This content has been downloaded from IOPscience. Please scroll down to see the full text.
Download details:
IP Address: 140.113.38.11
This content was downloaded on 24/04/2014 at 23:46
Please note that terms and conditions apply.
Improvement of quantum efficiency in green light-emitting diodes with pre-TMIn flow treatment
View the table of contents for this issue, or go to the journal homepage for more 2011 J. Phys. D: Appl. Phys. 44 224015
(http://iopscience.iop.org/0022-3727/44/22/224015)
J. Phys. D: Appl. Phys. 44 (2011) 224015 (6pp) doi:10.1088/0022-3727/44/22/224015
Improvement of quantum efficiency in
green light-emitting diodes with
pre-TMIn flow treatment
Ya-Ju Lee
1, Yi-Ching Chen
1and Tien-Chang Lu
21Institute of Electro-Optical Science and Technology, National Taiwan Normal University, 88, Sec.4, Ting-Chou Road, Taipei 116, Taiwan
2Department of Photonics, National Chiao Tung University, 1001 Ta Hsueh Road, Hsinchu, Taiwan E-mail:[email protected]
Received 29 November 2010, in final form 28 January 2011 Published 13 May 2011
Online atstacks.iop.org/JPhysD/44/224015 Abstract
The effects of pre-trimethlyindium (TMIn) flow on the improved electrical characteristics and highly stable temperature properties of InGaN green light-emitting diodes (LEDs) are discussed. For the LED sample with a pre-TMIn flow treatment, the tunnelling of injected carriers associated with threading defects is significantly reduced, which promotes the diffusion–recombination of injected carriers, as well as the overall emission efficiency of the LED. In addition, the pre-TMIn flow treatment evidently reduces the dependence of external quantum efficiency on temperature and efficiency droop of green LEDs. As a result, we conclude that the pre-TMIn flow treatment is a promising scheme for the improvement of output performance of InGaN-based green LEDs.
(Some figures in this article are in colour only in the electronic version)
1. Introduction
Light-emitting diodes (LEDs) have attracted considerable interest as candidates for next-generation lighting because they promise to reduce energy consumption enormously. LEDs directly convert electrical energy into light, and over the years their efficiency has rapidly increased to become comparable to that of fluorescent lamps [1]. In particular, the development of InGaN alloys is of primary concern in the production of LEDs that are highly efficient right across the visible spectrum. The state-of-the-art InGaN-based high-power blue LEDs can deliver an external quantum efficiency (EQE) as high as 56% [2]. Yet, the peak EQE of green LEDs is significantly lower than that of InGaN-based blue and AlGaInP-based red LEDs, which is characteristically called the ‘green gap’ [3]. Additionally, the efficiency of green LEDs falls more rapidly than that of blue LEDs as the injected current increases [4,5]. Therefore, to harness the full potential of LEDs for next-generation lighting, further increases in the quantum efficiency of InGaN-based green LEDs are required. Generally, to engineer a narrow bandgap suited to emission wavelengths from λ= 500 nm to λ = 570 nm, high indium compositions
of up to 30% in the quantum wells (QWs) of LEDs are essential. Such high indium compositions are associated with a complex array of properties, which can adversely affect the crystalline quality and output performance of LEDs [6]. To overcome these issues, a few studies have considered improving the optical properties and the interface abruptness of InGaN-based green LEDs by supplying a flow of trimethlyindium (TMIn) prior to the growth of the QWs [7–9]. Nevertheless, the influences of such pre-TMIn flow treatment on the output performance of LEDs have rarely been discussed. In this paper, the experimental results on the improved performance and highly stable temperature characteristics of an InGaN green LED with a pre-TMIn flow treatment are presented. In the proposed scheme, several essential properties of the LED, including the electrical characteristics and the efficiency droop, are significantly improved. The potential physical causes of the improved performance are also discussed.
