Hybrid Tunnel Junction–Graphene Transparent Conductive

Department of Electrical and Computer Engineering, John Hopkins ... that low tunnel resistance can be achieved in the n+-GaN/u-InGaN/p+-GaN TJ, ... Th...
0 downloads 0 Views 3MB Size
Download PDF
Research Article www.acsami.org

Hybrid Tunnel Junction−Graphene Transparent Conductive Electrodes for Nitride Lateral Light Emitting Diodes Liancheng Wang,*,†,§ Yan Cheng,†,∥ Zhiqiang Liu,*,† Xiaoyan Yi,† Hongwei Zhu,*,‡ and Guohong Wang† †

Semiconductor Lighting Technology Research and Development Center, Institute of Semiconductors, Chinese Academy of Sciences, Beijing 100083, China ‡ School of Materials Science and Engineering, State Key Lab of New Ceramic & Fine Processing, Tsinghua University, Beijing 100084, China § Mind Star (Beijing) Technology Co., Ltd., Zhongguancun South Street, Haidian District, No. 45 Hing Fat Building 1001, Beijing 100872, China ∥ Department of Electrical and Computer Engineering, John Hopkins University, Baltimore, Maryland 21218, United States S Supporting Information *

ABSTRACT: Graphene transparent conductive electrode (TCE) applications in nitride light emitting diodes (LEDs) are still limited by the large contact resistance and interface barrier between graphene and p-GaN. We propose a hybrid tunnel junction (TJ)−graphene TCE approach for nitride lateral LEDs theoretically and experimentally. Through simulation using commercial advanced physical models of semiconductor devices (APSYS), we found that low tunnel resistance can be achieved in the n+-GaN/u-InGaN/p+-GaN TJ, which has a lower tunneling barrier and an enhanced electric field due to the polarization effect. Graphene TCEs and hybrid graphene−TJ TCEs are then modeled. The designed hybrid TJ−graphene TCEs show sufficient current diffusion length (Ls), low introduced series resistance, and high transmittance. The assembled TJ LED with the triple-layer graphene (TLG) TCEs show comparable optoelectrical performance (3.99 V@20 mA, LOP = 10.8 mW) with the reference LED with ITO TCEs (3.36 V@20 mA, LOP = 12.6 mW). The experimental results further prove that the TJ−graphene structure can be successfully incorporated as TCEs for lateral nitride LEDs. KEYWORDS: graphene, tunnel junction, light emitting diodes, transparent conductive electrodes, gallium nitride, current spreading

■

INTRODUCTION Graphene, with attractive electrical (with reported values in excess of 15 000 cm2 V−1 s−1), optical (with ∼97.7% transmission for single layer graphene, SLG), and thermal properties (thermal conductivity of approximately 5300 W m−1 K−1), is considered to be an ideal candidate for TCEs in optoelectronic devices, including III−nitride based LEDs.1−4 We have recently reviewed the research progress of the graphene TCEs for GaN LEDs.5 For the conventional lateral LEDs, the large work function (WF) discrepancy between pGaN (∼7.5 eV) and graphene (∼4.8 eV for light p-doped) creates an interface barrier that hinders the movement of electrons, resulting in an increased forward voltage (VF), as shown in the band diagram in Figure 1a. The reported countermeasures include p-GaN/graphene interface engineering and graphene chemical doping approaches. The interface engineering approach suggests insertion of metal or ITO in the form of dots, nanowires (NWs), or an ultrathin film between the p-GaN and graphene, as shown in Figure 1a. However, this approach complicates the fabrication process. For the chemical © 2015 American Chemical Society

doping approach, the improvement is limited and not stable in the long term. We proposed the tunnel junction (TJ) approach previously and demonstrated the proof-of-principle of the first experimental result on TJ−graphene TCEs for blue LEDs.5 It exhibits a lower VF (3.58 V@20 mA) and higher light output power (LOP, 8.18 mW@20 mA), than LEDs with bare graphene TCEs (5.02 V, 3.85 mW@20 mA), whereas it is still quite inferior to the reference LED with ITO TCEs (3.08 V, 12.39 mW@20 mA). By further optimizing the TJ−graphene TCEs, adjusting the metal organic chemical vapor deposition (MOCVD) growth process, and improving the quality of graphene, TJ LEDs with graphene TCEs show comparable optoelectrical performance (3.99 V, LOP = 10.8 mW@20 mA) with the reference LED with ITO TCEs (3.36 V, LOP = 12.6 mW@20 mA). This proves the availability of our proposal and Received: October 5, 2015 Accepted: December 23, 2015 Published: December 23, 2015 1176

