Appl. Phys. Lett. 103, 081107 (2013); https://doi.org/10.1063/1.4819737 103, 081107

© 2013 AIP Publishing LLC.

Tunneling-based carrier regeneration incascaded GaN light emitting diodes toovercome efficiency droopCite as: Appl. Phys. Lett. 103, 081107 (2013); https://doi.org/10.1063/1.4819737Submitted: 16 July 2013 . Accepted: 09 August 2013 . Published Online: 23 August 2013

Fatih Akyol, Sriram Krishnamoorthy, and Siddharth Rajan

ARTICLES YOU MAY BE INTERESTED IN

InGaN/GaN tunnel junctions for hole injection in GaN light emitting diodesApplied Physics Letters 105, 141104 (2014); https://doi.org/10.1063/1.4897342

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

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

Tunneling-based carrier regeneration in cascaded GaN light emittingdiodes to overcome efficiency droop

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

1Electrical and Computer Engineering, The Ohio State University, Columbus, Ohio 43210, USA2Material Science and Engineering, The Ohio State University, Columbus, Ohio 43210, USA

(Received 16 July 2013; accepted 9 August 2013; published online 23 August 2013)

In this work, we demonstrate visible wavelength transparent GaN/GdN/GaN tunnel junction

interconnects to cascade multiple p-n junctions with low resistance (5.7 � 10�4 X � cm2). We model

a device structure using cascaded light emitting diodes (LEDs) with tunnel junction-based carrier

regeneration to create high-power LEDs operating at low current density. Experimental LED

characteristics are used to model the use of these low resistance tunnel junctions in cascaded multiple

active region LEDs that can effectively provide high power by operating under low current and high

forward voltage. The adoption of cascaded LED structures can enable high power LEDs while

maintaining high efficiency. VC 2013 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4819737]

The III-nitride material system has exceptional potential

for semiconductor emitters, detectors, and photovoltaics due

to the large range of direct band gaps available in this mate-

rial system, and has already been commercialized for violet,

blue, green, white light emitting diodes (LEDs) and laser

diodes (LDs).1–5 The maximum external efficiency (EQE)

has reached 70% at 410 nm emission,6 which exceeds tradi-

tional forms of lighting such as incandescent or compact flu-

orescent, making LED or solid state lighting very promising

for future large-scale lighting applications.7 For significant

impact on worldwide lighting energy consumption, LEDs

would need to be operated at very high output power density

while maintaining high efficiency.7 However, while the peak

efficiency values are high, the efficiency of emitters

decreases dramatically as the input power increases. This ef-

ficiency droop impacts the efficiency and affordability of

solid state lighting, preventing large-scale adoption of LED

lighting technology. The origin of efficiency droop is still

being debated, and the phenomenon is attributed to various

causes including Auger recombination processes, electron

overflow, and inefficient hole injection and transport

mechanisms.8–14 While many elegant solutions have been

proposed to overcome efficiency droop,8–12,14,15 none has

been completely successful until now, and it may be that the

inherent properties of the III-nitride material system will

preclude complete elimination of this effect.

The underlying cause of efficiency droop is that for

higher operating power density, LEDs are driven at higher

currents. In this paper, we suggest a solution to the efficiency

droop problem by proposing cascaded structures where high

optical output power can be obtained at low current density.

We show that it is possible to cascade multiple p-n junctions

using visible transparent tunnel junctions (TJs) with low re-

sistance and negligible voltage drop. Based on performance

parameters of current commercial LEDs, we show that the

efficiency droop problem can be almost completely circum-

vented by using tunneling-based carrier regeneration in

cascaded multiple active region LED structures. Higher

luminous output power can be achieved by increasing the

operating voltage, rather than the current, leading to unprece-

dented power density and efficiency for III-nitride LEDs.

To demonstrate the feasibility of cascading GaN emit-

ters, we designed an experiment using multiple (1, 2, and 4)

p-n junctions connected epitaxially in series (Fig. 1) with

recently demonstrated GdN-based visible transparent tunnel

junctions.16,17 The top and bottom of this structure are both

n-type since the tunnel junction eliminates the need for a top

p-contact. The energy band diagrams of this structure at

equilibrium and forward bias are shown in Figs. 2(a) and

2(b), respectively. As the p-n junction is forward biased, the

tunnel junction gets reverse biased, electrons in the valence

band of pþ GaN layer of the tunnel junction tunnel into

empty states available in the nþ GaN layer, leaving behind a

hole in the pþ GaN layer. These electrons and holes regener-

ated at pþ and nþ side of the tunnel junctions are injected

into the active region which in this case is a p-n junction. In

the case of a multi-quantum well active region LED, the

injected hole and electron would recombine radiatively in

the quantum wells.

