Open Access
Add to BibSonomy
Abstract
The GaN-based light emitting diodes (LEDs) have a great potential for visible light communication (VLC) due to their ubiquitous application in general lighting, but the modulation bandwidth of conventional c-plane LEDs is limited by carrier recombination rate in InGaN quantum wells (QWs) due to the polarization-field-induced quantum confined Stark effect (QCSE). Furthermore, the high modulation bandwidth on c-plane sapphire substrates can only be achieved at high current densities. Here, blue LEDs with ultra-thin InGaN QWs (1nm) and GaN barriers (3nm) are grown on c-plane sapphire substrate to suppress QCSE and extend the cut-off frequency from 214 MHz for conventional LEDs to 536 MHz at a current density of 2.5 kA/cm2, which is comparable to devices grown on semi-polar substrates.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Visible light communication (VLC) employs the light intensity variation in visible light emitting diodes (LEDs) to transfer data [1]. In recent years, GaN-based LEDs that become ubiquitous in general lighting, are considered as the ideal light sources for VLC [2–9]. The small-signal modulation bandwidth of an LED device is inversely related to the total carrier lifetime of the device. The total lifetime includes transport-related lifetime (RC constant) and the lifetime in the active region [10, 11]. Much effort has been made to reduce the total carrier lifetime in GaN-based LEDs [12–18], and it has been revealed that the carrier lifetime inside quantum well (QW) is crucial for improving modulation bandwidth in conventional LED devices.
The carrier lifetime in QW (τ) is related to the radiative (τr) and non-radiative (τnr) recombination lifetimes: 1/τ = 1/τr + 1/τnr, and a shorter τ can be achieved by reducing both τr and τnr. The relation between τr and the spontaneous emission rate (Rsp) in QW can be written as [19]:
where n and Rsp are the effective carrier density and spontaneous emission rate in the active region, respectively, while is the electron-hole (e-h) wavefunction overlap.In conventional c-plane GaN-based blue LEDs, the e-h wavefunction overlap decreases due to polarization-induced quantum confined Stark effect (QCSE) in the active region. It has been shown that screening of QCSE at high current density is important to achieve high modulation bandwidths [20]. The influence of polarization can be reduced or even eliminated by changing the growth plane to semi-polar or non-polar planes. Rashidi et. al. have demonstrated LEDs grown on non-polar free-standing GaN substrate that could achieve a record modulation bandwidth of 1.5 GHz [21]. However, the high-cost and small-size of non-polar substrates limit the mass production of high-speed LEDs in this way. Therefore, how to suppress the QCSE and improve the e-h wavefunction overlap in conventional and industry-compatible c-plane LEDs is still a critical issue at present.
According to Eq. (1), the thin-QWs tend to have shorter τr values compared to thick ones due to the higher e-h wavefunction overlap. Moreover, the thin quantum barriers (QBs) in multiple-QWs (MQWs) can also play an important role in reducing the lifetime. The thin-QBs can reduce polarization fields inside QWs [22], increase hole injection into the active region, and result in more uniform carrier distribution throughout the active region [23]. By increasing radiative recombination and reducing QCSE in the active layers, a higher bandwidth at low current densities is expected for LEDs with thin-QWs and QBs. In this work, we have grown blue LEDs with ultra-thin InGaN QWs (1 nm) and GaN barriers (3 nm). The small-signal modulation bandwidth of device enhanced from 214 MHz to 536 MHz at 2.5 kA/cm2, as compared with a conventional LED with 3 nm/10 nm InGaN/GaN MQWs. Moreover, the modulation bandwidth at low current densities is improved significantly. The experimental results reveal a different relation between cut-off frequency (fc) and applied current density (j) in thin QWs/QBs, wherein the origin of this difference is investigated experimentally.