2. Experiments
Figure 1(a) schematically depicts the green LED structure. The green LED was grown on c-plane (0 0 0 1) sapphire
J. Phys. D: Appl. Phys. 44 (2011) 224015 Y-J Lee et al
Figure 1. (a) Schematics of the InGaN-based green LED with pre-TMIn flow treatment during the MQWs’ growth. (b) Time chart of the reactant sources in MQWs’ growth. Inset: photograph of the green LED with pre-TMIn flow treatment taken at I = 20 mA.
substrates by low-pressure metal-organic chemical vapour deposition. Following the growth of a GaN nucleation layer at 520◦C, a 2 µm-thick undoped GaN layer and a 2.5 µm-thick Si-doped n-type (5× 1018cm−3) GaN cladding layer were
grown at 1050◦C. The active region composed of six periods of a nominally undoped 3 nm-thick In0.3GaN0.7N/14 nm-thick
GaN multiple quantum wells (MQWs) was grown at 750◦C. The growth temperatures for the InGaN well and GaN barrier were 750◦C and 900◦C, respectively. Figure1(b) plots the time chart of the reactant sources for the growth of MQWs. An interruption between the growths of well and barrier layers was used to stabilize the reactor temperature and gas flows, as shown in figure 1(b). During the interruption of growth, a TMIn flow of 60 sccm was introduced into the reactor prior to the growth of an InGaN well for 10 s. Furthermore, during the growth temperature ramping (from 750◦C to 900◦C), a 2 nm-thick GaN capping layer was grown on the well layer to prevent the decomposition of the InGaN well layer. Finally, a 15 nm-thick Mg-doped (1× 1018cm−3)
Al0.15Ga0.85N electron-blocking layer (EBL) and a 300
nm-thick p-type (1× 1018cm−3)GaN cladding layer were grown
on top of the QWs. The grown wafer was then fabricated into LED chips with a mesa size of 300 × 300 µm2 by
a standard photolithographic process, which were used for subsequent measurements. A 200 nm-thick indium tin oxide layer was deposited as a current spreading layer and Cr/Au was then deposited as n and p electrodes. The photograph of the green LED with the pre-TMIn flow treatment taken at I = 20 mA is shown in the inset of figure 6(b). For comparison, a control LED sample obtained by not introducing the TMIn flow prior to the well layers was also grown and fabricated. In this study, the optical properties of the green LED were measured using a TRIAX 190 spectrometer. The LED sample was mounted on the copper block of a cold cryostat to vary the ambient temperature over the range 77–400 K. The LED sample was biased using a ILX LDP-3811 pulsed current source with a pulse width of 6.5 µs and a duty cycle of 0.1%.
3. Results and discussion
Figure 2 shows the cross-sectional transmission electron microscopic (TEM) images of the MQWs for LED samples (a) without and (b) with pre-TMIn flow treatment, respectively. The boundaries of the InGaN/GaN QWs in both samples are clear and well defined, suggesting that the pre-TMIn flow treatment slightly affects interface abruptness. Figures2(c) and (d) present the high-resolution TEM images of the LED samples without and with pre-TMIn flow treatment, respectively. Clearly, indium atoms are segregated along the (0 0 0 1) growth direction in both samples (blurred dark stripe), suggesting that the desorption of indium and its subsequent penetration into the GaN barrier layer remain in the sample that underwent the pre-TMIn flow treatment. Figure 3 plots the atomic force microscopy (AFM) images of the LED samples (a) without and (b) with pre-TMIn flow treatment, respectively. Both LED samples were etched with phosphoric acid to form etch pits on their surfaces [10]. The densities of etch pit estimated from figures3(a) and (b) were 5.7× 109cm2 and 7.5× 109cm2, respectively. The slightly higher density of the threading dislocation in the sample that underwent the pre-TMIn flow treatment could be due to its higher indium content in the QW compared with the conventional LED.
Next, we discuss the temperature dependence of the electrical characteristics of the LED samples. Figure4displays the forward-bias I –V curves on a semi-log scale of the LED samples (a) without and (b) with pre-TMIn flow treatment for temperatures ranging from 77 to 400 K. Generally, the
I–V curves of both samples at all temperatures exhibit the typical current rectification of InGaN-based LEDs. As shown in figures4(a) and (b), IPs of both LED samples are insensitive
to temperature and have similar values (IP ∼ 10−11A). This
can be attributed to the fact that the leakage paths of the injected carriers that bypass the junction region of the LED are barely changed or influenced by varying the ambient temperature of the LEDs. More importantly, almost equal values of IPfor both
LED samples suggest that the fabrication of LEDs is stable and reproducible. To further examine the I –V characteristics of 2
Figure 2. Bright-field TEM images of InGaN/GaN MQWs grown (a) without and (b) with pre-TMIn flow treatment. Enlarged TEM images of samples (c) without and (d) with pre-TMIn flow treatment.
