GaN Based High Brightness LEDs and UV LEDs

8/4/2019 GaN Based High Brightness LEDs and UV LEDs

http://slidepdf.com/reader/full/gan-based-high-brightness-leds-and-uv-leds 1/4

GaN Based High Brightness LEDs and U V LEDs

S . P. DenBaars, T. Katona, P. Cantu, A. Hanlon, S. Keller, M. Schmidt, T.Margalith, M . Pattisson,C.Moe, J. Speck, and S. Nakamura

Materials Department,Solid-state Lighting and Display Center

University of C alifornia, Santa Barbara, California 93106 USA

Abstract

This talk will summ arize the important materialsand device results in gallium nitride based lightemitter technology. GaN has emerged as the mostpromising material for high brightness LEDs withcolors ranging from W , blue, green, and white.Recent progress on ultra-violet 0 mittingLEDs using AlGaN single quantum wells indicateswavelengths as short as 2 9 2 m are achievable. UV

LEDs are of great interest for solid state whitelighting due to the high conversion efficiencies oftypical phosphors in the UV specbum. This paperwill focus on recent progress in improving theproperties of W LEDs.

Introduction

Recently, GaN based LEDs have shownremarkable progress for efficient light generationspanning the blueigreen through W light spectrum.When combining blue or W GaN LED chips withphosphor eMicient white light generation is nowpossible. Currently, the efficiency of GaN basedwhite HB-LEDs is now 30lumensiwatt inproduction and 74lumensiwatt in R&D[I], whichnow exceeds that of incandescent lighting and hasled to new applications in solid state lighting anddisplays (See figure 1). Using MOCVD we haveachieved GaN based LEDs with external quantumefficiencies ranging from 20% to sub 1% as thewavelength is decreased from 4SOnm o 290nm (asshown in Figure 2) . MOCVD growth of InGaN

quantum well structures and the effect ofdislocations are shown to dominate the LEDefficiency.W LEDs have application in solid-state white

lighting due to the high conversion efficiencies oftypical phosphors in the W spectrum. W LEDsare also of interest due to their applications forbiological and chemical weapons detection and

non-line of sight communication. Currently, hulkAIN substrates are not available. These substrateswould provide a low threading dislocation density(TDD), W transparent substrate allowing

heteroepitaxial AlGaN based LEDs to be grown incompression. An alternative to hulk AIN substratesis AIN or high composition AlGaN grown directlyon sapphire. It is difficult tn limit the dislocationdensity using this approach because the highsticking coefficient and low surface mobility of AIcompared to Ga causes rapid AlGaN islandcoalescence. This leads to formation of a largenumber of edge dislocations at coalescenceboundaries [I]. A large amount of work currently

focuses on growing AIN on sapphire via MOCV Dwith a low threading dislocation density (TDD) [Z-

31. In the absence of low dislocation densitytransparent substrates, we have grown UV LEDs onseveral microns of GaN on sapphire reducing theTDD to low to mid IO8 cm2. The effect of TDD onnon-radiative recombination in nitride basedemitters is well documented [4-61. Unfortunately,

the use of a GaNkapphire substrate causes theAlGaN based LED to be grown in tension. The

average composition of the devices grown in theseexperiments is 17.7%. We have inserted a 300A

AIN interlayer to relieve the stress and allow forcrack-kee growth of the UV LED [7-91. In thiswork we have investigated several different growthconditions for the AIN interlayer and show that thedevice performance improves as the long rangeorder of the AIN interlayer decreases.

