Appl. Phys. Lett. 105, 141104 (2014); https://doi.org/10.1063/1.4897342 105, 141104

© 2014 AIP Publishing LLC.

InGaN/GaN tunnel junctions for holeinjection in GaN light emitting diodesCite as: Appl. Phys. Lett. 105, 141104 (2014); https://doi.org/10.1063/1.4897342Submitted: 12 March 2014 . Accepted: 25 September 2014 . Published Online: 06 October 2014

Sriram Krishnamoorthy, Fatih Akyol, and Siddharth Rajan

ARTICLES YOU MAY BE INTERESTED IN

Low resistance GaN/InGaN/GaN tunnel junctionsApplied Physics Letters 102, 113503 (2013); https://doi.org/10.1063/1.4796041

Interband tunneling for hole injection in III-nitride ultraviolet emittersApplied Physics Letters 106, 141103 (2015); https://doi.org/10.1063/1.4917529

Polarization-engineered GaN/InGaN/GaN tunnel diodesApplied Physics Letters 97, 203502 (2010); https://doi.org/10.1063/1.3517481

InGaN/GaN tunnel junctions for hole injection in GaN light emitting diodes

Sriram Krishnamoorthy,1,a) Fatih Akyol,1 and Siddharth Rajan1,2,a)

1Department of Electrical and Computer Engineering, The Ohio State University, Columbus, Ohio 43210, USA2Department of Materials Science and Engineering, The Ohio State University, Columbus, Ohio 43210, USA

(Received 12 March 2014; accepted 25 September 2014; published online 6 October 2014)

InGaN/GaN tunnel junction contacts were grown using plasma assisted molecular beam epitaxy (MBE)

on top of a metal-organic chemical vapor deposition (MOCVD)-grown InGaN/GaN blue (450 nm) light

emitting diode. A voltage drop of 5.3 V at 100 mA, forward resistance of 2� 10�2 X cm2, and a higher

light output power compared to the reference light emitting diodes (LED) with semi-transparent p-con-

tacts were measured in the tunnel junction LED (TJLED). A forward resistance of 5� 10�4 X cm2 was

measured in a GaN PN junction with the identical tunnel junction contact as the TJLED, grown com-

pletely by MBE. The depletion region due to the impurities at the regrowth interface between the MBE

tunnel junction and the MOCVD-grown LED was hence found to limit the forward resistance measured

in the TJLED. VC 2014 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4897342]

This paper describes the incorporation of tunnel junc-

tions (TJs) on blue light emitting diodes (LEDs) with mini-

mal electrical loss introduced by the tunnel junction. Tunnel

junctions1 in the gallium nitride material system enable

n-type tunneling contacts to p-GaN, multi-junction devices

(multiple wavelength LEDs connected in series, multi-

junction solar cells), and reverse polarization structures2

(p-down LEDs). Tunnel junctions also provide a pathway to

circumvent efficiency droop in LEDs using a cascaded LED

design run at low current density.3 However, low tunneling

resistance is a prerequisite for the incorporation of TJs in

commercial devices. Recently, tunnel junction device

designs exploiting polarization in nitrides and embedded rare

earth nitride nanoislands resulted in low resistance GaN tun-

nel junctions with resistivity as low as 10�4 X cm2.1,4–7 In

the case of the polarization1,4–6,8–11 based approach, a

pþGaN/InGaN/nþGaN structure is used where the depletion

field of the pþ/nþ GaN junction is aligned with the polariza-

tion field due to the sheet charge at InGaN/GaN interfaces.