Figure 3. AFM images of the surfaces of LED samples prepared (a) without and (b) with pre-TMIn flow treatment. the samples, the turn-on voltage (Von)and the series resistance
(RS)of both samples for all temperatures are extracted and
presented in figure 4(c). Clearly, the values of Von and RS
for both samples slowly decline as the temperature increases.
Von of both samples decreases from 4.5 V at 77 K to 2.5 V at 400 K, mainly because of a shrinkage of the energy bandgap of (In)GaN material as the temperature increases, reducing
the magnitude of the potential barrier for injected carriers to surpass and prompting the injected carriers to flow into the active region at elevated temperatures [11]. Similarly, RS
of both samples decreases from 32 to 15 with the same change in temperature, mainly because of the improvement in the p-type transport properties as the temperature increases, promoting the activation of acceptors at elevated temperatures.
J. Phys. D: Appl. Phys. 44 (2011) 224015 Y-J Lee et al
Figure 4. ln|I| versus V for the sample (a) without and (b) with pre-TMIn flow treatment from T = 77 K to T = 400 K. (c) Turn-on voltage (Von)and series resistance (RS)versus temperature for both LED samples.
Moreover, as shown in figure 4(c), at a given temperature, the values of Von and RS for both samples are similar,
suggesting that the pre-TMIn flow treatment barely affected the transport property of injected carriers under forward bias injection.
In general, the I –V curve of the InGaN-based LED is well described by the Shockley equation, which takes into account the defect-assisted leakage and diffusion– recombination currents, as well as the parasitic effect [12–15].
Accordingly, the total injected current of LEDs, I , is given by
I = IP+ IJ= IP+ Ileakage+ Idiffusion, (1)
where IP represents the current that bypasses the junction
region and is given by IP= (V − IRS)/RP, and RPdenotes the
parallel resistance of the LED. IJis the current that flows across
the junction, and comprises the defect-assisted leakage current,
Ileakage, and the diffusion–recombination current, Idiffusion.
Both Idiffusionand Ileakageare given by
Ileakage= IO1 exp e(V− IRS) n1kT − 1 , (2) Idiffusion= IO2 exp e(V − IRS) n2kT − 1 , (3) where IO1 and IO2 are saturation currents produced by
the defect-assisted leakage and the diffusion–recombination processes, respectively, n1 and n2 are the ideality factors for
the defect-assisted leakage and the diffusion–recombination processes, respectively. Substituting equations (2) and (3) into equation (1) and considering IP = (V − IRS)/RP, the I –V
curve of an InGaN LED can be rewritten as
I =(V − IRS) RP + IO1 exp e(V− IRS) n1kT − 1 + IO2 exp e(V− IRS) n2kT − 1 . (4)
As compared with the I –V curve of the LED with pre-TMIn flow treatment shown in figure4(b), an obvious defect-assisted leakage process is observed in the LED without the pre-TMIn flow treatment from T = 77 K to T = 250 K (figure4(a)). Such a defect-assisted leakage process becomes less significant as the temperature is further increased to 400 K. In contrast, no defect-assisted leakage process is observed in the LED with the pre-TMIn flow treatment at any temperature. Researchers have reported that indium atoms can preferentially bond at the core of a threading dislocation, pinning it, and thus preventing gliding during growth [16]. Therefore, when pre-TMIn flow is supplied, the tunnelling of injected carriers that are associated with threading defects is significantly reduced, prohibiting the recombination of injected carriers via the defect-assisted leakage process. As a result, for the LED with the pre-TMIn flow treatment, it is mainly the diffusion–recombination current that flows through the depletion region and dominates the LED. Notably, although the I –V characteristics of both LED samples at and above room temperature are very similar, the influence of threading defects on the sample that underwent the pre-TMIn flow treatment is reduced, hence improving its internal quantum efficiency, which leads to a higher light output power compared with the conventional LED.
We theoretically fitted the I –V curves plotted in figures 4(a) and (b) to obtain the saturation currents (IO1
and IO2) of both LED samples using equation (4), where
the ideality factor and series resistance of both LED samples were determined using nideal = (q/kT )(∂ ln I/∂V )−1[17] and
RS= ∂V /∂I, respectively. In addition, the parallel resistances
of both LED samples are set to be constants (RP ∼ 150 G)
for all temperatures, because IPs of both LED samples are
Figure 5. IO1, IO2, n1and n2versus temperature for LED samples (a) without and (b) with pre-TMIn flow treatment.
