A p-i-n structure with near-UV(n-UV) emitting InGaN/GaN multiple quantum well(MQW) structure stacked on a green unipolar InGaN/GaN MQW was epitaxially grown at the same sapphire substrate. Photon recycling green light-emitting diodes(LEDs) with vertical-conduction feature on silicon substrates were then fabricated by wafer bonding and laser lift-off techniques. The green InGaN/GaN QWs were pumped with n-UV light to reemit low-energy photons when the LEDs were electrically driven with a forward current. Efficiency droop is potentially insignificant compared with the direct green LEDs due to the increase of effective volume of active layer in the optically pumped green LEDs, i.e., light emitting no longer limited in the QWs nearest to the p-type region to cause severe Auger recombination and carrier overflow losses.
© 2015 Optical Society of America
White light-emitting diodes (LEDs) present promising next-generation general lighting sources . Commercial white LEDs are mainly constructed by two methods. The most popular method for producing white LEDs involves bringing together GaN-based blue LEDs and yellow phosphor . InGaN LEDs emitting blue wavelength exhibit relatively high external quantum efficiency (EQE), but green LEDs, which are achieved by increasing the indium concentration in the InGaN active region, show a notable drop in efficiency, especially for operating at high injection current density. In addition to the Auger recombination issue, this efficiency loss could be attributed to a number of material issues, such as high-density dislocation and strong polarization field . An alternative approach for combining a blue LED and green phosphor together to construct a green emitter was demonstrated . This down-conversion pumping could prevent the green gap of InGaN-based LEDs and alleviate the effect of efficiency droop at high current densities. However, the down-converted green emission from phosphors exhibited a significantly broad emission peak compared with direct electroluminescence (EL) green LEDs. Considering the LEDs’ application for projection, this broader emission spectrum may lead to cross talk and thereby limit the efficiency of the light source. Broad emission could be alleviated by using InGaN green MQW to replace green phosphor, while the blue MQW as the pumping source was integrated with the green MQW using a bonding technique . In this article, optical properties of photon recycling vertical green GaN-based LEDs were systematically studied. Considering the spectrum of photon recycling LEDs (PRLEDs), the green peak originates from n-UV light generated by electrical pumping to optically pump the epitaxially stacked green MQW. In other words, a p-i-n structure with n-UV emitting InGaN/GaN MQW structure stacked on a green unipolar InGaN/GaN MQW was epitaxially grown on the same sapphire substrate. Next, the proposed green GaN-based LEDs featuring a vertical conduction structure on silicon substrate were fabricated by wafer bonding and laser lift-off (LLO) techniques . When the n-UV MQW was electrically driven with forward current, the green InGaN/GaN QWs were therefore pumped by n-UV light to reemit photons with lower energies. As a result, the emission exhibited a resultant spectrum including two peaks at approximately of 400 and 510 nm. In contrast to the previous reports , our proposed device structure would prevent a problem of UV light and heat possibly damaging the adhesion layer, which combines UV pumping LED and MQW converter, especially as the devices are operated with high input power. The detailed fabrication procedures and characterizations of experimental LEDs are presented in the following sections.