Even with a thin InGaN layer (4 nm In0.25Ga0.75N), the band

edges can be aligned to enable inter-band tunneling.1 GdN

nanoislands embedded in GaN pþ/nþ junction provide inter-

mediate states for tunneling across the depletion region.7

With reduced tunnel barrier width for each of the two tunnel-

ing steps, a resistance of 5� 10�4 – 1� 10�3 X cm2 was

reported. Realization of such low resistance GaN-based tun-

nel junctions has generated interest in tunnel junction

enabled devices such as tunnel junction LEDs,12–15 tunnel

junction laser diodes,16 and multi-junction solar cells.17,18

One of the immediate applications of a GaN-based tun-

nel junction is to replace the high resistance p-GaN layer in

commercial LEDs with an n-type top contact. Since the

n-type layer has low spreading resistance, the metal elec-

trode coverage can be greatly minimized on the top surface

and the light can be extracted from the top surface. Such an

approach would have the advantage of avoiding the free car-

rier absorption losses in ITO current spreading layers in the

case of top-emitting thin film LEDs and the complex

fabrication processes involved in the case of flip-chip LEDs.

A tunnel junction draws electrons from the valence band of

the p-type material into the n-type material through inter-

band tunneling, and holes in the valence band on the p-side

are in turn injected into the active region.

Previous reports on tunnel junction LEDs have demon-

strated the advantage in using a low spreading resistance

n-type top contact layer. These tunnel junctions were primar-

ily based on degenerately doped PN junctions (pþ/nþ tunnel

junctions),19,20 current spreading by two dimensional elec-

tron gas,21 strained layer super lattices,22 and transparent

conducting ZnO nanoelectrodes.23 Recently, polarization-

based GaN/InGaN/GaN tunnel junctions on LEDs have also

been demonstrated using metal organic chemical vapor dep-

osition (MOCVD) technique.12–15 Although a higher output

power was measured in all the aforementioned devices, volt-

age drop across the tunnel junction was high, and lead to

high losses in LEDs. This was mainly due to the issues with

the activation of the buried p-type layers grown by MOCVD.

An ideal tunnel junction device should conduct with minimal

bias (voltage drop) across it, while the LED is in forward

operation. The on-resistance of the tunnel junction should also

be minimal. Tunnel junctions satisfying this criteria, with a re-

sistance of 10�3–10�4 X cm2 were recently demonstrated by

using molecular beam epitaxy (MBE) technique.1,7 In this

work, we demonstrate MBE grown tunnel junctions on com-

mercial LED wafers with a low forward voltage compared to

the tunnel junction LEDs reported in the literature.12,14,19,20

Furthermore, previous demonstrations of MBE-grown InGaN/

GaN TJs focused on structures along the N-polar orienta-

tion.1,4–6 In this work, polarization engineered InGaN/GaN TJ

with a low tunneling resistance (5� 10�4 X cm2) is reported

on the technologically more important Ga-polar orientation.

Three structures are compared in this work—a reference

MOCVD-grown LED (Fig. 1(a)) (REFLED), a MOCVD-

grown LED with a MBE-grown tunneling p-contact

(MOCVD LED þ MBE TJ) (Fig. 1(b)) (TJLED), and a

MBE-grown PN junction with a tunneling p-contact (MBE

PN junction þMBE TJ) (Fig. 1(c)) (TJPN). REFLED consists

of a commercial blue (450 nm) InGaN/GaN LED wafer with

thin semi-transparent Ni/Au p-type contacts. For the TJLED

a)Authors to whom correspondence should be addressed. Electronic addresses:

krishnamoorthy.13@osu.edu and rajan@ece.osu.edu.

0003-6951/2014/105(14)/141104/4/$30.00 VC 2014 AIP Publishing LLC105, 141104-1

APPLIED PHYSICS LETTERS 105, 141104 (2014)

(Fig. 1(b)), GaN/InGaN/GaN TJ-based contacts were grown

by MBE on the same commercial InGaN/GaN-based blue

LED as the reference LED. MBE regrowth on MOCVD LED

leads to interfacial impurity effects. To understand the effects

of the regrowth interface, the TJPN sample was grown com-

pletely by MBE without any interruption or regrowth.

REFLED and TJLED share the same MOCVD active region,

while the tunnel junction in TJLED and TJPN was identical.

