSEL. The laser emission wavelength is nm with a linewidth of about.17 nm. down to 6 K using liquid nitrogen. The chamber has a window for laser pumping and output monitoring during the experiment. One polarizer was placed before the entrance of the monochromator to investigate the degree of polarization of the laser emission. 5. Results and Discussion 5.1 Characteristics of hybrid DBR VCSEL The light emission intensity from GaN-based VCSEL as a function of the pumping energy is shown in Fig. 1. A distinct threshold characteristic was observed at the threshold pumping energy (E th ) of about 55 nj corresponding to an energy density of 7.8 mj/cm 2. Then the laser output increased linearly with the pumping energy beyond the threshold. A dominant laser emission line at nm appearing above the threshold pumping energy is shown in the Fig. 11. The lower threshold density of this hybrid DBR VCSEL in comparison to our previous report 19) should be due to the improvement of the reflectivity of the epitaxially grown AlN/GaN DBR inserted with SL layers. The laser emission spectral linewidth reduces as the pumping energy above the threshold energy and approaches.17 nm at the pumping energy of 1:5E th. In order to understand the spontaneous coupling factor of this cavity, we normalized the vertical scale of Fig. 1 and re-plotted it in a logarithm scale as shown in Fig. 12. According to the reference, 2) the

7 Normalized intensity (arb.unit) β~6x Pumping energy(nj/pulse) Fig. 12. Laser emission intensity versus pumping energy in logarithmic scale for hybrid DBR VCSEL. The difference between the heights of the emission intensities before and after the threshold corresponds roughly to the value of. The dash lines are guides for the eye. The value of the hybrid VCSEL estimated from the figure is about difference between the heights of the emission intensities before and after the threshold corresponds roughly to the value of. The value of our VCSEL estimated from Fig. 12 is about We also estimated the value based on the approximation equation used in ref. 21: F p ¼ 3 Q 4 2 V c =ð=nþ 3 ; and ¼ F p ; 1 þ F p Normalized Intensity (arb.unit) I max ~17.7I min experiment data fitting curve (cos 2 θ) Angle of Polarizer (degree) Fig. 13. The polarization characteristic of the laser emission at the pumping energy of 1:5E th for the hybrid DBR VCSEL. The solid dot shows the experiment data and the solid line is the fitting curve. ln(e th ) experiment data Linear fit of experiment data Temperature (K) where F p is the Purcell factor, Q is the cavity quality factor, is the laser wavelength, V c is the optical volume of laser emission, and n is the refractive index. Since the photoluminescence spectrum of our hybrid DBR VCSEL showed a narrow emission peak with full width at half maximum of.61 nm, cavity quality factor was estimated to be 76. The refractive index is 2.45 for the GaN cavity. For the estimation of the optical volume, we used the spot size of the laser emission image we reported previously 19) which was about 3 mm and the cavity length of about 9:5 with considering the penetration depth of the DBRs. By using these parameters, the Purcell factor of about 2:9 1 2 was obtained and we estimated the value to be about 2:8 1 2, which has the same order of magnitude as the above value estimated from Fig. 12. This value is three order of magnitude higher than that of the typical edge emitting semiconductor lasers (normally about 1 5 ) indicating the enhancement of the spontaneous emission into a lasing mode by the high quality factor microcavity effect in the VCSEL structure. Figure 13 shows the laser emission intensity as a function of the angle of the polarizer at the pumping energy of 1:5E th. The variation of the laser emission intensity with the angle of the polarizer shows nearly a cosine square variation. The degree of polarization (P) is defined as P ¼ I max I min I max þ I min ; 543 Fig. 14. The semi natural-logarithm threshold energy as a function of the operation temperature for the hybrid DBR VCSEL. The solid dot shows the experiment data and the solid line is the linear fit of the experiment data. where I max and I min are