2. Device fabrication and experiment methods
InGaN-based LED structures used in this study were grown on c-face sapphire substrates by a metal–organic vapor-phase epitaxy reactor. The layer structure consisted of a 30-nm-thick low-temperature (530 °C) GaN nucleation layer, followed by a 3-µm-thick undoped GaN (u-GaN) grown at 1000 °C with chamber pressure at 500 torr. Thereafter, a 30-pair green InGaN/GaN MQW converter containing a 2.5-nm-thick In0.28Ga0.72N well separated by a 9-nm-thick GaN barrier in each pair was grown on the u-GaN layer. Next, a top p–i–n heterostructure consisting of a 4-µm-thick Si-doped n+-GaN (n ≈3 × 1018 cm−3), a 12-pair InGaN/AlGaN MQW containing a 3-nm-thick InxGa1-xN well separated by a 12-nm-thick Al0.1Ga0.9N barrier in each pair, a 20-nm-thick Mg-doped p-Al0.15Ga0.85N (p ~1 × 1017 cm−3), and a 0.2-µm-thick Mg-doped p-GaN (p ≈5 × 1017 cm−3) were sequentially grown on the green InGaN/GaN MQW converter. In the light of photoluminescence (PL) spectra, the typical green MQW converter emitted a wave length of approximately 510 nm. The typical peak wavelength of the PL spectrum obtained from InxGa1-xN/Al0.1Ga0.9N MQW was approximately 400 nm when the x was 0.05, denoting n-UV MQW. Figure 1(a) shows the schematic layer structure of the above-mentioned epitaxial heterostructure. After epitaxial growth, a bilayer metal of Ni/Ag (1/200 nm) was deposited onto the p-GaN top layer to serve as ohmic/reflector contact layer . After the formation of the reflector layer, the barrier layer, which was configured to alleviate the diffusion of Ag from reflector layer to bonding layer and consists of a 200-nm-thick TiW layer and a 50-nm-thick Pt layer, was deposited between the reflector and bonding layers. To fabricate GaN-based LEDs with vertical conduction, a bilayer metal of Ti/Au (20 nm /1500 nm) and a 3-µm-thick In layer were deposited onto the Si substrates to serve as ohmic contact and bonding layer, respectively. These Si wafers served as receptors for the wafer bonding process. After the bonding process, the sapphire substrate was removed by laser lift-off technique to expose the u-GaN layer. The InxGa1 − xN/GaN-based heteroepitaxial layers were thus transferred to the Si substrate with the u-GaN top layer. Then, the samples were treated in potassium hydroxide (concentration of 3 M) solution at an elevated temperature of 60 °C to texture the u-GaN layer for enhancing light extraction efficiency. After the surface texturing process, selective-area dry etching was performed to expose the n+-GaN layer. Ti/Al/Ni/Au (20/30/150/2000 nm) metal layers were then deposited onto the exposed n+-GaN layer to form n-type ohmic contacts (cathode electrodes) on the wafers. Finally, the Si substrates were thinned to 150 µm and coated with a bilayer metal of Ti/Au (50 nm /500 nm) metals to serve as back ohmic contact layer. Figure 1(b) shows a typical top-view microscope image obtained from the proposed vertical InGaN/GaN/Si green PRLEDs. The dashed line in Fig. 1(b) indicates where the cross section corresponds to a schematic layer structure of the vertical green PRLEDs, as shown in Fig. 1(c). All experimental LED chips used in the present study possessed an area of 1.25 × 1.25 mm2. Notably, conventional vertical UV and green LEDs with emission wavelengths of 400 and 510 nm, respectively, were also prepared for reference and were labeled as direct-UV and direct-green LEDs, respectively. In other words, the light emission of LEDs originates from direct electrical injection, unlike the green emission of PRLEDs that originate from optical pumping. Figure 1(d) shows the schematic layer structure of conventional vertical InGaN/GaN/Si green LED. Direct-UV LEDs with schematic layer structure similar to direct-green LEDs was not shown here. Room-temperature current–voltage (I–V) characteristics of experimental LEDs were measured using the HP-4156C semiconductor parameter analyzer. The light output power–current (L–I) characteristics of the LEDs were measured using a calibrated integrating sphere combined with a current source meter. Photoluminescence (PL) measurements of the PRLED wafers were performed at 300 K using a diode laser emitting at 375 nm as excitation source.
3. Results and discussion
Figure 2 illustrates the typical I–V characteristics of vertical InGaN/GaN/Si direct-UV LEDs and green PRLEDs. With a forward driving current of 350 mA, typical forward voltages were 2.76 and 2.75 V for direct-UV LEDs and green PRLEDs, respectively. At a reverse bias of 20V, typical leakage currents measured from the direct-UV and green PRLEDs were 0.9 and 0.3 μA, respectively. This result implied that the green InGaN/GaN MQW inserted between the sapphire substrate and the In0.05Ga0.95N/Al0.1Ga0.9N n-UV MQW did not cause a significantly negative effect on electrical properties. The absorption of the InGaN/GaN MQW converter depends on band gap, thickness, (i.e., number of InGaN QWs in the MQW converter), and excitation light wavelength (λexcit).
