Open Access
Add to BibSonomy
Abstract
The heterogeneous integration of III-V semiconductors with the Silicon platform enables the merging of photon sources with Silicon electronics while allowing the use of Silicon mature processing techniques. However, the inherent sufficient quality of III-Vs’ native oxides made imperative the use of deposited interfacial oxide layers or adhesives to permit the bonding. Here we present a novel approach enabling the heterogeneous integration of structured III-V semiconductors on silicates via molecular bonding at 150 °C, much below the CMOS degradation temperature, is presented. The transfer of 235 nm thick and 2 mm long InGaP waveguides with widths of 4.65, 2.6 and 1.22 μm on 4 μm thick Si thermal oxide, with optional SX AR-N 8200.18 cladding, has been experimentally verified. Post-processing of the 1.20 and 0.60 μm input/output tappers has allowed the implementation of double-inverse tapers. The minimal processing requirements and the compatibility with transferring non-cladded structures of the presented technique are demonstrated. The quality of the transferred waveguides bonding interface and their viability for non-linear optics applications has been tested by means of the surface contribution to the optical non-linearity via modal phase-matched second-harmonic generation.
Published by Optica Publishing Group under the terms of the Creative Commons Attribution 4.0 License. Further distribution of this work must maintain attribution to the author(s) and the published article's title, journal citation, and DOI.
1. Introduction
The relevance of heterogeneous integration of III-V-on-Silicon (Si) platform, an essential aspect of Si photonics (SiP), has enjoyed a tremendous growth in the last decade [1]. Since the profitable commercialization of the first market-available heterogeneously integrated product by Intel in 2016 [2], wafer bonding (WB) technologies are becoming indispensable to be able to offer SiP devices at low processing cost [3]. Although the surface and cleanliness conditions needed for successful WB might be relatively trivial for industrial foundries [4], the bonding of full III-V wafers on Si/SiO$_2$ substrates is not practical for most academic labs. The high cost of bonding-ready III-V wafers and WB tools prevents the development of full WB technologies in academia. The practical inconveniences of WB are commonly circumvented by flip-chip integration with micro-transfer printing-like processing [5], or the implementation of suspended structures [6]. Material combinations exhibiting high enough etching selectivity to withstand the relatively long sacrificial layer etchings become imperative with those approaches. As a result, implementing sacrificial etching holes and sustaining tethers on the hard mask increases the design complexity and processing steps [7]. Additionally, the essential sacrificial layer etching performed during coupon and structure release leads to a convex profile as the edges of the structures remain in contact with the etching solution longer than the center part, more distant to the etching access holes [8]. Nevertheless, III-V heterogeneous integration technologies are not fully mature, inhibiting the implementation of III-V materials to a broad application range. InGaP, exhibits substantially larger second-order nonlinear coefficient [9] and refractive index [10] than other commonly used nonlinear materials such as LiNbO$_3$. Although wafer-scale III-V films have been successfully transferred [11], the impossibility of commercially acquiring III-V on insulator prevents the utilization of this attractive material candidates for nonlinear optics applications.
Here, we present a non-cladded compatible heterogeneous integration technique to transfer tens of nm thick III-V layers to silicates which features various advantages. By structuring the III-V layers prior to transfer, the cleanliness requirements are relaxed as <3${\% }$ of the contact area contains structures. Particles within the bonding interface lead to the deficient bonding of a single structure rather than a interfacial void of diameter d $\approx$ 2000 h on the wafer, being h the particle height [12]. Furthermore, although interfacial oxide deposition is imperative for III-V WB due to the inherent sufficient quality of III-Vs’ native oxides [13], this technique allows the bonding to silicates via molecular bonding of the III-V’s native oxide and the silicate. The material-independent bonding process, following the reaction in Eq. (1) (being M a large electronegativity metal) [14,15], does not require additional adhesive layers which could induce stress and inhomogeneities [16,17].