the maximum and minimum intensity of the nearly cosine square variation, respectively. The result shows that the laser beam has a degree of polarization of about 89%, suggesting a near linear polarization property of the laser emission. Figure 14 shows the semi natural-logarithm plots of the dependence of the threshold pumping energy (ln E th ) on the operation temperature (T). The threshold energy gradually increased as the operation temperature rose from 12 to 3 K. The relationship between the threshold energy and the operation temperature could be characterized by the equation E th ¼ E e T=T, where T is the characteristic temperature and E is a constant. Therefore, we obtain a high characteristic temperature of about 244 K by linear fitting the experiment data. This T value is close to the T value of 3 K for GaN-based VCSELs predicted by Honda et al., 22) and higher than the reported T of 17 K 23) or 235 K 24) for the GaN-based edge-emitting laser diode. The high T of our sample could be due to various factors. We measured the lasing wavelength variation with the temperature. The result showed a slight red shift of about 1.6 nm as the temperature rose from 12 to 3 K. Normally, the temperature dependent lasing wavelength of a VCSEL device is mainly

8 5 E=.25Eth Emission intensity (a.u.) Emission intensity (a.u.) 1.13Eth Eth.25Eth.25 nm Intensity (a.u.) Excitation energy (nj/pulse) Fig. 15. Laser emission intensity as a function of the pumping energy at room temperature conditions for the dielectric DBR VCSEL Fig. 16. Emission spectra from the dielectric DBR VCSEL at various pumping energy. The lasing emission wavelength is 414 nm with a linewidth of.25 nm. The inset shows the rescaled emission spectrum under pumping power of :25E th. determined by the refractive index change along with the temperature change. Siozade et al. reported that the refractive index change of GaN from 1 to 3 K at 43 nm can be estimated to be about 3.5 nm, 25) which is larger than our measurement result. The discrepancy could be due to the relative high carrier density pumped into the 1-pair In :2 Ga :8 N/GaN MQW region at higher temperature in our VCSEL that modifies the change of refractive index. Further investigation on the index change as a function of carrier density would be needed to quantify this change. On the other hand, the PL emission of the MQW was also measured and the result showed a red shift of about 2.9 nm over the same temperature range. These results suggest that the gain peak almost align with the cavity mode within this temperature variation range. Therefore the high characteristics temperature obtained from our GaN VCSEL sample could be partly attributed to the good alignment of the gain peak and cavity mode. Furthermore the 1-pair In :2 Ga :8 N/ GaN MQW structure could help to suppress the carrier leakage from the MQW active layers to the cladding layers. 5.2 Characteristics of dielectric DBR VCSEL Figure 15 shows the laser emission intensity from the dielectric DBR VCSEL as a function of the pumping energy at room temperature conditions. A clear evidence of threshold condition occurred at the pumping energy of E th ¼ 27 nj corresponding to an energy density of 21.5 mj/cm 2. The output laser intensity from the sample increased linearly with the pumping energy level beyond the threshold energy. The estimated carrier density at the threshold is in the order of 1 2 cm 3 assuming that the pumping light with the emission wavelength of 355 nm has experienced a 6% transmission through the SiO 2 /Ta 2 O 5 DBR layers and undergone a 98% absorption in the thick GaN layer. Furthermore, in order to simplify the calculation, the carrier was assumed to distribute uniformly over the 1 pairs MQWs. According to the report by Park, 26) the gain coefficient of InGaN at this carrier density level is about 1 4 cm 1. We estimated the threshold gain (g th ) value of our VCSEL using the equation g th 1=ðL a Þlnð1=R 1 R 2 Þ, where is the axial enhancement factor, L a is the total thickness of the InGaN MQWs, and R 1 and R 2 are the reflectivity of the dielectric DBRs. Since the active region 544 covers half of the emission wavelength, is