The epitaxial structure (Fig. 1) was grown by plasma

assisted molecular beam epitaxy on free-standing n-type GaN

substrates, and consisted of repeats (N¼ 1, 2, and 4) of p-GaN/

n-GaN junction diodes, and pþGaN/GdN nano-island/nþGaN

tunnel junctions. Further details of GdN tunnel junctions,17

FIG. 1. Epitaxial design of the cascaded p-n junctions.a)Electronic mail: rajan@ece.osu.edu. Tel.: 614-247-7922.

0003-6951/2013/103(8)/081107/4/$30.00 VC 2013 AIP Publishing LLC103, 081107-1

APPLIED PHYSICS LETTERS 103, 081107 (2013)

and the growth dynamics of GdN nano-islands are discussed

elsewhere.16 The device was processed using contact lithogra-

phy, mesa isolation was done using Cl2/BCl3 etching, and

ohmic contacts (to top and bottom n-type regions) were made

using Al/Au. Electrical characteristics of 100 lm2 diodes of

the cascaded p-n junction stack (number of p-n junctions

N¼ 1, 2, and 4) are shown in Figs. 3(a) and 3(b). The turn-on

voltage of the cascaded device increased with the number of

p-n junctions stacked epitaxially. All the structures showed

rectifying behavior, although the device with 4 junctions does

have higher leakage current. This effect needs further investi-

gation, and optimization of the growth conditions is expected

to mitigate these effects. The series resistance (estimated from

the linear portion of the IV curve) was found to scale with N,

and based on this, we estimate �5.7� 10�4 X � cm2 resistiv-

ity for each tunnel junction (inset of Fig. 3(b)). This resistance

would result in relatively low voltage drop of 57 mV for a cur-

rent density of 100 A/cm2. This report of cascaded III-nitride

p-n junctions with low series resistance could enable several

devices including multi-junction solar cells, photodetectors,

and multiple active region LEDs.

Since our experiment demonstrates the feasibility of cas-

cading active regions, we can use the measured results to

model cascaded multiple active region LEDs. To estimate

the efficiency of cascaded LED structures, we combine our

estimates of tunneling resistance with the experimental data

from the commercial LED published elsewhere so that non-

idealities of the active region are taken into account.18 The

modeled reference LED emits at 470 nm and consists of a

1.02 mm2 mesa area with ZnO layer as contact to p-GaN.18

The differential series resistance (Rs) of the reported LED

was �0.02 X� cm2.18 We fit the experimental data to the

diode equation

I ¼ C x expq x ðV � I x RsÞ

nkT

� �; (1)

which is modified for the case of “N” identical LEDs cas-

caded to

I ¼ C x expq x ðVN � I x RsÞ

nkT

� �; (2)

where “N” is total number of the cascaded junctions and “n”

is the ideality factor of a single junction device. The series

resistance of the tunnel junctions in the structure is taken

into account in Rs term. The tunnel junction resistance

used for these calculations is the experimental value of

RsðTJÞ ¼ 6:4� 10�4 X cm2 which was the highest measured

resistance value for a single TJ in the cascaded p-n junctions

discussed above. We note that lower series resistance would

be achieved with further optimization of the tunnel junctions.

The modeled I-V curves (Fig. 4(a)) for the cascaded structure

(N¼ 5, 20, 50) behave as expected. The input voltage at

turn-on increases as the number of junctions is increased.

The external quantum efficiency was modeled by curve fit-

ting the experimental data using ABC model.19 As shown in

Fig. 5(b), the model fit the measured data accurately. We note

here that this model was used as an empirical fit, rather than to

convey any physical information. Combining this data with the

I-V characteristics of the reference LED, the wall plug effi-

ciency (WPE) as a function of current for single and cascaded

LEDs was obtained (Fig. 5(a)). We can now estimate the output

power, efficiency, and loss for cascaded LEDs including the

active region and tunnel junction losses.

FIG. 2. Energy band diagram of a cascaded p-n junction under (a) equilib-

rium (b) forward bias showing tunneling of electrons from p GaN to n GaN

through GdN nano-islands.

FIG. 3. The I-V characteristic of

the cascaded p-n junctions with N¼ 1,

3, and 4 (a) in linear scale (b) semi-log

scale. The inset to (b) shows the

change in series resistance with the

number of p-n junctions (N).