Figure 6. (a) Light output power versus injected current for both LED samples at T = 300 K. Inset: EL spectra with peak wavelengths. (b) EQE versus injected current for both LED samples from T = 77 to T = 400 K.
insensitive to temperature and have similar values as previously mentioned. The fitted IO1and IO2, and measured n1, and n2
as functions of temperature for both LED samples without and with pre-TMIn flow treatment are plotted in figures5(a) and (b), respectively. In figure 5(a), at low temperatures (77 K < T < 250 K), the saturation current of the defect-assisted leakage process (IO1)is remarkably larger than that
of the diffusion–recombination process (IO2), suggesting that
most of the injected carriers recombine initially at the deep-level traps that are associated with dislocation defects, without any diffusion–recombination process. However, the difference in magnitude of saturation current between the defect-assisted leakage and the diffusion–recombination processes becomes less significant as the temperature increases, indicating that the diffusion–recombination current eventually dominates the LED at high temperatures. Figure5(a) also suggests that the saturation current is an important index of the domination of the recombination of injected carriers in an InGaN-based LED. Also, the ideality factor is strongly temperature-dependent for both the defect-assisted leakage (n1)and the diffusion–
recombination (n2) processes. Both n1 and n2 decrease as
the temperature is increased. At a given temperature, n1 is
generally larger than n2, causing the defect-assisted leakage
current to be lower than the diffusion–recombination current,
although IO1is significantly larger than IO2. In stark contrast,
as shown in figure 5(b), one major recombination process of injected carriers dominates the I –V curve of the LED with pre-TMIn flow treatment. The calculated values of saturation current (IO2)and the ideality factor (n2)of the LED
with the pre-TMIn flow treatment are similar to those of the LED without the pre-TMIn flow treatment. Restated, this suggests that when pre-TMIn flow is supplied, the influence of threading defects on the injected carriers is significantly reduced, promoting the diffusion–recombination of injected carriers, as well as the overall emission efficiency of the LED. Finally, the temperature dependence of EQE of the LED samples is considered. Figure6(a) shows the output power as a function of injected current for both LED samples of interest at T = 300 K. Clearly, the output power of the sample that underwent the pre-TMIn flow treatment is higher than that of the control sample. On average, the output power is enhanced by 9.5%. The inset in figure6(a) plots the EL spectra for both samples for injected current I = 20 mA. Significantly, a stronger EL intensity with an additional peak is obtained at a longer wavelength (λ= 522 nm) from the LED sample with the pre-TMIn flow treatment, and could be due to its localized higher indium content in the QW compared with the conventional sample. Figure 6(b) plots the EQE
J. Phys. D: Appl. Phys. 44 (2011) 224015 Y-J Lee et al
as a function of injected current for both LED samples from
T = 77 K to T = 400 K. The maximum EQE is normalized
in figure6(b). The maximum EQE of both LED samples is observed to decrease with increasing temperature. In addition, for the LED without the pre-TMIn flow treatment, its EQE drops rapidly at low temperatures and the efficiency-droop behaviour is mitigated as the temperature increases. The rapid efficiency droop at low temperatures could originate from the low mobility and non-uniform distribution of injected holes in the LED [18,19]. Therefore, with the increase in injected current, holes are confined in the QWs closest to the p-side and electrons exhibit serious leakage into the p-type GaN, resulting in constrained recombination of injected carriers in this QW [20]. As a result, a severe efficiency-droop behaviour is observed in LEDs without the pre-TMIn flow treatment at low temperatures. In general, the temperature-dependent efficiency droop of the LED with the pre-TMIn flow treatment follows a similar behaviour, as shown in figure6(b). However, the EQE values of the sample that underwent the pre-TMIn flow treatment declines slightly as the temperature increases, demonstrating that the characteristics of this sample are more thermally stable. Studies have suggested that the injected carriers in the localized high indium content regions function excellently as radiative recombination centres, effectively suppressing the overflow of injected carriers outside the MQWs at elevated temperatures. This fact indicates that the pre-TMIn treatment can substantially reduce the dependence of EQE on temperature. In addition, the energy band profile was altered due to the incorporation of localized high indium content in QWs, which may reduce the influence of the quantum-confined stark effect. As a result, the overlap of electron–hole wave functions was significantly enhanced, also alleviating the efficiency droop [21,22].