The TJLED and TJPN samples were grown by MBE

using standard effusion cells for In, Ga, Si, and Mg, with

Veeco UNI-Bulb nitrogen plasma source. RF plasma power

of 350 W, corresponding to a nitrogen-limited growth rate of

260 nm/h, was used in this study. In the case of the TJLED

sample, magnesium delta doping was carried out to partially

compensate the positive charge (O, C, and Si impurities)24 at

the regrowth interface (Fig. 1(b)). This was followed by a

heavily doped pþ-GaN layer using nitrogen shutter pulsing

technique with a duty cycle of 33%. The pþ-GaN layer was

grown in Ga-rich condition to avoid polarity inversion, and

Ga polarity was confirmed through the absence of 3� 3

reconstructions25 in the post-growth RHEED spectrum.

Excess droplets were thermally desorbed after the pþ-GaN

layer to ensure a dry surface before InGaN growth. InGaN

was grown in In-rich conditions using Ga flux less than the

stoichiometric flux.26 By pulsing the nitrogen shutter with a

duty cycle of 50%, high doping was achieved in the nþ-GaN

(1.2� 1020 cm�3) layer. The structure was terminated with

an n-type GaN (260 nm, 2.5� 1019 cm�3) current spreading

layer grown in Ga-rich conditions. Identical growth condi-

tions were used in TJPN diode growth, to achieve identical

tunnel junction structures in both the cases. Figure 2

shows the atomic force microscope (AFM) image of the

surface of the InGaN/GaN TJLED, showing smooth sur-

face morphology with rms roughness in the range of

0.4–0.5 nm. Al (20 nm)/Ni (20 nm)/Au (200 nm) top con-

tact metal was deposited on nþ GaN by e-beam evapora-

tion on the TJPN diode and TJLED samples. Mesa isolation

was done using BCl3/Cl2 chemistry to reach the n-type GaN

bottom contact layer. Al (20 nm)/Ni (20 nm)/Au(200 nm)

contacts were deposited on the bottom n-GaN contact layer

by e-beam evaporation. The reference LED (Fig. 1(a)) was

fabricated using semi-transparent Ni (4 nm)/Au (6 nm) con-

tacts on p-GaN and Al (20 nm)/Ni (20 nm)/Au (200 nm) con-

tacts on n-type GaN. Electrical characterization was done

using a B1500 Agilent Semiconductor Device Analyzer.

Electroluminescence and on-wafer light output power meas-

urements were obtained using a calibrated Ocean Optics USB

2000 spectrometer with a coupled fiber optic cable without an

integrating sphere.

Electroluminescence measurements of the TJLED and

REFLED samples (pulsed measurement with a duty cycle of

0.1% to avoid joule heating effects) are shown in Figure 3. As

the current increased from 20 mA to 100 mA, a blue shift in

the peak emission wavelength from 443.8 nm to 442.4 nm was

measured in the REFLED (Fig. 3(a)). Electroluminescence

characteristics of the TJLED (Fig. 3(b)) were found to be sim-

ilar to that of the REFLED, with a blue shift in the peak wave-

length from 447.4 nm to 444.9 nm, corresponding to an

increase in the current from 20 mA to 100 mA. It should be

noted that the band gap of the InGaN layer used in the TJLED

FIG. 1. (a) A reference MOCVD-grown

LED (REFLED) epitaxial stack with a

semi-transparent top p-contact. (b)

Epitaxial stack of MBE-grown InGaN/

GaN tunnel junction on a MOCVD-

grown LED (TJLED). Regrowth inter-

face is marked with a red dotted line. (c)

Epitaxial stack of a MBE-grown PN

junction with a tunnel junction p-contact

(TJPN) grown completely by MBE. TJ

acts as a tunneling contact to p-GaN.

FIG. 2. AFM image of InGaN/GaN TJLED showing smooth morphology

indicating step flow growth with low rms roughness (0.4–0.5 nm).

FIG. 3. Electroluminescence characteristics of (a) REFLED and (b) InGaN/

GaN TJLED.