unity. We obtained an estimated gain coefficient of about 1 4 cm 1 which is consistent with the above g th value estimated from the carrier density. This also proved that the quality of the MQWs had been kept after the laser lift-off and lapping process. Figure 16 shows the evolution of the VCSEL emission spectrum with the pumping energy at room temperature. When the pumping energy is below the threshold, the spontaneous emission spectrum shows multiple cavity modes. The mode spacing is about 7 nm corresponding to a cavity length of 4.3 mm, which is nearly equal to the thickness of the epitaxial cavity. The linewidth of a single cavity mode is.8 nm as shown in the inset of Fig. 16. The cavity quality factor (Q factor) estimated from the linewidth is about 518. Considering the optical absorption of GaN layer, an estimated effective cavity reflectivity based on this Q factor is about 97%, which is close to the cavity reflectivity formed by the two dielectric DBRs. This result indicates the laser cavity structure was nearly intact after the laser lift-off process. As the pumping energy increased above the threshold, a dominant laser emission line appeared at 414 nm with a narrow linewidth of about.25 nm. The lasing wavelength is located at one of cavity modes near the peak emission wavelength of the InGaN MQWs. The laser emission polarization contrast between two orthogonal directions was measured as shown in Fig. 17. The difference between the two angles of minimum intensities is 18 and a degree of polarization of about 7% is estimated. The lower degree of polarization in comparison to the hybrid DBR VCSEL could be due to the smaller Q factor for this dielectric DBR VCSEL. The laser emission intensity as the function of pumping energy is shown in Fig. 18 in a logarithmic scale. The factor was estimated as about 1: Since the cavity volume of this dielectric DBR VCSEL is large than the above hybrid DBR VCSEL, the Purcell factor and the spontaneous coupling factor shall be lower accordingly. The temperature dependence of the lasing threshold of the VCSEL is shown in Fig. 19. The threshold pumping energy increased gradually with increasing temperature. We obtain a characteristic temperature of about 278 K for this dielectric type VCSEL for the temperature range of 58 to 322 K by

9 2 1 st spot 1 st spot 16 Intensity (a.u.) 12 8 (a) 5 µm (b) 2 nd spot Emission intensity (arb. unit) β~ 1.1x1-2 Polarizer angle ( ) Fig. 17. Polarization characteristic of the dielectric DBR VCSEL. The degree of polarization is calculated to be 7%. (c) Intensity (a.u.) E th 1.2E th Fig. 2. Emission pattern of the dielectric DBR VCSEL at pumping energy of (a) 1:15E th with single laser emission spot and (b) 1:2E th with two laser emission spots. The arrows indicate the position of the first and second emission spots. (c) Spectra of the first emission spot and second emission spot at pumping energy of 1:15E th and 1:2E th, respectively Pumping energy (nj/pulse) Fig. 18. Laser emission intensity versus pumping energy in logarithmic scale. The value estimated from the difference between the two dash lines is about 1: Ln(E o /E th ) Experimetal data Linear fit Temperature (K) Fig. 19. Temperature dependence of the lasing threshold pumping energy (E th ) and the estimated T of 278 K. linearly fitting the experimental result. This T value is close to the T value for the hybrid DBR VCSEL, showing that the potential of high temperature operation in GaN VCSELs. The laser emission patterns from the aperture show singlespot and multiple-spots emission patterns under different 545 pumping conditions as shown in Figs. 2(a) and 2(b) at pumping energy of 1:15E th and 1:2E th. The lasing wavelengths from the different emission spots are slightly different (about few nanometers). At low pumping energy, the emission pattern showed a single spot with a spot size of about 3 mm. The lasing spot could be considered as a small aperture which forms a small optical volume for the laser. The relatively high spontaneous emission factor could be due to the small lasing spot. As the pumping energy increased, a second spot appeared showing a double spot emission pattern