081107-2 Akyol, Krishnamoorthy, and Rajan Appl. Phys. Lett. 103, 081107 (2013)

In Fig. 5(a), the output power characteristics as a func-

tion of input power for cascaded LEDs and the conventional

LED are compared. The input power was calculated using

I-V characteristics as shown in Fig. 4(a). To estimate the

optical output power, we assume that a structure with N

cascaded stages has EQEN¼N�EQE(single LED). Although

absorption loss via free-carriers and InGaN QW layers are

expected to increase as “N” increases, these are ignored

here. As expected, the conventional LED output power

increases and saturates below 1 W under 10 W input power.

At the same input power level, relatively lower currents and

higher voltages are needed for the cascaded LED structures

(Fig. 4(a)). As the number of cascade regions is increased,

the saturation is pushed out further, with negligible satura-

tion in output power for the LEDs with N¼ 20 and N¼ 50

cases up to 3.5 W.

These results are more striking when shown as effi-

ciency versus power curves (Figs. 5(b) and 5(c)). From these

curves, it can be seen that for a cascaded structure with Nstages, the maximum efficiency point occurs at approxi-

mately N times higher power than a single LED. This implies

that the maximum device efficiency can be obtained at much

higher drive powers even with an LED that exhibits signifi-

cant droop. Under 10 W input power, modeled conventional

LED shows 83% droop and this value decreases as N

increases. For the LED with N¼ 50, the droop is 14% at the

same input power which is an enhancement of �420% in

WPE compared to conventional case.

The cascaded LED structure not only provides an

enhancement of EQE but also suppresses joule heating. The

total power dissipated due to series resistance of an LED is

PR ¼ I2x RS. For a given output power, Rs / N and I / 1/N,

and therefore, power dissipated is proportional to 1/N, which

implies that increasing N enables high power operation of

LEDs with higher efficiency while reducing overall heat gen-

eration. Resistive power dissipation as a function of input

power is shown in Fig. 4(b) for the modeled reference and

cascaded LEDs with N¼ 5, 20, and 50 revealing suppression

of Joule heating in cascaded LEDs by over an order of mag-

nitude over single active region LEDs, at the same output

power. This not only enhances efficiency but also can enable

LED operation at elevated output power, and could eliminate

thermal management issues.

While the proof of concept experiment and modeling for

cascaded LEDs provides a solution to the long standing effi-

ciency droop problem, there are some technological chal-

lenges that need to be overcome. First, GaN-based tunnel

junctions based on MBE22–24 and metalorganic chemical

vapor deposition (MOCVD) growth20,21,28 are still relatively

less efficient than the results described here and reported pre-

viously using MBE growth.24–27 Another challenge is the

activation of buried Mg-atoms (p-type dopant) in MOCVD/

MOVPE growth,20,21,28 which has been addressed previously

using sidewall activation.29 Finally, re-absorption of photons

for LEDs with a large number of active regions, and exces-

sive heating due to un-extracted photons could both reduce

the efficacy of these LEDs. Further work is required to

address these challenges.

In conclusion, low resistance visible transparent GdN

tunnel junctions can enable cascaded multiple GaN p-n junc-

tion diodes. I-V measurements showed the expected shift in

diode turn on voltage as N increases up to 4 junctions. The

average resistance of each tunnel junction was measured as

5.7� 10�4 X� cm2. We show that cascaded p-n junctions

could circumvent efficiency droop of nitride LEDs through

identical active regions with LEDs operating under high volt-

age rather than current. Our calculations show that maximum

efficiency point can be pushed towards higher input power re-

gime as N increases, and cascading LEDs decrease joule

heating by operating at lower current regime. The use of tun-

nel junction cascaded LEDs could lead to unprecedented high

FIG. 4. (a) I-V characteristic and (a) the change in Joule Heating as a

function of input power of the modeled reference single junction LED and

cascaded LEDs with N¼ 5, 20, and 50.

FIG. 5. The change in (a) output power (b) WPE in logarithmic scale (c) WPE in linear scale of the modeled reference single junction LED and cascaded

LEDs with N¼ 5, 20, and 50. The experimental data from reference 18 and the modeled fit is also shown in Fig. 5(b).

081107-3 Akyol, Krishnamoorthy, and Rajan Appl. Phys. Lett. 103, 081107 (2013)

power emitters for various applications including lighting,

display, and lasers.

We acknowledge funding from NSF DMR 1106177 and

Office of Naval Research (ONR N00014-11-1-0721

(Dr. Paul A. Maki), and the OSU Wright Center for High

Performance Power Electronics (CHPPE).

1H. Kume, “Cree Sets New Record for White LED Efficiency,” Tech-On, 23

April 2012.2S. Nakamura, M. Senoh, S. Nagahama, N. Iwasa, T. Yamada, T.

