Applied Physics Express 6 (2013) 092105 http://dx.doi.org/10.7567/APEX.6.092105

Metal Organic Vapor Phase Epitaxy of Monolithic Two-Color Light-Emitting Diodes Using an InGaN-Based Light Converter Benjamin Damilano1
, Hyonju Kim-Chauveau1 y, Eric Frayssinet1 , Julien Brault1 , Sakhawat Hussain1;2 , Kaddour Lekhal1 , Philippe Venne´gue`s1 , Philippe De Mierry1 , and Jean Massies1 1 CRHEA-CNRS, Centre de Recherche sur l’He´te´ro-Epitaxie et ses Applications–Centre National de la Recherche Scientifique, Rue Bernard Gregory, 06560 Valbonne, France 2 University of Nice Sophia-Antipolis, 06103 Nice Cedex 02, France E-mail: [email protected]

Received June 4, 2013; accepted August 16, 2013; published online September 5, 2013 Monolithic InGaN-based light-emitting diodes (LEDs) using a light converter fully grown by metal organic vapor phase epitaxy are demonstrated. The light converter, consisting of 10–40 InGaN/GaN quantum wells, is grown first, followed by a violet pump LED. The structure and growth conditions of the pump LED are specifically adapted to avoid thermal degradation of the light converter. Electroluminescence analysis shows that part of the pump light is absorbed by the light converter and reemitted at longer wavelength. Depending on the emission wavelength of the light converter, different LED colors are achieved. In particular, for red-emitting light converters, a color temperature of 2100 K corresponding to a tint between warm white and candle light is demonstrated. # 2013 The Japan Society of Applied Physics

ne of the great successes of group III–nitride semiconductors has been to open the way for solid-state lighting. Since the mid-90s and the first demonstration of a white light emitting diode (LED), based on a blue-emitting InGaN LED covered by a yellow-emitting cerium-doped yttrium aluminium garnet dispersed in epoxy, there has been huge progress in the performances of these devices.1) Record efficiencies exceeding 200 lm/W are now reported.2,3) Despite these impressive achievements, phosphor-converted white LEDs intrinsically suffer from some drawbacks compared with simple blue InGaN LEDs. The use of phosphor adds in the LED process complexity, cost, problems of color stability with time and temperature, and induces reliability issues compared with blue LEDs.4) It is also possible to make phosphor-free monolithic white LEDs by taking advantage of the ability of group III–nitrides to emit light in the whole visible range.5) By mixing blueand yellow-emitting InGaN quantum wells (QWs) inside the p–n junction of the LED, white light can be obtained as already reported by several groups.6–14) The same principle has also been applied to three-dimensional layers such as facetted surfaces or vertical nanowires.15–17) However, some problems are encountered with these approaches: the carrier distribution inside the different QWs strongly evolves with the injected current producing a color change,8,18,19) and the luminous efficiency is smaller than that for phosphorconverted LEDs.18) In order to overcome these problems, we have recently proposed to use a monolithically integrated light converter made of InGaN/GaN multiple (M)QWs. The long-wavelength (yellow) part of the white LED is no longer electrically injected but optically excited by a shortwavelength pump (blue-violet).20–22) Optical pumping instead of electrical pumping presents a double advantage: the carrier distribution inside the light converter is not dependent on current injection efficiency and the carrier density inside each QW of the light converter is smaller compared with current-injected monolithic white LEDs.21) This latter property will limit the impact of the efficiency droop at large current densities, particularly pronounced for green-yellow-emitting InGaN QWs.23)

O

y

Present address: Saint-Gobain Crystals 2720, Chemin Saint Bernard, Les Moulins I, F-06220 Vallauris, France.