To avoid degradation of material quality ascribed to strain release and achieve PRLEDs with the maximum possible efficiency, λexcit was set at around 400 nm for exciting the 30-pair green MQW converter. In general, InGaN/GaN epitaxial layers grown on sapphire present dark gray appearance after epitaxial growth if total thickness exceeds critical thickness for InGaN layers, indicating indium segregation inside the InGaN layers. As shown in the Fig. 3(a), the typical mapping diagram of PL peak wavelength at around 510 nm taken from a typical green PRLED wafer indicated that the uniformity of emission wavelength(i.e., the spatial distribution of indium content in the QWs) is acceptable. The inset of Fig. 3(b) is a photograph taken from the PRSLED wafers with semitranslucent appearance. Figure 3(b) is a representative PL spectrum taken from the green PRLED wafers.
Figure 4(a) shows the typical emission spectra of green PRLEDs driven at various forward currents. The typical EL spectra of green PRLEDs exhibited two peaks at around 400 and 510 nm. These dual-peak spectra indicated that excited source photons were not completely absorbed by the green MQW converter in the PRLEDs. Direct-UV LEDs and direct-green LEDs showed single-peak spectra at around 400 and 510 nm, respectively, as shown in the Fig. 4(b). Based on the current-dependent EL measurements, the light conversion percentages from n-UV to green decreased from 55 to 45% as the driving current densities were increased from 1 to 100 A/cm2, as shown in the Fig. 4(c). Since the green emission from direct-green LEDs and green PRLEDs originate from different pumping mechanisms, the optical properties are different between these LEDs. Notably, the growth conditions and layer structure of PRLEDs are not optimized at present. Therefore, the typical output power of green PRLEDs remains lower than those of direct-green LEDs, as shown in Fig. 5(a). In addition, the effective area of quantum well of the green PRLEDs was lower than the direct-green LEDs because a part of quantum well was etch away to form the top electrodes, as shown in Figs. 1(c) and 1(d). The reduced quantum well region may cause a reduction in light output power. The green PRLEDs and direct-green LEDs may have comparable light outputs as the efficiency of pumping source(UV MQW) and green MQW converter in PRLED could be further raised. In addition to the use of high-efficiency UV-emitting MQW, the loss of Stokes shift should be able to be further reduced if a blue-emitting MQW acts as pumping source to replace the present n-UV-emitting MQW. Figure 5(b) shows the EQE as a function of current density of the aforesaid green PRLEDs. To evaluate the droop effect on efficiency as a function of current density, normalized EQE was also depicted in Fig. 5(b). In addition, conventional vertical direct-green LEDs with MQW active layers similar to the MQW converter inside the proposed green PRLEDs were also prepared for comparison, and the emission wavelengths of the LEDs were around 510 nm. Direct-green LEDs and green PRLEDs showed peak EQEs at around 50 and 150 mA, respectively. Evidently, the efficiency droop observed in direct-green LEDs is more severe than that in green PRLEDs. The efficiency droop observed from the n-UV peak of green PRLEDs resembled the green peak, as shown in Fig. 5(b). Compared with the n-UV emission in PRLEDs, the significant efficiency droop in direct-green LEDs could be attributed to the fact that built-in polarization increases with increasing indium composition in the InGaN QWs, which is due to the strain-induced piezoelectric polarization. This increase enhances the separation of electron–hole pairs in QWs and/or the electron overflow to the p-type region . On another hand, the Auger recombination loss in direct-green LEDs is higher than that of green PRLEDs because of the carrier accumulation in QWs closest to the p-side layer. In other words, electrically injected carriers readily accumulate in QWs close to the p-type region, especially for LEDs with high In-containing InGaN/GaN heterostucture and/or operated under high current density. Although the origin of the efficiency droop remains controversial , a number of reports ascribe this phenomenon to Auger recombination loss [8,9]. Considering green PRLEDs, the Auger recombination loss is expected to be insignificant compared with that of direct-green LEDs because of the increased effective volume of the active layer in PRLEDs, i.e., light emitting is no longer limited in QWs nearest to the p-type region. As shown in the Fig. 5(b), one may ask why n-UV emission shows a minor efficiency droop compared with green emission in PRLEDs if Auger recombination loss dominates the efficiency droop for electrically pumped n-UV LED. In fact, the result presents a good example explaining that the effect of built-in polarization in GaN-based LEDs on the efficiency droop could not be ignored . In other words, strain-induced polarization field would lead to more carrier overflow outside the green MQW converter compared with the n-UV MQW.