DOI: 10.1021/acsami.5b09419 ACS Appl. Mater. Interfaces 2016, 8, 1176−1183

Research Article

ACS Applied Materials & Interfaces

Figure 1. Comparison of the graphene and hybrid graphene−tunnel junction (TJ) TCEs for GaN LEDs. (a) reference LED with graphene TCEs. The band diagram is shown on the left and shows that the formed graphene/p-GaN interface barrier will hinder the movement of carriers. Graphene structure on the LED imprinted with the circuit model is shown on the right. It illustrates that the interface engineering approach reduces the graphene/p-GaN resistance. (b) LED with hybrid TJ−graphene TCEs. The band diagram is shown on the left, and the graphene/TJ LEDs’ structure imprinted with the circuit is shown on the right. TJ is realized through in situ growth of p+-GaN/u-Al(In)GaN/n+-GaN. (c) Scanning electron microscope (SEM) images of reference LED (upper part) and TJ LED with graphene (lower part).

as is inferred from the band diagram in Figure 1b. Degenerately doping raises the EF into the conduction band (CB) at the Al(In)GaN side and lowers the EF into the valence band (VB) at the p+-GaN side at the n+-Al(In)GaN/p+-GaN interface. When the LED is forward biased, the n-AlInGa N/p+-GaN TJ is actually reverse biased, and the electrons’ probability from VB of p+-GaN to tunnel across the TJ increases. Large-density intrinsic dislocations and defects (108/cm) in the state-of theart GaN grown on sapphire also provides multiple midgap states for the carrier to tunnel across. In addition, the relatively low growth temperature of p+-GaN (∼960 °C) and the degenerated doping result in even poorer material quality of p+GaN, which makes the tunnel process much easier. Our first attempt in ref 5 shows that the TJ/graphene TCE based LEDs had a higher VF, indicating that the tunnel resistance of the TJ is still large. Here we show our further efforts to optimize the TJ−graphene TCEs.

takes a step forward for the graphene TCEs’ ultimate application in nitride LEDs and optoelectrical devices.

■

TUNNEL JUNCTION DESIGN AND GRAPHENE TCE MODELING The TJ approach proposes to locally grow multiple layers of heavily doped Alx(In)Ga1−xN layers following the p-GaN layer during the MOCVD process, as illustrated in Figure 1b. In a comparison with the above-mentioned interface engineering approach, this eliminates the additional steps for metal or ITO deposition in the chip fabrication process before transfer of graphene. Figure 1b also sketches the graphene/n+-Al(In)GaN/p+-GaN band diagram. Therefore, two TJ interfaces are formed, including graphene/n-Al(In)GaN interface and nAl(In)GaN/p+-GaN interface. A nearly ohmic (linear) current−voltage (I−V) characteristic was observed by Esaki6 in the reverse biased TJ, which first revealed the quantum tunnel across the depletion region of the junction. Tunnel resistance is determined by the equivalent energy and momentum states of the electrons and holes at both sides of the TJ. With reference to our former experimental results, a low ρc is expected between graphene and the top n+-Al(In)GaN layer.7 This is due to the high holes’ (electrons’) probability to tunnel from graphene (Al(In)GaN) to Al(In)GaN (graphene),

■

TUNNEL JUNCTION DESIGN First, we optimized the TJ structure. We performed numerical simulations using APSYS,8 which self-consistently solves the Poisson equation, continuity equation, and Schrö dinger equation with proper boundary conditions. 40% of the theoretical polarization induced sheet charge density was 1177

DOI: 10.1021/acsami.5b09419 ACS Appl. Mater. Interfaces 2016, 8, 1176−1183

Research Article

ACS Applied Materials & Interfaces

Figure 2. TJ structure design. (a) The energy band diagram of the n+-GaN/p+-GaN, n+-Al0.15 Ga0.85N/p+-GaN, and n+-In0.15Ga0.85N/p+-GaN TJs on the equilibrium and (b) reverse biased (−7.5 V) status. (d) The energy band diagram for n+-GaN/p+-GaN, n+-GaN/u-Al0.15Ga0.85N/p+-GaN, and n+GaN/u-In0.15Ga0.85N/p+-GaN TJs at the equilibrium. (e) The calculated electric fields within the n+-GaN/p+-GaN, n+-GaN/u-Al0.15Ga0.85N/p+-GaN, and n+-GaN/u-In0.15Ga0.85N/p+-GaN TJs at the equilibrium. (f) The sketched polarization induced electric field and sheet charge for the n+Al0.15Ga0.85N/p+-GaN, n+-GaN/u-In0.15Ga0.85N/p+-GaN, and n+-GaN/u-Al0.15Ga0.85N/p+-GaN TJs. (c) The calculated I−V characteristics of all the TJs.