with a spatial separation of about 22 mm apart. The corresponding emission spectra of these two spots are shown in Fig. 2(c). The wavelength of the single emission spot is about nm. For the double emission pattern, there is second emission wavelength at about nm in addition to the nm emission line. Since the separation between these two spots is large compared to the axial mode spacing, the difference of the emission wavelengths could be caused by either the spatial nonuniformity in InGaN MQWs or the non-uniformity in the DBR cavity. To clarify the origin of these emission wavelength variations through the aperture, we conducted the micro-pl intensity mapping of the VCSEL using a scanning optical microscopy. Figure 21(a) shows the intensity mapping of the entire aperture of the VCSEL. With a fine scan inside the square area in Fig. 21(a), Fig. 21(b) shows the non-uniform PL emission intensity across the aperture has patches of bright areas with about 2 4 mm in size. The bright areas have about 1.8-times higher intensity than the dark areas. Fig. 21(c) shows the PL spectra of bright

10 (a) 3 5 (b) B 5 µm A nm, we can estimate the In composition difference of about.2% (2) to make 1 nm difference in wavelength. This amount of In composition difference is possible in a InGaN MQWs. The effects of carrier localization and change in refractive index caused by In inhomogeneity could also induce different lasing wavelength. In addition, the similar inhomogeneity was observed for our hybrid DBR VCSEL. Therefore, we believe the indium inhomogeneity in the VCSEL MQWs could be responsible for the appearance of spatially separated lasing spots within the mesa aperture and the difference in the emission wavelength could be due to the slight variation in the indium content of the MQW. The non uniform spatial emission patterns of the laser emission seem to be a peculiar character of GaN-based VCSEL with InGaN/GaN MQWs. (c) PL intensity (a.u.) PL intensity (a.u.) Bright area (A) 3 Dark area (B) Fig. 21. (a) Micro-PL intensity mapping image of the dielectric DBR VCSEL aperture. (b) Fine micro-pl scan inside the square area in (a). (c) PL spectra of bright (point A) and dark (point B) areas. (marked as A) and dark (marked as B) areas. Nevertheless, spatial inhomogeneity in cavity loss due to potential micrometer-scale imperfection of the DBRs could also cause different threshold gains in spatial distribution. However, micro-pl measurement results in Fig. 21(c) show similar linewidth of the spontaneous emission for bright and dark areas, indicating no significant spatially non-uniformity in the DBRs across the circular mesa. On the other hand, the different micro-pl intensities across the VCSEL aperture imply a non-uniform material gain distribution existed in InGaN/GaN MQW layers. In fact, the indium fluctuation was commonly observed for the expiaxially grown InGaN MQWs and also the subsequent carrier localization effect had been reported. 27) If the In inhomogeneity was contributed to the lasing wavelength difference between and Conclusions We have reviewed the design considerations and fabrication technology of two types of GaN-based VCSEL structures: hybrid DBR VCSELs and dielectric DBR VCSELs with similar InGaN/GaN MQW active layer. For the hybrid DBR VCSEL, we have fabricated a crack-free, high reflectivity AlN/GaN DBR by insertion of SL layers and use the high reflective Ta 2 O 5 /SiO 2 as dielectric mirror and achieved laser operation under optical pumping at room temperature. The laser has a threshold energy of 55 nj and emission wavelength at nm with a linewidth of.17 nm. The laser emission showed a high degree of polarization of 89% and high spontaneous emission coupling factor of For the dielectric DBR VCSEL, a key fabrication technique, laser lift-off was used for formation of dielectric DBR cavity. The dielectric DBR VCSEL also achieved laser action under optical pumping at room temperature condition. The laser has a threshold pumping energy of 27 nj and emitted wavelength at 414 nm with a linewidth of.25 nm. The laser also has a large spontaneous emission coupling factor of about 1:1 1 2 and degree of polarization of 