2. Device structure and fabrication
The LED structures were grown on 2-inch c-plane sapphire substrates by Aixtron 2000HT metal-organic chemical vapor deposition (MOCVD). The layer structure consisted of a 1 µm-thick undoped GaN buffer layer, a 2.5 µm-thick Si-doped n-GaN, the MQW active region, a 30 nm-thick p-AlGaN electron-blocking layer, and a 75 nm-thick p-GaN contact layer. Before growing the QW structure, 30 periods of low-indium InGaN/GaN (2 nm/2 nm) pre-strain superlattice (PS-SL) was grown to improve the crystalline quality of the active region. In the active region, a thin GaN cap layer is grown for 15 seconds at the same temperature of InGaN well layer (770°C) to protect QW layer and prevent the diffusion of indium into the barriers [24]. Then the temperature was raised to 890°C for GaN barrier growth.
Two samples labeled as S1 and S2 are grown for comparison. The only difference between them is the active region, which is listed in Table 1. The total thickness of MQW region was kept similar for both samples and the photoluminescence (PL) emission wavelength was also tuned to around 440 nm by adjusting the indium composition in the QWs.
Table 1. The Structure of the Active Region and Dislocation Density in the Samples
The X-ray diffraction (XRD) measurements of (002) and (102) planes were performed to verify the quality of the samples, wherein the screw and edge dislocation densities (nd) were estimated according to the full width at half maximum (FWHM) of the rocking curves [25]. The value of FWHM is calculated by fitting a Gaussian equation on the data, and nd is obtained from nd = (FWHM)2/(4.35 × βs2), where βs is Burgers vector [26]. According to Table 1, the crystalline quality of the samples is very close to each other.
LED chips with a size of 300 × 300 μm2 are manufactured by the standard LED fabrication process, and the conventional 5 mm epoxy packaging is used for DC electrical and light output power measurements.
For modulation bandwidth measurements, the LED chips were fabricated with the process techniques described in [5]. The ring-shaped contact layers were used for uniform current spreading across the device, and the diameter of the ring is reduced to 50~175 μm to allow high current densities.
3. Characterization and discussion
3.1 Structural and optical characterization
The high angle annular dark field scanning transmission electron microscopy (HAADF-STEM) showed uniform layers for both samples. But for S1, the sharp interface between QWs and barriers is somewhat hard to distinguish and the thickness of barriers seems not fully the same for all QWs [Fig. 1(a)]. This is due to the higher growth temperature of QBs compared to QWs; and considering the thin layers in the active region of S1, a rapid decrease (increase) of temperature is required to grow the QWs (QBs), which can lead to some fluctuations in the thickness.
Fig. 1 (a) HAADF-STEM cross-sectional images of S1 with 1 nm-thick QWs and 3 nm-thick barriers. (b) The room-temperature TRPL at the PL peak emission wavelength for the samples. The solid lines are the fitting of double-exponential decay on experimental data.
Room-temperature time-resolved PL (TRPL) is performed using a pulsed semiconductor laser with 399.6 nm excitation wavelength and 70 ps FWHM pulse width. The repetition rate and average energy of each pulse was 2.5 MHz and 550 pJ, respectively, and the spot diameter on the sample was around 1 mm. The results are shown in Fig. 1(b), where a fast decay of 1.41 ns obtained for S1 by double-exponential fitting. This value increases to 8.16 ns for S2, implying faster radiative and/or non-radiative recombination process in S1. Considering the weak excitation of TRPL and low carrier number (~1.1 × 109) for each pulse, the dominant recombination channel can shift to the non-radiative recombination in the samples, and the fast decay for S1 can be partially due to defect-assisted non-radiative recombination in the active region.
3.2 DC electrical characterization
For electroluminescence (EL) measurement the applied current was controlled by a low-noise DC power supply (Keithley 2400), and the spectrum obtained from a CCD array spectrometer (CDS 1100) equipped with an integrating sphere (Labsphere LMS-100).