Spontaneous and strain-induced polarization fields are known to cause a blue shift of emission spectra of InGaN/GaN-based LEDs with increasing injected carrier density because of the screening of quantum-confined Stark effect (QCSE) [11–14]. Considering direct-green LEDs and green PRLEDs, as shown in Fig. 6(a), the blue shift of peak wavelengths with an increase in injection current could be attributed to the possible mechanism in which increasing injection current causes the screening of QCSE. For direct-green LEDs, the screening of QCSE originated from the increase in injected carriers and external bias voltage. A similar phenomenon was also observed in the n-UV emission peak of PRLEDs. However, the blue shift of green emission peak of PRLEDs could be mainly attributed to the increase in optical pumping intensity because the external bias was not directly applied on the green MQW converter. In general, it accompanies a reduction in full width at half maximum (FWHM) of emission spectra if screening effect contributes to the transition process of carriers. Figure 6(b) shows the FWHM of emission spectra as a function of injection currents. The FWHM of the green peak observed from PRLEDs decreased with increasing injection current. This trend is consistent with the aforesaid contention that the blue shift of the emission spectra accompanying a reduction in FWHM could be due to the screening of QCSE. Although direct-green LEDs and the n-UV peak of PRLEDs showed a similar trend in the reduction in FWHM when the injection current was lower than 50 mA, as shown in the inset of Fig. 6(b), the FWHM of the emission spectra observed from direct-green LEDs and the n-UV peak of PRLEDs both increased with increasing injection current when the injection current was higher than 50 mA. This result implies that the shift of the emission spectra observed from direct-green LEDs and the n-UV peak of PRLEDs may be caused by other mechanisms except the screening of QCSE. In addition to different trends in current-dependent FWHM of emission spectra, one can see that the FWHMs of the emission spectra obtained from green PRLEDs are larger than those obtained from direct-green LEDs. This result is consistent with the aforesaid contention that green emission from direct-green LEDs is mainly limited in minor QWs nearest to the p-type region, unlike the green emission of PRLEDs that originated from most of QWs in green MQW converter. Given that the inhomogeneity of indium composition and layer thickness fluctuation of InGaN MQWs, self-organized In-rich clusters are generated in InGaN well layers, which result in potential fluctuation of the energy band gap [15–17], the variations of transition energy between different QWs would cause significant broadening of the emission spectrum. Considering PRLEDs, as shown in Fig. 6(a), one can see that the FWHM of the emission spectra exhibited different trends with increasing injection current between n-UV and green peaks, which originated from electrical and optical pumping, respectively. Although the band-filling effect of localized states (induced by In-rich clusters) appears to explain that the blue shift with increasing injection current accompanied the broadened FWHM of emission spectra from n-UV MQW, indium -rich clusters are not likely to be formed in n-UV MQW because of the low indium content in InGaN layers. Therefore, the blue shift of emission spectra with increasing injection current into n-UV