We previously demonstrated monolithic white LEDs with a light converter either by using molecular beam epitaxy (MBE) only or by a combination of metal organic vapor phase epitaxy (MOVPE) for the light converter and MBE for the pump LED.20,21) MBE has the advantage of lower growth temperatures than MOVPE (typically 200–300  C lower) for n- and p-type GaN. Hence, the choice of growing the pump LED by MBE instead of MOVPE was motivated by the goal of avoiding thermal degradation of the high-In-composition InGaN QWs of the light converter during the growth of the pump LED. However, blue LEDs grown by MBE still have a lower efficiency than those grown by MOVPE and, furthermore, MBE is not used for the production of InGaN-based LEDs.3,24) Therefore, it is of prime importance for the monolithic white LED efficiency and the cost reduction related issue to develop a growth process using MOVPE only. The samples were grown by MOVPE on commercially available n-type GaN-on-sapphire templates. Trimethylgallium or triethylgallium, trimethylaluminium, trimethylindium, bis(cyclopentadienyl)magnesium, silane, and ammonia were used as precursors for gallium, aluminium, indium, magnesium, silicon, and nitrogen, respectively. The In composition and QW thicknesses of the LED structures are deduced from InGaN/GaN MQW calibration samples characterized by high resolution X-ray diffraction and transmission electron microscopy. Photoluminescence (PL) was measured at room temperature (RT) using a frequencydoubled Ar laser at 244 nm. For electrical contacts, one edge of the LED samples surface is scratched with a diamond tip in order to expose the n-type GaN layer and then In is used to contact both n- and p-type GaN layers. The LEDs are measured on wafer at RT under continuous wave conditions. The light output was collected by a 200 m core UV/visible optical fiber, dispersed by a BWTEK spectrometer equipped with a thermoelectrically cooled linear CCD array. The output power of the LEDs is measured through the sapphire substrate using a calibrated Si photodiode. The structure of the monolithic LEDs with a light converter is schematically drawn in Fig. 1. The growth is initiated by 0.5 m of Si-doped GaN at 1080  C. Then, Iny Ga1y N/GaN MQWs constituting the light converter are grown. The growth temperature of Iny Ga1y N is set between 715 and 780  C, for the series of different samples.

092105-1

# 2013 The Japan Society of Applied Physics

Appl. Phys. Express 6 (2013) 092105

B. Damilano et al. Table I. Compositions and thicknesses of the layers constituting the monolithic LEDs with a light converter.

Sample

Pump LED active zone

A

B

C

D

GaN:Mg thickness (nm)

150

150

150

150

Al0:2 Ga0:8 N:Mg thickness (nm)

15

15

15

15

Number of QWs

5

5

5

5

Inx Ga1x N thickness (nm)

1.4

1.4

1.4

1.4

In composition (x)

0.1

0.1

0.1

0.1

9

9

9

9

GaN barrier thickness

Structure of the monolithic LEDs with a light converter. The light converter is constituted by 10- to 40-period Iny Ga1y N/GaN MQWs. On top of the light converter is grown a pump LED with a 5-period Inx Ga1x N/ GaN MQW. The In composition x is smaller than y.

PL intensity (norm.)

1.0

20

20

20

20

10

40

40

Iny Ga1y N thickness (nm) In composition (y)

1.6 3.4 3.1 3.1 0.26 0.18 0.22 0.28

GaN:Si barrier thickness (nm)

T=300K

0.8 0.6 0.4

21

18

18

18

(b)

300

400

500

600

700

Wavelength (nm)

0.2 0.0 350

Fig. 2.

Light converter

20

Number of QWs

EL intensity (arb. unit)

Fig. 1.

GaN:Si thickness (nm)

450 550 650 Wavelength (nm)

Fig. 3. Photograph (a) and RT EL spectrum (b) at an injection current of 20 mA of a violet LED (i.e., a pump LED without light converter).

750

RT PL of a series of InGaN/GaN MQWs emitting from blue to

red.

