Although intrinsic hydrogenated amorphous silicon (a-Si:H) wire waveguides clad with normal SiO2 layers have low propagation loss of 2.7 ± 0.1 dB/cm for transverse electric (TE) mode in the 1550-nm range, the transparency degrades when interfaced with other dielectrics (e.g., air) and/or exposed to elevated temperatures due to degradation of surface passivation in the a-Si:H waveguides. The thermal stability of a-Si:H wire waveguides with various cladding layers is systematically investigated, showing that the a-Si:H wire waveguides are stable at annealing temperature lower than ~350°C, while they degrade quickly when annealed at a higher temperature. It indicates that the thermal stability is mainly determined by the annealing temperature rather than the annealing time, which may be attributed to quick evolution of weakly bonded hydrogen in the a-Si:H waveguides. A thin Si3N4 intercladding layer between SiO2 cladding and a-Si:H waveguide core may degrade transparency due to N-H bond absorption and is of no benefit to the thermal stability, thus its overall effect on the a-Si:H waveguides is detrimental.
© 2012 OSA
Hydrogenated amorphous silicon (a-Si:H) has distinct advantages such as high refractive index, low absorption loss at telecommunication wavelengths of 1550 nm , capability of low-temperature (~200-400°C) plasma-enhanced chemical vapor deposition (PECVD) on almost any substrates , and even possibility for active modulation and detection [3–6]. Recently, it has emerged as an important material for integrated Si photonics, especially for backend integration of Si photonics above the complementary metal-oxide-semiconductor (CMOS) chip and/or monolithic integration of multilayer photonics and electronics on a same chip . A very low propagation loss of ~2-3 dB/cm at 1550 nm has been reported for a-Si:H wire waveguides [8–10], which is comparable to the crystalline Si counterparts with the same dimensions. However, large propagation loss has also been reported for a-Si:H wire waveguides, e.g., ~14 dB/cm in Ref . It is well known the low optical loss of a-Si:H waveguides is mainly resulted from sufficient passivation of dangling bonds in a-Si, while the degree of passivation may be affected by many factors such as deposition condition, cladding layer, and post-deposition process etc [11–13], thus resulting in large discrepancy for optical losses of a-Si:H wire waveguides reported in literature. The effect of deposition conditions such as RF power, gas flow, pressure, and substrate temperature, etc. on optical loss has been studied and the optimized deposition condition for low-loss a-Si:H waveguides has been obtained [10, 14]. In these previous studies, the a-Si:H films are deposited on a SiO2 surface and the waveguides are covered by a thick SiO2 cladding layer. However, in the manufacturing of Si integrated photonics, the a-Si:H waveguides may also be deposited on and/or covered by dielectrics other than SiO2, for example Si3N4, — which is a dielectric usually incorporated in the process flow for process control  and also is an important material for Si photonics . It has been reported that a thin Si3N4 intercladding layer may improve the transparency of a-Si:H wire waveguides because of its mediate refractive index between a-Si:H and SiO2 , however, this claim may be questionable because the control waveguide (i.e., the a-Si:H wire waveguide with normal SiO2 cladding) for comparison in that study has a large optical loss of 12 dB/cm. Moreover, for some applications such as optical sensing , it may need to leave the a-Si:H waveguide uncovered (i.e., with air cladding). It is expected that the different cladding layer may affect the surface passivation (thus transparency) of a-Si:H waveguides. However, it has not been reported yet. Another important issue for implementation of a-Si:H waveguides in Si photonics is their thermal stability during the subsequent processing steps such as additional photonic layer fabrication for multilayer photonics and/or contact annealing for active devices. It has been reported that a-Si:H waveguides will degrade substantially after annealing at temperatures larger than 400°C , — which is a typical temperature for backend processing in CMOS technology. Therefore, it is extremely important to know the temperature limitation of a-Si:H waveguides in order to estimate and/or minimize the degradation of a-Si:H waveguides during the subsequent processing steps.
In this paper, a-Si:H wire waveguides with various cladding layers (i.e., SiO2, Si3N4, and air) are fabricated, and rapid thermal annealing (RTA) with different temperature and time is carried out to simulate thermal process in CMOS backend processing steps. Their effect on the optical loss of a-Si:H waveguides is revealed, and is explained by degradation of surface passivation in a-Si:H waveguides at elevated temperatures.
