We report the fabrication of quantum dot infrared photodetectors (QDIPs) on silicon (Si) substrates by means of metal wafer bonding and an epitaxial lift-off process. According to the photoluminescence (PL) and x-ray diffraction measurements, the QDIP layer was transferred onto the Si substrate without degradation of the crystal quality or residual strain. In addition, from the PL results, we found that an optical cavity was formed because Pt/Au of the bonding material was served as the back mirror and the facet of the GaAs/air was served as the front mirror. The device performance capabilities were directly compared and peak responsivity was enhanced by nearly twofold from 0.038 A/W to 0.067 A/W.
© 2017 Optical Society of America
InAs/GaAs quantum dot infrared photodetectors (QDIPs) have been the subject of much attention due to their potential for high-temperature operation and their inherent response characteristics of normal infrared (IR) light incidence under an IR detector field [1–4]. Among the various IR photodetectors currently available, HgCdTe detectors have excellent photosensitivity capabilities, but these devices are very expensive and have limitations of the yield of array. Comparing to HgCdTe detectors, QDIPs have the advantage of a relatively low cost, mature III-V materials and a similar thermal expansion coefficient mismatch with silicon (Si). These properties make QDIPs promising alternatives for the monolithic integration with Si readout integrated circuits (ROICs). Conventionally, QDIPs have been integrated with Si ROICs, using indium bumps, to create hybrid chips with which to realize IR imaging. However, the fabrication process with indium bumps is very difficult to conduct and highly complicated. There are several steps when creating hybrid chips with indium bumps, including under bump metallization (UBM), the deposition of a thick indium film, an indium reflow process, flip-chip bonding, the use of under-fill epoxy, and a chemical mechanical planarization (CMP) process to remove the GaAs substrate. First, the deposition of indium film more than 2 μm thick is essential to form the high required height of indium bumps through a reflow process . In addition, the under-fill epoxy process is also unavoidable to overcome the weak material properties of indium. These complicated processes call for an easy and cost-efficient integration method instead of indium bumps.
QDIP directly grown on a Si substrate can reduce these complicated fabrication steps and increase the yield rate. Recently, several groups reported the fabrication of QDIPs on Si substrates via a direct monolithic growing technique using molecular beam epitaxy (MBE) . However, these techniques are difficult to conduct on Si  exact wafers, as off-cut Si wafers have small atomic steps, acting as helpful crystallization seeds during the III-V material epitaxy on the Si substrate. Additionally, numerous variables, such as the substrate temperature, the thickness, and the period of the dislocation filter layer, make it difficult to realize reproducible results using different growth equipment. Furthermore, high thermal treatment above 900 °C inside the MBE chamber to remove native oxide on the surface of Si substrate will damage to actual Si ROICs.
To integrate a QDIP onto a Si substrate practically and monolithically, we studied a metal wafer bonding (MWB) process and an epitaxial lift-off (ELO) technique. In our previous works, we demonstrated p-i-p dots-in-a-well (DWELL) IR photodetectors operating beyond 200 K due to their low dark-current level . A high effective mass of the hole reduces the transport of carriers in the valance band of the p-type QDIPs, leading to a low dark current level. In addition, with regard to the energy band, p-type QDIPs are more apt to generate photo-electrons compared to n-type QDIPs because the hole has three states (heavy, light, and spin-orbit split-off) in the valance band [8,9]. Furthermore, p-type GaAs have good ohmic properties when Pt/Au is used as the bonding material on a Si substrate [10,11]. Therefore, in this work, p-i-p QDIPs on Si substrates are investigated for the first time using the MWB and ELO methods.
We investigated the material and device characteristics of bonded QDIPs on Si substrates (B-QDIPs). The material properties of the B-QDIP layers were evaluated by the photoluminescence (PL) and high-resolution X-ray diffraction (HRXRD) methods, and the device performance capabilities of the fabricated B-QDIP were characterized by measuring the dark-current density voltage (J-V) and spectral response outcomes.