The light-current density-voltage (L-j-V) characteristic of both samples is shown in Fig. 2(a), where the light output power of S1 is lower than that of S2. The external quantum efficiency (EQE) measurement results are shown in Fig. 2(b), where a significant reduction in EQE is observed for S1 compared with S2. The maximum EQE value of 8.3% and 22.4% is recorded for S1 and S2, respectively, and the efficiency droop in S1 was lower than S2 (about 3% for S1 and 7% for S2 at 110 A/cm2 current density).
Fig. 2 (a) L-j-V characteristics for S1 and S2. (b) The EQE measurement of the samples. (c) The EL spectrum for S1 and S2, where the applied current density is increased from 4 to 110 A/cm2. The inset in (c) is the peak wavelength changes with applied current density.
The EL spectrum is shown in Fig. 2(c), and S1 exhibits a smaller wavelength shift with applied current density, which can be due to the weaker QCSE in the active layers due to using thinner QWs/QBs [22].
3.3 Frequency response
To measure the frequency response of the samples, we used a network analyzer (Agilent E5062A) with a frequency coverage from 300 kHz to 1.5 GHz. The small-signal modulation voltage from Port 1 of the network analyzer as well as a DC bias was applied to the LED samples. The LED response detected by a silicon PIN photodetector (Graviton SPD-2) was fed back to Port 2 of the network analyzer.
The results show that fc of S1 is much higher than that of S2, and a value of 536 MHz is obtained at 2.5 kA/cm2 applied current density [Fig. 3(a)]. The pad capacitance for the samples was so small (around 1.7 pF) and the measured forward bias RC lifetime was below the differential carrier lifetime (τ∆n) in the range of applied current densities.
Fig. 3 (a) The normalized power response of S1 and S2 at 2.5 kA/cm2 applied current density. The inset in (a) shows a schematic view of the ring-shaped contact layers, where the mesa diameter is reduced to achieve high current densities. (b) The cut-off frequency vs. applied current density.
The experimentally measured values of fc in terms of j are shown in Fig. 3(b), where each data point is the averaged value of three samples. The obtained values of fc for S1 is high at low applied current densities, and a value of 120 MHz is obtained at 56 A/cm2 current density. This can be due to lower QCSE in the active layers of S1. Moreover, due to the lower density of states in thin QWs, the coulomb-enhanced capture process can increase the recombination rate in the active region at high applied current densities [18]. The slope of increasing (dfc / dj) is larger for S2, indicating higher QCSE compared with S1. At high current densities, the screening of QCSE causes a higher slope for S2 [20].
The differential carrier lifetime τ∆n can be obtained from fc according to τ∆n = 1/2πfc [Fig. 4(a)]. Then the carrier density (n) in the active region can be calculated in terms of applied current density considering perfect injection efficiency [27,28]:
where G = j/ed is the total recombination rate. The lower limit of the integral corresponds to the recombination at zero j, but the measured values of τ∆n do not give zero recombination rate at the lowest measurable j values [28]. By extrapolating differential carrier lifetime data in Fig. 4(a) and calculating y-intercept, an estimation of Shockley-Read-Hall lifetime (τSRH) at low current densities can be obtained for the samples. The value of τSRH for S1 and S2 is 2.8 ns and 24.8 ns, respectively. The value of differential radiative lifetime (τ∆n,R) can be estimated by τ∆n,R = τSRH × τ∆n/(τSRH - τ∆n) at low applied current densities, assuming that Auger recombination is negligible in this range [Fig. 4(b)]. The results show about five fold reduction τ∆n,R for S1 compared with S2, which is attributed to the higher e-h wavefunction overlap in thin QWs.Fig. 4 (a) The differential carrier lifetime in terms of applied current density. (b) The differential radiative lifetime for the low-j range. (c) The calculated carrier density in the active region in terms of applied current density using the method introduced in [28].