The bonding is performed at 150 $^{\circ }$C, much below the CMOS degradation temperature, making it compatible with the Si platform while circumventing the $\approx$ 90 ${\% }$ mismatch in thermal expansion coefficients of InGaP and Si/SiO$_2$ [18]. Similar reduction of the common $\approx$ 500 $^{\circ }$C post-bonding annealing treatments have been reported in other works when performing surface plasma treatment before the bonding [19]. The thermal strain accumulated during bonding or annealing processes is commonly relaxed through thermal cracks [20], limiting the temperature at which those processes can be performed. The trapping of the gas by-products, shown in Eq. (1), leads to interfacial void formation, resulting in defective device operation [21]. High-temperature annealing, surface preparation techniques [22] or outgassing-efficient structures at the bonding interface [23] are used to prevent void formation. However, in the presented study, no interfacial void formation has been observed when bonding at 150$^{\circ }$C due to the hydrophilic and porous properties of the thick layer of AZ-4562 photoresist (PR) and the prestructuring of the layers prior to transfer.
The quality of the transferred InGaP waveguides (WGs) has been tested via second-harmonic generation (SHG), the process is depicted in Fig. 1. In particular, the surface induced optical non-linearity has allowed the evaluation of the top surface and the bonding interface quality of the WG [24]. Considering a TM polarized fundamental mode propagating along the crystallographic69 axis, the surface- and bulk contribution to SHG process can be distinguished by polarization70 measurement of the generated light [25]. The generation of a TM polarized mode, in contrast with the TE polarized mode which is obtained from the bulk contribution, is allowed by exploiting the surface contribution to $\chi ^{(2)}$ [26]. Dispersion studies to achieve modal phase matching have been performed, allowing equiphase velocity of the pump and SHG mode. As a result, modal phase matching of the surface contribution to SHG is achieved. The generated signal spectra, in perfect agreement with dispersion studies, has been measured.
Fig. 1. Graphical representation of the surface induced SHG, where two pump photons (red) upconvert to a signal photon (blue). The propagation axis, z, also corresponds to the crystallographic axis of the structure. The interfaces that contribute to d$_S^{11}$ are highligthed in green.
2. Experimental
The WGs are structured on the top layer of GaAs lattice-matched epitaxially grown Ga$_{0.51}$In$_{0.49}$P and GaAs films on a (100) GaAs wafer. The epitaxially grown layers and its thicknesses are shown in Table 1. After etching the 20 nm GaAs cap in a H$_3$PO$_4$:H$_2$O$_2$:H$_2$O (1:1:20) mixture a 900 nm ma-N 2410 layer has been spin coated for later electron beam (EB) exposure and developement in MaD 525. The first 150 nm of the WGs have been etched in a dry manner with standard chlorine chemistry, shown in Table 2 with all the other used recipes. In order to obtain a smooth WG wall profile, the final 80 nm have been etched in a H$_3$PO$_4$:HCl (1:1) mixture. Finally, a 400 nm thick SX AR-N 8200.18 EB resist cladding with a width of 10 $\mu$m has been obtained after EB exposure and developement in AR-300-40. The patterned WGs, whose dimension are shown in Fig. 2(a), are then coated with a $\approx$ 12 $\mu$m layer of AZ 4562 PR and bonded to a 1.5 cm x 1.5 cm SiO$_2$ etching scaffold. The etching scaffold is prepared by dicing a 500 $\mu$m fused SiO$_2$ wafer for later spin-coating $\approx$ 8 $\mu$m of PMMA and $\approx$ 12 $\mu$m of AZ 4562 PR. After a complete substrate removal in a H$_3$PO$_4$:H$_2$O$_2$ (1:5) mixture, the InGaP and GaAs remaining layers are etched in H$_3$PO$_4$:HCl (1:1) and H$_3$PO$_4$:H$_2$O$_2$:H$_2$O (1:1:20), respectively. As a result, AZ 4562 PR suspended WGs are obtained, as shown in Fig. 2(b), for non-cladded WGs.
Fig. 2. Microscope images of 235 nm thick InGaP WGs: On the GaAs wafer with AR-N 8200.18 cladding (a) and on AZ-4562 PR after substrate back-etching (b). The fabrication outline to suspend the InGaP WGs is shown on top.