Depending on the Iny Ga1y N growth temperature (larger In compositions are obtained at lower temperatures) and the QW thickness, blue to red PL (up to 650 nm) at RT can be achieved for 10-period Iny Ga1y N/GaN MQWs (Fig. 2). The PL peaks display small intensity oscillations due to light interferences in the cavity formed by the GaN-based layers sandwiched by the sapphire and the air. We observe a broadening of the PL peaks when y increases probably due to the increase of the Iny Ga1y N alloy fluctuations and the QW interface roughness. It is generally preferable to avoid PL peak broadening to maximize the spectral purity. However, in the case of the application targeted here, i.e., for a white light source, it is preferable to have as broad PL peaks as possible to improve the color rendering index. Four different samples were grown. Their characteristics are presented in Table I. The number of QWs is 20 and 10 for samples A and B, respectively, and 40 for samples C and D. The targeted colors for the different light converters are blue, green, yellow-green, and red for samples A–D, respectively. To make the whole LED structure, a violet LED (pump LED) is grown on top of the light converter. First, a 20 nm n-type GaN layer is deposited followed by a 5-period Inx Ga1x N/GaN MQW, a 20 nm p-type Al0:2 Ga0:8 N electron-blocking layer, and a 150 nm p-type GaN layer. The Inx Ga1x N alloy is grown at 800  C, the corresponding In composition is 0.1, smaller than y and, therefore, the emission wavelength of the pump LED is

shorter than that of the MQW light converter. This is a condition for having light absorption of the pump LED by the light converter. If the growth temperature of the n- and ptype (Al)GaN layers of the pump LED is kept at 1080  C, we observe, after growth, that the samples become dark indicating metal clustering inside the high-In-composition Iny Ga1y N QWs of the light converter. As a consequence, the light emission intensity is strongly degraded. We have then progressively decreased the growth temperature of these layers in order to avoid the thermal degradation of the QWs in the light converter. We have found that a growth temperature of 970  C was low enough. It should be mentioned that keeping the Mg concentration and the doping profile of p-type layers optimized for a growth temperature of 1080  C down to 970  C, results in LEDs with a very poor efficiency. Therefore, these parameters were re-optimized on standard violet LED structures (i.e., pump LEDs without light converter) with a growth temperature of 970  C for the p-type layers in order to obtain efficient hole injection and to maximize the electroluminescence (EL) intensity.25) Actually, the output power measured on these LEDs is the same as that for LEDs for which the p-type layers are grown at 1080  C (2 mW at 20 mA). A typical RT EL spectrum at 20 mA and the corresponding photograph obtained on such LED are shown in Fig. 3. The EL peak wavelength is 400 nm, corresponding to a violet color. We now discuss the EL properties of samples A–D, i.e., the monolithic LEDs with different light converters. The photographs and EL spectra at an injection current of 20 mA are shown in Figs. 4 and 5, respectively. It is evident that the

092105-2

# 2013 The Japan Society of Applied Physics

Appl. Phys. Express 6 (2013) 092105

B. Damilano et al.

0.8

y

0.6 0.4 0.2 0.0 0.0

(a)–(d) Photographs of the monolithic LEDs with light converters corresponding to samples A–D.

(a)

300

E L intensity (arb. unit)

EL intensity (arb. unit)

Fig. 4.

490 nm

400

500

(b)

526 nm

700 300

600

E L intensity (arb. unit)

EL intensity (arb. unit)

Wavelength (nm) (c)

300

551 nm

400

500

600

700

Wavelength (nm)

400

500

600

700

Wavelength (nm)

(d)

300

605 nm

400

500

600

700

Wavelength (nm)

Fig. 5. (a)–(d) RT EL spectra under continuous wave conditions at a current of 20 mA corresponding to samples A–D.

LED color is modified by the presence of the light converter and by its emission wavelength. While violet light is observed when the pump LED is grown alone, blue, green, yellow-green, and white-orange are observed for samples A–D, respectively. For all these samples, we can observe in the EL spectra the presence of two distinct peaks, one originating from the pump LED at 380–400 nm, and the other from the light converter at longer wavelength. The nominal growth conditions of the active zone of the pump LED are the same, but some differences in the EL wavelength of the pump LED are observed due to on-wafer inhomogeneities and run to run variations. The light converter emits at 490, 526, 551, and 605 nm for samples A–D, respectively. We define R as the integrated EL intensity of the light converter divided by the total integrated EL intensity. The R values are 0.60, 0.55, 0.70, and 0.43 for samples A–D, respectively. This variation can be understood by considering two different parameters: the internal quantum efficiency and the absorption of the light converter. The internal quantum efficiency of InGaN/GaN QWs generally decreases when the emission wavelength increases.7) This is