Experimental Details

The device design is shown schematically in Figure 3. Growth was performed in a vertical close-spaced

showerhead MOCVD reactor, with growthconditions for all layers except the AIN interlayerreported elsewhere 191. Th e growth temperatures ofthe AIN interlayer were 700,900, and 1100°C. Th eAIN grown at 900°C was grown both with andwithout TMIn flow. Th e TMAl and TMIn flowswere 10.9, and 16.1 pnoVmin respectively for alltemperatures, with a ViIII ratio of 6896 for the AIN

layer grown with TMIn. A semitransparent Pd A u(3O.&/SOA) p-contact and TilAVNilAu(IOOA /2000~200A/3000.k) n-contact was

0-7803-7872-5/03/$17.00 02 00 3 IEEE

16.1.1

IEDM 03-385



deposited followed by a thick TUAu (200Al5000A)probe pad. Spectral measurements were madeusing a UV-Visible CCD array spectrometer withthe light couple d in through an optical fiber. DCLIV testing was done on chip using a HP 4145Bparameter analyzer with light detected by a largearea W enhanced Si photodiode located - 6m m

above the wafer.

The AlGaN total thickness of 4210 A was chosenbecause it is approximately the critical thicknessthat could be achieved before cracking occurred. Itshould be noted that if the 5000 A n-GaN contactlayer is removed from on top of the AIN interlayer,a much thicker device design (up to 1 pm

attempted) can be grown without cracking but thedevice performance degraded. We expect thedegradation resulted from a combination of highern-contact resistance, enhanced current crowding inthe n-contact layer resulting in non-unifonn currentinjection, and an increased TDD in the device. Thecritical thickness was the same for all four of theAIN interlayer growth conditions. We also expectthat for the interlayer grown with TMIn flow, theindium is acting almost solely as a surfactant as theincorporation is low due to the relatively highgrowth temperatures for indium containingcompounds [IO].

Results

The tum-on voltage of the LEDs is approximately3.7 V as is expected for diodes emitting at 334 nm,but there is a significant amount of diode leakage atlow currents that can be seen by the absence ofoutput power up to almost 5 mA injection current.The maximum efficiency was 0.8% at 340nm and20 mA of current. However, as the LED active

layer wavelength is decreased to 290- theexternal quantum efficiency drops to 0.01%. Webelieve the drop in efficiency is related to theincreasing non-radiative rate due to either higherdefect density or higher piezoelectric inducedcharge separation. The threading dislocationdensity was estimated from atomic forcemicroscopy images taken of the surface after thedevice growth [ I I]. The dislocation densities wereestimated to be 3.2, 4.3, and 5.2 x IO 9 cm.2 for

LEDs on AIN interlayers grown at 700 "C, 900 "C,and 900 OC with indium doping respectively. Thisindicates that the AIN surface morphologyinfluences the dislocation density of the n-GaNgrown on top of it, and as the AIN long-range orderdecreases, the resulting TDD is lower. Under

pulsed testing conditions the power saturated at ahigher injection current and the maximum outputpower was up to 9 times greater under pulsedconditions. Self-heating is expected with therelatively low quantum efficiency when comparedto commercial InGaN LEDs The quantumefficiency reac hes it's ma xim um value betw een 20 -

25 mA and decreases by - 25% from it's peak valueto the efficiency at 100 mA .

Electroluminescence (EL) spectral measurementsare shown in figure 4. A sharp peak, 8nm full

width half maximum, is observed at 292 nmattributed to emission due to recombination in the

quantum wells. There are also broad longwavelength peaks previously observed by othergroups [13,14]. It has been observed that the Mgconcentration in MOCVD grown films does notreach its terminal value until up to 1000 A afterinjection of Cp2Mg into the reactor. These UVLEDs have a relatively thin, 1460 A, p-type layercompared to - 5000 b, for typical InCaN basedvisible LEDs to reduce the series resistance

Conclusions

We have demonstrated that although AlGaN basedUV LEDs are grown in tension on GaN base layers,an AIN interlayer can be inserted allowing thegrowth of crack-free devices. Despite the LEDbeing grown on an absorbing epitaxial base layer,we have demonstrated approximately 0.8% externalefficiency at 340nm and 0.01% at 290nm, however,much work is needed to improve the performanceof deep UV emitting LE Ds.