A 1.5 cm long and 2040 $\mu$m wide target substrate is prepared by evaporating a 250 nm thick Cr hard mask on a 500 $\mu$m (100) Si wafer after growth of a 4 $\mu$m thick wet thermal oxide. A standard SiO$_2$ dry etch in C$_4$F$_8$ chemistry followed by a wafer through Bosch process, whose etching parameters can be found in Table 2, have been used to structure the target substrate. Then, the PR suspended WGs and the target substrate are pressed together after removing the Cr mask and performing an oxygen plasma treatment to hydroxilaze the surfaces, greatly suppressing interfacial void the creation without decreasing the silicate covalent bonding strength [21]. The temperature is then increased to 150 $^{\circ }$C for 10 min with a rate of 0.1 $^{\circ }$C/s and decreased to room temperature at the same rate. As a result, a bonded stack, shown in Fig. 3(a) is obtained. The PMMA layer is dissolved in an acetone bath and the remaining heavily cured AZ 4562 PR is ashed in oxygen plasma. After a successful transfer, cladded and non-cladded WG arrays on 4 $\mu$m of SiO$_2$ are obtained as shown in Fig. 3(b) and Fig. 3(c), respectively.
Fig. 3. Microscope images of molecularly bonded InGaP WGs on wet thermal oxide. Before SiO$_2$ scaffold separation (a). After SiO$_2$ scaffold separation and O$_2$ plasma cleaning for non-cladded (b) and with SX AR-N 8200.18 cladding (c). The fabrication route to molecularly bond the InGaP WGs is shown on top.
In order to be able to in/out couple light, a lithography process using AZ 4562 PR and AZ 400K:H$_2$O (1:4) as developer is performed to expose the WG ends. Dry etching in a Cl$_2$:Ar (1:2) during 4 minutes at 10 mTorr with 75 W and 150 W of RF and ICP power has been performed followed by 4 $\mu$m of SiO$_2$ and 35 $\mu$m of Si dry etching. Finally, an isotropic Si etching using SF$_6$ has been performed for 30 s, suspending the final 10 $\mu$m of the WGs, as shown in Fig. 4(a). This chip finishing route yields double-inverse tappered WGs, as shown in Fig. 4(b) for a 3 $\mu$m wide WG transfered on 1 $\mu$m of SiO$_2$ for imaging purposes. The tappered end of a non-claded WG before the isotropic Si etching is shown in Fig. 4(c).
Fig. 4. Microscope image of SX AR-N 8200.18 cladded InGaP WG array transferred to wet thermal oxide after an isotropic Si etching process (a). SEM image of a 235 nm thick and 3 $\mu$m wide InGaP WG on 0.9 $\mu$m of wet thermal oxide (b). SEM image of a 235 nm thick InGaP WG with 4.65 $\mu$m width tappered to 800 nm on 4 $\mu$m of wet Si thermal oxide (c). The fabrication route to expose the WG ends is shown on top.
Finally, the WGs have been tested in transmission mode by exciting the fundamental TM mode with a SC400-X supercontinuum-generation broad-band source after blocking wavelengths below 1400 nm with a FELH1400 longpass filter and selecting the vertical polarization with a LPNIRA050-MP2 polarizer. An average power of 300 mW with 6 ps pulses and a 40 MHz repetition rate have been used for the measurements. The spectrum of the generated SHG light, is measured by blocking the pump with a FESH1000 shortpass filter and selecting the desired polarization with a LPNIRA050-MP2 polarizer. In concordance with the performed dispersion studies, a 2.6 nm FWHM peak with a central wavelength of 711 nm has been measured for cladded WGs with 1.22 $\mu$m width while the non-cladded WGs exhibit a 5.4 nm FWHM at a central wavelength of 703 nm, as shown in Fig. 5(a). Within the spectral region of interest the pulse peak power is 6.48 W and 2.12 W for the non-cladded and cladded WGs, respectively. The efficiency for the surface induced SHG process is 10$^{-7}$ and 10$^{-6}$ SHG photons per pump photon for the non-cladded and cladded WGs, respectively. A 5 ${\% }$ WG coupling efficiency, 70 ${\% }$ detection efficiency and 9.5 ${\% }$ transmission through the detection system has been measured. The coupling efficiency of the source to the optical fiber has been neglected to provide a lower bound estimation.
