High-power laser structures grown on bulk GaN crystals

High-power laser structures grown on bulk GaN crystals

ARTICLE IN PRESS Journal of Crystal Growth 272 (2004) 274–277 www.elsevier.com/locate/jcrysgro High-power laser structures grown on bulk GaN crystal...

189KB Sizes 2 Downloads 46 Views

ARTICLE IN PRESS

Journal of Crystal Growth 272 (2004) 274–277 www.elsevier.com/locate/jcrysgro

High-power laser structures grown on bulk GaN crystals Pawel Prystawkoa,, Robert Czernetzkia,c, Lucy Gorczycaa,b, Grzegorz Targowskia, Przemek Wisniewskia, Piotr Perlina, Marcin Zielinskid, Tadeusz Suskia, Mike Leszczynskia,c, Izabella Grzegorya,c, Sylwester Porowskia a

Unipress, High Pressure Research Center, PAS, Sokolowska 29/37, Warsaw 01-142, Poland University of Science and Technology, Material Sciences and Ceramics, Al Mickiewicza 30, Krakow, Poland c TopGaN Ltd, Sokolowska 29/37, Warsaw 01-142, Poland d University of Montpellier II, GES-CNRS, Case courrier 074, Montpellier 34095, France

b

Abstract High-pressure-grown bulk GaN crystals have the best reported structural quality and the lowest defect density among all available GaN substrates. We report on MOVPE growth, device processing and properties of high-power laser structures on such substrates. In this work, we describe very high optical output power per laser facet of 1.9 W from single LD device with 15 mm  500 mm cavity. The measured value is up to our knowledge the highest nitride laser power reported. Also, the electrical and optical power densities, yet pulsed, are in the range of the highest reported for broad area lasers what confirms that nitride-based wide bandgap structures are capable of handling very high power. From thermally accelerated electrical ageing tests we extracted the activation energy of one of degradation mechanisms to be 0.32 eV. Catastrophic optical damage was not observed up to power density of 40 MW/cm2. Demonstrated results show that the availability of very low defect density GaN substrates is clearly the key issue in fabrication of highly reliable devices and high power LDs. r 2004 Elsevier B.V. All rights reserved. PACS: 42.55.Px; 81.15.Kk Keywords: A3. Metalorganic chemical vapor deposition; B1. GaN

1. Introduction

Corresponding author. Tel.: +48-22-876-0309; fax: +48-

22-632-4218. E-mail address: [email protected] (P. Prystawko).

Recently, there has been great interest in GaNbased wide bandgap highly stable semiconductors due to their applications for optoelectronic as well as electronic (high-power and high-frequency) devices. However, reliability of these devices is

0022-0248/$ - see front matter r 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.jcrysgro.2004.08.098

ARTICLE IN PRESS P. Prystawko et al. / Journal of Crystal Growth 272 (2004) 274–277

obstructed by high densities of threading dislocations and point defects, especially for laser diodes and critically for those of high power. It is induced by the use of the lattice-parameters and thermal expansion mismatched substrates. Improvement in the quality of GaN layers using low-temperature buffer layers has some limitation (dislocation density 108 cm 2) and further improvement has been developed based on the various selective growth methods. Unfortunately, the low defect density area is essentially restricted even using mentioned methods, since the buffer layers originate from different never perfectly oriented nucleation sites. The high-performance LD devices occupy relatively large area, therefore they are very sensitive to the number of defects in the cavity surroundings. We report on some properties of high-power laser devices grown on bulk GaN crystal substrates obtained by the high pressure solution method (HPS) [1]. The main advantage of these crystals is the lowest available dislocation density of 100 cm 2.

2. Epitaxial growth The test laser structures were grown in lowpressure MOVPE home-made vertical reactor. The sources were ammonia, trimethylgallium, triethylgallium, trimethylaluminium, trimethylindium, monosilane and bis-cyclopentadienyl magnesium. The reactor pressure was 100–900 mbar and the carrier gas was nitrogen and/or hydrogen. The other details on the growth conditions are given in Ref. [2]. The structure was grown on slightly misoriented (0 0 0 1) GaN surface of 60-mm-thick n-type substrate [3] starting from Si-doped GaN buffer (Fig. 1). This buffer was followed by a 0.6mm-thick n-type strained layer superlattice (SLS) cladding layer consisting of 120 periods of Sidoped GaN and Al0.16Ga0.84N heterostructures, a 0.1-mm thick n-type GaN optical guiding layer, MQW active region consisting of five periods of 8.5 nm-thick In0.02Ga0.98N:Si barriers and 4.5-nmthick In0.08Ga0.92N wells, a 1.5 nm undoped GaN, a 20-nm-thick p-type Al0.22Ga0.78N electronblocking layer, 0.1-mm-thick p-type GaN optical guiding layer, a 0.38-mm-thick p-type strained