141104-2 Krishnamoorthy, Akyol, and Rajan Appl. Phys. Lett. 105, 141104 (2014)

would suggest absorption of blue photons. However, the

photo-generated electron-hole pair would experience a very

high electric field in the InGaN layer. As a result, the gener-

ated hole is swept back to the p-GaN layer. The photo-

generated electron is swept to the top contact, and is injected

back in to the n-GaN layer through the power supply. Thus, if

the internal quantum efficiency of the device is high, which is

the case for blue LEDs, the photo-generated carriers are re-

injected into the active region and absorption losses due to the

low bandgap InGaN tunnel barrier is not expected to be an

appreciable loss mechanism. Hence, in spite of the low band

gap of the InGaN barrier, the electroluminescence spectrum

of the TJLED does not show additional peaks compared to the

REFLED, indicating that there is no measurable absorption/

re-emission due to the tunnel junction.

On-wafer pulsed output power measurements (0.1%

duty cycle) show an increased power output from the

TJLED compared to the REFLED (Fig. 4(a)). This is sim-

ply because of less electrode coverage on the top surface of

the TJLED. Optical micrograph (Fig. 4(b)) shows excellent

current spreading in the n-GaN layer. These results are con-

sistent with the previous reports of higher output power in

TJ LEDs.12–14,19–21,23

The main metric for a TJLED is the forward voltage and

the forward resistance. Current-voltage characteristics of the

TJ LED device and the reference LED (350 lm� 350 lm) are

shown in Figure 5. Forward voltage measured in the TJLED

at 100 lA (500 lA) current is 2.625 V (2.9 V), which is

slightly larger compared to 2.5 V (2.675 V) in the REFLED.

However, after the device turn-on the forward resistance of

the TJLED is comparable to that of the REFLED. 20 mA

(100 mA) current drive, relevant for LED device operation, is

achieved at 3.9 V (5.35 V) in InGaN/GaN TJLED compared

to 3.775 V (5.975 V) in the REFLED. Forward resistance of

the TJ LED device was extracted to be 2� 10�2 X cm2.

Electrical characteristics of the TJPN diode device

(50 lm� 50 lm) are shown in Figure 6. The TJPN sample

showed rectification, with a forward voltage of 2.85 V meas-

ured at a current density of 20 A/cm2, compared to 3 V meas-

ured in a GaN PN junction27 reported in the literature. This

indicates that there is no significant voltage drop across the

tunnel junction. A total forward resistance of 5 � 10�4 X cm2

was measured in the linear region, with the top and bottom

contact resistances measured to be 1� 10�6 X cm2 and 2

� 10�6 X cm2, respectively. Hence, a tunnel junction resist-

ance of 5 � 10�4 X cm2 was extracted, which is the difference

between the total resistance and the contact resistances.

The forward voltage in the TJLED structure is the sum

of the voltage drop across the active region, the regrowth

interface, and the tunnel junction. The TJLED and the

TJPN samples had the identical tunnel junction structure.

TJPN diode was grown completely by MBE, and hence the

regrowth interface effects are eliminated. TJPN diode

showed low tunneling resistance as well as no significant

voltage drop across the tunnel junction; hence, the voltage

drop measured in the TJ LED structure can be attributed to

the regrowth interface depletion, which acts as a barrier for

hole injection.

The regrowth interface (Fig. 1(b)) is generally found to

contain oxygen, carbon, and silicon impurities, which act as

n-type dopants on the surface. Although chemical treatment

can reduce the amount of impurities, complete removal has

been a challenge, as reported previously for high electron

mobility transistors (HEMT).24 The positive impurity sheet

FIG. 4. (a) On-wafer output power of TJLED and REFLED, showing higher

output from TJLED due to less metal footprint on the top contact surface.

(b) Optical micrograph showing uniform light emission at high current den-

sity (100 mA) due to low spreading resistance n-type top layer enabled by

the tunnel junction.

FIG. 5. Linear I-V characteristics of the TJLED and the REFLED (350 lm

� 350 lm). Forward resistance of the TJLED was measured to be 2

� 10�2 X cm2. TJLED shows lowest reported voltage drop of 5.3 V at 100 mA

current drive. Inset: Log I-V characteristics of TJ LED and REFLED.