QWs of PRLEDs should be due to QCSE screening. However, the blue shift observed from the n-UV emission peak in PRLEDs did not exhibit reduction of FWHM but the blue shift accompanied an increase in FWHM instead when the injection current was higher than 50 mA. We can tentatively suggest that another mechanism to cause the broadening of n-UV emission peak, and its impact surpassed the screening effect of QCSE on the reduction of FWHM. On one hand, the increased FWHM of the emission spectra observed from the n-UV peak could be tentatively attributed to the fact that junction temperatures increased with injection currents applied to n-UV MQW in PRLEDs, thereby inducing thermal broadening in the EL spectra. On the other hand, the green emission originated from optical pumping of n-UV light without external currents to directly drive the green MQW converter in PRLEDs. Consequently, the effect of thermal broadening on the green emission peak would be minor in comparison with the n-UV emission peak of PRLEDs. To clarify this point, temperature-dependent FWHM of emission spectra obtained from PRLEDs were performed. Figure 6(c) shows the FWHM as a function of temperature obtained from green and n-UV peaks of PRLEDs. The LEDs were driven at a relative lower current of 175 mA to alleviate junction heating. Evidently, FWHMs increased with increasing temperature because of a larger energy spread of the carrier distribution and phonon scattering . This result is consistent with the aforesaid contention that increased junction heating causes the thermal broadening of EL spectrum emitting from the n-UV MQW and green MQW converter. At a given driving current and an ambient temperature in the PRLEDs, however, junction heating caused by current injection would not induce a significant effect on the broadening of emission spectrum emitting from the green MQW converter, as shown in Fig. 6(b). In other words, the increased currents into PRLED induced junction heating mainly affected the n-UV MQW rather than the green MQW converter to cause thermal broadening on n-UV emission spectra, and the effect of junction heating on emission spectra FWHM surpassed the screening of QCSE as the injection currents were raised. However, the blue shift of emission spectra with increasing injection currents was mainly governed by QCSE screening and/or band filling. Similar results were also observed on direct-green LEDs, as shown in Fig. 6(a).
InGaN/GaN-based green photon recycling LEDs (PRLEDs) with vertical conduction feature on silicon substrates have been demonstrated using wafer bonding and laser lift-off techniques. PRLEDs featured an n-UV MQW as pumping source, and a green MQW converter was epitaxially grown on the same sapphire substrate. In contrast with previous reports, our proposed device structure avoids a problem of damaging the organic-based and transparent adhesion layer by generated heat and UV light, which combines UV pumping LED and MQW converter, especially as the devices are operated with high input power. For the proposed PRLEDs, the larger effective active volume and lower thermal heating effect on the emitting layer led to an alleviation of efficiency droop compared with conventional direct green LEDs.
This work was supported from the Ministry of Science and Technology Taiwanfor the financial support under contract Nos. NSC-101-2221-E-218-012-MY3, NSC-101-2221-E-006-171-MY3, NSC-100-2112-M-006-011-MY3 and NSC-100-3113-E-006-015.