8-inch Si wafers with 2-μm PECVD-SiO2 were used as substrates. A 220-nm a-Si:H film was deposited by PECVD in an Applied MaterialsTM parallel plate reactor using the optimized recipe as follows : SiH4 flow 100 sccm, N2 flow 1500 sccm, temperature 400°C, pressure 4.2 torr, and RF power 100 W. The deposition rate is ~1.58 nm/s. Wire waveguides with inverted taper structures at both input/output terminals were patterned by 248-nm deep UV lithography and dry etched down to the bottom SiO2 using a thin SiO2 layer as the hard mask. The fabrication details have been described elsewhere . Total 8 samples were fabricated with different bottom and upper cladding layers, as listed in Table 1 . Si3N4 and SiO2 layers (except S3 sample) are also deposited in the Applied MaterialsTM parallel plate reactor. The recipe for Si3N4 deposition is: SiH4 flow 110 sccm, NH3 flow 38 sccm, N2 flow 2500 sccm, temperature 400°C, pressure 4.2 torr, and RF power 410 W. The deposition rate is ~5 nm/s. The recipe for SiO2 deposition is: SiH4 flow 115 sccm, N2O flow 2000 sccm, temperature 400°C, pressure 4.2 torr, and RF power 270 W. The deposition rate is ~3.4 nm/s. For S3 sample, the upper cladding SiO2 is deposited at 150°C in a STS-CVD tool. Figure 1 shows cross sectional transmission electron microscopy (XTEM) images for S4, S6 and S8 samples. The a-Si:H waveguide core has almost rectangular profile and has no difference due to the different cladding layer. A thin Si3N4 layer covers the a-Si:H core conformally. The Si3N4 thicknesses indicated in Figs. 1(b) and 1(c) are measured from the enlarged XTEM images (not shown here). The sidewall Si3N4 is proportionally thinner than the top Si3N4. Other samples have similar cross section for a-Si:H waveguide cores, thus they are not shown here for simplicity. The surface roughness (root-mean-square (RMS) within 5-μm × 5-μm area) measured by atomic force microscope (AFM) is ~0.47 nm for the initial bottom SiO2 layer, ~1.90 nm after 220-nm a-Si film deposition, and ~1.83 nm after additional 20-nm Si3N4 deposition, as shown in Fig. 2 . It indicates that the thin Si3N4 deposition may smooth the a-Si film slightly.
After dicing, the chips were performed by RTA at N2 ambient with different temperature and time. In an isochronal RTA experiment, the annealing temperature ranges from 250°C to 500°C in 50°C increments while the annealing time keeps for 20 min. In an isothermal RTA experiment, the annealing time ranges from 20 s to 30 min while the annealing temperature keeps at 400°C or 500°C.
The conventional cutback method is used to extract the propagation loss. A set of 7 wire waveguides located on the same chip is measured at room temperature. The waveguide length ranges from 0.69 cm to 2.68 cm. Because the propagation loss of a-Si:H waveguides depends slightly on the waveguide dimensions and the input light, for simplicity, we fix the waveguide dimension to 220-nm (height) × 500-nm (width) and the input light to transverse electric (TE)-polarized mode in 1550 nm. The detailed measurement setup has been described elsewhere . Figure 3 shows the row measurement results for some samples. For waveguides with relatively low optical loss (e.g., S4, S6, and S8 in Fig. 2), the measured output power exhibits good linearity with the waveguide length, thus enabling linear fitting the experimental data points to extract the propagation loss and the coupling loss between the waveguide and the fiber. To minimize the experimental error, three chips are measured for each condition. For waveguides with propagation loss ranging from ~2–15 dB/cm, the coupling loss is in the range of ~3-5 dB/facet, and there is a rough tendency that the waveguide with larger propagation loss has a slightly larger coupling loss, probably because (1) the weak output light may have relatively large measurement error due to the leakage of light besides that transports through the waveguide and/or (2) the factors (such as insufficient passivation) that degrade the transparency of a-Si:H wire waveguides may also degrade their coupling loss. For waveguides with large loss (e.g., S1 and S2 in Fig. 2), the output power through the long waveguide is too weak to be measured (i.e., smaller than the detection limit of our system of ~-55 dBm). The propagation loss is then roughly estimated from the output powers measured on the short waveguides by a simple assumption that their coupling loss is only slightly larger than that extracted from low loss waveguides. Be noted that the propagation loss estimated by this method has a large experimental error.