FIG. 6. Linear current voltage characteristics of Ga-polar GaN TJPN diode

showing a low tunnel junction resistance of 5 � 10�4 X cm2. Inset: Log J-V

characteristics of TJPN diode.

141104-3 Krishnamoorthy, Akyol, and Rajan Appl. Phys. Lett. 105, 141104 (2014)

charge concentration at the regrowth interface can lead to a

parasitic PN junction in series with the tunnel junction and

the LED. With higher impurity sheet charge density, the re-

sultant undesirable band bending at the regrowth interface

can be large (Fig. 7), and can cause additional voltage offset

for the TJLED device. Even with the detrimental effect of

the regrowth interface, the forward voltage measured in this

work is lower than the reported values.12,14

Regrowth interface is therefore a major limitation in uti-

lizing low resistance tunnel junctions for light emitters.

Surface treatment strategies previously demonstrated for

electronic devices28 could be adapted for regrown tunnel

junctions. Growing the entire structure without growth inter-

rupts, such as in an all-MOCVD process would naturally

eliminate the regrowth impurities. Mg out-diffusion and acti-

vation of buried p-type layers are also a challenge for inte-

gration of MOCVD tunnel junctions,12,13 and solutions to

these are under active investigation.

In summary, InGaN/GaN tunnel junctions were incor-

porated in 450 nm LEDs to realize a p-contact free LED

design with a low spreading resistance n-GaN layer as the

top contact layer. An operating voltage of 5.3 V at 100 mA

is reported. Higher output power was achieved in TJ LEDs

on account of the low metal footprint on the top surface

enabled by the low spreading resistance of the top n-type

GaN contact layer. MBE grown TJPN diode along the

Ga-polar orientation, without any regrowth interface,

resulted in a low TJ resistance of 5 � 10�4 X cm2. The

elimination of regrowth interface impurities through an all

MOCVD process, or surface treatment procedures will

allow TJLEDs to harness the full potential of polarization

engineered tunnel junctions for n-type tunneling contacts

and multi-active region structures.

The authors would like to acknowledge Professor James

Speck at the Solid State Lighting Center, University of

California, Santa Barbara, for providing the MOCVD LED

wafers used for tunnel junction regrowth in this study. We

would like to acknowledge funding from Office of Naval

Research under the DATE MURI program (Program

manager: Paul Maki, ONR N00014-11-1-0721) and the

National Science Foundation (DMR-1106177).

1S. Krishnamoorthy, F. Akyol, P. S. Park, and S. Rajan, Appl. Phys. Lett.

102, 113503 (2013).2M. L. Reed, E. D. Readinger, H. Shen, M. Wraback, A. Syrkin, A. Usikov,

O. V. Kovalenkov, and V. A. Dmitriev, Appl. Phys. Lett. 93,133505 (2008).3F. Akyol, S. Krishnamoorthy, and S. Rajan, Appl. Phys. Lett. 103, 081107

(2013).4S. Krishnamoorthy, D. N. Nath, F. Akyol, P. S. Park, M. Esposto, and S.

Rajan, Appl. Phys. Lett. 97, 203502 (2010).5S. Krishnamoorthy, P. S. Park, and S. Rajan, Appl. Phys. Lett. 99, 233504

(2011).6T. A. Growden, S. Krishnamoorthy, D. N. Nath, A. Ramesh, S. Rajan, and

P. R. Berger, in Proceedings of the 70th Annual Device ResearchConference (DRC) (2012), pp. 163–164.

7S. Krishnamoorthy, T. F. Kent, J. Yang, P. S. Park, R. C. Myers, and S.

Rajan, Nano Lett. 13, 2570–2575 (2013).8M. J. Grundmann and U. K. Mishra, Phys. Status Solidi C 4, 2830 (2007).9J. Simon, Z. Zhang, K. Goodman, H. Xing, T. Kosel, P. Fay, and D. Jena,

Phys. Rev. Lett. 103, 026801 (2009).10C. Wetzel, T. Takeuchi, H. Amano, and I. Akasaki, J. Appl. Phys. 85,

3786 (1999).11M. F. Schubert, Phys. Rev. B 81, 035303 (2010).12M. Kaga, T. Morita, Y. Kuwano, K. Yamashita, K. Yagi, M. Iwaya, T.