Using τSRH for the lower limit, the carrier density in the active region can be calculated from Eq. (2) [Fig. 4(c)]. The value for S1 is about 10-times lower than that of S2 at low j values, which can be due to higher recombination rate in S1. Furthermore, the lower n at low j values can also be due to higher defect-assisted non-radiative recombination in S1, which can also explain the low EQE. Due to the close values of dislocation density from XRD experiment for the samples, the higher defect density (due to Ga vacancies and alloy fluctuations [29]) at the interface of QW and QB can be the cause of high non-radiative recombination in S1. The number of these interfaces in S1 is three times higher than S2, and the higher indium concentration of the well layer in S1 can increase the interface defects significantly [30].
4. Conclusion
By using thinner QWs/QBs, a cut-off frequency of 536 MHz at 2.5 kA/cm2 applied current density is achieved for c-plane GaN-based blue LEDs. These devices also show 120 MHz cut-off frequency at 56 A/cm2 applied current densities, which is a record for GaN devices grown on c-plane substrates. The improvements are due to the shorter differential carrier lifetime and weaker QCSE in the active layers, which is attributed to the higher e-h wavefunction overlap in QWs and lower polarization effects of thin barriers on QW layers. In spite of improvements in modulation bandwidth, the light output power shows a considerable reduction due to the high non-radiative recombination. The non-radiative recombination mainly originates from the defects at the interface of QWs and QBs, which is expected due to the high-indium-contrast between QW and QB layers and the higher number of the interfaces compared with the reference structure.
Funding
National Key R&D Program of China (2017YFB0403100, 2017YFB0403101); Science Challenge Project (TZ2016003); National Natural Science Foundation of China (51561165012, 61621064, 61574082); Tsinghua University Initiative Scientific Research Program (2015THZ02-3).
Acknowledgment
This work was supported by Collaborative Innovation Center of Solid-State Lighting and Energy-Saving Electronics.
References
1. S. Dimitrov and H. Haas, Principles of LED Light Communications: Towards Networked Li-Fi (Cambridge U., 2015).
2. J. W. Shi, H. Y. Huang, J. K. Sheu, C. H. Chen, Y. S. Wu, and W. C. Lai, “The improvement in modulation speed of GaN-based green light-emitting diode, (LED) by use of n-type barrier, doping for plastic optical fiber (POF) communication,” IEEE Photonics Technol. Lett. 18(15), 1636–1638 (2006). [CrossRef]
3. J. W. Shi, C. W. Lin, W. Chen, J. E. Bowers, J. K. Sheu, C. L. Lin, Y. L. Li, J. Vinogradov, and O. Ziemann, “Very high-speed GaN-based cyan light emitting diode on patterned sapphire substrate for 1 Gbps plastic optical fiber communication,” in OFC/NFOEC Technical Digest (2012), pp. 1–3.