Fig. 5. Measured spectra of the generated SHG signal on 235 nm thick and 1.22 $\mu$m wide InGaP WGs on 4 $\mu$m of wet Si thermal oxide for non-cladded (blue) and 400 nm thick SX AR-N 8200.18 cladding (red) (a). Vertical polarization has been selected at the collection objective. AFM scan of the 1.22 $\mu$m wide WG measured (b) showing the substrate and WG top surface RMS roughness. Thickness profile of the WGs at the center of the substrate and 500 $\mu$m from the edges (c).
Table 1. Epitaxial GaAs-lattice-matched layers grown on a 500 $\mu$m thick GaAs (001) wafer.
Table 2. Dry etching parameters used for etching InGaP, SiO$_2$ and Si. The InGaP etching has been performed in a RIE Oxford Plasmalab 100 and the Si and silicates in a RIE Oxford PlasmaPro 100 Cobra 300 ICP.
3. Results and discussion
The previously described heterogeneous integration technique features several advantages when compared to micro-transfer printing or free-standing structures for the implementation of III-V WGs on silicates. Starting from the PR suspension process, by performing a substrate back-etching uniform contact with the etching solution is achieved, avoiding the characteristic convex profile obtained when etching a sacrificial release layer. Additionally, by greatly reducing the etching time of the sacrificial layer, which in our case is 60 s for 200 nm of GaAs, material combinations which do not exhibit large etching selectivity and would not withstand a conventional release layer etching through etching access holes have been successfully transferred, evidencing the versatility of technique. The design complexity and processing requirements are greatly reduced as no sacrificial etching access holes, tethers or hard masks are required, making the technique compatible with the transfer of non-cladded structures. Critical point drying or device anchors to prevent the structures collapse due to surface tension are not required either, minimizing design constraints and processing requirements.
However, a suitable PR combination has to be found. Both PR should be positive, due to the need of thermal cross-linking to achieve a robust stack in contrast with negative PR, which would require UV exposure. It is imperative that the solvents of both PR slightly dissolve the other PR to be able to achieve inter-cross-linking between both layers of the SiO$_2$ scaffold after curing the stack. The PR used as separation layer in the back-etching scaffold should not be heavily cured after the bonding, performed at 150 $^{\circ }$C, to be able to gently separate the target substrate from the etching scaffold. The direct mechanical separation of both samples after the bonding can lead to structure breakage, specially in zones where the features are smaller. The combination of PMMA and AZ 4562 has been found suitable for this purpose. The reflowing properties of AZ 4562 PR ensure a proper contact of the WGs and the target substrate during the transfer. Even when contamination particles are present within the bonding interface, the reflowing properties of AZ 4562 allow for conformal deformation around the particle, ensuring a proper contact between the sample and the substrate. Using the described route, particle contamination might lead to the deficient bonding of a single structure rather than a large interfacial void. However, less than a 3${\% }$ of the bonding surface contains structures, further minimizing the likelihood of unsuccessful structure bonding. Moreover, the hydrophilicity and permeability of the PR provide an efficient outgassing route of the water byproducts shown in Eq. (1).