3. Results and discussion
Table 1 lists the propagation losses of a-Si:H wire waveguides with various cladding layers. The a-Si:H wire waveguide with normal SiO2 cladding (S4) is used as reference, which has very loss propagation loss of 2.7 ± 0.1 dB/cm, similar to our previous report  and also close to the lowest propagation loss reported in literature [8, 9]. Firstly, we find that the a-Si:H waveguide without SiO2 cladding (S1 sample, air cladding) has a very large propagation loss of ~43 dB/cm. For comparison, the crystalline Si waveguide with air cladding has similar optical loss as that covered by SiO2 cladding. The effect of possible moisture absorption in the a-Si:H core during the dicing process could be ruled out because the propagation loss remains very large after annealing. We suspect that the as-fabricated a-Si:H waveguide may contain a large amount of dangling bonds at the surface, probably induced by the dry etching process, thus leading to the very large optical loss. During the subsequent SiO2 or Si3N4 deposition, these surface dangling bonds are passivated by active hydrogen ions dissociated from SiH4. The amount of passivation depends on time and temperature during the cladding layer deposition. For S4 sample, these surface dangling bonds are sufficiently passivated, thus leading to a very low propagation loss. For S2 sample, these surface dangling bonds are only partially passivated because the deposition time (for 10-nm Si3N4) is very short (i.e., ~2 s), thus its propagation loss is smaller than S1 while much larger than S4. For S3 sample, its large propagation loss cannot be attributed to the SiO2 absorption because the crystalline Si waveguide with such low-temperature SiO2 cladding remains low propagation loss as that with the normal SiO2 cladding. Therefore, the large loss of S3 sample is preferably attributed to the insufficient surface passivation in the a-Si:H waveguide due to the low SiO2 deposition temperature (i.e., 150°C).
Secondly, we find that a thin Si3N4 intercladding layer degrade the propagation loss as compared to that without the Si3N4 intercladding layer (i.e., S4 sample). For S5, S6, and S7 samples which have a thin S3N4 intercladding layer between upper cladding SiO2 and a-Si:H waveguide core, the propagation loss is ~5-7 dB/cm, close to that reported in Ref . for a-Si:H wire waveguides with an as-deposited Si3N4 intercladding layer. For S8 sample which has thin S3N4 intercladding layers both in upper and bottom interfaces, the propagation loss is even larger (~13.9 dB/cm). The large propagation loss of these samples cannot be attributed to the surface roughness of Si3N4 film as reported in Ref , but be preferably attributed to N-H bond absorption. It has been reported that the N-H concentration in PECVD Si3N4 can be reduced by optimization of Si3N4 deposition parameters  or using an in situ N2/Ar plasma treatment . It indicates that the thin Si3N4 intercladding layer has two completing effects on the optical loss: increasing the loss due to N-H absorption and decreasing the loss due to its mediate refractive index between a-Si:H and SiO2 (which can reduce the waveguide roughness induced optical loss). For samples reported in Ref , the N-H absorption is reduced by the in situ N2/Ar plasma treatment and the reference a-Si:H waveguides have large loss probably due to large sidewall roughness, thus the overall effect of the thin Si3N4 intercladding layer is beneficial. In contrast, for our samples, the reference a-Si:H waveguide (i.e., S4) already has a very low propagation loss (it implies that the waveguide sidewall of our samples is quite smooth) and the thin Si3N4 film is as-deposited, thus the overall effect of the thin Si3N4 intercladding layer is detrimental. S7 sample has similar optical loss as S6 sample (within the measurement error) and is only slightly larger than S5 sample. The very weak dependence of optical loss on the intercladding Si3N4 thickness may also be attributed to the above two completing effects.
Figure 4 plots propagation losses for various a-Si:H wire waveguides after isochronal RTA in N2 ambient for 20 min. The annealing temperature ranges from 250°C to 550°C. The data for S1 and S2 samples are not included because their propagation losses are too large to be accurately measured. We can see in Fig. 4 that all samples exhibits similar thermal behavior: the propagation loss keeps almost unchanged after annealing at temperature <350°C, degrades substantially at 400°C, and becomes very larger after annealing at temperature >450°C. It indicates that the annealing temperature is a key factor that determines the thermal stability of a-Si:H waveguides.