2. Device fabrication
The B-QDIP fabrication process is depicted in Fig. 1. First, the p-i-p QDIP structure was grown on a semi-insulating GaAs (100) substrate using a Veeco GEN 930 MBE. The p-i-p QDIPs consisted of an AlAs 10 nm sacrificial layer, 400-nm-thick P + GaAs contact layers with doping concentration of 5 × 1018 cm−2, 75 nm spacer GaAs layers with a 3 nm p modulation doped GaAs layer, and ten stacks of DWELL active layers. The doping level of the p modulation layer is 3 × 1017 cm−2 for providing two holes per QD. In the DWELL active layer, 1.8 monolayer of InAs QDs, grown in the Stranski-Krastanov growth mode at 480 °C, were sandwiched between a 7.1 Å of In0.18Ga0.82As layer and a 42.7 Å of In0.18Ga0.82As layer. The AlAs sacrificial layer between the QDIP structure and the GaAs wafer was for the ELO process, during which the GaAs wafer was removed after the MWB step. After epitaxial growth, Pt/Au (10/10 nm) was deposited onto both the QDIPs and Si substrates as a bonding material. The deposited Pt/Au on the P + GaAs also served as a bottom electrode on the Si substrates after the MWB process. After deposited Pt/Au on the QDIP sample, pre-patterning step was conducted for the ultra-fast ELO process  by dipping H3PO4:H2O2:DI (1:1:5) solution. Subsequently, Ar plasma was irradiated onto two substrates for surface activation, followed by MWB at room temperature (RT) with a uniaxial pressure of approximately 20 kgf/cm2. After the MWB step, the GaAs donor substrates were separated from the QDIP/Si substrates by dipping each sample in a HF:acetone solution [12,13]. After wafer bonding and ELO steps, the optical properties and changes in the crystal quality were evaluated by the PL and HRXRD methods, respectively. The PL spectra were obtained using a 532-nm pumping laser and an InGaAs photodetector at RT. The double-crystal (DC) ω − 2θ was measured with an HRXRD. The detector device of the B-QDIP was fabricated by depositing Pt/Ti/Pt/Au (150/50/150/200 nm) on the top of the B-QDIP, followed by low-temperature annealing with a furnace at 200°C for two hours in N2 ambient. The as-grown detector was fabricated through a standard fabrication flow. The top and the bottom electrodes were deposited with Pt/Ti/Pt/Au (150/50/150/200 nm). Low-temperature thermal annealing was also conducted at 200 °C for two hours. The spectral response was measured at 77 K with detector window dimensions of 800 × 800 μm2 using a Bruker Vertex 80v Fourier transform infrared spectrometer (FTIR). In our FTIR system, we used a global source which emits mid-infrared radiation along with a KBr beam splitter. The signal of the cooled detector was amplified by a low-noise current amplifier (Keithley 428) and then embedded in the spectrometer. The blackbody responsivity was measured with 500 K of blackbody radiation, and the peak responsivity can be calculated from the relationship between the blackbody responsivity and the cut-off wavelength . The spectral responsivity was calibrated with the calculated the peak responsivity. The noise current was measured at 500 Hz using a HP3562A signal analyzer. The dark J-V characteristics were measured by a HP4156A semiconductor parameter analyzer at 77 K.
3. Results and discussion
3.1 Evaluation of the material properties of the bonded p-i-p QDIPs on Si substrates
Figure 2 shows the PL spectra of the as-grown QDIPs and the B-QDIPs. The as-grown p-i-p QDIP showed a broad PL spectrum with an emission peak at 1206 nm. On the other hand, four sharp peaks of the PL spectra were observed at the B-QDIP. The PL peaks of the B-QDIP were at 992 nm, 1070 nm, 1162 nm, and 1262 nm. The PL peak shift outcomes are due to the optical cavity of the B-QDIP layer . As the reflectance of the Pt/Au layer is high, above 70%, the optical cavity of the B-QDIP was formed after the MWB and ELO process due to the Pt/Au/GaAs as the bottom mirror and the GaAs/air as the front mirror. This is schematically depicted in the inset of Fig. 2. To verify that this shift phenomenon is due to the optical cavity, the emission spectra of a B-QDIP layer on a Si substrate was calculated. The emission spectra can be obtained by multiplying the transmission of the B-QDIP layer and the PL emission spectra of the as-grown QDIPs. Because the absorption of QDs varies among published reports and is very sensitive to the layer structure and growth conditions, it is difficult to assign. For this reason, the transmission of the B-QDIP layer was simply calculated through subtraction of the reflectance from 100% while ignoring the absorption. The reflectance was calculated based on Fresnel’s equation. In Fig. 2, the calculated transmission and emission spectra of the B-QDIP layer are shown. The calculated emission peaks of the B-QDIP show excellent agreement with the experimental PL emission peaks of the B-QDIP sample. This indirectly indicates that the crystal quality and the strain do not change after the MWB and ELO process. Furthermore, from these interesting results, the performance of the B-QDIP can be further improved using a resonant cavity [16,17]. The slight difference in the PL shape between the PL emission spectra of the B-QDIP and the calculated emission spectra of the B-QDIP from the wavelength of 1162 nm to the wavelength of 1262 nm may be due to the absorption of the DWELL active layer in the QDIP.