Continuing with the silicate bonding, although the inherent poor quality of the native oxide of III-V’s native oxides made the use of interfacial oxide layers compulsory for wafer bonding [13], pre-structuring the layers allows for molecular bonding of the native oxide to silicates. The deposition of high density interfacial oxide layers to achieve large enough out-gassing efficiencies to adsorb the byproducts of Eq. (1) is optional. The subsequent polishing to achieve the required 0.5 nm root mean squared roughness for molecular bonding [27], is thus unnecessary, greatly reducing the processing time and cost while enabling to optimize the oxide thickness for device performance purposes if required. Furthermore, adhesive interfacial layers are not required, providing the full optical window of the materials used and being able to disregard polymer absorption, leading to thermal expansion, specially predominant at large pumping powers. The transfer of > 10 $\mu$m wide layers to fussed SiO$_2$ has been recently demonstrated [28]. Here, we present further developments enabling the transfer to thermal oxides which exhibit compressive stress, more porosity and less density than fussed SiO$_2$ [29], difficulting the bonding while enhancing the technique’s Si platform compatibility. As shown in Fig.3b thermal oxide layers of 0.9 $\mu$m are enough to achieve efficient outgassing. Consequently, post-processing of the transferred structures with accuracy limited by the lithographic process ($\approx$ 0.5 $\mu$m) in contrast with standard bonding tools where $\approx$ 5 $\mu$m can be achieved.
The transferred WGs have been tested in SHG process. By in-coupling a TM mode propagating along the crystallographic axis of the 235 nm thick InGaP WGs, shown in Fig. 5(c), two main components of the second-order nonlinear-polarization density have been excited [30]:
Fig. 6. Mode profiles of the pump mode (a) and the SHG mode (b) for non-cladded 235 nm thick and 1.22 $\mu$m InGaP WG on SiO$_2$. Chromatic dispersion of the TM fundamental (red) and first order (blue) modes in 235 nm thick and 1.22 $\mu$m wide InGaP WGs without (dashed) and with (solid) SX AR-N 8200.18 cladding (c). The expected PM point for 233 nm and 237 nm thickness is also shown.
The electric quadrupole contribution to surface SHG, arising from the strong gradient of the electric field at interfaces is justifying the choice of InGaP for implementing the aforementioned optical configuration as it exhibits a substantially larger refractive index when compared to traditional nonlinear materials, greatly enhancing the electric quadrupole contribution to the surface nonlinear coefficient as d$_{11}^s\;{\propto }[(\frac {\epsilon _{InGaP}}{\epsilon _{Subs}})^2-1]$ [32]. The chromatic dispersions of the fundamental and first order TM modes in 235 nm thick WGs of 1.22 $\mu$m width have been computed for non-cladded WGs and WGs with a 400 nm SX AR-N 8200.18 cladding. The InGaP and SiO$_2$ refractive index has been obtained from the data reported by Schubert et al. [33] and Paliks’ database, respectively. For modelling the SX AR-N 8200.18 refractive index, the Cauchy equation n($\lambda$)= n$_0$ + n$_1$/$\lambda ^2$ with coefficients n$_0$=1.461 and n$_1$=72 $\mu$m$^2$ reported in the product data-sheet have been used. The phase-matching condition for SHG in WGs is satisfied when the effective refractive index of the pump and the second-harmonic modes are equal. Such condition can be used to estimate the wavelength at which SHG is more efficient. As shown in Fig. 6(c), a TM$_0$ to TM$_1$ conversion at 701 nm and 707 nm is expected for non-cladded and cladded WGs, respectively. The expected wavelengths, in perfect agreement with the measurements shown in Fig. 5(a), evidence the SHG of light from the surface contribution to $\chi ^{(2)}$. The 4 nm miss-match between the predicted and measured SHG wavelengths for cladded WGs is mainly attributed to the refractive index overestimation of the Cauchy’s equation at the infrared, slightly increasing the effective refractive index of the fundamental TM mode at the pump wavelengths, hence leading to a slight underestimation of the modal matched wavelength.
4. Conclusion
An adhesive-free low-temperature heterogeneous integration technique allowing to experimentally verify the surface contribution to the optical nonlinearity is presented. The transfer of 2 mm long and 235 nm thick InGaP WGs with optional SX AR-N 8200.18 cladding on silicates via molecular bonding of the III-V native oxide and thermally grown SiO$_2$ has been experimentally demonstrated. The bulk and surface contribution to SHG are distinguished by means of the polarization of the generated signal. Modal-matched SHG arising from the surface contribution to $\chi ^{(2)}$ has been measured at 703 nm and 711 nm with FWHM of 5.4 nm and 2.6 nm, in perfect agreement with dispersion studies, for non-cladded and SX AR-N 8200.18 cladded WGs, respectively.