Acknowledgements

This work was funded by LTC. John Carrano underthe DARPA SUVOS program awarded throughCREE Inc. and the Solid State Lighting and DisplayCenter at UCSB.

References

[ I 1 CREE press release. (see

hap://-w.cree.comR\(ewsincws I75.asp)H.M. Wang, 1. P. Zhang. C. Q. Chcn, Q. Farecd, 1.Yang,

Y. Kida, T. Shibata, H. Naoi. H. Miyake, K H i m a t s u , M.

H.Marchand, J. P. Ibbetson, P. T. ini, S. Chiehibu, S. I.Rosner, S . Kcller, S. P. DenBaars, I. S. Speck. U. K.Mishra, Pmseeding9 of the lnfematlonal Symposium On

Compound Semiconducton (1999). p. 68 14 .S . Nakamura, ElCE Transactions on Electronics E83C

(4), 529-535 2000) .M. ansen, P. Fmi, L. Zhao, A. C . Abara, L. A. Colden, J.

S. Speck, S. P. DenBaars, Appl. Phys. Lett. (USA) 76 (5),

[7] M. waya, S. Terao, N. Hayashi, T. Kashima, T.Detchprohm, H. mano, 1 Akasaki, A. Himno, C. Pemot,ho ceed i nm of the Matends Rcsearch SafieW Vol. 595.

[Z]

131

14)

M. A. man,~ p p i . hys. ~eu.USA) 81 (41,604 (znnz).

T ~ W . ~ ,hys. stat. SO]. A. 194 (2). 498 (znnz).

[ 5 ]

[6 ]

529-531 (2nnn).

(znno), p< W I .in.16.181 1. Han, K E. Waldrip, S. R. Lee, I. 1.Figiel, S. I. H a m e ,

G. A. Petenen, S. M . M y e n , Appi. Phys. Lett. (USA) 78~

(1),67 (2001).T. Katona, M. C. Schmidt, T. Margalith, C. Mop, H.

Tamura, H. Sato, C. Funaoka, R. Underwood, S .

Nakamura, J. S. Speck, S. P. DenBaars. Submilled Io

Physic0 Sfalu Solidi A . J u n ~003.S. Keller, S. Heikman. 1 Ben-Yaacov, L. Shen, S. P.

DenBaars, U. K. Mishra. Phys. Stat. Sol. A. 188 (Z),775 - 8 (2001).

16.1.2

3 8 6 - IEDM 0 3



H. Marchand, J. P. lbbetson, P. T. Fini, P. Kozodoy, S .

Keller, S. DeoBaars, J. S . Speck, U. K. Mishra. MR S

lntemet 1.Nitride Semi. Res. 3 (3), I (1998).A. Chimis, I. P. Zhang, V. Adivarahan, W. Shuai, J.Sun, M. Shatalov, I. W. Yang, G. Simin, M. A. Khan.Jpn. 1.Appl. Phys. 41 (48) L450-51 2002).T. Nishida. N. Kabayashi, Phys. Stat. Sol. A. 188 (I).113-6(2001).

J. P. Zhang, V. Adivarahan, H. M. W a g , Q. Fareed, E.Kuakstis, A. Chimir, M. Shatalov, I. W. Yang, G.Simin, M. A. Khan,M. Shur, R. Gaska, Jpn. J. Appl.

Phyr. 40 L9 214 (2001).

GaN.green ' .IGa N- whi t e

R

Figure 1. HBLED luminous efficiency VS.Year

Figure 2. Extemal Quanrum Effkicncy of GaN devices as a

function of wavelength. Comparison of all reported W LEDsindicate a sharp fall off n efficiency h m lue to UV.

Figure 3. SrmcNre of UV LEDs containing high AI allay s and 7QWs.

1 nm)

Figure 4. DeepW c m i s io n speccm h m aN MQ W LED

16.1.3

IEDM 03-387



3 8 8 - I E D M 0 3

GaN Based High Brightness LEDs and UV LEDs

Documents