Figure 5 plots propagation losses for various a-Si:H wire waveguides after isothermal RTA at 400°C. The annealing time ranges from 20 s to 30 min. All samples exhibit similar lifetime behavior: the propagation loss increases substantially after RTA for 20 s, then keeps almost constant over a long annealing time, and then increases slowly with annealing time further increasing. For 500°C RTA, the propagation loss becomes very large (>20 dB/cm) after annealing for 20 s for all a-Si:H wire waveguides, as shown in Fig. 6 . It indicates that the degradation of a-Si waveguides occurs within a very short period of time during annealing, namely, the annealing time plays a minor role to the thermal stability of a-Si:H waveguides.
The thermal stability of a-Si:H films for applications such as solar cells and thin-film transistors has been extensively investigated for decades [11–13], and is usually related with degradation of passivation at elevated temperatures. Concentrations of Si-H vibration bonds and dangling bond in a-Si films as a function of annealing temperature have been reported in many published papers . The dynamics of passivation and de-passivation of dangling bonds in a-Si:H films depends on many factors such as initial a-Si:H condition and it is very complex. For example, it has been reported that a-Si:H film may contain weakly bonded hydrogen which will evaluate at a relatively low temperature (e.g., ~365°C) and strongly bonded hydrogen which will evaluate at a relatively high temperature (e.g., ~696°C) . The thermal behavior of a-Si:H wire waveguides observed in this work can be explained within the same framework. Some key points are highlighted as follows. Firstly, the a-Si:H wire waveguides are stable at low temperature over very long period of time. Some samples are re-measured after storage at room temperature over two years or after annealing at 300°C for tens hours, no considerable change in propagation loss is observed. Secondly, a-Si:H waveguides degrade very quickly when annealed at temperature > ~450°C. Ref . reports that the density of Si-H bonds starts to decrease significantly at this annealing temperature. It indicates that the level of degradation is mainly determined by the annealing temperature rather than the annealing time, simply because the quick evolution of weekly bonded hydrogen during initial annealing already makes the a-Si:H waveguide too lossy (e.g., >20 dB/cm) to be useful. Thirdly, the critical temperature for a-Si:H waveguide thermal stability is ~400°C. At this temperature, although the propagation loss increases from 2.7 ± 0.1 dB/cm to 4.9 ± 0.2 dB/cm after 20-s annealing, and to 6.1 ± 0.2 dB/cm after 20-min annealing, the optical loss in this level is still acceptable for many applications. Therefore, to maintain the optical loss of a-Si:H wire waveguides low enough for applications, the temperature for further processing steps should be preferably less than 400°C, while there is no limit for the time of the further processing steps. Fourthly, all samples studied in this work exhibit similar thermal behavior, implying that a thin Si3N4 intercladding layer between SiO2 cladding and a-Si:H waveguide core cannot improve the thermal stability of a-Si:H waveguides. A possible approach for thermal stability improvement is to try to reduce the density of weakly bonded hydrogen (increase the density of strongly bonded hydrogen simultaneously) by optimizing a-Si:H waveguide fabrication processes. Alternatively, low-temperature forming gas annealing may partially re-passivate the dangling bonds formed during the high-temperature thermal treatment, thus improving the thermal stability of a-Si:H waveguides apparently.
The intrinsic a-Si:H wire waveguide with normal SiO2 cladding provides a very low propagation loss of 2.7 dB/cm and is stable at temperature less than ~350°C, thus it is a suitable waveguide material for backend integration above the CMOS chip. In this work, we have shown that the SiO2 cladding plays a key role for low optical loss because it provides sufficient surface passivation during deposition. The thermal stability of a-Si:H wire waveguides is strongly dependent on the annealing temperature rather than the annealing time, and is independent on the different cladding layer. In order to maintain low optical loss of a-Si:H waveguides, it is recommended that the temperature for further processing steps should not exceed 400°C, while the time for further processing steps can be relatively long.
This work was supported by Singapore A*STAR Infuse Exploratory Grant I02-0331-12.