Matsushita, Y. Sugimoto and H. Kiyoku, Appl. Phys. Lett. 69, 4056 (1996).3S. Nakamura, M. Senoh, N. Iwasa, S. Nagahama, T. Yamada, and

T. Mukai, Jpn. J. Appl. Phys. 34, L1332 (1995).4T. Mukai, H. Marimatsu, and S. Nakamura, Jpn. J. Appl. Phys. 37, L479

(1998).5S. Nakamura and G. Fasol, The Blue Laser Diode (Springer-Verlag,

Heidelberg, 1997).6M. J. Cich, R. I. Aldaz, A. Chakraborty, A. David, M. J. Grundman, A.

Tyagi, M. Zhang, F. M. Steranka, and M. R. Krames, Appl. Phys. Lett.

101, 223509 (2012).7US DOE Solid state lightening research and development multi-year pro-

gram plan, April 2013.8Y. C. Shen, G. O. Mueller, S. Wanabe, N. F. Gardner, A. Munkholm, and

M. R. Krames, Appl. Phys. Lett. 91, 141101 (2007).9H. Pyu, H. Kim, and J. Shim, Appl. Phys. Lett. 95, 081114 (2009).

10M. F. Schubert, S. Chhajed, J. K. Kim, E. F. Schubert, D. D. Koleske, and

M. H. Crawford, Appl. Phys. Lett. 91, 231114 (2007).11X. Ni, Q. Fan, R. Shimada, U. Ozgur, and H. Markoc, Appl. Phys. Lett.

93, 171113 (2008).12C. H. Wang, C. C. Ke, C. Y. Lee, S. P. Chang, J. C. Li, Z. Y. Li, H. C. Yang,

H. C. Kuo, T. C. Lu, and C. Wang, Appl. Phys. Lett. 97, 261103 (2010).

13J. Iveland, L. Martinelli, J. Peretti, and J. S. Speck, Phys. Rev. Lett. 110,

177406 (2013).14M. H. Kim, M. F. Schubert, Q. Dai, J. K. Kim, E. F. Schubert, J. Piprek,

and Y. Park, Appl. Phys. Lett. 91, 183507 (2007).15F. Akyol, D. N. Nath, S. Krishnamoorthy, P. S. Park, and S. Rajan, Appl.

Phys. Lett. 100, 111118 (2012).16T. F. Kent, J. Yang, L. Yang, M. J. Mills, and R. C. Myers, Appl. Phys.

Lett. 100, 152111 (2012).17S. Krishnamoorthy, T. F. Kent, J. Yang, P. S. Park, R. C. Myers, and S.

Rajan, Nano Lett. 13, 2570–2575 (2013).18A. H. Reading, J. J. Richardson, C. C. Pan, S. Nakamura, and S. P.

Denbaars, Opt. Express 20, A13 (2012).19G.-B. Lin, D. Meyaard, J. Cho, E. F. Schubert, H. Shim, and C. Sone,

Appl. Phys. Lett. 100, 161106 (2012).20T. Takeuchi, G. Hasnain, S. Corzine, M. Hueschen, R. P. Schneider, 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).21S.-R. Jeon, Y.-H. Song, H.-J. Jang, G. M. Yang, S. W. Hwang, and S. J.

Son, Appl. Phys. Lett. 78, 3265 (2001).22M. J. Grundmann and U. K. Mishra, Phys. Status Solidi C 4, 2830 (2007).23J. Simon, Z. Zhang, K. Goodman, H. Xing, T. Kosel, P. Fay, and D. Jena,

Phys. Rev. Lett. 103, 026801 (2009).24M. F. Shubert, Phys. Rev. B 81, 035303 (2010).25S. Krishnamoorthy, D. N. Nath, F. Akyol, P. S. Park, M. Esposto, and S.

Rajan, Appl. Phys. Lett. 97, 203502 (2010).26S. Krishnamoorthy, P. S. Park, and S. Rajan, Appl. Phys. Lett. 99, 233504

(2011).27S. Krishnamoorthy, F. Akyol, P. S. Park, and S. Rajan, Appl. Phys. Lett.

102, 113503 (2013).28Z.-H. Zhang, S. T. Tan, Z. Kyaw, Y. Ji, W. Liu, Z. Ju, N. Hasanov, X. W.

Sun, and H. V. Demir, Appl. Phys. Lett. 102, 193508 (2013).29Y. Kuwano, M. Kaga, T. Morita, K. Yamashita, T. Takeuchi, M. Iwaya, S.

Kamiyama, and I. Akasaki, paper presented at the International Workshop

on Nitride Semiconductors (IWN), Sapporo, Japan, 2012.

081107-4 Akyol, Krishnamoorthy, and Rajan Appl. Phys. Lett. 103, 081107 (2013)