Further, III−nitride epitaxial layers grown along the corientation (0001) are well-known to exhibit strong spontaneous and piezoelectric polarization,9,12 which induces sheet charges with relatively high density at the heterojunction interfaces. These charges are able to generate strong electric field, resulting in band bending. Consequently, tunneling probability is also affected by the polarization field. We compared LEDs with n + -GaN/p + -GaN, n + -GaN/uAl0.15Ga0.85N/p+-GaN, and n+-GaN/u-In0.15Ga0.85N/p+-GaN TJs. Figure 2d shows the energy band diagram at the equilibrium status. Similar to the results shown in Figure 2a, n+-GaN/u-In0.15Ga0.85N/p+-GaN shows the lowest tunneling barrier. Footprints of the introduced InGaN or AlGaN are observed, as denoted in Figure 2d. Figure 2e shows the calculated electric fields within the TJ at equilibrium. The electric field is ∼10.4 MV/cm in the n+-GaN/p+-GaN TJ, which equals to the internal built-in electric fields (Ebi) . With the InGaN (AlGaN) insertion, the electric field is increased (decreased) to ∼13.8 MV/cm (∼9.3 MV/cm). It is inferred that the polarization induced electric field (Esp+pz) in the compressive-strained InGaN (tensile stressed AlGaN) layer is added as the additional electric field component. Figure 2f shows the polarization induced electric field and sheet charge. The enhanced electric field in the n+-GaN/u-In0.15Ga0.85N/p+GaN TJ results in an increased carrier tunneling probability, and therefore boosts the tunneling current. We found that the n+-GaN/u-In0.15Ga0.85N/p+-GaN TJ combines the advantages of a lower tunneling barrier and an enhanced polarization electric field, thus serving as the best candidate for TJ. Figure 2c shows the calculated I−V characteristics of all the TJs. The n+GaN/u-In0.15Ga0.85 N/p+-GaN TJ (n+-GaN/u-Al0.15Ga0.85N/p+GaN) exhibits the highest (lowest) tunneling current with 0.43A/m@-5 V (2.9 × 10−9A/m @-5 V) among all the TJs discussed above. This is due to the opposite polarization field,

assumed to be due to the crystal relaxation through dislocation generation during the growth.9 The other parameters used can be found elsewhere.10 The equilibrium and reverse biased (−7.5 V) energy band diagrams of the n+-GaN/p+-GaN, n+Al0.15GaN/p+-GaN, n+-In0.15GaN/p+-GaN TJs are shown in Figure 2a,b, respectively. Under reverse biased (the n side is positively biased) status, the hole quasi-Fermi level (EF,p) for p+-GaN was significantly raised above the electron quasi-Fermi level (EF,n) for n+-Al(In)GaN; thus, electrons in the VB of the p+-GaN layer were able to tunnel into the conduction band of the n+-Al(In)GaN layer through the forbidden band, as indicated in Figure 2b. Tunneling resistance is inversely related to the potential barrier (Vb) and tunnel distance d. The tunnel probability in the triangular potential approximation is related to the band gap Eg and the electric field E, and is roughly described by the following equation11 ⎡ ⎤ 1 4 (2m)1/2 ≈ exp⎢ − (eVb)1/2 d ⎥ ℏ TR 3 ⎣ ⎦ 3/2 ⎤ ⎡ 4 (2m)1/2 (Eg ) ⎥ ≈ exp⎢ − ⎢⎣ ℏ eE ⎥⎦ 3

(1)

where ℏ is the reduced Planck constant and m is the carrier effective mass. Due to a lower Eg of InGaN than AlGaN and GaN, carriers in the n+-InGaN/p+-GaN have the highest carrier tunnel capability, according to eq 1. The n+-InGaN/p+-GaN TJ exhibits the lowest tunneling current among the n+-AlGaN/p+GaN, n+-GaN/p+-GaN, and n+-InGaN/p+-GaN bilayer TJs, as shown in Figure 2c. Also, the electric field E and the density of states (DOE) available for tunneling increase when reverse bias (LED’s forward operate bias) increases. DOE can be integrated E from the EF,n to EF,p with DOE = ∫ EF,p D(E)dE.6 F,n 1178

DOI: 10.1021/acsami.5b09419 ACS Appl. Mater. Interfaces 2016, 8, 1176−1183

Research Article

ACS Applied Materials & Interfaces

Figure 3. Graphene TCEs modeling. (a) The E−k relations for single layer graphene (SLG), bilayer graphene (BLG), and trilayer graphene (TLG) near the Dirac point. The inset enlarges the circled region. (b) The calculated transmission (T) at the visible wavelength range. Inset of part b shows transmission (T) as a function of layer number n. (c) The sheet resistance (Rsh) for graphene TCEs as a function of layer number n. (d) Rsh versus WF for graphene TCEs with various layer number n. (e) Summary of the theoretically and experimentally derived Rsh and T values for graphene TCEs. (f) The hole concentration distribution by APSYS simulation with corresponding graphene with different Rsh.