Takeuchi, S. Kamiyama, and I. Akasaki, Jpn. J. Appl. Phys., Part 1 52,

08JH06 (2013).13Y. Kuwano, M. Kaga, T. Morita, K. Yamashita, K. Yagi, M. Iwaya, T.

Takeuchi, S. Kamiyama, and I. Akasaki, Jpn. J. Appl. Phys., Part 1 52,

08JK12 (2013).14Z. H. Zhang, S. T. Tan, Z. Kyaw, Y. Ji, W. Liu, Z. Ju, N. Hasanov, X. Wei

Sun, and H. V. Demir, Appl. Phys. Lett. 102, 193508 (2013).15M. C. Tsai, B. Leung, T. C. Hsu, and Y. K. Kuo, J .Light Technol. 31(22),

3575 (2013).16M. X. Feng, J. P. Liu, S. M. Zhang, D. S. Jiang, Z. C. Li, K. Zhou, D. Y.

Li, L. Q. Zhang, F. Wang, H. Wang, P. Chen, Z. S. Liu, D. G. Zhao, Q.

Sun, and H. Yang, Appl. Phys. Lett. 103, 043508 (2013).17O. V. Khokhlev, K. A. Bulashevich, and S. Y. Karpov, Phys. Status Solidi

A 210(7), 1369 (2013).18J. Y. Chang, S. H. Yen, Y. Chang, B. T. Liou, and Y. K. Kuo, IEEE Trans.

Eleclectron Devices 60(12), 4140 (2013).19T. Takeuchi, G. Hasnain, S. Corzine, M. Hueschen, R. P. Schneider, Jr., C.

Kocot, M. Blomqvist, Y. Chang, D. Lefforge, M. R. Krames, L. W. Cook,

and S. A. Stockman, Jpn. J. Appl. Phys., Part 2 40, L861 (2001).20S.-R. Jeon, M. S. Cho, M.-A. Yu, and G. Mo. Yang, IEEE J. Selected Top.

Quantum Electron. 8, 739 (2002).21J. H. Lee and J. H. Lee, IEEE Electron Device Lett. 31, 455 (2010).22J. K. Sheu, J. M. Tsai, S. C. Shei, W. C. Lai, T. C. Wen, C. H. Kou, Y. K.

Su, S. J. Chang, and G. C. Chi, IEEE Electron Device Lett. 22, 460 (2001).23S. J. An, Appl. Phys. Lett. 100, 223115 (2012).24G. Koblm€uller, R. M. Chu, A. Raman, U. K. Mishra, and J. S. Speck,

J. Appl. Phys. 107(4), 043527 (2010).25A. R. Smith, R. M. Feenstra, D. W. Greve, M. S. Shin, M. Skowronski, J.

Neugebauer, and J. E. Northrup, Appl. Phys. Lett. 72, 2114 (1998).26D. N. Nath, E. G€ur, S. A. Ringel, and S. Rajan, J. Vac. Sci. Tech. B 29(2),

021206 (2011).27C. A. Hurni, O. Bierwagen, J. R. Lang, B. M. McSkimming, C. S.

Gallinat, E. C. Young, D. A. Browne, U. K. Mishra, and J. S. Speck, Appl.

Phys. Lett. 97, 222113 (2010).28S. Chowdhury, M. H. Wong, B. L. Swenson, and U. K. Mishra, IEEE

Electron Device Lett. 33(1), 41 (2012).

FIG. 7. Effect of the positive regrowth interface charge on the equilibrium

band diagram of a regrown tunnel junction on p-GaN. The higher the

regrowth interface sheet charge sheet density, higher is the barrier for holes

injected from the tunnel junction into the LED. Inset: Energy barrier for

holes at the regrowth interface.

141104-4 Krishnamoorthy, Akyol, and Rajan Appl. Phys. Lett. 105, 141104 (2014)