275

p-electrode (Ni/Au) p-Al0.16Ga 0.84 N/GaN SLS clad (0.38 µm) p-GaN guide (0.1 µm)

p-Al0.22Ga0.78 N (0.02 µm)

n-GaN guide (0.1 µm)

MQW active layer

n-Al0.16Ga0.84N/GaN SLS clad (0.6 µm)

n-GaN (1 µm, MOVPE)

n-GaN Substrate (60 µm, HPS) n-electrode (Ti/Au)

Fig. 1. Schematic structure of the test device.

layer superlattice (SLS) cladding layer consisting of 80 periods of Mg-doped GaN and Al0.16Ga0.84N heterostructures and a 0.03 mm-thick ptype GaN contact layer. The p-type layers of our structure were thermally activated in the reactor. SCH ridge waveguide laser diode cavities were formed with the length ranging from 200 to 600 mm and width between 3 and 20 mm. A pelectrode Ni/Au contact on top of relatively shallow mesa stripe (about 0.28 mm deep) was deposited. Both cleaved mirrors were symmetrically coated with two pairs of SiOX/ZrO2 quarter wave layers. For high-power tests, chips with the 500-mm  15 mm structure was mounted on diamond heatsink.

3. Results and discussion Laser devices were electrically tested with pulsed current source at room temperature. For 30 ns pulses and repetition rate of 100 kHz on 10–20 mm wide, 500-mm-long stripes, we obtained the threshold current densities of 3.5 kA/cm2 and voltage at the threshold of 8 V. The lasing wavelength was 412 nm in this case. For 500-mm-long cavities processed from the same wafer we observed increase of the threshold current density with the narrowing of the ridge width as shown in Fig. 2. One can explain this result assuming that our LD design suffered from the strong current spreading within the upper SLS cladding layer. With the gain guided structure shown schematically in Fig. 1, one can observe 5-fold increase in the threshold current density using 3-mm-wide ridge in compar-

ARTICLE IN PRESS P. Prystawko et al. / Journal of Crystal Growth 272 (2004) 274–277

276

35 30 25 20 15

1.4

1x104 life time[h]

current max 1.3A 100 kHz. 30 ns Ea~0.32eV

2.1

Ld382b928 Ld421b956

40

Optical power [a.u.]

2

Threshold current density [kA/cm ]

45

11 375 h - 20° C

1x103

1x102

T=55°C

20

0.7

30

40 50 60 70

T [ °C]

T=65°C T=70°C

10

0.0

5 0

525 530 535 540 545 550 555 560 565 570 575 580 585 590 595 0

2

4

6

8

10

12

14

16

18

20

22

Time [h]

24

Stripe width [µm] Fig. 2. Threshold current densities for 500 mm long cavity, various width stripes.

Fig. 3. Thermally accelerated ageing tests. Inset shows expected lifetime fit for 100 mW output power at 20 1C.

2000 1800 1600

Optical power [mW]

ison to the broad ridge. One would expect about 2fold higher current density at the threshold of those gain-guided structures in comparison to the index-guided devices with step-like refractive index in waveguide plane. Therefore, by applying different processing scheme and proper device area for certain application, it is possible to achieve lower electrical power needed at lasing threshold. This would also allow for the tailoring of the emitted beam geometry [4]. Next, the 15-mm-wide 500-mm-long cavities with the threshold current of 400 mA were tested at elevated temperatures above the threshold. For constant current conditions a slow decrease in the emitted optical power was observed (Fig. 3). From this accelerated test measurement, we estimated the lifetime of the device at 20 1C to be about 11 000 h for emitted peak power of 100 mW per facet. Extracted activation energy for output intensity degradation is of 0.32 eV. Similar laser structures mounted using diamond heat spreaders were tested with high current driver circuit up to 10.5 A in the peak. L–I characteristic of that device is presented in Fig. 4. For output power below 300 mW, the differential efficiency is of about 0.33 W/A. For the measurement setup limit of 10.5 A, we achieved 1.9 W emitted power in peak per one facet (both mirrors were symmetrical). This is the highest reported power emitted from single cavity. We also estimated the optical power