To verify the crystal quality and strain change of the B-QDIP, DC ω − 2θ measurements of the as-grown QDIPs and the B-QDIP layers were taken. The HRXRD near the GaAs (004) reflection peak is depicted in the Fig. 3. The Si peak was not observed in the B-QDIP because the Si substrate was covered with Pt/Au of bonding material. The full-width-at-half-maximum (FWHM) values of the GaAs peak of both the as-grown QDIP and the B-QDIP are close to the values of 38 arcsec and 39 arcsec, respectively. A periodic superlattice peak is observed near the GaAs (004) reflection peak in both the as-grown QDIP and the B-QDIP. This is due to the multi-stack structure of the DWELL active layers of the QDIP , and the good agreement of the superlattice peaks strongly confirms that the transferring of the QDIP layers onto the Si substrates using the MWB and ELO process does not induce any degradation of the nano scale crystal quality of the InAs QD or residual strain inside the film.
3.2 Device performance
Figure 4 shows the spectral responsivity of the as-grown QDIP and B-QDIP. The as-grown QDIP shows a broad spectral response from 2 μm to 12 μm. This broad spectral response characteristic is due to the many hole states of the p-type QDIP and the DWELL structure [7,8]. The peak wavelength of the as-grown QDIP is 7.8 μm. The overall spectral response profile of the as-grown QDIP follows the responsivity rule, which indicates that the responsivity is linearly proportional to the wavelength. For the B-QDIP, the spectral response shows two narrow response bands at 2.5-3.5 and 4-5 μm and one primary response from 6 μm to 12 μm, having five peaks at wavelengths of 2.98 μm, 4.45 μm, 7.15 μm, 8.23 μm, and 10.35 μm. The peak responsivity values of the as-grown QDIP and the B-QDIP are 0.038 A/W and 0.067 A/W, respectively. The level of the responsivity of the as-grown QDIP is similar to those of QDIPs in other reports [8, 9]. The responsivity of the B-QDIP is nearly two times higher than that of the as-grown QDIP at the primary response band (6-12 μm). It is currently challenging to find the causes of the changed spectral response, as a simulation of the absorption spectrum of a B-QDIP is not easy. The lack of information about the refractive index about 3 dimensional QDs and other materials in mid-IR spectrum at cryo-temperature levels makes this difficult to calculate. However, from previous PL results, we consider that this interesting change in the spectral responsivity can be attributed to the interference effect inside the optical cavity of the B-QDIP. The steep rise of the spectral response at 6 μm is one of evidence which supports our contention. The spectral response of the detector should show a steeper slope of the drop at the cut-off wavelength (λc) than that of the rise at wavelengths shorter than λc. We strongly believe that this unusual spectral profile of the steep rise is altered from the original spectral response of the as-grown QDIP detector due to extrinsic factors such as the interference of light, not because of any change in the inherent material or device characteristics, because the previous PL and XRD results show that the material characteristics of the B-QDIP did not change after the MWB and ELO processes. Furthermore, the increased responsivity suggests that the back reflector of Pt/Au as a bonding material reflects light and thus is more likely to absorb the light, leading to an enhancement of the detector performance. For these reasons, we suggest that the B-QDIP has wavelength-selective properties due to the optical cavity. The related issue of the cavity effects of the B-QDIP will be studied in detail in a future publication. The dark J-V curves of the as-grown QDIP and B-QDIP at 77 K are plotted in the inset of Fig. 4. The dark-current density levels of the as-grown QDIP and B-QDIP show negligible differences and the values of the dark-current density are 1.03 × 10−8 A/cm2 and 8.82 × 10−9 A/cm2 at 0 V, respectively. Note that there was no dark current blocking layer in p-i-p QDIP, indicating that there are still large rooms to reduce the dark current level of p-i-p QDIP. The small difference of dark current density levels stems from device-to-device variations, suggesting that the MWB and ELO processes can be used successfully to integrate a QDIP on a Si substrate, with good ohmic characteristics between the p-GaAs and an n-Si substrate even at cryo-temperature levels.