0.2

0.4

x

0.6

Fig. 6. Chromatic coordinates of the LEDs with light converters. The black line corresponds to the chromatic coordinates of the black body curve. The triangle, circle, square, and diamond symbols correspond to samples A–D, respectively.

ascribed to the degradation of the InGaN material quality with increasing In composition and also to the consequence of the quantum-confined Stark effect. Indeed, the presence of the internal electric field in these (0001)-oriented InGaN QWs induces a decrease of the oscillator strength of the optical transition and the reduction of the internal quantum efficiency when non-radiative recombination paths are present. The QW number and the QW thickness of the light converter are not the same for all the samples, therefore the amount of light absorbed by the light converter varies for the different samples. The most remarkable consequence is that the R value of sample C is larger than that of sample B despite a longer emission wavelength. The internal quantum efficiency decrease is, in this case, compensated by the larger absorption of the 40 QWs of the light converter of sample C compared to the 10 QWs of sample B. The EL wavelength of the pump LED is varying between 380 and 400 nm for the different LEDs and this could influence the absorption of the light converter and modify the ratio R. However, according to published data on the absorption of InGaN/GaN MQWs, the pump LED wavelength range in the present study is above the effective band-edge of the MQW, and therefore, only minor absorption variations are expected.26) The luminous power at an injection current of 50 mA is 86, 264, 67, and 5 mlm for samples A–D, respectively. The optimum luminous power is obtained for the green emitting light converter of sample B. This result is the consequence of a limited decrease of the internal quantum efficiency with the increase of the emitted wavelength (see above) for InGaN QWs emitting in the green. In addition, the eye sensitivity approaches its maximum value in this spectral region (683 lm/W at 555 nm). As a comparison, the best luminous power obtained previously for similar monolithic LED structures, including a light converter emitting in the green grown by MOVPE and a pump LED grown by MBE, was 200 mlm.21) Compared to this previous work, the luminous power of sample B is 32% higher. Even if this improvement is not huge, it already shows the potential of a full-MOVPE approach for monolithic LEDs with a light converter. We have calculated the CIE-1931 chromatic coordinates of the samples and reported the values, together with those of the black body temperature curve in the chromaticity diagram of Fig. 6. For sample D, the chromatic coordinates

092105-3

# 2013 The Japan Society of Applied Physics

Appl. Phys. Express 6 (2013) 092105

B. Damilano et al.

are close to those of the black body temperature curve. The corresponding color temperature is 2100 K, which indicates that a tint between warm white and candle light is achieved. In conclusion, we have shown that it is possible to make monolithic LEDs with a light converter entirely grown by MOVPE. Depending on the QW number and the emission wavelength of the light converter, the color of the monolithic LEDs can be tuned in a large spectral range. Some improvements of the structure proposed in this work can be done to achieve better color control. For example, instead of using only one emission in the light converter, we can mix several emissions. This can be of prime importance in view of improving the color rendering index of a white light source emitting at a desired color temperature. Acknowledgments The authors would like to thank M. Leroux and B. Vinter for the critical reading of the manuscript, O. Daniel from FIST S.A., P. Chalbet, J. M. Chauveau, and B. Poulet for their support and interest in this work. This work was partly funded by the French agency for research ANR 2011 EMMA 004 01 project ‘‘DELMONO’’.

9)

10) 11) 12) 13) 14)

15) 16) 17) 18) 19)

1) S. Nakamura and G. Fasol: The Blue Laser Diode (Springer, Berlin, 1997). 2) Y. Narukawa, M. Sano, M. Ichikawa, S. Minato, T. Sakamoto, T. Yamada,

and T. Mukai: Jpn. J. Appl. Phys. 46 (2007) L963.

20) 21)

3) Y. Narukawa, M. Ichikawa, D. Sanga, M. Sano, and T. Mukai: J. Phys. D

43 (2010) 354002.

22)

4) M. Meneghini, A. Tazzoli, G. Mura, G. Meneghesso, and E. Zanoni: IEEE

Trans. Electron Devices 57 (2010) 108.