References and links
1. A. Harke, M. Krause, and J. Mueller, “Low-loss single mode amorphous silicon waveguides,” Electron. Lett. 41(25), 1377–1379 (2005). [CrossRef]
2. R. A. Street, Hydrogenated Amorphous Silicon (Cambridge University Press, 1991).
4. Y. Shoji, T. Ogasawara, T. Kamei, Y. Sakakibara, S. Suda, K. Kintaka, H. Kawashima, M. Okano, T. Hasama, H. Ishikawa, and M. Mori, “Ultrafast nonlinear effects in hydrogenated amorphous silicon wire waveguide,” Opt. Express 18(6), 5668–5673 (2010). [CrossRef] [PubMed]
5. F. G. Della Corte, S. Rao, G. Coppola, and C. Summonte, “Electro-optical modulation at 1550 nm in an as-deposited hydrogenated amorphous silicon p-i-n waveguiding device,” Opt. Express 19(4), 2941–2951 (2011). [CrossRef] [PubMed]
7. A. Biberman, K. Preston, G. Hendry, N. Sherwood-Droz, J. Chan, and K. Bergman, “Photonic network-on-chip architectures using multilayer deposited silicon materials for high-performance chip multiprocessors,” ACM J. on Emerging Technologies in Computing Systems 7(2), DOI 10.1145 (2011).
8. R. Sun, K. McComber, J. Cheng, D. K. Sparacin, M. Beals, J. Michel, and L. C. Kimerling, “Transparent amorphous silicon channel waveguides with silicon nitride intercladding layer,” Appl. Phys. Lett. 94(14), 141108 (2009). [CrossRef]
9. R. Sun, J. Cheng, J. Michel, and L. Kimerling, “Transparent amorphous silicon channel waveguides and high-Q resonators using a damascene process,” Opt. Lett. 34(15), 2378–2380 (2009). [CrossRef] [PubMed]
10. S. Y. Zhu, G. Q. Lo, and D. L. Kwong, “Low-loss amorphous silicon wire waveguide for integrated photonics: effect of fabrication process and the thermal stability,” Opt. Express 18(24), 25283–25291 (2010). [CrossRef] [PubMed]
11. P. K. Lim, W. K. Tam, L. F. Yeung, and F. M. Lam, “Effect of hydrogen on dangling bond in a-Si thin film,” J. of Phys.: Conference Series 61, 708–712 (2007). [CrossRef]
12. T. A. Li, F. W. Chen, A. Cuevas, and J. E. Cotter, “Thermal stability of microwave PECVD hydrogenated amorphous silicon as surface passivation for n-type heterojunction solar cells,” European Photovoltaic Solar Energy Conference 2007, ed. Conference Program Committee, WIP-Renewable Energies, Germany, 1326–1331 (2007).
13. C. J. Arendse, D. Knoesen, and D. T. Britton, “Thermal stability of hot-wire deposited amorphous silicon,” Thin Solid Films 501(1-2), 92–94 (2006). [CrossRef]
14. S. K. Selvaraja, E. Sleeckx, M. Schaekers, W. Bogaerts, D. V. Thourhout, P. Dumon, and R. Baets, “Low-loss amorphous silicon-on-insulator technology for photonic integrated circuitry,” Opt. Commun. 282(9), 1767–1770 (2009). [CrossRef]
15. S. Y. Zhu, G. Q. Lo, and D. L. Kwong, “Components for silicon plasmonic nanocircuits based on horizontal Cu-SiO₂-Si-SiO₂-Cu nanoplasmonic waveguides,” Opt. Express 20(6), 5867–5881 (2012). [CrossRef] [PubMed]
16. J. Song, Y. Z. Li, X. Zhou, and X. Li, “A highly sensitive optical sensor design by integrating a circular-hole defect with an etched diffraction grating spectrometer on an amorphous-silicon photonic chip,” IEEE Photon. J. 4(2), 317–326 (2012). [CrossRef]
17. J. Kang, Y. Atsumi, M. Oda, T. Amemiya, N. Nishiyama, and S. Arai, “Low-loss amorphous silicon multilayer waveguides vertically stacked on silicon-on-insulator substrate,” Jpn. J. Appl. Phys. 50, 120208 (2011). [CrossRef]
18. S. C. Mao, S. H. Tao, Y. L. Xu, X. W. Sun, M. B. Yu, G. Q. Lo, and D. L. Kwong, “Low propagation loss SiN optical waveguide prepared by optimal low-hydrogen module,” Opt. Express 16(25), 20809–20816 (2008). [CrossRef] [PubMed]