absorption that forms near the Van Hove singularity at the saddle point of the band structure and couples to the Dirac continuum.14 The calculated result shows a linear dependence of T with n, T = 1−2.3% × n, as shown in the inset of Figure 3b, which is consistent with the experimental result.14 Multiple layer graphene (MLG) is optically equivalent to a simple stack of SLG, with little perturbation of each adjacent layer. As n increases, Rsh is significantly decreased (Figure 3c). Graphene is chemically or electrically doped, resulting in WF variation.15,16 We also investigated the T and Rsh as a function of WF (the intrinsic WF was assumed to be 4.7 eV). T remained almost unvaried, while Rsh obviously decreased in the p-doping region when WF increased due to the increased hole concentration, regardless of the layer number (Figure 3d). For a better visualization, we summarized the theoretically and experimentally derived results shown in Figure 3e. The experimental Rsh and T values were adapted from our and other group’s works.5 The experimentally obtained Rsh is still higher than the theoretical value due to the still high density of defects in the state-of-art CVD graphene, which decreases the carrier charge carrier mobility.17,18 Figure 1a also presents the circuit model for LED with bare graphene TCEs. Holes (electrons) are simultaneously laterally diffused through graphene TCEs (n-GaN) and vertical flow through the diode junction. ρpl, ρc, ρp, and ρJ represent the parallel resistivity, contact resistivity, p-GaN resistivity, and diode junction resistivity, respectively. RG, Rn, and Rp represent the lateral resistance of graphene TCEs, n-GaN, and p-GaN, respectively. We further denote ρs to be the total series resistivity (ρs = ρc + ρp + ρJ). ρ has units of Ω cm2, and R has the unit of Ω. We reasonably simplified the circuit model into two vertical current branches (path 1 and path 2 as denoted in Figure 1a). ρpl and Rp are considered to be quite large; the voltage drop between A and B (VAB) should be equal across either through path 1 or 219,20

which significantly accelerates and inhibits the carriers’ tunneling process in n+-GaN/u-InGaN/p+-GaN and n+-GaN/ u-AlGaN/p+-GaN TJ, respectively (Figure 2f). It is also found that the n+-AlGaN/p+-GaN TJ exhibits a higher tunneling current, compared with n+-GaN/p+-GaN TJ. This should be attributed to the piezoelectric polarization field across the AlGaN/GaN interface (left panel in Figure 2f). In all the TJs’ simulations, the doping concentration and thickness of the n+ and p+ component is set to be 5 × 1019 cm−3 and 40 nm. The sandwiched u-AlGaN and u-InGaN is set to be 3 nm in thickness.

■

GRAPHENE AND GRAPHENE−TJ TCES MODELING The major concerning parameters associated with TCEs include sheet resistance (Rsh, Ω/□) and transmittance (T), both of which are related to the number of layers (n) for the specific graphene TCEs. We used the classic electromagnetic field theory in our modeling to describe the graphene optical response and the imaginary part of the complex dielectric constant. The optical conductivity is calculated according to the Kubo formula.13 Rsh is calculated according to eq 2, where μe and μh are the electron and hole mobility, with Ne and Nh being the electron and hole densities in graphene, respectively. R sh =

1 e(μe Ne + μ h Nh)

(2)

The E−k dispersion curves were obtained first using a tightbinding approach. The details for the E−k relation derivation process are given in part I of the Supporting Information. Figure 3a plots the E−k relations for single layer graphene (SLG), bilayer graphene (BLG), and trilayer graphene (TLG) near the Dirac point, and the inset of Figure 3a enlarges the circled part. Different from SLG, BLG and TLG no longer show linear characteristics near the Dirac point. The carrier state density as a function of energy, carrier density, and optical conductivity is subsequently obtained. Figure 3b plots the transmittance of graphene in the visible range. It is almost flat above 400 nm. The dip at ∼280 nm is due to the resonant

VAB = I1(ρs /S) + I1R n = I2R G + I2(ρs /S) 1179

(3)

DOI: 10.1021/acsami.5b09419 ACS Appl. Mater. Interfaces 2016, 8, 1176−1183

Research Article

ACS Applied Materials & Interfaces

Figure 4. Characterization of the reference LED and TJ LED. (a) XRD (0001) ω/2θ scan profiles for the reference LED and TJ-LED, the inset of part a shows the (0001) ω-rocking curves. (b) The atom concentration of TJ LED versus depth profile revealed the secondary ion mass spectrometry (SIMS). (c) PL results of the reference LED and TJ LED. (d) Electroluminescence (EL) of the reference LED and (f) TJ LED. (e) The extracted peak EL intensity versus current for both LEDs.

where S is the same order with the device area. From eq 3, we obtain R G + ρs /S I1 = I2 R n + ρs /S

structure is presented in Figure 1b. The TJ insertion added a series vertical resistivity ρT and parallel lateral resistance RT. The doping concentration ND of a 40 nm-thick n+-GaN is 5 × 1019 cm−3, and RT was calculated to be ∼200 Ω. The total Ls is considered to be the sum of the diffusion length in the graphene and TJ. Ls was calculated to be larger than 200 μm.