23)

3976.

24)

5) T. Mukai, M. Yamada, and S. Nakamura: Jpn. J. Appl. Phys. 38 (1999) 6) B. Damilano, N. Grandjean, C. Pernot, and J. Massies: Jpn. J. Appl. Phys.

40 (2001) L918.

25)

L246.

26)

7) M. Yamada, Y. Narukawa, and T. Mukai: Jpn. J. Appl. Phys. 41 (2002) 8) S. Dalmasso, B. Damilano, C. Pernot, A. Dussaigne, D. Byrne, N.

092105-4

Grandjean, M. Leroux, and J. Massies: Phys. Status Solidi A 192 (2002) 139. S. J. Chang, L. W. Wu, Y. K. Su, C. H. Kuo, W. C. Lai, Y. P. Hsu, J. K. Sheu, J. F. Chen, and J. M. Tsai: IEEE Trans. Electron Devices 50 (2003) 519. A. Dussaigne, J. Brault, B. Damilano, and J. Massies: Phys. Status Solidi C 4 (2007) 57. C.-F. Huang, C.-F. Lu, T.-Y. Tang, J.-J. Huang, and C. C. Yang: Appl. Phys. Lett. 90 (2007) 151122. S.-N. Lee, H. S. Paek, H. Kim, T. Jang, and Y. Park: Appl. Phys. Lett. 92 (2008) 081107. C.-F. Lu, C.-F. Huang, Y.-S. Chen, W.-Y. Shiao, C.-Y. Chen, Y.-C. Lu, and C.-C. Yang: IEEE J. Sel. Top. Quantum Electron. 15 (2009) 1210. A. F. Tsatsulnikov, W. V. Lundin, A. V. Sakharov, E. E. Zavarin, S. O. Usov, A. E. Nikolaev, N. V. Kryzhanovskaya, M. A. Synitsin, V. S. Sizov, A. L. Zakgeim, and M. N. Mizerov: Semiconductors 44 (2010) 808. M. Funato, T. Kondou, K. Hayashi, S. Nishiura, M. Ueda, Y. Kawakami, Y. Narukawa, and T. Mukai: Appl. Phys. Express 1 (2008) 011106. H.-W. Lin, Y.-J. Lu, H.-Y. Chen, H.-M. Lee, and S. Gwo: Appl. Phys. Lett. 97 (2010) 073101. W. Guo, A. Banerjee, P. Bhattacharya, and B. S. Ooi: Appl. Phys. Lett. 98 (2011) 193102. M. Yamada, Y. Narukawa, H. Tamaki, Y. Murazaki, and T. Mukai: IEICE Trans. Electron. E88-C (2005) 1860. R. Charash, P. P. Maaskant, L. Lewis, C. McAleese, M. J. Kappers, C. J. Humphreys, and B. Corbett: Appl. Phys. Lett. 95 (2009) 151103. B. Damilano, A. Dussaigne, J. Brault, T. Huault, F. Natali, P. Demolon, P. De Mierry, S. Chenot, and J. Massies: Appl. Phys. Lett. 93 (2008) 101117. B. Damilano, P. Demolon, J. Brault, T. Huault, F. Natali, and J. Massies: J. Appl. Phys. 108 (2010) 073115. B. Damilano, N. Trad, J. Brault, P. Demolon, F. Natali, and J. Massies: Phys. Status Solidi A 209 (2012) 465. D. Schiavon, M. Binder, A. Loeffler, and M. Peter: Appl. Phys. Lett. 102 (2013) 113509. K. Johnson, V. Bousquet, S. E. Hooper, M. Kauer, C. Zellweger, and J. Heffernan: Electron. Lett. 40 (2004) 1299. K. Ko¨hler, T. Stephan, A. Perona, J. Wiegert, M. Maier, M. Kunzer, and J. Wagner: J. Appl. Phys. 97 (2005) 104914. L. Siozade, P. Disseix, A. Vasson, J. Leymarie, B. Damilano, N. Grandjean, and J. Massies: Phys. Status Solidi A 183 (2001) 139.

# 2013 The Japan Society of Applied Physics