4. J.-M. Wun, C.-W. Lin, W. Chen, J. K. Sheu, C.-L. Lin, Y.-L. Li, J. E. Bowers, J.-W. Shi, J. Vinogradov, R. Kruglov, and O. Ziemann, “GaN-based miniaturized cyan light-emitting diodes on a patterned sapphire substrate with improved fiber coupling for very high-speed plastic optical fiber communication,” IEEE Photonics J. 4(5), 1520–1529 (2012). [CrossRef]
5. C. L. Liao, Y. F. Chang, C. L. Ho, and M. C. Wu, “High-speed GaN-based blue light-emitting diodes with Gallium-doped ZnO current spreading layer,” IEEE Electron Device Lett. 34(5), 611–613 (2013). [CrossRef]
6. L. Chien-Lan, H. Chong-Lung, C. Yung-Fu, W. Chi-Hung, and W. Meng-Chyi, “High-speed light-emitting diodes emitting at 500 nm with 463-MHz modulation bandwidth,” IEEE Electron Device Lett. 35(5), 563–565 (2014). [CrossRef]
7. R. X. G. Ferreira, E. Xie, J. J. D. McKendry, S. Rajbhandari, H. Chun, G. Faulkner, S. Watson, A. E. Kelly, E. Gu, R. V. Penty, I. H. White, D. C. O’Brien, and M. D. Dawson, “High bandwidth GaN-based micro-LEDs for multi-Gb/s visible light communications,” IEEE Photonics Technol. Lett. 28(19), 2023–2026 (2016). [CrossRef]
8. M. H. Kane, N. Dietz, I. T. Ferguson, T.-C. Lin, Y.-F. Yin, W.-Y. Lan, and J. Huang, “High optical bandwidth GaN based photonic-crystal light-emitting diodes,” in Fifteenth International Conference on Solid State Lighting and LED-based Illumination Systems (2016). [CrossRef]
9. Z. Quan, D. V. Dinh, S. Presa, B. Roycroft, A. Foley, M. Akhter, D. O’Mahony, P. P. Maaskant, M. Caliebe, F. Scholz, P. J. Parbrook, and B. Corbett, “High bandwidth freestanding semipolar (11–22) InGaN/GaN light-emitting diodes,” IEEE Photonics J. 8(5), 1–8 (2016). [CrossRef]
10. K. Ikeda, S. Horiuchi, T. Tanaka, and W. Susaki, “Design parameters of frequency response of GaAs—(Ga,Al)As double heterostructure LED’s for optical communications,” IEEE Trans. Electron Dev. 24(7), 1001–1005 (1977). [CrossRef]
11. J. W. Shi, J. K. Sheu, C. H. Chen, G. R. Lin, and W. C. Lai, “High-speed GaN-based green light-emitting diodes with partially n-doped active layers and current-confined apertures,” IEEE Electron Device Lett. 29(2), 158–160 (2008). [CrossRef]
12. J. W. Shi, K. L. Chi, J. M. Wun, J. E. Bowers, Y. H. Shih, and J. K. Sheu, “III-Nitride-based cyan light-emitting diodes with GHz bandwidth for high-speed visible light communication,” IEEE Electron Device Lett. 37, 894–897 (2016).
13. A. David, B. Moran, K. McGroddy, E. Matioli, E. L. Hu, S. P. DenBaars, S. Nakamura, and C. Weisbuchb, “GaN/InGaN light emitting diodes with embedded photonic crystal obtained by lateral epitaxial overgrowth,” Appl. Phys. Lett. 92(11), 113514 (2008). [CrossRef]
14. D. V. Dinh, Z. Quan, B. Roycroft, P. J. Parbrook, and B. Corbett, “GHz bandwidth semipolar (112¯2) InGaN/GaN light-emitting diodes,” Opt. Lett. 41(24), 5752–5755 (2016). [CrossRef] [PubMed]
15. E. K. Lau, A. Lakhani, R. S. Tucker, and M. C. Wu, “Enhanced modulation bandwidth of nanocavity light emitting devices,” Opt. Express 17(10), 7790–7799 (2009). [CrossRef] [PubMed]
16. A. Rashidi, M. Monavarian, A. Aragon, S. Okur, M. Nami, A. Rishinaramangalam, S. Mishkat-Ul-Masabih, and D. Feezell, “High-speed nonpolar InGaN/GaN LEDs for visible-light communication,” IEEE Photonics Technol. Lett. 29(4), 381–384 (2017). [CrossRef]