4. Summary
In summary, a pre-TMIn flow treatment was incorporated into an InGaN-based green LED to minimize the temperature dependence of EQE and alleviate the efficiency droop. For the LED sample with the pre-TMIn flow treatment, the tunnelling of injected carriers associated with threading defects is significantly reduced, promoting the diffusion–recombination of injected carriers, as well as the overall emission efficiency of the LED. Moreover, the presented device exhibits less efficiency droop at all measured temperatures. We therefore conclude that the pre-TMIn flow treatment is a promising scheme for the improvement of output performance of InGaN-based green LEDs and bridging the ‘green gap’ in solid-state lighting.
Acknowledgments
The authors gratefully acknowledge the financial support from the National Science Council of Republic of China (ROC) in Taiwan under contract No NSC–98–2112–M–003–001–MY2.
References
[1] Narukawa Y, Sano M, Ichikawa M, Minato S, Sakamoto T, Yamada T and Mukai T 2007 Japan. J. Appl. Phys.
46 L963
[2] Krames M R, Shchekin O B, Mueller-Mach R, Mueller G O, Zhou L, Harbers G and Craford M G 2007 J. Disp. Technol.
3 160
[3] Crawford M H 2009 IEEE J. Sel. Top. Quantum Electron.
15 1028
[4] Mukai T, Yamada M and Nakamura S 1999 Japan. J. Appl. Phys.38 3976–81
[5] Lee Y J, Chen C H and Lee C J 2010 IEEE Photon. Technol. Lett.22 1506
[6] Chichibu S F et al 2006 Nature Mater.5 810
[7] Liu J P, Jin R Q, Zhu J J, Zhang J C, Wang J F, Wu M, Chen J, Wang Y T and Yang H 2004 J. Cryst. Growth264 53
[8] Leem S-J, Kim M H, Shin J, Choi Y and Jeong J 2001 Japan. J. Appl. Phys.40 L371
[9] Kumar M S, Park J Y, Lee Y S, Chung S J, Hong C-H and Suh E-K 2008 Japan. J. Appl. Phys. 47 839
[10] Chen J, Wang J F, Wang H, Zhu J J, Zhang S M, Zhao D G, Jiang D S, Yang H, Jahn U and Ploog K H 2006 Semicond. Sci. Technol.21 1229
[11] Li Y, Zhao W, Xia Y, Zhu M, Senawiratne J, Detchprohm T, Schubert E F and Wetzel C 2007 Phys. Status Solidi c
4 2784
[12] Cao X A, LeBoeuf S F, Kim K H, Sandvik P M, Stokes E B, Ebong A, Walker D, Kretchmer J, Lin J Y and Jiang H X 2002 Solid-State Electron.46 2291
[13] Cao X A, Teetsov J M, D’Evelyn M P, Merfeld D W and Yan C H 2004 Appl. Phys. Lett.85 7
[14] Shah J M, Li Y L, Gessmann Th and Schubert E F 2003 J. Appl. Phys.94 2627
[15] Lee S W et al 2006 Appl. Phys. Lett.89 132117
[16] Yamaguchi S, Kariya M, Nitta S, Amano H and Akasaki I 2000 Japan. J. Appl. Phys.39 2385
[17] Zhu D, Xu J, Noemaun A N, Kim J K, Schubert E F, Crawford M H and Koleske D D 2009 Appl. Phys. Lett.
94 081113
[18] Cao X A and LeBoeuf S F 2007 IEEE Trans. Electron Devices
54 3414
[19] Wang C H, Chen J R, Chiu C H, Kuo H C, Lee Y L, Lu T C and Wang S C 2010 IEEE Photon. Technol. Lett.22 236–8
[20] Chang J Y, Tsai M C and Kuo Y K 2010 Opt. Lett.
35 1368–70
[21] Park S-H, Ahn D and Kim J-W 2009 Appl. Phys. Lett.
94 041109
[22] Zhao H, Liu G, Li X-H, Hiang G S, Poplawsky J D, Penn S T, Dierolf V and Tansu N 2009 Appl. Phys. Lett.95 061104