7%. A non uniform spatial laser intensity emission pattern was observed for both lasers. Although both dielectric and hybrid DBR VCSEL structures are applicable for electrical injection VCSEL, there are few technical issues needed to be resolved before such electrical injection GaN-based VCSEL can be realized. These include the design of a transparent and good current spreading contact structure for carrier injection, improvement of the spatial uniformity of InGaN MQW active region because of relative small light emission aperture of the current injection device, enhancement of p- type GaN layer conductivity, and increase of the high reflectivity of DBRs for reduction of threshold current. The successful operation of optically pumped GaN-based VCSELs should be a useful and helpful guide for the design and fabrication of electrically pumped GaN-based VCSELs in the near future. Acknowledgements The authors would like to acknowledge Professor K. Iga of Tokyo Institute of Technology for helpful suggestion. The work was supported by the MOE ATU program and the National Science Council of the Republic of China under contract nos. NSC M-9-8, NSC E- 9-7-PAE, and NSC E

11 1) M. Pelton, C. Santori, J. Vuckovic, B. Zhang, G. S. Solomon, J. Plant, and Y. Yamamoto: Phys. Rev. Lett. 89 (22) ) H. Deng, G. Weihs, C. Santori, J. Bloch, and Y. Yamamoto: Science 298 (22) ) T. Kobayashi, T. Segawa, A. Morimoto, and T. Sueta: presented at 43rd Fall Meet. Japan Society of Applied Physics, Tokyo, September, 1982 [in Japanese]. 4) J. M. Redwing, D. A. S. Loeber, N. G. Anderson, M. A. Tischler, and J. S. Flynn: Appl. Phys. Lett. 69 (1996) 1. 5) T. Someya, K. Tachibana, J. Lee, T. Kamiya, and Y. Arakawa: Jpn. J. Appl. Phys. 37 (1998) L ) Y.-K. Song, H. Zhou, M. Diagne, I. Ozden, A. Vertikov, A. V. Nurmikko, C. Carter-Coman, R. S. Kern, F. A. Kish, and M. R. Krames: Appl. Phys. Lett. 74 (1999) ) J.-F. Carlin, J. Dorsaz, E. Feltin, R. Butté, N. Grandjean, M. Ilegems, and M. Laügt: Appl. Phys. Lett. 86 (25) ) J. T. Chu, T. C. Lu, H. H. Yao, C. C. Kao, W. D. Liang, J. Y. Tsai, H. C. Kuo, and S. C. Wang: Jpn. J. Appl. Phys. 45 (26) ) K. E. Waldrip, J. Han, J. J. Figiel, H. Zhou, E. Makarona, and A. V. Nurmikko: Appl. Phys. Lett. 78 (21) ) J. Dorsaz, J.-F. Carlin, C. M. Zellweger, S. Gradecak, and M. Ilegems: Phys. Status Solidi A 21 (24) ) H. M. Ng, T. D. Moustakas, and S. N. G. Chu: Appl. Phys. Lett. 76 (2) ) N. Nakada, H. Ishikawa, T. Egawa, and T. Jimbo: Jpn. J. Appl. Phys. 42 (23) L ) J. P. Zhang, H. M. Wang, M. E. Gaevski, C. Q. Chen, Q. Fareed, J. W. Yang, G. Simin, and M. Asif Khan: Appl. Phys. Lett. 8 (22) ) E. Feltin, B. Beaumont, M. Laugt, P. Vennegues, H. Lahreche, M. Leroux, and P. Gibart: Appl. Phys. Lett. 79 (21) ) M. Kurimoto, T. Nakada, Y. Ishihara, M. Shibata, T. Honda, and H. Kawanishi: Jpn. J. Appl. Phys. 38 (1999) L ) Y. Ishihara, J. Yamamoto, M. Kurimoto, T. Takano, T. Honda, and H. Kawanishi: Jpn. J. Appl. Phys. 38 (1999) L ) G. S. Huang, T. C. Lu, H. H. Yao, H. C. Kuo, S. C. Wang, C. W. Lin, and Li Chang: Appl. Phys. Lett. 88 (26) ) W. S. Wong, T. Sands, N. W. Cheung, M. Kneissl, D. P. Bour, P. Mei, L. T. Romano, and N. M. Johnson: Appl. Phys. Lett. 75 (1999) ) Chih-Chiang Kao, Y. C. Peng, H. H. Yao, J. Y. Tsai, Y. H. Chang, J. T. Chu, H. W. Huang, T. T. Kao, T. C. Lu, H. C. Kuo, and S. C. Wang: Appl. Phys. Lett. 87 (25) ) R. J. Horowicz, H. Heitmann, Y. Kadota, and Y. Yamamoto: Appl. Phys. Lett. 61 (1992) ) S. Kako, T. Someya, and Y. Arakawa: Appl. Phys. Lett. 8 (22) ) T. Honda, H. Kawanishi, T. Sakaguchi, F. Koyama, and K. Iga: MRS Internet Nitride Semicond. Res. 4S1 (1999) G ) C. Skierbiszewski, P. Perlin, I. Grzegory, Z. R. Wasilewski, M. Siekacz, A. Feduniewicz, P. Wisniewski, J. Borysiuk, P. Prystawko, G. Kamler, T. Suski, and S. Porowski: Semicond. Sci. Technol. 2 (25) ) M. Ikeda and S. Uchida: Phys. Status Solidi A 194 (22) ) L. Siozade, S. Colard, M. Mihailovic, J. Leymarie, A. Vasson, N. Grandjean, M. Leroux, and J. Massies: Jpn. J. Appl. Phys. 39 (2) 2. 26) S. H. Park: Jpn. J. Appl. Phys. 42 (23) L17. 27) K. Okamoto, A. Kaneta, Y. Kawakami, S. Fujita, J. Choi, M. Terazima, and T. Mukai: J. Appl. Phys. 98 (25)