(4)

Equation 4 shows that, if RG ≫ Rn, then I1 ≫ I2, which indicates that the current tends to be crowded under the pelectrode. Conversely, if RG ≪ Rn, then I1 ≪ I2, and the carriers are instead more prone to accumulate and get recombined near the mesa edge. The hole concentration distribution by APSYS simulation confirmed our analysis, as shown in Figure 3f. When RG is 20 Ω (3000 Ω), holes tended to diffuse far to the mesa edge (crowded under the p-electrode). As the generated photons under the p-electrode are partly absorbed,21 a lower RG than Rn is expected to reach a “lateral resistance balance” for improved efficiency.19,20 This “interactive double diffusion mechanism” of current in lateral LEDs is different from that of vertical LEDs.22 For a typical LED sized 254 × 585 μm2, ρs/S (actually is the series resistance), is 10−15 Ω, and Rn is ∼80 Ω. For the state-of-art graphene TCEs, RG is ∼400−1000 Ω for SLG and ∼200 Ω for TLG after chemical doping (Figure 3e), which is still substantially larger than Rn. With the assumption of ρs being ∼2 × 10−2 Ω cm2, the current diffusion length Ls in graphene TCEs can be calculated to be ∼44.7−100 μm (Ls = ρs /R G (eq 5)),22 which is insufficient for current spreading. Actually, the ρc for p-GaN/graphene contact is large (2.2 × 10−1 Ω cm2),7 which boosts the Ls to be ∼300 μm due to the blocking effect,22 but increases the VF (ΔVF ∼ ΔRI = 0.32 V, @I = 20 mA). By increasing the layer number to n = 10, combined with further chemical doping, RG can be reduced to ∼100 Ω (Figure 3e). However, this induced a quantitative transmission loss (ΔT ∼ 20%, Figure 3b). The hurdle brought from the high ρc of graphene/p-GaN and short Ls for the graphene TCEs is solved by inserting a designed TJ. The circuit model for LED−TJ−graphene

Ls =

(ρs + ρT )/R G +

(ρs + ρT )/RT

(5)

ρT includes the ρc of graphene/n+-GaN junction and intrinsic ρT, which is estimated to be ∼2 × 10−2 Ω cm2, and ρT has same order with the original ρs. The low graphene/n+-GaN ρc (∼10−5−10−6 Ω cm2 order) is expected due to the degenerate doping of the GaN and the inherently low WF mismatch between the constituent graphene and n-GaN. The intrinsically low ρT was achieved through the optimally designed TJ (n+GaN/u-In0.15Ga0.85N/p+-GaN), brought from the InGaN midgap states with lower Eg and enhanced piezoelectric polarization field, as discussed above. Through the above analysis, we know that the hybrid graphene−TJ TCEs satisfies the requirements as TCEs for GaN LEDs.

■

EXPERIMENTAL RESULTS AND DISCUSSION Figure 4a shows the XRD (0001) ω/2θ scan profiles for both samples. The strongest peak stemming from the GaN buffer substrate is calibrated as the benchmark (ω = 0°). The zero order satellite peaks are identical, indicating the same indium incorporation. The additional peaks on the left side of the zero order MQWs satellite peak stem from the interference peaks of the MQWs. The peaks on the right stem from the prestrained wells. All the satellite peaks are identical, indicating that the material qualities of the MQWs are not influenced by the TJ in situ growth in TJ LED. The inset of Figure 4a shows the (0001) ω-rocking curve, with full width at half-maximum (fwhm) to be ∼230 arcsecs and 258 arcsecs for the reference and TJ LED, suggesting that negligent screw dislocations were 1180