1400 1200 1000 800 600 400 200 0 0

2

4

6

8

10

12

Current [A]

Fig. 4. L–I characteristics for broad stripe laser diode, 412 nm.

densities on the mirrors to be about 40 MW/cm2 . Catastrophic Optical Damage (COD) was not observed for this devices. Already published value for COD of 57 MW/cm2 for similar mirrors [5] allow for use asymmetric high-reflecting coatings what should result in further increase in the optical power emitted from one cavity mirror. Despite the huge heating effect what resulted in the observed lasing wavelength shift by 3 nm towards the longer wavelength and reduction in differential efficiency to 0.2 W/A, our broad stripe LDs are capable of handling very high power. In our opinion, it is possible only for very low defect density in the epi

ARTICLE IN PRESS P. Prystawko et al. / Journal of Crystal Growth 272 (2004) 274–277 6

intensity [cps]

24

Mg

Al N Ga In

+

20

10

3

105

21

10

10 4

10

10 3

10

102

10

10 10

19

18

17

1

16

10

0

10 -1

0.0

Mg concentration [at/cm ]

10

15

10

0.5

1.0

1.5

1014

2.0

depth (µm)

277

demonstrated from single LD device. From thermally accelerated electrical ageing tests, activation energy for output intensity degradation was found to be 0.32 eV. Catastrophic optical damage was not observed up to 40 MW/cm2 of power density. The mentioned above results were demonstrated on very low defect density GaN substrates. The high power of our devices (though still not fully optimized) confirm that the key issue in fabrication of highly reliable devices and highpower LDs is the ultra-low defect density in the GaN substrates and epi layers. From this study, we can also conclude that very low defect density substrates are necessary to improve already demonstrated high-power LDs [7,8] and there is still room for that.

Fig. 5. Magnesium SIMS profile of the SCH MQW laser diode.

Acknowledgements structures [6]. The relatively low differential efficiency is probably caused both by the poor light guiding in gain-guided design and limited holes delivery rate what translates to the broad and weaker gain spectrum of the MQW. To confirm that, we measured the magnesium concentration in p-type layers using SIMS technique. The result shown in Fig. 5 indicate that the magnesium content below 1  1019 cm 3 in electron blocking layer and 2  1019 cm 3 not uniformly distributed in p-type cladding region of our device is still not the optimum one as determined by Hall effect measurement on bulk Mg-doped AlGaN layers. Therefore, further careful optimization of p-type doping is required.

4. Summary SCH ridge waveguide MQW InGaN laser diode was grown using MOVPE technique and subsequently processed to form 15 mm  500 mm cavity. Using pulsed current measurement, the voltage and current at the threshold were determined to be 8 V and 400–500 mA, respectively, for lasing wavelength of 412 nm. Very high optical output power per laser facet of 1.9 W (3.8 W total) was

This work was partially supported by the European Commission, Grant ‘‘Support for Centers of Excellence’’ No. ICA1-CT-2000-70005 and Sixth European Framework Programme Priority ‘‘3’’ Proposal/Contract no.: STREP 505641-1 ‘‘GaNano’’.

References [1] Izabella Grzegory, J. Phys.: Condens. Matter 13 (2001) 6875. [2] P. Prystawko, et al., Phys. Stat. Sol. (a) 192 (2002) 320. [3] M. Sarzynski, et al., Special Issue of Physica Status Solidi, The Proceedings of EXMATEC04, in 2004. [4] Shigetoshi Ito, Yukio Yamasaki, Susumu Omi, Kunihiro Takatani, Toshiyuki Kawakami, Tomoki Ohno, Masaya Ishida, Yoshihiro Ueta, Takayuki Yuasa, Mototaka Taneya, Phys. Stat. Sol. (a) 200 (1) (2003) 131. [5] M. Ikeda, T. Mizuno, M. Takeya, S. Goto, S. Ikeda, T. Fujimoto, Y. Ohfuji, T. Hashizu, Phys. Stat. Sol. (c) 1 (6) (2004) 1461. [6] P. Perlin, et al., MRS Internet J. Nitride Semicond. Res. 9 (2004) 3. [7] Shu Goto, Makoto. Ohta, Yoshifumi Yabuki, Yukio Hoshina, Kaori Naganuma, Koshi Tamamura, Toshihiro Hashizu, Masao Ikeda, Phys. Stat. Sol. (a) 200 (1) (2003) 122. [8] R. Czernetzki, et al., Phys. Stat. Sol. (a) 200 (1) (2003) 9.