Funding
Wallenberg Center for Quantum Technology, Chalmers University of Technology; Vetenskapsrådet (2018-03457).
Acknowledgments
The authors acknowledge professor Anand Srinivassan and Esteban Toledo Carrillo for fruitful discussions to refine the fabrication process.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. D. Thomson, A. Zilkie, J. E. Bowers, et al., “Roadmap on silicon photonics,” J. Opt. 18(7), 073003 (2016). [CrossRef]
2. R. Jones, P. Doussiere, J. B. Driscoll, et al., “Heterogeneously integrated inpsilicon photonics: Fabricating fully functional transceivers,” IEEE Nanotechnol. Mag. 13(2), 17–26 (2019). [CrossRef]
3. D. Liang and J. E. Bowers, “Recent progress in heterogeneous iii-v-on-silicon photonic integration,” Light. Adv. Manuf. 2(1), 59 (2021). [CrossRef]
4. U. Gösele and Q.-Y. Tong, “Semiconductor wafer bonding,” Annu. Rev. Mater. Sci. 28(1), 215–241 (1998). [CrossRef]
5. B. Haq, S. Kumari, K. Van Gasse, et al., “Micro-transfer-printed iii-v-on-silicon c-band semiconductor optical amplifiers,” Laser Photonics Rev. 14(7), 1900364 (2020). [CrossRef]
6. A. Chopin, A. Barone, I. Ghorbel, et al., “Ultra-efficient generation of time-energy entangled photon pairs in an ingap photonic crystal cavity,” Commun. Phys. 6(1), 77 (2023). [CrossRef]
7. J. Zhang, G. Muliuk, J. Juvert, et al., “Iii-v-on-si photonic integrated circuits realized using micro-transfer-printing,” APL Photonics 4(11), 110803 (2019). [CrossRef]
8. A. D. Groote, P. Cardile, A. Z. Subramanian, et al., “Transfer-printing-based integration of single-mode waveguide-coupled iii-v-on-silicon broadband light emitters,” Opt. Express 24(13), 13754–13762 (2016). [CrossRef]
9. I. Shoji, T. Kondo, A. Kitamoto, et al., “Absolute scale of second-order nonlinear-optical coefficients,” J. Opt. Soc. Am. B 14(9), 2268–2294 (1997). [CrossRef]
10. D. E. Aspnes and A. A. Studna, “Dielectric functions and optical parameters of si, ge, gap, gaas, gasb, inp, inas, and insb from 1.5 to 6.0 ev,” Phys. Rev. B 27(2), 985–1009 (1983). [CrossRef]
11. E. J. Stanton, J. Chiles, N. Nader, et al., “Efficient second harmonic generation in nanophotonic gaas-on-insulator waveguides,” Opt. Express 28(7), 9521–9532 (2020). [CrossRef]