In this equation, Ri is the responsivity (A/W), Ad is the detector area, ∆f is the noise-equivalent bandwidth, and in is the noise current. The D* of the as-grown QDIP and the B-QDIP was obtained from the peak responsivity at 7.8 μm and 7.15 μm, respectively. The noise current in was measured at 77 K. The maximum D* levels of the as-grown QDIP and the B-QDIP were 3.77 × 109 cm·Hz1/2/W at −0.05 V and 6.66 × 109 cm·Hz1/2/W at −0.05V, respectively. The inset of Fig. 5 shows the noise current level of the as-grown QDIP and the B-QDIP. As the noise current level of the B-QDIP was almost identical to that of the as-grown QDIP, the enhanced detectivity of the B-QDIP is mainly due to the increased responsivity. These results indicate that the detector performance can be enhanced further by adjusting the cavity length of the detector. According to the results presented here, photodetectors based on the GaAs material, including quantum well and quantum dot infrared photodetectors, can be integrated with Si substrates easily and inexpensively using the MWB and ELO techniques. Furthermore, as mentioned above, we suggest that such detectors can still realize many improvements to their performance capabilities when adopting a resonant cavity mode [16,17]. Finally, these fabrication skills can also be used to create array devices with a patterned ELO technique, and they are quite promising to deploy on a practical basis as well .
We demonstrate for the first time the fabrication of a QDIP on a Si substrate using the MWB and ELO techniques instead of an indium bump, which is the complex method for the integration of a QDIP onto a Si substrate. The transferred material quality on the Si substrate was evaluated by the PL and HRXRD methods. According to the PL results, we noted the formation of a cavity due to the bonding material of Pt/Au and the facet of the GaAs/air. From HRXRD results, there was no degradation of crystal quality or any residual strain inside the transferred QDIP layer. The detector performance capabilities were directly compared between the as-grown QDIP and the B-QDIP by spectral responsivity, dark J-V characteristic and detectivity measurements. The dark J-V outcomes show identical dark-current levels, indicating that the fabrication of the QDIP by MWB and ELO was successful. The spectral responsivity of the as-grown QDIP shows a broad response from 2 μm to 12 μm due to the numerous hole states and the DWELL active layer. On the other hand, the B-QDIP shows three windows with two narrow response bands and one primary response band at 6-12 μm. In addition, the responsivity of the B-QDIP was enhanced by approximately two times compared to that of the as-grown QDIP at the primary response band. This changed performance can be attributed to the interference effect within the optical cavity of the B-QDIP, as confirmed by the PL results. The detectivity of the B-QDIP was also enhanced due to the increased responsivity. These results indicate that the MWB method can be used to integrate QDIPs and Si substrates inexpensively and practically and that the detector performance capabilities can be enhanced further by the MWB and ELO techniques considering the effects of the aforementioned cavity.
This work is supported by the KIST institutional program of Flag-ship (2E26420), Brain Korea 21 Plus project in 2017, and partly supported by the National Research Foundation of Korea (NRF) grant (Grant No.2015004870).