References and links
2. G. Verzellesi, D. Saguatti, M. Meneghini, F. Bertazzi, M. Goano, G. Meneghesso, and E. Zanoni, “Efficiency droop in InGaN/GaN blue light-emitting diodes: Physical mechanisms and remedies,” J. Appl. Phys. 114(7), 071101 (2013). [CrossRef]
3. H. Wu, X. Zhang, C. Guo, J. Xu, M. Wu, and Q. Su, “Three-band white light from InGaN-based blue LED chip precoated with green/red phosphors,” IEEE Photon. Technol. Lett. 17(6), 1160–1162 (2005). [CrossRef]
4. D. Schiavon, M. Binder, A. Loeffler, and M. Peter, “Optically pumped GaInN/GaN multiple quantum wells for the realization of efficient green light-emitting devices,” Appl. Phys. Lett. 102(11), 113509 (2013). [CrossRef]
5. J.-K. Sheu, F.-B. Chen, S.-H. Wu, M.-L. Lee, P.-C. Chen, and Y.-H. Yeh, “Vertical InGaN-based green-band solar cells operating under high solar concentration up to 300 suns,” Opt. Express 22(S5), A1222–A1228 (2014). [CrossRef] [PubMed]
6. B. Galler, P. Drechsel, R. Monnard, P. Rode, P. Stauss, S. Froehlich, W. Bergbauer, M. Binder, M. Sabathil, B. Hahn, and J. Wagner, “Influence of indium content and temperature on Auger-like recombination in InGaN quantum wells grown on (111) silicon substrates,” Appl. Phys. Lett. 101(13), 131111 (2012). [CrossRef]
7. U. Ozgur, H. Liu, X. Li, X. Ni, and H. Morkoc, “GaN-based light-emitting diodes: efficiency at high injection levels,” Proc. IEEE 98(7), 1180–1196 (2010). [CrossRef]
8. D. Schiavon, M. Binder, M. Peter, B. Galler, P. Drechsel, and F. Scholz, “Wavelength‐dependent determination of the recombination rate coefficients in single‐quantum‐well GaInN/GaN light emitting diodes,” Phys. Status Solidi. B250(2), 283–290 (2013).
9. F. Bertazzi, M. Goano, and E. Bellotti, “Numerical analysis of indirect Auger transitions in InGaN,” Appl. Phys. Lett. 101(1), 011111 (2012). [CrossRef]
10. M. F. Schubert, J. Xu, J. K. Kim, E. F. Schubert, M. H. Kim, S. Yoon, S. M. Lee, C. Sone, T. Sakong, and Y. Park, “Polarization-matched GaInN/AlGaInN multi-quantum-well light-emitting diodes with reduced efficiency droop,” Appl. Phys. Lett. 93(4), 041102 (2008). [CrossRef]
11. J. K. Sheu, G. C. Chi, Y. K. Su, C. C. Liu, C. M. Chang, W. C. Hung, and M. J. Jou, “Luminescence of an InGaN/GaN multiple quantum wells light-emitting diode,” Solid-State Electron. 44(6), 1055–1058 (2000). [CrossRef]
12. T. Mukai, M. Yamada, and S. Nakamura, “current and temperature dependences of electroluminescence of InGaN-Based UV/blue/green light-emitting diodes,” Jpn. J. Appl. Phys. 37(2), L1358–L1360 (1998). [CrossRef]
13. Y. L. Li, T. Gessmann, E. F. Schubert, and J. K. Sheu, “Carrier dynamics in nitride-based light-emitting p-n junction diodes with two active regions emitting at different wavelengths,” J. Appl. Phys. 94(4), 2167–2172 (2003). [CrossRef]
14. S. C. Shei, J. K. Sheu, C. M. Tsai, W. C. Lai, M. L. Lee, and C. H. Kuo, “Emission mechanism of mixed-color InGaN/GaN multi-quantum well light-emitting diodes,” Jpn. J. Appl. Phys. 45(4A), 2463–2466 (2006). [CrossRef]
15. E. Kuokstis, J. W. Yang, G. Simin, M. Asif Khan, R. Gaska, and M. S. Shur, “Two mechanisms of blueshift of edge emission in InGaN-based epilayers and multiple quantum wells,” Appl. Phys. Lett. 80(6), 977 (2002). [CrossRef]
16. J. H. Na, R. A. Taylor, K. H. Lee, T. Wang, A. Tahraoui, P. Parbrook, A. M. Fox, S. N. Yi, Y. S. Park, J. W. Choi, and J. S. Lee, “Dependence of carrier localization in InGaN/GaN multiple-quantum wells on well thickness,” Appl. Phys. Lett. 89(25), 253120 (2006). [CrossRef]
17. F. K. Yam and Z. Hassan, “InGaN: An overview of the growth kinetics, physical properties and emission mechanisms,” Superlattices Microstruct. 43(1), 1–23 (2008). [CrossRef]
18. E. F. Schubert, Light-Emitting Diode (Second Edition, Cambridge University Press, 2006.), pp.88–90.