12. Q. Tong and U. Gösele, Semiconductor Wafer Bonding (John Wiley & Sons, New York, 1998).
13. T. Plach, K. Hingerl, S. Tollabimazraehno, et al., “Mechanisms for room temperature direct wafer bonding,” J. Appl. Phys. 113(9), 094905 (2013). [CrossRef]
14. D. Liang, A. W. Fang, H. Park, et al., “Low-temperature, strong sio2-sio2 covalent wafer bonding for iii–v compound semiconductors-to-silicon photonic integrated circuits,” J. Electron. Mater. 37(10), 1552–1559 (2008). [CrossRef]
15. D. Liang, G. Roelkens, R. Baets, et al., “Hybrid integrated platforms for silicon photonics,” Materials 3(3), 1782–1802 (2010). [CrossRef]
16. N. Poulvellarie, C. M. Arabi, C. Ciret, et al., “Efficient type ii second harmonic generation in an indium gallium phosphide on insulator wire waveguide aligned with a crystallographic axis,” Opt. Lett. 46(7), 1490–1493 (2021). [CrossRef]
17. W. Cheng, Z. Geng, Z. Yu, et al., “Wafer-scale inverted gallium phosphide-on-insulator rib waveguides for nonlinear photonics,” Opt. Lett. 48(14), 3781–3784 (2023). [CrossRef]
18. J.-S. Park, M. Tang, S. Chen, et al., “Heteroepitaxial growth of iii-v semiconductors on silicon,” Crystals 10(12), 1163 (2020). [CrossRef]
19. D. Pasquariello and K. Hjort, “Plasma-assisted inp-to-si low temperature wafer bonding,” IEEE J. Sel. Top. Quantum Electron. 8(1), 118–131 (2002). [CrossRef]
20. S. Oh, D. H. Jun, K. W. Shin, et al., “Control of crack formation for the fabrication of crack-free and self-isolated high-efficiency gallium arsenide photovoltaic cells on silicon substrate,” IEEE J. Photovoltaics 6(4), 1031–1035 (2016). [CrossRef]
21. Q.-Y. Tong, G. Fountain, and P. Enquist, “Room temperature sio2/sio2 covalent bonding,” Appl. Phys. Lett. 89(4), 042110 (2006). [CrossRef]
22. Y.-L. Chao, Q.-Y. Tong, T.-H. Lee, et al., “Ammonium hydroxide effect on low-temperature wafer bonding energy enhancement,” Electrochem. Solid-State Lett. 8(3), G74 (2005). [CrossRef]
23. D. Liang and J. E. Bowers, “Highly efficient vertical outgassing channels for low-temperature InP-to-silicon direct wafer bonding on the silicon-on-insulator substrate,” J. Vac. Sci. & Technol. B: Microelectron. Nanometer Struct. Process. Meas. Phenom. 26(4), 1560–1568 (2008). [CrossRef]
24. R. Sanatinia, M. Swillo, and S. Anand, “Surface second-harmonic generation from vertical gap nanopillars,” Nano Lett. 12(2), 820–826 (2012). PMID: 22214365. [CrossRef]
25. R. Sanatinia, S. Anand, and M. Swillo, “Modal engineering of second-harmonic generation in single gap nanopillars,” Nano Lett. 14(9), 5376–5381 (2014). PMID: 25157424. [CrossRef]
26. E. D. Luca, D. Visser, S. Anand, et al., “Gallium indium phosphide microstructures with suppressed photoluminescence for applications in nonlinear optics,” Opt. Lett. 44(21), 5117–5120 (2019). [CrossRef]
27. G. Roelkens, J. Van Campenhout, J. Brouckaert, et al., “Iii-v/si photonics by die-to-wafer bonding,” Mater. Today 10(7-8), 36–43 (2007). [CrossRef]
28. A. Peralta Amores and M. Swillo, “Low-temperature bonding of nanolayered ingap/sio2 waveguides for spontaneous-parametric down conversion,” ACS Appl. Nano Mater. 5(2), 2550–2557 (2022). [CrossRef]
29. W. A. Pliskin, “Comparison of properties of dielectric films deposited by various methods,” J. Vac. Sci. Technol. 14(5), 1064–1081 (1977). [CrossRef]
30. A. Yariv and P. Yeh, Photonics: Optical Electronics in Modern Communication (Oxford University Press; 6th edition, 2006).
31. E. D. Luca, R. Sanatinia, S. Anand, et al., “Focused ion beam milling of gallium phosphide nanostructures for photonic applications,” Opt. Mater. Express 6(2), 587–596 (2016). [CrossRef]
32. D. Epperlein, B. Dick, G. Marowsky, et al., “Second-harmonic generation in centro-symmetric media,” Appl. Phys. B 44(1), 5–10 (1987). [CrossRef]
33. M. Schubert, V. Gottschalch, C. M. Herzinger, et al., “Optical constants of GaxIn(1 x)P lattice matched to GaAs,” J. Appl. Phys. 77(7), 3416–3419 (1995). [CrossRef]