DOI: 10.1021/acsami.5b09419 ACS Appl. Mater. Interfaces 2016, 8, 1176−1183

Research Article

ACS Applied Materials & Interfaces

was then calculated to be ∼2.2 × 10−1 and ∼1.1 × 10−6 Ω cm2, respectively. Figure S2b shows the Raman spectra of SLG and TLG with excitation wavelength of 633 nm. Two typical peaks at 1589 cm−1 (G peak) and 2661 cm−1 (2D peak) are clearly observed. The peak intensity contrast ratio between G and 2D peak indicates its single layer and multiple layer characteristics,23 respectively, which is further confirmed by the transmittance value (part II of Supporting Information, Figure S2a, ∼97% for SLG, 92% for TLG).11 The D peak in TLG at ∼1350 cm−1 was due to defects and possible interlayer disoriented stacking in TLG. Rsh for SLG and TLG was ∼1100 and ∼520 Ω/□, respectively. With further doping in nitric acid vapor for about 2 min, the Rsh decreased to about ∼500 and ∼30 Ω/□, respectively. The optical transparency was invariant after doping. To get a comprehensive comparison, the reference and TJ LEDs with various TCEs (ITO, SLG, TLG) were

induced for TJ LED. The atom concentration of TJ LED versus depth was investigated by secondary ion mass spectrometry (SIMS), as shown in Figure 4b. The doping levels for n-GaN and p-GaN were accessed as ∼1.8 × 1018 and ∼4.3 × 1019 cm−3. High doping was successfully realized in the TJ region, with the p+ doping peak being astonishingly as high as 2.4 × 1020 cm−3, and n+ doping was ∼4.65 × 1019 cm−3. Due to the “pause” growth strategy, an abrupt interface was achieved between the p+ and n+ regions, which is critical for TJ realization. The thin sandwiched InGaN layer is hinted from the rising In element profile. Photoluminescence (PL) results shown in Figure 4c reveal identical PL features with two observable peaks having approximately the same intensity. The main peak is located at 466 nm, and the shoulder peak is at 444 nm, which is associated with the MQWs and the prestrained wells, respectively. The reabsorption from the TJ constituent uIn0.15Ga0.85N is neglected. PL results further prove the preserved material quality of MQWs in TJ LED. In contrast, the reference and TJ LED showed apparent distinctions in electroluminescence (EL), as shown in Figure 4d,f, respectively. TJ LED showed nearly equal EL intensity at low current ( 40 mA). The abnormal negative resistance phenomenon (V is abnormally decreased when I increases) is assumed to be the breakdown (Zener or Avalanche breakdown) of the TJ. Comparing Vth and VRV, we found that, for the reference LED with graphene TCEs, it is the graphene/p-GaN interface barrier instead of the introduced graphene/p-GaN contact resistance, which dominated the ΔVth part (ΔVth = Vth − 2.6 eV); conversely, the TJ LED with graphene TCEs had these two factors’ weight approximately in the same order. For example, ΔVth was equal to 1.62 eV (0.29 eV) for the reference LED with SLG TCEs (TJ LED with SLG TCEs), and the interface barrier and series resistance contribution were 1.5 eV (0.218 eV), 0.12 eV (0.072 eV), respectively. The tunneling process in our TJ LED is enabled unambiguously from the high doping concentration, assistance of the high density defect, and polarization field. More discussions and comparative analysis of the GaN based TJ and TJ LED can be found in part III of the Supporting Information. Most importantly, it was found that the TJ LED with TLG TCEs (3.99 V@20 mA, 4.67 V@60 mA, 5.46 V@100 mA) were comparable with the reference LED with ITO TCEs (3.36 V@20 mA, 4.49 V@60 mA, 5.3 V@100 mA) in terms of the optoelectrical performance. This is attributed to the optimized design and fine experimental realization, to achieve TJ− graphene TCEs with low resistance and Rs. The typical I− LOP curves were plotted in Figure 5b and also summarized in Table 1. The LOP of the TJ LED with TLG TCEs (LOP = 10.8 mW@20 mA) is comparable with that of the reference LED with ITO TCEs (LOP = 12.6 mW@20 mA). The microscope emission photographs are shown in Figure 5c, corresponding to the TJ LED with TLG TCEs and the reference LED with ITO TCEs at 20, 100 mA, respectively. The emission of the TJ LED with graphene TCEs is uniform, indicating a uniformly distributed current. Our experimental

results prove the availability of the hybrid TJ−graphene TCEs in GaN LEDs. It can be further optimized to act as an alternative choice beyond the conventional ITO TCEs.

■

CONCLUSIONS

■

EXPERIMENTAL SECTION

To conclude, we proposed the TJ−graphene TCEs for nitride LEDs theoretically and experimentally. TJ based on n+-GaN/uIn0.15Ga0.85N/p+-GaN structure is the best candidate, combining the advantages of a lower tunneling barrier and an enhanced polarization electric field. Further combined with graphene, the hybrid TJ−graphene structure effectively functions as TCEs with sufficient current diffusion length and low introduced series resistance simultaneously. The experimental results demonstrated that TJ LED with TLG TCEs show comparable optoelectrical performance (3.99 V@20 mA, LOP = 10.8 mW) with the reference LED with ITO TCEs (3.36 V@20 mA, LOP = 12.6 mW). This proves the availability of our proposed TJ− graphene TCE approach for nitride LEDs. Our work is significantly meaningful for graphene TCEs’ application in nitride based LEDs and other optoelectrical devices.