17. M. Monavarian, A. Rashidi, A. A. Aragon, S. H. Oh, A. K. Rishinaramangalam, S. P. DenBaars, and D. Feezell, “Impact of crystal orientation on the modulation bandwidth of InGaN/GaN light-emitting diodes,” Appl. Phys. Lett. 112(4), 041104 (2018). [CrossRef]
18. M. Monavarian, A. Rashidi, A. A. Aragon, M. Nami, S. H. Oh, S. P. DenBaars, and D. Feezell, “Trade-off between bandwidth and efficiency in semipolar (20-2-1) InGaN/GaN single- and multiple-quantum-well light-emitting diodes,” Appl. Phys. Lett. 112(19), 191102 (2018). [CrossRef]
19. S. L. Chuang and C. S. Chang, “A band-structure model of strained quantum-well wurtzite semiconductors,” Semicond. Sci. Technol. 12(3), 252–263 (1997). [CrossRef]
20. S. Zhu, S. Lin, J. Li, Z. Yu, H. Cao, C. Yang, J. Li, and L. Zhao, “Influence of quantum confined Stark effect and carrier localization effect on modulation bandwidth for GaN-based LEDs,” Appl. Phys. Lett. 111(17), 171105 (2017). [CrossRef]
21. A. Rashidi, M. Monavarian, A. Aragon, A. Rishinaramangalam, and D. Feezell, “Nonpolar m-plane InGaN/GaN micro-scale light-emitting diode with 1.5 GHz modulation bandwidth,” IEEE Electron Device Lett. 39(4), 520–523 (2018). [CrossRef]
22. G.-B. Lin, D.-Y. Kim, Q. Shan, J. Cho, E. F. Schubert, H. Shim, C. Sone, and J. K. Kim, “Effect of quantum barrier thickness in the multiple-quantum-well active region of GaInN/GaN light-emitting diodes,” IEEE Photonics J. 5(4), 1600207 (2013). [CrossRef]
23. C. K. Wang, Y. Z. Chiou, S. J. Chang, C. Y. Chang, T. H. Chiang, T. K. Lin, and X. Q. Li, “On the effect of quantum barrier thickness in the active region of nitride-based light emitting diodes,” Solid-State Electron. 99, 11–15 (2014). [CrossRef]
24. R. A. Oliver, F. C. P. Massabuau, M. J. Kappers, W. A. Phillips, E. J. Thrush, C. C. Tartan, W. E. Blenkhorn, T. J. Badcock, P. Dawson, M. A. Hopkins, D. W. E. Allsopp, and C. J. Humphreys, “The impact of gross well width fluctuations on the efficiency of GaN-based light emitting diodes,” Appl. Phys. Lett. 103(14), 141114 (2013). [CrossRef]
25. M. A. Moram and M. E. Vickers, “X-ray diffraction of III-nitrides,” Rep. Prog. Phys. 72(3), 036502 (2009). [CrossRef]
26. T. Metzger, R. Höpler, E. Born, O. Ambacher, M. Stutzmann, R. Stömmer, M. Schuster, H. Göbel, S. Christiansen, M. Albrecht, and H. P. Strunk, “Defect structure of epitaxial GaN films determined by transmission electron microscopy and triple-axis X-ray diffractometry,” Philos. Mag. A 77(4), 1013–1025 (1998). [CrossRef]
27. D. Yevick and W. Streifer, “Radiative and nonradiative recombination law in lightly doped InGaAsP lasers,” Electron. Lett. 19(24), 1012 (1983). [CrossRef]
28. A. David and M. J. Grundmann, “Droop in InGaN light-emitting diodes: A differential carrier lifetime analysis,” Appl. Phys. Lett. 96(10), 103504 (2010). [CrossRef]
29. S. Marcinkevičius, K. M. Kelchner, L. Y. Kuritzky, S. Nakamura, S. P. DenBaars, and J. S. Speck, “Photoexcited carrier recombination in wide m-plane InGaN/GaN quantum wells,” Appl. Phys. Lett. 103(11), 111107 (2013). [CrossRef]
30. J. T. Kobayashi, N. P. Kobayashi, X. Zhang, P. D. Dapkus, and D. H. Rich, “Structural and optical emission characteristics of InGaN thin layers and the implications for growing high-quality quantum wells by MOCVD,” J. Cryst. Growth 195(1-4), 252–257 (1998). [CrossRef]