References and links
1. D. Pal and E. Towe, “Characteristics of high-operating-temperature InAs/GaAs quantum-dot infrared detectors,” Appl. Phys. Lett. 88(15), 153109 (2006). [CrossRef]
2. S. Chakrabarti, A. D. Stiff-Roberts, P. Bhattacharya, S. Gunapala, S. Bandara, S. B. Rafol, and S. W. Kennerly, “High-temperature operation of InAs-GaAs quantum-dot infrared photodetectors with large responsivity and detectivity,” IEEE Photonics Technol. Lett. 16(5), 1361–1363 (2004). [CrossRef]
4. S. Raghavan, P. Rotella, A. Stintz, B. Fuchs, S. Krishna, C. Morath, D. A. Cardimona, and S. W. Kennerly, “High-responsivity, normal-incidence long-wave infrared (λ ∼ 7.2 μm) InAs/In0.15Ga0.85As dots-in-a-well detector,” Appl. Phys. Lett. 81(8), 1369–1371 (2002). [CrossRef]
5. Y. Huang, C. Lin, Z. H. Ye, and R. J. Ding, “Reflow flip-chip bonding technology for infrared detectors,” J. Micromech. Microeng. 25(8), 085009 (2015). [CrossRef]
6. J. Wu, Q. Jang, S. Chen, M. Tang, Y. I. Mazur, Y. Maidaniuk, M. Benamara, M. P. Semtsiv, W. T. Masselink, K. A. Sablon, G. J. Salamo, and H. Liu, “Monolithically integrated InAs/GaAs quantum dot mid-infrared photodetectors on silicon substrates,” ACS Photonics 3(5), 749–753 (2016). [CrossRef]
7. M. S. Park, V. Jain, E. H. Lee, S. H. Kim, H. Pettersson, Q. Wang, J. D. Song, and W. J. Choi, “InAs/GaAs p-i-p quantum dots-in-a-well infrared photodetectors operating beyond 200 K,” Electron. Lett. 50(23), 1731–1733 (2014). [CrossRef]
8. Y. F. Lao, S. Wolde, A. G. U. Perera, Y. H. Zhang, T. M. Wang, H. C. Liu, J. O. Kim, T. S. Sandy, Z. B. Tian, and S. S. Krishna, “InAs/GaAs p-type quantum dot infrared photodetector with higher efficiency,” Appl. Phys. Lett. 103(24), 241115 (2013). [CrossRef]
9. A. G. U. Perera, Y. F. Lao, S. Wolde, Y. H. Zhang, T. M. Wang, J. O. Kim, T. S. Sandy, Z. B. Tian, and S. S. Krishna, “InAs/GaAs quantum dot and dots-in-well infrared photodetectors based on p-type valence-band intersublevel transitions,” Infrared Phys. Technol. 70, 15–19 (2015). [CrossRef]
11. J. R. Bickford, D. Qiao, P. K. L. Yu, and S. S. Lau, “Electrical characterization of GaAs metal bonded to Si,” Appl. Phys. Lett. 89(1), 012106 (2006). [CrossRef]
12. D. M. Geum, M. S. Park, J. Y. Lim, H. D. Yang, J. D. Song, C. Z. Kim, E. Yoon, S. Kim, and W. J. Choi, “Ultra-high-throughput production of III-V/Si wafer for electronic and photonic applications,” Sci. Rep. 6(1), 20610 (2016). [CrossRef] [PubMed]
13. F. L. Wu, S. L. Ou, R. H. Horng, and Y. C. Kao, “Improvement in separation rate of epitaxial lift-off by hydrophilic solvent for GaAs solar cell applications,” Sol. Energy Mater. Sol. Cells 112, 233–240 (2014). [CrossRef]
14. E. L. Dereniak and G. D. Boreman, Infrared Detectors and Systems (Wiley, 1996), Chap 6.
15. J. I. Maeda, Y. Sasaki, N. Dietz, K. Shibahara, S. Yokoyama, S. Miyazaki, and M. Hirose, “High-rate GaAs epitaxial lift-off technique for optoelectronic integrated circuits,” Jpn. J. Appl. Phys. 36(3), 1554–1557 (1997). [CrossRef]
16. C. C. Wang and S. D. Lin, “Resonant cavity-enhanced quantum-dot infrared photodetectors with sub-wavelength grating mirror,” J. Appl. Phys. 113(21), 213108 (2013). [CrossRef]
17. T. Asano, C. Hu, Y. Zhang, M. Liu, J. C. Campbell, and A. Madhukar, “Design consideration and demonstration of resonant-cavity-enhanced quantum dot infrared photodetectors in mid-infrared wavelength regime (3-5 μm),” IEEE J. Quantum Electron. 46(10), 1484–1494 (2010). [CrossRef]
18. Y. H. Kim, K. Y. Ban, and C. B. Honsberg, “Multi-stacked InAs/GaAs quantum dots grown with different growth modes for quantum dot solar cells,” Appl. Phys. Lett. 106(22), 222104 (2015). [CrossRef]
19. S. H. Kim, D. M. Geum, M. S. Park, H. S. Kim, J. D. Song, and W. J. Choi, “Fabrication of high-quality GaAs-based photodetector arrays on Si,” Appl. Phys. Lett. 110(15), 153505 (2017). [CrossRef]