Material Growth. The materials are epitaxially grown on a sapphire substrate using MOCVD method. The reference LED sample consisted of unintentionally doped u-GaN(∼3 μm), n-GaN:Si (∼3−4 μm, n ∼ 1 × 1018 cm−3), three pairs of In0.1Ga0.9N(∼2 nm)/GaN (∼8−12 nm) prestrained wells, seven pairs of In0.15Ga0.85N (∼2 nm)/ GaN(∼8−12 nm) multiple quantum wells (MQWs), and a p-GaN layer: Mg (∼100 nm, p ∼ 3 × 1017cm−3) sequentially. The p-GaN layer was grown to about 80 nm for the TJ LED sample. Then, a heavily Mg-doped p+-GaN (∼20 nm), an undoped In0.15Ga0.85N layer (∼3 nm), and a heavily Si doped n+-GaN (∼20 nm) were grown in sequence on the p-GaN. The flow rates of Cp2Mg and TMGa were 1.5 and 22.0 μmol/min, respectively, and the Mg dopant ionization ratio at room temperature was considered to be lower than 1% in p+-GaN. The growth condition for In0.15Ga0.85N was the same as that for the quantum wells, with TEGa as the precursor and N2 as the ambience. n+-GaN was grown under the same conditions as p-GaN, except that the Cp2Mg precursor was replaced by SiH4. Considering the compensation effect of the diffused Mg dopants, the In0.15Ga0.85N and n+-GaN layers were intentionally grown by pausing 2 min after p+GaN layer growth. Device Fabrication. The devices were fabricated with a standard LEDs fabrication process: inductively coupled plasma (ICP) mesadefining etching, SiO2 deposition for sidewall passivation, graphene TCEs transfer or ITO deposition (100 nm thick for TJ LED and 200 nm thick for the reference LED), and metal deposition. 1182

DOI: 10.1021/acsami.5b09419 ACS Appl. Mater. Interfaces 2016, 8, 1176−1183

Research Article

ACS Applied Materials & Interfaces

■

(14) Chae, D. H.; Utikal, T.; Weisenburger, S.; Giessen, H.; Klitzing, K. V.; Lippitz, M.; Smet, J. Excitonic Fano Resonance in Free-standing Graphene. Nano Lett. 2011, 11 (3), 1379−1382. Nair, R. R.; Blake, P.; Grigorenko, A. N.; Novoselov, K. S.; Booth, T. J.; Stauber, T.; Geim, A. K. Fine Structure Constant Defines Visual Transparency of Graphene. Science 2008, 320 (5881), 1308−1308. (15) Liu, H.; Liu, Y.; Zhu, D. Chemical Doping of Graphene. J. Mater. Chem. 2011, 21 (10), 3335−3345. (16) Novoselov, K. S.; Geim, A. K.; Morozov, S. V.; Jiang, D.; Zhang, Y.; Dubonos, S. A.; Firsov, A. A. Electric Field Effect in Atomically Thin Carbon Films. Science 2004, 306 (5696), 666−669. (17) Kholmanov, I. N.; Magnuson, C. W.; Aliev, A. E.; Li, H.; Zhang, B.; Suk, J. W.; Ruoff, R. S.; et al. Improved Electrical Conductivity of Graphene Films Integrated with Metal Nanowires. Nano Lett. 2012, 12 (11), 5679−5683. (18) Jeong, C.; Nair, P.; Khan, M.; Lundstrom, M.; Alam, M. A. Prospects for Nanowire-doped Polycrystalline Graphene Films for Ultra-transparent, Highly Conductive Electrodes. Nano Lett. 2011, 11 (11), 5020−5025. (19) Kim, H.; Lee, J. M.; Huh, C.; Kim, S. W.; Kim, D. J.; Park, S. J.; Hwang, H. Modeling of a GaN-based Light-emitting Diode for Uniform Current Spreading. Appl. Phys. Lett. 2000, 77 (12), 1903− 1904. (20) Kim, H.; Park, S. J.; Hwang, H.; Park, N. M. Lateral Current Transport Path, a Model for GaN-based Light-emitting Diodes: Applications to Practical Device Designs. Appl. Phys. Lett. 2002, 81 (7), 1326−1328. (21) Zhang, Y.; Zheng, H.; Guo, E.; Cheng, Y.; Ma, J.; Wang, L.; Li, J.; et al. Effects of Light Extraction Efficiency to the Efficiency Droop of InGaN-based Light-Emitting Diodes. J. Appl. Phys. 2013, 113 (1), 014502. (22) Wang, L.; Zhang, Z. H.; Wang, N. Current Crowding Phenomenon: Theoretical and Direct Correlation with the Efficiency Droop of Light Emitting Diodes by a Modified ABC Model. IEEE J. Quantum Electron. 2015, 51 (5), 1−9. (23) Ferrari, A. C.; Meyer, J. C.; Scardaci, V.; Casiraghi, C.; Lazzeri, M.; Mauri, F.; Geim, A. K.; et al. (2006). Raman Spectrum of Graphene and Graphene Layers. Phys. Rev. Lett. 2006, 97 (18), 187401.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsami.5b09419. Graphene modeling method, transmittance characteristics, Raman spectra of graphene TCEs, and more discussions on tunneling mechanism in GaN (PDF)

■

AUTHOR INFORMATION

Corresponding Authors

*E-mail: [email protected]. *E-mail: [email protected]. *E-mail: [email protected]. Author Contributions

L.W. and Y.C. contributed equally to this work. L.W. wrote the whole manuscript. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported by the National High Technology Program of China (Grant 2013AA03A101). We show our great appreciation for help from Zihui Zhang on the TJ simulation part, and Haiyang Zheng on the graphene modeling part. The epitaxial material was provided by Xiaoyan Yi and Guohong Wang. Graphene was purchased from the Graphenea Co., Ltd.

■

REFERENCES

(1) Ellmer, K. Past Achievements and Future Challenges in the Development of Optically Transparent Electrodes. Nat. Photonics 2012, 6 (12), 809−817. (2) Bonaccorso, F.; Sun, Z.; Hasan, T.; Ferrari, A. C. Graphene Photonics and Optoelectronics. Nat. Photonics 2010, 4 (9), 611−622. (3) Bae, S.; Kim, H.; Lee, Y.; Xu, X.; Park, J. S.; Zheng, Y.; Iijima, S.; et al. Roll-to-roll Production of 30-in. Graphene Films for Transparent Electrodes. Nat. Nanotechnol. 2010, 5 (8), 574−578. (4) Chandramohan, S.; Kang, J. H.; Katharria, Y. S.; Han, N.; Beak, Y. S.; Ko, K. B.; Hong, C. H.; et al. Work-function-tuned Multilayer Graphene as Current Spreading Electrode in Blue Light-emitting Diodes. Appl. Phys. Lett. 2012, 100 (2), 023502. (5) Wang, L.; Liu, W.; Zhang, Y.; Zhang, Z. H.; Tan, S. T.; Yi, X.; Demir, H. V.; et al. Graphene-based Transparent Conductive Electrodes for GaN-based Light Emitting diodes: Challenges and Countermeasures. Nano Energy 2015, 12, 419−436. (6) Esaki, L. New Phenomenon in Narrow Germanium p-n Junctions. Phys. Rev. 1958, 109 (2), 603. (7) Wang, L.; Zhang, Y.; Li, X.; Liu, Z.; Guo, E.; Yi, X.; Wang, G.; et al. Interface and Transport Properties of GaN/graphene Junction in GaN-based LEDs. J. Phys. D: Appl. Phys. 2012, 45 (50), 505102. (8) Zhang, Z. H.; Tan, S. T.; Kyaw, Z.; Ji, Y.; Liu, W.; Ju, Z.; Demir, H. V.; et al. InGaN/GaN Light-emitting Diode with a Polarization Tunnel Junction. Appl. Phys. Lett. 2013, 102 (19), 193508. (9) Fiorentini, V.; Bernardini, F.; Ambacher, O. Evidence for Nonlinear Macroscopic Polarization in III−V Nitride alloy Heterostructures. Appl. Phys. Lett. 2002, 80 (7), 1204−1206. (10) Vurgaftman, I.; Meyer, J. R. Band Parameters for Nitrogencontaining Semiconductors. J. Appl. Phys. 2003, 94 (6), 3675−3696. (11) Sze, S. M.; Ng, K. K. Physics of Semiconductor Devices, 3rd ed.; John Wiley Publications, 2006. (12) Romanov, A. E.; Baker, T. J.; Nakamura, S.; Speck, J. S. Straininduced Polarization in Wurtzite III-nitride Semipolar Layers. J. Appl. Phys. 2006, 100 (2), 023522. (13) Stauber, T.; Peres, N. M. R.; Geim, A. K. Optical Conductivity of Graphene in the Visible Region of the Spectrum. Phys. Rev. B: Condens. Matter Mater. Phys. 2008, 78 (8), 085432. 1183

DOI: 10.1021/acsami.5b09419 ACS Appl. Mater. Interfaces 2016, 8, 1176−1183